-
We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.
This will never be settled! Do you use an all SCSI or all IDE based sytem and why?
Page 13 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
I go all IDE cause I was raised using IDE 😛
And I don't want to spend that much more money to go SCSI, which does not give so much bang for the buck! SCSI = Expensive, but fast. IDE = Cheap and Cheerful, and ATA100 when everything is optimised is not far off SCSI I think?! 😕
Albert.
And I don't want to spend that much more money to go SCSI, which does not give so much bang for the buck! SCSI = Expensive, but fast. IDE = Cheap and Cheerful, and ATA100 when everything is optimised is not far off SCSI I think?! 😕
Albert.
"Then why is it that I'm perfectly satisfied with an IDE setup in Windows 2000?"
Who said you shouldn't be? Everyone has different needs. If you're happy with your IDE system, stick with it, no none said you need to be using SCSI.
"I've messed around with SCSI systems"
That's real descriptive, what's that supposed to mean?
"but what are you doing that makes the extra cost of SCSI not matter to you?"
If you can't notice the difference in system responsiveness when any disk is necessary, you're not a SCSI candidate.
Who said you shouldn't be? Everyone has different needs. If you're happy with your IDE system, stick with it, no none said you need to be using SCSI.
"I've messed around with SCSI systems"
That's real descriptive, what's that supposed to mean?
"but what are you doing that makes the extra cost of SCSI not matter to you?"
If you can't notice the difference in system responsiveness when any disk is necessary, you're not a SCSI candidate.
I presently have a hybrid setup, IDE drives and CD-burner(TDK VeloCD) with SCSI CD-roms and DVD. I also have a DEC(Digital) SCSI-2 1.6 gig. harddrive that has been with me since 1994. Its noisy, runs hot and is only 5400rpm with 512k cache memory. But its still running.
Some people drives Porshe for a living, while some for travelling. They take a laugh at Toyota because they are less safe, accelerate like a turtle, won't go above 200MPH, too toy-like and too dirt cheap. Toyota is probably good for their spare car.
Those who drive Toyota say it is plain stupid to paid 400% more just for 20% better acceleration and a Porshe won't go any faster in urban. It is more than enough to drive them work and their kids school. Though they agree that Porshe does attract girls.
Those who drive Toyota say it is plain stupid to paid 400% more just for 20% better acceleration and a Porshe won't go any faster in urban. It is more than enough to drive them work and their kids school. Though they agree that Porshe does attract girls.
I think the best + fairest approach the VALUE question is:
quantum atlas 10K II vs ibm 75gxp .. using benchmarked (not manufacturers spec) access times: 8.1ms vs 12.2ms.
(storagereview says seek is the most important metric .. but, in real life, *access* is the thing we actually use/experience.)
these numbers differ by (exactly) a factor of 50% (altho some benches show up a much higher percentage, especially when using the benchmark favored by SR - Intel's IOMeter).
note that if we used calc'ed *seek* times (from benched access'es), we'd get numbers even *more* favorable to SCSI:
5.1 (manufacturer spec 4.7) vs 8.0 (manu spec 8.5) .. yields a slightly better 57% improvement.
using Modus' own cost for the ide drive, we have $100 .. & $144 for the 9GB quantum (with THIS rebate (valid 'til end of March) .. which is more than enuf room to run the OS, apps & swap/page file .. (where it'll do the most good) .. that's a diff of 44% (not very 'astronomical' or 'phenomenal', is it?) .. based on drives prices HERE.
so we have 50-57% better perf a 44% premium in cost .. even using Modus' mythical *linear* model (perf/price models are based on *curves*, they're rarely linear, as any first-year Economics major will tell you) .. it wins .. by between 6 & 13%. I don't care what Modus says .. in my book, that's called a bargain.
but you still need a controller.
altho you can get by with cheaper controller, (like the tekram dc-390u2b, for $90, which Modus suggested) i prefer the dc-390u2w (~us$125) cuz it allows you to run non-lvd devices (like burners, scanners, zip & cd-rom drives) without degrading perf of devices on the u2w/lvd bus (hdd's) ..
.. and, since no single drive yet comes *close* to saturating the (80mb/s) u2w bus, it's nowhere close to being a bottleneck .. for the foreseeable future, anyway (this even includes the recently announced, but still unavailable 15krpm drive from ibm - 36z15). both pariah & myself have this card. i've personally used it with 7 or 8 diff OS'es (great compatibility).
the u2b cannot run non-lvd devices without degrading perf of your hdd's, which is why i don't like it, cuz i have a scsi burner & cd-rom, but for sake of arguing hdd perf VALUE, it would be valid to use the $90 figure. that card will run the X15, or any other 15krpm drive without being the bottleneck for either access'es or sustained x-fers.
speaking of 15krpm drives .. another thing scsi buys you .. is future upgradability .. when prices of 15krpm drives come down, you'll be able to slip into one of those badboys for cheap .. and they've been dropping steadily .. altho an 18-ggiger still costs ~$400, and no 9-gig is available .. which, with a 6.5ms ACCESS, would make our price/perf debate more interesting .. but it won't be long until they're close to the current prices of 10krpm drives.
so what does the scsi card buy you, except for future upgradability?
* better irq management (1 for 15 devices)
* the ability to add devices other than hdd's & optical drives
* the ability to have more than a combined number of 4 hdd's & optical drives without add-on card$ or additional irq's.
This is prolly of no importance to the person who has only 1 hdd, but it's *definitely* of significance to the person who has more than 2 .. tho, admittedly, this is not a large # of ppl, which is one of the reasons why we say, <disclaimer|caveat> SCSI is not for everyone </disclaimer|caveat> ..
* the ability to use more reliable hard drives to run your system & apps .. this is a significant factor, FOR ME
* the biggest reason, tho, is a wonderful sense of *responsiveness* that is impossible to describe if you've never experienced it
it what makes the entire computing experience more enjoyable. Modus claims this is nothing more than perf perceived .. and, to a point, he's correct .. but, in it's perceived state, it's nevertheless the thing that does it for so many scsi users.
What I'm saying is that, we don't *think* about seek or access times when we're using our PCs. FOR ME, it was a *dramatic* difference .. so much so that the first thought upgrading (after saying, 'wow') was that i shoulda done it sooner (i had the same thought when i got my cable modem).
note that i'm not *selling* scsi .. i don't benefit in any financial way from promoting scsi (altho i should 🙂 ). my only reward is when someone sends me a note saying, 'dang, you were right.'
it's also widely discussed that current benchmarks - even the best ones, like intels' IOMeter - do not adequately/accurately test the true capabilities of a scsi drive or the scsi interface .. which is why so many users are so surprised at how much better their system responds/performs with the upgrade. Unfortunately, there's no way to prove this (that i know of).
so, the person who sees VALUE in those advanatges, will likely feel that the cost was worth it .. while the person whop doesn't, may not. That's the VALUE issue of SCSI in a nutshell. Obviously, there's much more. But also note that I used retail numbers .. you can get even better deals by purchasing USED drives & adapters.
so to answer NFS4's Q, I'm just asking what is it that you do that makes SCSI so much better .. it's kinda like the Q, what's it like to fall in love .. it's something u have to experience for yourself. 🙂
quantum atlas 10K II vs ibm 75gxp .. using benchmarked (not manufacturers spec) access times: 8.1ms vs 12.2ms.
(storagereview says seek is the most important metric .. but, in real life, *access* is the thing we actually use/experience.)
these numbers differ by (exactly) a factor of 50% (altho some benches show up a much higher percentage, especially when using the benchmark favored by SR - Intel's IOMeter).
note that if we used calc'ed *seek* times (from benched access'es), we'd get numbers even *more* favorable to SCSI:
5.1 (manufacturer spec 4.7) vs 8.0 (manu spec 8.5) .. yields a slightly better 57% improvement.
using Modus' own cost for the ide drive, we have $100 .. & $144 for the 9GB quantum (with THIS rebate (valid 'til end of March) .. which is more than enuf room to run the OS, apps & swap/page file .. (where it'll do the most good) .. that's a diff of 44% (not very 'astronomical' or 'phenomenal', is it?) .. based on drives prices HERE.
so we have 50-57% better perf a 44% premium in cost .. even using Modus' mythical *linear* model (perf/price models are based on *curves*, they're rarely linear, as any first-year Economics major will tell you) .. it wins .. by between 6 & 13%. I don't care what Modus says .. in my book, that's called a bargain.
but you still need a controller.
altho you can get by with cheaper controller, (like the tekram dc-390u2b, for $90, which Modus suggested) i prefer the dc-390u2w (~us$125) cuz it allows you to run non-lvd devices (like burners, scanners, zip & cd-rom drives) without degrading perf of devices on the u2w/lvd bus (hdd's) ..
.. and, since no single drive yet comes *close* to saturating the (80mb/s) u2w bus, it's nowhere close to being a bottleneck .. for the foreseeable future, anyway (this even includes the recently announced, but still unavailable 15krpm drive from ibm - 36z15). both pariah & myself have this card. i've personally used it with 7 or 8 diff OS'es (great compatibility).
the u2b cannot run non-lvd devices without degrading perf of your hdd's, which is why i don't like it, cuz i have a scsi burner & cd-rom, but for sake of arguing hdd perf VALUE, it would be valid to use the $90 figure. that card will run the X15, or any other 15krpm drive without being the bottleneck for either access'es or sustained x-fers.
speaking of 15krpm drives .. another thing scsi buys you .. is future upgradability .. when prices of 15krpm drives come down, you'll be able to slip into one of those badboys for cheap .. and they've been dropping steadily .. altho an 18-ggiger still costs ~$400, and no 9-gig is available .. which, with a 6.5ms ACCESS, would make our price/perf debate more interesting .. but it won't be long until they're close to the current prices of 10krpm drives.
so what does the scsi card buy you, except for future upgradability?
* better irq management (1 for 15 devices)
* the ability to add devices other than hdd's & optical drives
* the ability to have more than a combined number of 4 hdd's & optical drives without add-on card$ or additional irq's.
This is prolly of no importance to the person who has only 1 hdd, but it's *definitely* of significance to the person who has more than 2 .. tho, admittedly, this is not a large # of ppl, which is one of the reasons why we say, <disclaimer|caveat> SCSI is not for everyone </disclaimer|caveat> ..
* the ability to use more reliable hard drives to run your system & apps .. this is a significant factor, FOR ME
* the biggest reason, tho, is a wonderful sense of *responsiveness* that is impossible to describe if you've never experienced it
it what makes the entire computing experience more enjoyable. Modus claims this is nothing more than perf perceived .. and, to a point, he's correct .. but, in it's perceived state, it's nevertheless the thing that does it for so many scsi users.
What I'm saying is that, we don't *think* about seek or access times when we're using our PCs. FOR ME, it was a *dramatic* difference .. so much so that the first thought upgrading (after saying, 'wow') was that i shoulda done it sooner (i had the same thought when i got my cable modem).
note that i'm not *selling* scsi .. i don't benefit in any financial way from promoting scsi (altho i should 🙂 ). my only reward is when someone sends me a note saying, 'dang, you were right.'
it's also widely discussed that current benchmarks - even the best ones, like intels' IOMeter - do not adequately/accurately test the true capabilities of a scsi drive or the scsi interface .. which is why so many users are so surprised at how much better their system responds/performs with the upgrade. Unfortunately, there's no way to prove this (that i know of).
so, the person who sees VALUE in those advanatges, will likely feel that the cost was worth it .. while the person whop doesn't, may not. That's the VALUE issue of SCSI in a nutshell. Obviously, there's much more. But also note that I used retail numbers .. you can get even better deals by purchasing USED drives & adapters.
so to answer NFS4's Q, I'm just asking what is it that you do that makes SCSI so much better .. it's kinda like the Q, what's it like to fall in love .. it's something u have to experience for yourself. 🙂
Hard_Boiled
Golden Member
Radboy, I think you're referring to the Atlas 10K for $169, but there is no rebate for that one. Also, does that coupon work anywhere or just at the Quantum authorized stores? Still, I think $169 is a good price for that drive, I own the 10K and it is real good. I don't really care about all these numbers of the optimal price/performance junk. Excluding networking devices, the hard drive is the slowest component in a computer(and I know floppy drives and cdroms are slower, but they arent constanly in use). It just seems that having a really fast hard drive should be desirable. Like you've been saying the whole thread a small 10K boot drive is great. Hell I've only got 1 small drive because thats all I need. When I need more space I'll go buy a 75GXP and throw junk on there. But it seems that for a person who only has IDE at the moment, unless they find some good deals, it costs at least $250 to get a good 10K drive and adapter. I paid more than that over a year ago for my Atlas 10K alone, so it doesn't seem that bad to me.
Jeez,
Radboy I don't really understand you, Modus has responded to your seek time and "wow affect" posts about a DOZEN times now and yet you keep coming back with "WHAT ABOUT THE SEEK TIME!". Are you even reading his posts?
Also, It still seems like the majority of the SCSI people (or just a select few) still don't get it because they keep posting responses comparing SCSI to expensive cars. Which, of course, assumes that EIDE users either can't afford nice cars or don't want them. What you don't seem to understand is that we (or just I) think our EIDE setups compareable to a nice car in such analogies. I don't think just because I have an EIDE setup that it's like driving a dodge colt and I can't afford anything better. If you want a pretty car analogy try this one: It's like having 2 cars that perform almost exactly the same, except some people decided to take the time to find out they performed similar before leaping on the one that costed more assuming it was that much better. Because they paid more they like to claim the wow factor and point at the name brand and smile. I am not criticizing, as I like to do the same sometimes. Like when you go to CVS and you know the CVS brand of everything is exactly the same as the name brands, yet you still buy the name brand because it helps you sleep better at night.
I am not saying scsi and eide are exactly alike, but in most home setups it will seem as such. I still have a hard time believing most people are doing true multitasking with their scsi setups, as only 1 person has stepped up and reported what they were actually running. The only reasons for scsi in my mind are multitasking, hot swapability, and massive raid setups. Take those things away from scsi and you've just got a high end EIDE drive. What is even more interesting is that my work is starting to look at EIDE drives in networkable storage units for some of their storage...
Note, I understand that a SCSI drive bought 3 years ago would have been a great investment. I would have done the same had I not been buying the cheapest damn EIDE's back then. There will always be exceptions.
Cya,
skace
Radboy I don't really understand you, Modus has responded to your seek time and "wow affect" posts about a DOZEN times now and yet you keep coming back with "WHAT ABOUT THE SEEK TIME!". Are you even reading his posts?
Also, It still seems like the majority of the SCSI people (or just a select few) still don't get it because they keep posting responses comparing SCSI to expensive cars. Which, of course, assumes that EIDE users either can't afford nice cars or don't want them. What you don't seem to understand is that we (or just I) think our EIDE setups compareable to a nice car in such analogies. I don't think just because I have an EIDE setup that it's like driving a dodge colt and I can't afford anything better. If you want a pretty car analogy try this one: It's like having 2 cars that perform almost exactly the same, except some people decided to take the time to find out they performed similar before leaping on the one that costed more assuming it was that much better. Because they paid more they like to claim the wow factor and point at the name brand and smile. I am not criticizing, as I like to do the same sometimes. Like when you go to CVS and you know the CVS brand of everything is exactly the same as the name brands, yet you still buy the name brand because it helps you sleep better at night.
I am not saying scsi and eide are exactly alike, but in most home setups it will seem as such. I still have a hard time believing most people are doing true multitasking with their scsi setups, as only 1 person has stepped up and reported what they were actually running. The only reasons for scsi in my mind are multitasking, hot swapability, and massive raid setups. Take those things away from scsi and you've just got a high end EIDE drive. What is even more interesting is that my work is starting to look at EIDE drives in networkable storage units for some of their storage...
Note, I understand that a SCSI drive bought 3 years ago would have been a great investment. I would have done the same had I not been buying the cheapest damn EIDE's back then. There will always be exceptions.
Cya,
skace
Sir Fredrick
Guest
>>>>> Modus has responded to your seek time and "wow affect" posts about a DOZEN times now and yet you keep coming back with "WHAT ABOUT THE SEEK TIME!". Are you even reading his posts?<<<<<
Actually, I've been wondering if Modus has been reading our posts, he doesn't respond to any of our points (or he at least conveniently ignores the best ones), and just repeats what he's been saying all along.
>>>> It's like having 2 cars that perform almost exactly the same, <<<<
Except one has significantly better handling (lower seek times), and faster speeds. Really though I agree, the car analogy is pretty dumb.
>>>>> Because they paid more they like to claim the wow factor and point at the name brand and smile. <<<<<
We have pointed to benchmarks and storage review time and time again, but it gets no response.
>>>>>I am not saying scsi and eide are exactly alike, but in most home setups it will seem as such.<<<<<
If a user can tell the diff between 128MB and 256MB of RAM, then they will feel a difference with SCSI.
Actually, I've been wondering if Modus has been reading our posts, he doesn't respond to any of our points (or he at least conveniently ignores the best ones), and just repeats what he's been saying all along.
>>>> It's like having 2 cars that perform almost exactly the same, <<<<
Except one has significantly better handling (lower seek times), and faster speeds. Really though I agree, the car analogy is pretty dumb.
>>>>> Because they paid more they like to claim the wow factor and point at the name brand and smile. <<<<<
We have pointed to benchmarks and storage review time and time again, but it gets no response.
>>>>>I am not saying scsi and eide are exactly alike, but in most home setups it will seem as such.<<<<<
If a user can tell the diff between 128MB and 256MB of RAM, then they will feel a difference with SCSI.
Sir Fredrick,
I thought Modus addressed the issues, but maybe I am wrong. This is what I thought until you posted your last reply:
A) The "Wow Factor" is strictly emotional and CANNOT be used as a test of whether something is better or not. Anytime you buy something new you get a wow factor, whether or not it is truly better.
B) The seek times didn't ammount to jack in real world tests, as the benchmarks Modus pasted proved.
Maybe I misread something, but those are the two things I keep seeing come up.
Also, the memory thing barely applies here, as 128 megs is the bare minimum for win2k. EIDE is not the 'bare minimum'. I guess the only comparison would be ram with ECC and ram without. An added feature that might be useful in certain situations. Can you tell which one has the ECC?
Cya,
skace
I thought Modus addressed the issues, but maybe I am wrong. This is what I thought until you posted your last reply:
A) The "Wow Factor" is strictly emotional and CANNOT be used as a test of whether something is better or not. Anytime you buy something new you get a wow factor, whether or not it is truly better.
B) The seek times didn't ammount to jack in real world tests, as the benchmarks Modus pasted proved.
Maybe I misread something, but those are the two things I keep seeing come up.
Also, the memory thing barely applies here, as 128 megs is the bare minimum for win2k. EIDE is not the 'bare minimum'. I guess the only comparison would be ram with ECC and ram without. An added feature that might be useful in certain situations. Can you tell which one has the ECC?
Cya,
skace
damn it. i can't keep my mouth shut.
<< Well yes, access times are the most important factor, but by no means the only factor. STR still accounts for something, as shown by WB99 and other real world tests. I used seek times earlier because of the confidence I have in SCSI's laughable inability to show enough of a performance advantage. If you want to use seek times and throw out every other benchmark (I can understand this as it's the most SCSI-friendly strategy), it's not really fair but I'll allow for the time being because I know SCSI still won't have a chance. >>
laughable? a 50% difference in seek time is not laughable. str doesn't account for much at all too. look at the eide raid 0 setups that are trounced by single disk solutions in the eide camp. higher seek times, but much higher str's than single disk solutions, so seek time is by far more important than str if you are NOT doing video editing which is mostly str dependent. and the majority of us do not do video editing. some eide raid setups are trounced by single disks capable of about 1/2 of the str. scsi vs. eide is exactly like raid 0 vs non-raid, wth do you need raid 0 for? it's more expensive, the performance is proven to be lower than single disk solutions in most cases. but many people insist that it "feels" fast, when in fact it isn't. compare this to scsi though, and you have something that is proven to be fast, not something that "feels" fast. there's also the issue that some of us prefer quality over quantity.
<< "Have you ever actually owned a SCSI drive or controller?"
Absolutely. But that's irrelevant. I don't need to own a GeForce2 Ultra to know its exact price/performance ratio and to know that it is an utter waste of money. >>
no modus, this isn't irrelevant. again, you have yet to disclose to us what scsi system you have owned, and for how long. again, you make an invalid comparison. a geforce 2 ulra's speed can be quantified purely on scores in games because that game is the real world application that provides some means to measure the performance of the graphics card, and at the same time is the actual application one would be using. benchmarks for hard drives are not balanced as well as a timedemo in quake 3 to demonstrate fillrate etc. because most benchmarks are synthetic. want to really run a real world demonstration of hard disk speed? go to adobe photoshop and limit the memory it uses to 10 megs. open up a 50-100 meg image, apply a couple of filters back to back to get some good swapping going on an identical setup with the only changes being a scsi and eide hard drive. guaranteed the scsi hard disk will run circles around the eide.
<< I am not saying scsi and eide are exactly alike, but in most home setups it will seem as such. I still have a hard time believing most people are doing true multitasking with their scsi setups, as only 1 person has stepped up and reported what they were actually running. >>
i run photoshop and open images and edit them that usually maxes out my ram. since most photoshop filters aren't smp'd, my second cpu allows me to do other things while applying a filter/resizing/etc. since i've limited photoshop to take all but enough memory for me to do rudimentary things like email, mp3's, this forum, and since swapping is going on my second hard drive, i can still run win2k on my os drive just fine. and i capture video occasionally, although not as often as i use to a year or so ago because of time constraints. on top of this, sometimes i'm burning to my 8x hp, playing mp3's, and serving files off of the same hard drive without a hiccup. my justification for scsi. a couple of years ago when i did purchase these drives, that heavy of a load on an eide drive would've been laughable and wouldn't even compare.
<< The seek times didn't ammount to jack in real world tests, as the benchmarks Modus pasted proved >>
actually the contrary is true. seek times are the most important factors in real world tests, as you hardly get sequential reads and writes to a hard drive, making str's more irrelevant and seek times more important than modus makes them out to be. case in point, rambus. there's also an article somewhere on sharkyextreme.com that compare eide raid vs. single eide solution, and the single eide hard disk comes out on top in "real world tests." i'm looking for it right now. if i find it, i'll link it to you.
//edit
said seek times are more important than seek times...duh
<< Well yes, access times are the most important factor, but by no means the only factor. STR still accounts for something, as shown by WB99 and other real world tests. I used seek times earlier because of the confidence I have in SCSI's laughable inability to show enough of a performance advantage. If you want to use seek times and throw out every other benchmark (I can understand this as it's the most SCSI-friendly strategy), it's not really fair but I'll allow for the time being because I know SCSI still won't have a chance. >>
laughable? a 50% difference in seek time is not laughable. str doesn't account for much at all too. look at the eide raid 0 setups that are trounced by single disk solutions in the eide camp. higher seek times, but much higher str's than single disk solutions, so seek time is by far more important than str if you are NOT doing video editing which is mostly str dependent. and the majority of us do not do video editing. some eide raid setups are trounced by single disks capable of about 1/2 of the str. scsi vs. eide is exactly like raid 0 vs non-raid, wth do you need raid 0 for? it's more expensive, the performance is proven to be lower than single disk solutions in most cases. but many people insist that it "feels" fast, when in fact it isn't. compare this to scsi though, and you have something that is proven to be fast, not something that "feels" fast. there's also the issue that some of us prefer quality over quantity.
<< "Have you ever actually owned a SCSI drive or controller?"
Absolutely. But that's irrelevant. I don't need to own a GeForce2 Ultra to know its exact price/performance ratio and to know that it is an utter waste of money. >>
no modus, this isn't irrelevant. again, you have yet to disclose to us what scsi system you have owned, and for how long. again, you make an invalid comparison. a geforce 2 ulra's speed can be quantified purely on scores in games because that game is the real world application that provides some means to measure the performance of the graphics card, and at the same time is the actual application one would be using. benchmarks for hard drives are not balanced as well as a timedemo in quake 3 to demonstrate fillrate etc. because most benchmarks are synthetic. want to really run a real world demonstration of hard disk speed? go to adobe photoshop and limit the memory it uses to 10 megs. open up a 50-100 meg image, apply a couple of filters back to back to get some good swapping going on an identical setup with the only changes being a scsi and eide hard drive. guaranteed the scsi hard disk will run circles around the eide.
<< I am not saying scsi and eide are exactly alike, but in most home setups it will seem as such. I still have a hard time believing most people are doing true multitasking with their scsi setups, as only 1 person has stepped up and reported what they were actually running. >>
i run photoshop and open images and edit them that usually maxes out my ram. since most photoshop filters aren't smp'd, my second cpu allows me to do other things while applying a filter/resizing/etc. since i've limited photoshop to take all but enough memory for me to do rudimentary things like email, mp3's, this forum, and since swapping is going on my second hard drive, i can still run win2k on my os drive just fine. and i capture video occasionally, although not as often as i use to a year or so ago because of time constraints. on top of this, sometimes i'm burning to my 8x hp, playing mp3's, and serving files off of the same hard drive without a hiccup. my justification for scsi. a couple of years ago when i did purchase these drives, that heavy of a load on an eide drive would've been laughable and wouldn't even compare.
<< The seek times didn't ammount to jack in real world tests, as the benchmarks Modus pasted proved >>
actually the contrary is true. seek times are the most important factors in real world tests, as you hardly get sequential reads and writes to a hard drive, making str's more irrelevant and seek times more important than modus makes them out to be. case in point, rambus. there's also an article somewhere on sharkyextreme.com that compare eide raid vs. single eide solution, and the single eide hard disk comes out on top in "real world tests." i'm looking for it right now. if i find it, i'll link it to you.
//edit
said seek times are more important than seek times...duh
Sir Fredrick
Guest
>>>>A) The "Wow Factor" is strictly emotional and CANNOT be used as a test of whether something is better or not.
Anytime you buy something new you get a wow factor, whether or not it is truly better.<<<<
Yeah, sometimes it's a "wow, this sucks" or "wow, what a rip off" see Radboy's repeated mentions of his CPU upgrade which left him significantly dissapointed.
>>>>B) The seek times didn't ammount to jack in real world tests, as the benchmarks Modus pasted proved. <<<<
The benchmark I assume you're referring to is winbench, which is not a real world test (no benchmark is truly a real world test), and the leading authority on benchmarks, storagereview, agrees that winbench is outdated and irrelevant. Storagereview also claims that seek times are the most important factor at this point. If you want to see low seek times in action, just start up a big program on an IDE drive and on a SCSI drive. Just starting windows requires thousands of files to be opened and closed, and I can guarantee you that those files are not all sequential, located on the hard drive in the order that they will be loaded, so seek time comes into play.
>>>>> Also, the memory thing barely applies here, as 128 megs is the bare minimum for win2k. EIDE is not the 'bare minimum'. I guess the only comparison would be ram with ECC and ram without. An added feature that might be useful in certain situations. Can you tell which one has the ECC? <<<<<
The performance increase you see by going from 128MB to 256MB *is* similar to the performance increase you see by going from IDE to SCSI, haven't we already established that it's about 40%? The comparison to ECC is irrelevant as ECC does not make any difference on performance, in fact it may slightly decrease performance.
IDE may not be a "bare minimum", although I don't know what would be below IDE. However, the hard drive is the slowest component in your system, so an increase in performance there WILL be felt on the system whenever there is disk access, unless the performance difference is minimal, which in the case of IDE vs SCSI, it is not.
If it comes down purely to a performance issue, SCSI wins hands down, no contest. However, some people do not care about performance enough to justify the purchase of a SCSI setup, but others value performance, and are willing to spend extra to achieve maximum performance. For them, SCSI is worth it. Have they made a poor decision? IMO, they have not, they simply purchased what was best for them.
I know that I do not *NEED* SCSI, but I also don't *need* 512MB of RAM, or an 8x burner (2x works just fine right?), or Dual PIII450s (a celeron 300 would do wouldn't it?), or a 17" monitor (15" will work). However I chose MY setup over a cheaper setup because I feel that using my computer should be an enjoyable experience, I value performance, and I had enough money to go with my setup.
Anytime you buy something new you get a wow factor, whether or not it is truly better.<<<<
Yeah, sometimes it's a "wow, this sucks" or "wow, what a rip off" see Radboy's repeated mentions of his CPU upgrade which left him significantly dissapointed.
>>>>B) The seek times didn't ammount to jack in real world tests, as the benchmarks Modus pasted proved. <<<<
The benchmark I assume you're referring to is winbench, which is not a real world test (no benchmark is truly a real world test), and the leading authority on benchmarks, storagereview, agrees that winbench is outdated and irrelevant. Storagereview also claims that seek times are the most important factor at this point. If you want to see low seek times in action, just start up a big program on an IDE drive and on a SCSI drive. Just starting windows requires thousands of files to be opened and closed, and I can guarantee you that those files are not all sequential, located on the hard drive in the order that they will be loaded, so seek time comes into play.
>>>>> Also, the memory thing barely applies here, as 128 megs is the bare minimum for win2k. EIDE is not the 'bare minimum'. I guess the only comparison would be ram with ECC and ram without. An added feature that might be useful in certain situations. Can you tell which one has the ECC? <<<<<
The performance increase you see by going from 128MB to 256MB *is* similar to the performance increase you see by going from IDE to SCSI, haven't we already established that it's about 40%? The comparison to ECC is irrelevant as ECC does not make any difference on performance, in fact it may slightly decrease performance.
IDE may not be a "bare minimum", although I don't know what would be below IDE. However, the hard drive is the slowest component in your system, so an increase in performance there WILL be felt on the system whenever there is disk access, unless the performance difference is minimal, which in the case of IDE vs SCSI, it is not.
If it comes down purely to a performance issue, SCSI wins hands down, no contest. However, some people do not care about performance enough to justify the purchase of a SCSI setup, but others value performance, and are willing to spend extra to achieve maximum performance. For them, SCSI is worth it. Have they made a poor decision? IMO, they have not, they simply purchased what was best for them.
I know that I do not *NEED* SCSI, but I also don't *need* 512MB of RAM, or an 8x burner (2x works just fine right?), or Dual PIII450s (a celeron 300 would do wouldn't it?), or a 17" monitor (15" will work). However I chose MY setup over a cheaper setup because I feel that using my computer should be an enjoyable experience, I value performance, and I had enough money to go with my setup.
<< a couple of years ago when i did purchase these drives, that heavy of a load on an eide drive would've been laughable and wouldn't even compare. >>
And I said several years ago I could have probably justified a SCSI drive =)
*edit*
Ok, Sir Fredrick, I ask you one thing, if EIDE is such a damn performance hit compared to SCSI how come I, being a huge performance freak, am still using EIDE. Or in other words, how much faster will my pc run if its already fast. I don't have slow down when opening things up, it is instant. I don't have disk access problems in video games, burning cds, using photoshop, or playing mp3s. Therefor, in my case, it isn't a performance thing or I would have already changed my opinion.
cya,
skace
Hi,
Well what you need to do is put a 2 comparable systems side by side, but one using scsi and one using IDE. Now here at my place, I do have exactly that. One of the computers has IDE(IBM 75GXP) and one system has SCSI(Atlas 10kII). Both systems are comparable in that they have about same amount of memory(256), same cpu(tbird 1 Gig), comparable mother board(abit, msi), etc. Both are tweaked to get better performance. Now load programs such as Microsoft word, or photoshop or gamespy or whatever and just notice the difference in loading time. The scsi system with atlas 10kii will load them all much more quickly then IDE. I must say though that when I went to scsi, I did not get much wow factor since my memory is very short term and did not remember how IDE was like but man when I go back to the IDE system comparing it to the scsi with both systems side by side, I must say "WOW" here....thats my take one the issue, although your system may be fast...scsi makes it go faster.
Well what you need to do is put a 2 comparable systems side by side, but one using scsi and one using IDE. Now here at my place, I do have exactly that. One of the computers has IDE(IBM 75GXP) and one system has SCSI(Atlas 10kII). Both systems are comparable in that they have about same amount of memory(256), same cpu(tbird 1 Gig), comparable mother board(abit, msi), etc. Both are tweaked to get better performance. Now load programs such as Microsoft word, or photoshop or gamespy or whatever and just notice the difference in loading time. The scsi system with atlas 10kii will load them all much more quickly then IDE. I must say though that when I went to scsi, I did not get much wow factor since my memory is very short term and did not remember how IDE was like but man when I go back to the IDE system comparing it to the scsi with both systems side by side, I must say "WOW" here....thats my take one the issue, although your system may be fast...scsi makes it go faster.
Sir Fredrick
Guest
>>>>> And I said several years ago I could have probably justified a SCSI drive =) <<<<<
SCSI drives are still a couple years ahead of IDE's performance curve. 🙂
>>>> Ok, Sir Fredrick, I ask you one thing, if EIDE is such a damn performance hit compared to SCSI how come I, being a huge performance freak, am still using EIDE.<<<<<
Because you've never used SCSI before? 🙂 Or because you don't have the cash, or you're too ornery and stubborn, how am I supposed to know why *you* do the crazy things you do? I'm not psychic, my friend. 🙂
>>>> Or in other words, how much faster will my pc run if its already fast.<<<<
40% faster on disk with anything disk intensive?
>>>>> I don't have slow down when opening things up, it is instant.<<<<
This is highly unlikely unless you are running everything from a RAM disk.
>>>> I don't have disk access problems in video games, burning cds, using photoshop, or playing mp3s. Therefor, in my case, it isn't a performance thing or I would have already changed my opinion.<<<<<
If you're satisfied with IDE's performance then by all means stick with it. I'm glad disk access isn't a *problem* for you, and it shouldn't be. I'm not saying that you will have problems if you don't use SCSI, just that you will see an improvement in performance with anything that accesses the hard drive.
SCSI drives are still a couple years ahead of IDE's performance curve. 🙂
>>>> Ok, Sir Fredrick, I ask you one thing, if EIDE is such a damn performance hit compared to SCSI how come I, being a huge performance freak, am still using EIDE.<<<<<
Because you've never used SCSI before? 🙂 Or because you don't have the cash, or you're too ornery and stubborn, how am I supposed to know why *you* do the crazy things you do? I'm not psychic, my friend. 🙂
>>>> Or in other words, how much faster will my pc run if its already fast.<<<<
40% faster on disk with anything disk intensive?
>>>>> I don't have slow down when opening things up, it is instant.<<<<
This is highly unlikely unless you are running everything from a RAM disk.
>>>> I don't have disk access problems in video games, burning cds, using photoshop, or playing mp3s. Therefor, in my case, it isn't a performance thing or I would have already changed my opinion.<<<<<
If you're satisfied with IDE's performance then by all means stick with it. I'm glad disk access isn't a *problem* for you, and it shouldn't be. I'm not saying that you will have problems if you don't use SCSI, just that you will see an improvement in performance with anything that accesses the hard drive.
<< And I said several years ago I could have probably justified a SCSI drive =) >>
and what makes it not justified now? scsi drive performance is still years ahead what eide will be. when you see 10000 rpm eide drives, scsi will have moved to 20000 rpm, if there will be such a thing. right now the fastest eide drive is 1/2 the spindle speed of x15.
<< if EIDE is such a damn performance hit compared to SCSI how come I, being a huge performance freak, am still using EIDE. >>
that's a good question. why aren't you using scsi? maybe you prefer polygons and pixels over disk performance? i have no clue.
Kind of late to the party, but what the hell. I'm all SCSI now. (well nearly, still use an IDE CDROM because I had it lying around). I do a fair amount of premiere, photoshop, tsumami and general video editing work so SCSI is a no brainer. I'm running a pair of Atlas V's on a TekRam DC390U2W. Like others have said when you get used to the responsiveness and multitasking abilities of SCSI, going back to IDE is painful!
My analogy is as follows. Your 1GHz CPU's process data in the order of gigabytes of information per second. Your ondie caches provide.. say 1GB/sec of bandwidth. Your RAM provides ~400-500MB/sec of bandwidth. Your hard drives provide what, 30-50MB/sec. There's a *small* bottleneck with the HD. So every bit of extra performance you can get from the HD's translates into magnitudes of performance at the CPU end of things.
My analogy is as follows. Your 1GHz CPU's process data in the order of gigabytes of information per second. Your ondie caches provide.. say 1GB/sec of bandwidth. Your RAM provides ~400-500MB/sec of bandwidth. Your hard drives provide what, 30-50MB/sec. There's a *small* bottleneck with the HD. So every bit of extra performance you can get from the HD's translates into magnitudes of performance at the CPU end of things.
Bah, I give up =). I don't really understand how something that has no slowdown, can have no slowdown but 40% faster. Also, I don't understand why anyone would even buy scsi for 'bootup speed'. Wtf? talk about blowing your money...
When new software comes out that makes my current system inadequate, I will upgrade to a 1.5ghz. When a new game comes out that makes my system run slowly, I will upgrade my video card and ram. When my sound card seems to pale in comparison to what I should be hearing, I will upgrade that also. When I start running out of the 60 gigs of hard drive space, which I almost am, I will throw on another hard drive. Thats the way I work. Until that happens, parts stay the same...
Just my view,
skace
When new software comes out that makes my current system inadequate, I will upgrade to a 1.5ghz. When a new game comes out that makes my system run slowly, I will upgrade my video card and ram. When my sound card seems to pale in comparison to what I should be hearing, I will upgrade that also. When I start running out of the 60 gigs of hard drive space, which I almost am, I will throw on another hard drive. Thats the way I work. Until that happens, parts stay the same...
Just my view,
skace
Sorry, it's been a long day. Been pulling my hair out cause I just destroyed a customer's motherboard with a BIOS flash from another bastard manufacturer that doesn't give a damn about end users. The ROM update was for exactly the model called for and the PCB revision matched up perfectly, but the flash program bombed out with a "Cannot identify chipset" error and now the board won't boot. To top it off, the board is old enough to be past warranty but not old enough to be worthless and replacable, meaning it's going to have to come out my pocket because he didn't specifically authorize the flash. Makes me want to scream. That's the second time I've done this. Maybe now I will realize that motherboard manufacturers are incompetent fools undeserving of the slightest trust.
OK, down to business:
Radboy,
<< But aren't you also one of the cake eaters? Haven't you also abandoned the 'masses' for SCSI? .. oh, Modus Robespierre? I was disappointed by the omission of your alleged & mysterious SCSI rig. >>
I simply mentioned that I had in the past owned and used a SCSI setup. As I said, I certainly did not pay market value for it. Actually, only the controller, an Adaptec 2940U2W, was bought used from a friend. I think I bought the 9G Quantum Atlas 10K2 new. I didn't have the system for very long before offloading it onto a customer. So yes, I do have the practical, first-hand experience you rail on and on about. I found what most people find: a slight increase in Windows responsiveness and a nice boost to application load times and disk swapping. Obviously, there was (and is) nothing in that modern SCSI setup that actually had the potential to increase my productivity; and as a jaded technologist, I was not thrilled by any suppoed "wow factor". Essentially, I'm not the kind of man who stakes his personal worth on the things he owns. So SCSI holds no allure for me.
But my personal, subjective experience (and anyone else's) is irrelevant to this discussion. Long before and after using a modern SCSI system for myself, I had much more valuable data: exact performance benchmarks by respected sites, and current price data by industry surveys. From these two things I could draw a perfect judgement without ever having to test the hardware for myself.
That is the sole reason for a site like AnandTech. It is the fundamental concept that allows us to purchase computer producst intelligently. If we had to personally test every piece of hardware before pronouncing a judgement on it, AnandTech would not and could not exist. Every single person would have to buy every single product and test it themselves. There would be no point for Mike of AnandTech or Eugene of StorageReview to benchmark something, because everyone would be just like you, reject the results and clamour for their own "hands on" subjective examination. This comes back to the "yoga reviewing" I mentioned before: lay your hands on the hardware, become ONE with the hardware, feel its warm, glowing, warming, glow, and wait for the hot fuzzy or cool prickly feeling to tell you whether it's a practical purchase.
No thanks. I'll take StorageReview's benchmarks and PriceWatch's surveys over Radboy's feelings and Pariah's opinions any day.
<< Uh, what do you call this?: "SCSI cannot hope to position itself as a smart, practical purchase. It's only market is the nerd-man who must have the biggest and most impressive computer penis to impress his nerd-friends." >>
A joke? Earth to Radboy, grow a sense of humour, please. And as long as we're on the topic of criticisms, maybe it'll refresh your memory to see how some SCSI advocates actually view IDE users -- with a thinly veiled sneer of disdain. Scroll down to my reply to Skace to read some nice, juicy quotes.
<< Need I remind you that the Storagerview says HERE that SEEK is the single most important metric in appraising hard drive performance? A short quote in case you're still in denial >>
I never denied that seek times were more important than other drive metrics. You, however, completely deny the importance of anything BUT seek times. Surely we must take a more holistic view. Perhaps a weighted average of several accepted benchmarks? Of course, that would simply serve to further hobble SCSI in any comparison, which nicely explains why you prefer a universe where hard drives are judged solely by their seek times. Understandable, but pitiable nonetheless.
<< For the person who (already) knows they want to upgrade to enterprise-class performance .. >>
It's a foregone conclusion! Can't you see that? We're talking about what's valuable, what's practical, what's sensible, and you're reduced to arguing that if some one has already decided they want top drawer performance, then SCSI is the best choice. Of course it is! The point I've been trying to drill into your head for the past week is that there is no logical, rational reason to pursure this level of performance when standard IDE technology yields adequate performance for PC applications while offering far superior value.
Essentially, what you're saying is, let's throw value out the window and see if SCSI wins. That's preposterous! Of course SCSI will win if money ceases to be a factor; the consumer has already decided that the computer is an end of means rather than a means to an end. He has decided to forgoe logic and reason and instead focus on his subjective emotional responses to a bunch of plastic and metal.
Here I must quote the great Jerry Seinfeld: "Not that there's anything wrong with that." These people cannot fairly be criticized for pursuing their passion. At the same time, you cannot seriously advocate a technology as a practical, sensible purchase for a person unless it can demonstrate a clear value advantage: a performance increase in line with its price increase. SCSI has thus far failed miserably to do that.
<< I'm also curious to hear your evaluation Matrox's recommended drive configurations posted HERE. >>
You know, Radboy, you're really cocky to throw that link out and bet that I won't follow it in detail. But a closer examination of the recommendations in question shows that they are for a D/V product that already comes with a SCSI controller. This completely destroys the validity of anything Matrox says in its hard drive recommendations because (1) they are merely trying to sell their product, which happens to have a SCSI controller built in and (2) since the money is now necessarily wasted to subsidize the on board controller, only the drive cost matters, which is totally contrary to the majority of real world scenarios. However, if you still want to hold this up as an example of a supposed SCSI value superority, be sure to factor in the cost of the Matrox DigiSuite as well. Then get back to me with the total cost of the SCSI solution. Mmkay?
Oh and pop quiz: what's the single most important hard disk performance factor in lossless digital video editing? Surprise, STR. Suddenly IDE doesn't look so terrible, does it?
<< so we have 50-57% better perf a 44% premium in cost >>
Your comparison is absolutely false, and so ridiculous it's almost as if you were deliberately handing me something I could disprove in my sleep. As I lapse into a boredom-induced comma, I will explain your errors and omissions whilst providing the correct comparison for your reference.
Your comparison has three main problems: (1) The SCSI drive received a rebate while the IDE drive did not, (2) at 9G the SCSI drive was less than half the size of the IDE drive -- far too small to be considered equivalent storage, and (3) your price premium purposely ignores the enormous proportional cost increase of the necessary SCSI controller.
Additionally, the link you provided to a store that magically offered the Atlas 10k2 for $20 less (before rebate) than the cheapest PriceWatch listing has now magically broken, leaving me scratching my head as to whether it is an innocent server outage or an actual case of falsifying evidence on your part. I tend toward the former explanation, which must of course beg the question of how reliable a price is from a company that can't even run its own web site properly.
I also take issue with the use of seek times as the ONLY drive metric. A more objective comparison would include several measures and weight them according to conventional wisdom, however, I don't have the time for such a project, so I will make do with your own completely biased method since I know your case is futile.
Thus, the correct comparison, using only verifiable data from PriceWatch and StorageReview:
18G Quantum Atlas 10K II: $295
Tekram DC-390U2B: $96
Total: $391, 7.9ms access time
20G Quantum Fireball LM Plus: $104, 11.5ms access time
Therefore, the SCSI setup offers 46% more performance in exchange for 276% more money. Decision: IDE. Now, I realize forcing the comparison to be fair, i.e. forcing the storage space to be equal, has totally destroyed SCSI, so I've found a nice compromise comparison using your own preferred "hybrid" setup:
9G Quantum Atlas 10K II: $189
20G Quantum Fireball LM Plus: $104
Tekram DC-390U2B: $96
Total: $389, 29G storage, 7.9ms access time
30G Quantum Fireball LM Plus: $116, 11.5ms access time
So using your supposedly "cost effective hybrid setup", SCSI still only offers a 46% performance increase in exchange for a whopping 235% price increase.
Thus, in the courtroom of value and sensible spending, SCSI's case is dismissed without the briefest hearing. It is banished to the realm of Judge Judy and the People's Court, where personal annecdotes and emotional assessments reign supreme and emperical evidence is frowned upon.
<< speaking of 15krpm drives .. another thing scsi buys you .. is future upgradability >>
LOL, how so? Please explain how a SCSI system gives you a better upgrade path than an IDE system. IDE specifications are all marketing and even the years old ATA/33 is still adequate for most drives. SCSI standards, on the other hand, require a completely new controller card to address. This expense is cripling to the value calculation that is so dearly important when trying to make a proper purchase.
<< so what does the scsi card buy you, except for future upgradability? better irq management (1 for 15 devices) >>
Nope. Ever heard of ACPI? Every device in your system shares one IRQ without conflicts. Even IDE controllers.
<< the ability to add devices other than hdd's & optical drives >>
Like what? Scanners, web cams, removable magnetic storage, professional FireWire devices? Nope, IDE systems can do all that with their free USB ports and dirt cheap ($29) IEEE 1394 controllers.
<< the ability to have more than a combined number of 4 hdd's & optical drives without add-on card$ >>
The best currently available motherboards from ASUS, Microstar and ABit for the best currently available platform, Socket A, all feature native support for either IDE RAID or at least a secondary ATA/100 controller to allow eight devices. Not that it's a concern for the vast majority of people, but it's nice to know you could go nuts and stack drives for a server if you had to. Especially when the ability is free and doesn't come at the insane price premium SCSI demands.
<< the ability to use more reliable hard drives to run your system & apps >>
We've already addressed this and seen how failure rates on modern IDE and SCSI drives are so low that the difference amounts to almost nothing for non-mission critical environments. And in those environments, the omnipresent redundant backup allows both IDE and SCSI to claim 100% uptime.
<< the biggest reason, tho, is a wonderful sense of *responsiveness* that is impossible to describe if you've never experienced it >>
So you acknowledge that SCSI cannot hope to compete on a rational, objective value basis? You admit that the decision to purchase a SCSI system is entirely emotional and not open to sensible debate?
<< What I'm saying is that, we don't *think* about seek or access times when we're using our PCs. >>
And we don't *think* about FPS in the middle of a deathmatch. Are you then advocating that Anand and Tom simply "feel" the performance of the latest video card? Just "eyeball" the speed? See how it makes them feel? Give me a break. Benchmarks are the only objective measures. The trick is simply finding the right ones.
<< so, the person who sees VALUE in those advanatges, will likely feel that the cost was worth it .. while the person whop doesn't, may not >>
Yes, Radboy, let's all gather 'round the campfire and sing folk songs while we create a sharing circle of acceptance and understanding where everyone is right and no one is wrong and no one questions mistakes or challenges assumptions.
<< it's also widely discussed that current benchmarks - even the best ones, like intels' IOMeter - do not adequately/accurately test the true capabilities of a scsi drive or the scsi interface >>
Oh, it is, eh? Is it also widely discussed that it's an absolute falacy to design a benchmark specifically to hilight the advantages of a certain technology? That the only valid benchmarks are real world tests blind to the hardware underpinnings they stress? In your circles, apparently not.
<< Unfortunately, there's no way to prove this >>
Isn't that the definition of false? 😉
Pariah,
<< Hot rods are style over substance where as computers are mostly substance over style >>
Not with SCSI. SCSI on the desktop PC is the ultimate case of style over substance. Well, maybe not. The iMAC still takes the crown in that department.
<< Hot rods are a hobby, what hobby is a constructive use of time that is money well spent from the viewpoint of someone who doesn't enjoy it? >>
No, we're not criticizing those who buy SCSI systems purely as a hobby or to fool around with cool stuff. If you're one of those, I salute you. What the IDE masses criticize are those who attempt to claim SCSI offers a tangible value advantage in practical circumstances.
<< If you can't notice the difference in system responsiveness when any disk is necessary, you're not a SCSI candidate. >>
Oh please. Now you have to be a special sort of person to achieve value in a SCSI purchase? This just delves further and further into the Twilight Zone.
Fredrick,
<< The benchmark I assume you're referring to is winbench, which is not a real world test (no benchmark is truly a real world test), and the leading authority on benchmarks, storagereview, agrees that winbench is outdated and irrelevant >>
Winbench is far, far closer to a real world than Threadmark or IOMeter. It's just that it tests applications that aren't so disk-limitted as they used to be. Still, it's emphasis on STR provides a welcome counterweight (albeit a small one) to IOMeter. If StorageReview truly thought Winbench was worthless, they wouldn't keep using it! As it is, they feature it in every single one of their reviews, and they point out that Quantum still views it as the best measure of disk performance.
We certainly agree that seek times are most important for real world usage. But that does not mean one can completely ignore less seek-dependent benchmarks simply because they paint SCSI performance in a less attractive light.
<< Because you've never used SCSI before? Or because you don't have the cash, or you're too ornery and stubborn, how am I supposed to know why *you* do the crazy things you do? >>
Again with this bigotry against IDE users: we're poor, we're cheap, we don't care about new technology. Please. Take it to a nerd convention. This is real life.
<< 40% faster on disk with anything disk intensive? >>
Which brings up a central point of a my last post that the SCSI advocates ignored: there are very, very few disk intensive applications left. Think about it. Video cards, CPU's, memory, all have common productivity and entertainment software that pushes them to the limit. But hard drives don't. There is no hard drive performance requirement on any modern software. Why? Because cheap RAM and fast chips have lead to systems where drive access is brief and contained in short bursts, where it has little effect on overall real world performance. Just another reason why SCSI's meager added performance gains in light of its massive added cost are not at all justified.
<< If you're satisfied with IDE's performance then by all means stick with it. I'm glad disk access isn't a *problem* for you, and it shouldn't be. >>
Yup, I can smell the marshmellows roasting right now: "Hi everybody. My name is Jeff, and I want to be accepted for what I put in my computer without the fear and recriminations of benchmarks and prices." A pause. "Hi Jeff!"
borealiss,
<< laughable? a 50% difference in seek time is not laughable >>
It is when it comes at a 300% price hike.
<< str doesn't account for much at all too >>
No, it only tells you how fast streams of data are read off your drive. That can't mean much.
<< no modus, this isn't irrelevant. again, you have yet to disclose to us what scsi system you have owned, and for how long. again, you make an invalid comparison >>
It is entirely irrelevant. However, to put this to rest and keep you people from stalking me and extracting the secrets of my SCSI machine from the brain of my cat, I have described it above in my response to Radboy. It doesn't mean anything though, since whatever SCSI system I've owned won't change the benchmarks on StorageReview or the prices on PriceWatch.
<< benchmarks for hard drives are not balanced as well as a timedemo in quake 3 to demonstrate fillrate etc. because most benchmarks are synthetic >>
You're correct, we do need better real world hard drive benchmarks. The problem, as I stated above, is that there is really no application in wide use that is hard disk-limitted. Drive performance forms only a small component of the total performance of modern applications, while CPU, memory, and video performance accounts for the vast bulk of it.
My solution to the hard drive benchmark question is to simply form a weighted average of StorageReview's benchmarks. Take each of the four areas and produce an average score. For STR, average the beginning and end rates. For access time, no change is needed. For Winbench 99, average the business and high end tests. And for IOMeter, average the three results. Then, normalize these four numbers against results from their contemporaries. Finally, assign a weight to each area. I think 50% IOMeter, 30% Winbench, 15% access time, and 5% STR would be a fair starting point open to debate. This kind of system would give the most holistic view of drive performance available with current benchmarking software.
<< want to really run a real world demonstration of hard disk speed? go to adobe photoshop and limit the memory it uses to 10 megs. open up a 50-100 meg image, apply a couple of filters back to back to get some good swapping going >>
Oh brother. Don't you see how stupid that is? It's not a real world test if you limit Photoshop's RAM to 10 megs!
Skace,
<< Radboy I don't really understand you, Modus has responded to your seek time and "wow affect" posts about a DOZEN times now and yet you keep coming back with "WHAT ABOUT THE SEEK TIME!". Are you even reading his posts? >>
Oh, he reads 'em. He's just getting bored with a loosing argument so he keeps returning to an area where he can show a SCSI number lower than an IDE number and gloss over the enormous price penalty incurred for those lower times.
<< What is even more interesting is that my work is starting to look at EIDE drives in networkable storage units for some of their storage... >>
Yeah, corporate IT is finally catching on that SCSI usually isn't cost effective, even in small to medium sized network servers. The low price of RAM and the steady advancement of IDE performance has brought us to the point where the majority of AnandTech's T-Bird webservers run IDE drives.
<< Also, It still seems like the majority of the SCSI people (or just a select few) still don't get it because they keep posting responses comparing SCSI to expensive cars. Which, of course, assumes that EIDE users either can't afford nice cars or don't want them. >>
You're absolutely right about the elitism of SCSI users, but it's more than that. It's a thinly veiled sentiment that people who choose IDE technology must be too poor to afford better. Here are just a few of the tastier bits:
"I find it amazing at a Computer Enthusiast Forum that people squawk about price performance. That should be a complaint for an AOL Forum or a tight wad like Modus" -- Red Dawn
"Again, SCSI is not for everyone. It's for the performance-minded, while IDE is for the value-minded. . . How can someone who's never driven a Ferrari say it's not worth it?" -- Radboy
"Just because you may be a student on a tight budget or you're tighter than bark on a stump with your money, don't condemn SCSI or write it off as a waste. I believe most, not all, but most of the negativity toward SCSI spanks of sour grapes" -- Toolman
"Sounds like sour grapes from someone who either can't afford it, or isn't willing to spend that type of money on their computer." -- Pariah
"If you worship the gods of price/performance so much, go back to the K62, get a motherboard with everything integrated, and a nice cheap winmodem. . . And you don't have to be rich to afford SCSI." -- Sir Fredrick
"SCSI is not for everyone, It's not even for most people .. and it's certainly not for the person who can't afford it. But many of the ppl who hangs here seem to be able to afford the finer things in life" -- Radboy
"BTW, this poll does not help to show which interface is superior at all, it just helps to show what most people are able to afford." -- Sir Fredrick
Modus
OK, down to business:
Radboy,
<< But aren't you also one of the cake eaters? Haven't you also abandoned the 'masses' for SCSI? .. oh, Modus Robespierre? I was disappointed by the omission of your alleged & mysterious SCSI rig. >>
I simply mentioned that I had in the past owned and used a SCSI setup. As I said, I certainly did not pay market value for it. Actually, only the controller, an Adaptec 2940U2W, was bought used from a friend. I think I bought the 9G Quantum Atlas 10K2 new. I didn't have the system for very long before offloading it onto a customer. So yes, I do have the practical, first-hand experience you rail on and on about. I found what most people find: a slight increase in Windows responsiveness and a nice boost to application load times and disk swapping. Obviously, there was (and is) nothing in that modern SCSI setup that actually had the potential to increase my productivity; and as a jaded technologist, I was not thrilled by any suppoed "wow factor". Essentially, I'm not the kind of man who stakes his personal worth on the things he owns. So SCSI holds no allure for me.
But my personal, subjective experience (and anyone else's) is irrelevant to this discussion. Long before and after using a modern SCSI system for myself, I had much more valuable data: exact performance benchmarks by respected sites, and current price data by industry surveys. From these two things I could draw a perfect judgement without ever having to test the hardware for myself.
That is the sole reason for a site like AnandTech. It is the fundamental concept that allows us to purchase computer producst intelligently. If we had to personally test every piece of hardware before pronouncing a judgement on it, AnandTech would not and could not exist. Every single person would have to buy every single product and test it themselves. There would be no point for Mike of AnandTech or Eugene of StorageReview to benchmark something, because everyone would be just like you, reject the results and clamour for their own "hands on" subjective examination. This comes back to the "yoga reviewing" I mentioned before: lay your hands on the hardware, become ONE with the hardware, feel its warm, glowing, warming, glow, and wait for the hot fuzzy or cool prickly feeling to tell you whether it's a practical purchase.
No thanks. I'll take StorageReview's benchmarks and PriceWatch's surveys over Radboy's feelings and Pariah's opinions any day.
<< Uh, what do you call this?: "SCSI cannot hope to position itself as a smart, practical purchase. It's only market is the nerd-man who must have the biggest and most impressive computer penis to impress his nerd-friends." >>
A joke? Earth to Radboy, grow a sense of humour, please. And as long as we're on the topic of criticisms, maybe it'll refresh your memory to see how some SCSI advocates actually view IDE users -- with a thinly veiled sneer of disdain. Scroll down to my reply to Skace to read some nice, juicy quotes.
<< Need I remind you that the Storagerview says HERE that SEEK is the single most important metric in appraising hard drive performance? A short quote in case you're still in denial >>
I never denied that seek times were more important than other drive metrics. You, however, completely deny the importance of anything BUT seek times. Surely we must take a more holistic view. Perhaps a weighted average of several accepted benchmarks? Of course, that would simply serve to further hobble SCSI in any comparison, which nicely explains why you prefer a universe where hard drives are judged solely by their seek times. Understandable, but pitiable nonetheless.
<< For the person who (already) knows they want to upgrade to enterprise-class performance .. >>
It's a foregone conclusion! Can't you see that? We're talking about what's valuable, what's practical, what's sensible, and you're reduced to arguing that if some one has already decided they want top drawer performance, then SCSI is the best choice. Of course it is! The point I've been trying to drill into your head for the past week is that there is no logical, rational reason to pursure this level of performance when standard IDE technology yields adequate performance for PC applications while offering far superior value.
Essentially, what you're saying is, let's throw value out the window and see if SCSI wins. That's preposterous! Of course SCSI will win if money ceases to be a factor; the consumer has already decided that the computer is an end of means rather than a means to an end. He has decided to forgoe logic and reason and instead focus on his subjective emotional responses to a bunch of plastic and metal.
Here I must quote the great Jerry Seinfeld: "Not that there's anything wrong with that." These people cannot fairly be criticized for pursuing their passion. At the same time, you cannot seriously advocate a technology as a practical, sensible purchase for a person unless it can demonstrate a clear value advantage: a performance increase in line with its price increase. SCSI has thus far failed miserably to do that.
<< I'm also curious to hear your evaluation Matrox's recommended drive configurations posted HERE. >>
You know, Radboy, you're really cocky to throw that link out and bet that I won't follow it in detail. But a closer examination of the recommendations in question shows that they are for a D/V product that already comes with a SCSI controller. This completely destroys the validity of anything Matrox says in its hard drive recommendations because (1) they are merely trying to sell their product, which happens to have a SCSI controller built in and (2) since the money is now necessarily wasted to subsidize the on board controller, only the drive cost matters, which is totally contrary to the majority of real world scenarios. However, if you still want to hold this up as an example of a supposed SCSI value superority, be sure to factor in the cost of the Matrox DigiSuite as well. Then get back to me with the total cost of the SCSI solution. Mmkay?
Oh and pop quiz: what's the single most important hard disk performance factor in lossless digital video editing? Surprise, STR. Suddenly IDE doesn't look so terrible, does it?
<< so we have 50-57% better perf a 44% premium in cost >>
Your comparison is absolutely false, and so ridiculous it's almost as if you were deliberately handing me something I could disprove in my sleep. As I lapse into a boredom-induced comma, I will explain your errors and omissions whilst providing the correct comparison for your reference.
Your comparison has three main problems: (1) The SCSI drive received a rebate while the IDE drive did not, (2) at 9G the SCSI drive was less than half the size of the IDE drive -- far too small to be considered equivalent storage, and (3) your price premium purposely ignores the enormous proportional cost increase of the necessary SCSI controller.
Additionally, the link you provided to a store that magically offered the Atlas 10k2 for $20 less (before rebate) than the cheapest PriceWatch listing has now magically broken, leaving me scratching my head as to whether it is an innocent server outage or an actual case of falsifying evidence on your part. I tend toward the former explanation, which must of course beg the question of how reliable a price is from a company that can't even run its own web site properly.
I also take issue with the use of seek times as the ONLY drive metric. A more objective comparison would include several measures and weight them according to conventional wisdom, however, I don't have the time for such a project, so I will make do with your own completely biased method since I know your case is futile.
Thus, the correct comparison, using only verifiable data from PriceWatch and StorageReview:
18G Quantum Atlas 10K II: $295
Tekram DC-390U2B: $96
Total: $391, 7.9ms access time
20G Quantum Fireball LM Plus: $104, 11.5ms access time
Therefore, the SCSI setup offers 46% more performance in exchange for 276% more money. Decision: IDE. Now, I realize forcing the comparison to be fair, i.e. forcing the storage space to be equal, has totally destroyed SCSI, so I've found a nice compromise comparison using your own preferred "hybrid" setup:
9G Quantum Atlas 10K II: $189
20G Quantum Fireball LM Plus: $104
Tekram DC-390U2B: $96
Total: $389, 29G storage, 7.9ms access time
30G Quantum Fireball LM Plus: $116, 11.5ms access time
So using your supposedly "cost effective hybrid setup", SCSI still only offers a 46% performance increase in exchange for a whopping 235% price increase.
Thus, in the courtroom of value and sensible spending, SCSI's case is dismissed without the briefest hearing. It is banished to the realm of Judge Judy and the People's Court, where personal annecdotes and emotional assessments reign supreme and emperical evidence is frowned upon.
<< speaking of 15krpm drives .. another thing scsi buys you .. is future upgradability >>
LOL, how so? Please explain how a SCSI system gives you a better upgrade path than an IDE system. IDE specifications are all marketing and even the years old ATA/33 is still adequate for most drives. SCSI standards, on the other hand, require a completely new controller card to address. This expense is cripling to the value calculation that is so dearly important when trying to make a proper purchase.
<< so what does the scsi card buy you, except for future upgradability? better irq management (1 for 15 devices) >>
Nope. Ever heard of ACPI? Every device in your system shares one IRQ without conflicts. Even IDE controllers.
<< the ability to add devices other than hdd's & optical drives >>
Like what? Scanners, web cams, removable magnetic storage, professional FireWire devices? Nope, IDE systems can do all that with their free USB ports and dirt cheap ($29) IEEE 1394 controllers.
<< the ability to have more than a combined number of 4 hdd's & optical drives without add-on card$ >>
The best currently available motherboards from ASUS, Microstar and ABit for the best currently available platform, Socket A, all feature native support for either IDE RAID or at least a secondary ATA/100 controller to allow eight devices. Not that it's a concern for the vast majority of people, but it's nice to know you could go nuts and stack drives for a server if you had to. Especially when the ability is free and doesn't come at the insane price premium SCSI demands.
<< the ability to use more reliable hard drives to run your system & apps >>
We've already addressed this and seen how failure rates on modern IDE and SCSI drives are so low that the difference amounts to almost nothing for non-mission critical environments. And in those environments, the omnipresent redundant backup allows both IDE and SCSI to claim 100% uptime.
<< the biggest reason, tho, is a wonderful sense of *responsiveness* that is impossible to describe if you've never experienced it >>
So you acknowledge that SCSI cannot hope to compete on a rational, objective value basis? You admit that the decision to purchase a SCSI system is entirely emotional and not open to sensible debate?
<< What I'm saying is that, we don't *think* about seek or access times when we're using our PCs. >>
And we don't *think* about FPS in the middle of a deathmatch. Are you then advocating that Anand and Tom simply "feel" the performance of the latest video card? Just "eyeball" the speed? See how it makes them feel? Give me a break. Benchmarks are the only objective measures. The trick is simply finding the right ones.
<< so, the person who sees VALUE in those advanatges, will likely feel that the cost was worth it .. while the person whop doesn't, may not >>
Yes, Radboy, let's all gather 'round the campfire and sing folk songs while we create a sharing circle of acceptance and understanding where everyone is right and no one is wrong and no one questions mistakes or challenges assumptions.
<< it's also widely discussed that current benchmarks - even the best ones, like intels' IOMeter - do not adequately/accurately test the true capabilities of a scsi drive or the scsi interface >>
Oh, it is, eh? Is it also widely discussed that it's an absolute falacy to design a benchmark specifically to hilight the advantages of a certain technology? That the only valid benchmarks are real world tests blind to the hardware underpinnings they stress? In your circles, apparently not.
<< Unfortunately, there's no way to prove this >>
Isn't that the definition of false? 😉
Pariah,
<< Hot rods are style over substance where as computers are mostly substance over style >>
Not with SCSI. SCSI on the desktop PC is the ultimate case of style over substance. Well, maybe not. The iMAC still takes the crown in that department.
<< Hot rods are a hobby, what hobby is a constructive use of time that is money well spent from the viewpoint of someone who doesn't enjoy it? >>
No, we're not criticizing those who buy SCSI systems purely as a hobby or to fool around with cool stuff. If you're one of those, I salute you. What the IDE masses criticize are those who attempt to claim SCSI offers a tangible value advantage in practical circumstances.
<< If you can't notice the difference in system responsiveness when any disk is necessary, you're not a SCSI candidate. >>
Oh please. Now you have to be a special sort of person to achieve value in a SCSI purchase? This just delves further and further into the Twilight Zone.
Fredrick,
<< The benchmark I assume you're referring to is winbench, which is not a real world test (no benchmark is truly a real world test), and the leading authority on benchmarks, storagereview, agrees that winbench is outdated and irrelevant >>
Winbench is far, far closer to a real world than Threadmark or IOMeter. It's just that it tests applications that aren't so disk-limitted as they used to be. Still, it's emphasis on STR provides a welcome counterweight (albeit a small one) to IOMeter. If StorageReview truly thought Winbench was worthless, they wouldn't keep using it! As it is, they feature it in every single one of their reviews, and they point out that Quantum still views it as the best measure of disk performance.
We certainly agree that seek times are most important for real world usage. But that does not mean one can completely ignore less seek-dependent benchmarks simply because they paint SCSI performance in a less attractive light.
<< Because you've never used SCSI before? Or because you don't have the cash, or you're too ornery and stubborn, how am I supposed to know why *you* do the crazy things you do? >>
Again with this bigotry against IDE users: we're poor, we're cheap, we don't care about new technology. Please. Take it to a nerd convention. This is real life.
<< 40% faster on disk with anything disk intensive? >>
Which brings up a central point of a my last post that the SCSI advocates ignored: there are very, very few disk intensive applications left. Think about it. Video cards, CPU's, memory, all have common productivity and entertainment software that pushes them to the limit. But hard drives don't. There is no hard drive performance requirement on any modern software. Why? Because cheap RAM and fast chips have lead to systems where drive access is brief and contained in short bursts, where it has little effect on overall real world performance. Just another reason why SCSI's meager added performance gains in light of its massive added cost are not at all justified.
<< If you're satisfied with IDE's performance then by all means stick with it. I'm glad disk access isn't a *problem* for you, and it shouldn't be. >>
Yup, I can smell the marshmellows roasting right now: "Hi everybody. My name is Jeff, and I want to be accepted for what I put in my computer without the fear and recriminations of benchmarks and prices." A pause. "Hi Jeff!"
borealiss,
<< laughable? a 50% difference in seek time is not laughable >>
It is when it comes at a 300% price hike.
<< str doesn't account for much at all too >>
No, it only tells you how fast streams of data are read off your drive. That can't mean much.
<< no modus, this isn't irrelevant. again, you have yet to disclose to us what scsi system you have owned, and for how long. again, you make an invalid comparison >>
It is entirely irrelevant. However, to put this to rest and keep you people from stalking me and extracting the secrets of my SCSI machine from the brain of my cat, I have described it above in my response to Radboy. It doesn't mean anything though, since whatever SCSI system I've owned won't change the benchmarks on StorageReview or the prices on PriceWatch.
<< benchmarks for hard drives are not balanced as well as a timedemo in quake 3 to demonstrate fillrate etc. because most benchmarks are synthetic >>
You're correct, we do need better real world hard drive benchmarks. The problem, as I stated above, is that there is really no application in wide use that is hard disk-limitted. Drive performance forms only a small component of the total performance of modern applications, while CPU, memory, and video performance accounts for the vast bulk of it.
My solution to the hard drive benchmark question is to simply form a weighted average of StorageReview's benchmarks. Take each of the four areas and produce an average score. For STR, average the beginning and end rates. For access time, no change is needed. For Winbench 99, average the business and high end tests. And for IOMeter, average the three results. Then, normalize these four numbers against results from their contemporaries. Finally, assign a weight to each area. I think 50% IOMeter, 30% Winbench, 15% access time, and 5% STR would be a fair starting point open to debate. This kind of system would give the most holistic view of drive performance available with current benchmarking software.
<< want to really run a real world demonstration of hard disk speed? go to adobe photoshop and limit the memory it uses to 10 megs. open up a 50-100 meg image, apply a couple of filters back to back to get some good swapping going >>
Oh brother. Don't you see how stupid that is? It's not a real world test if you limit Photoshop's RAM to 10 megs!
Skace,
<< Radboy I don't really understand you, Modus has responded to your seek time and "wow affect" posts about a DOZEN times now and yet you keep coming back with "WHAT ABOUT THE SEEK TIME!". Are you even reading his posts? >>
Oh, he reads 'em. He's just getting bored with a loosing argument so he keeps returning to an area where he can show a SCSI number lower than an IDE number and gloss over the enormous price penalty incurred for those lower times.
<< What is even more interesting is that my work is starting to look at EIDE drives in networkable storage units for some of their storage... >>
Yeah, corporate IT is finally catching on that SCSI usually isn't cost effective, even in small to medium sized network servers. The low price of RAM and the steady advancement of IDE performance has brought us to the point where the majority of AnandTech's T-Bird webservers run IDE drives.
<< Also, It still seems like the majority of the SCSI people (or just a select few) still don't get it because they keep posting responses comparing SCSI to expensive cars. Which, of course, assumes that EIDE users either can't afford nice cars or don't want them. >>
You're absolutely right about the elitism of SCSI users, but it's more than that. It's a thinly veiled sentiment that people who choose IDE technology must be too poor to afford better. Here are just a few of the tastier bits:
"I find it amazing at a Computer Enthusiast Forum that people squawk about price performance. That should be a complaint for an AOL Forum or a tight wad like Modus" -- Red Dawn
"Again, SCSI is not for everyone. It's for the performance-minded, while IDE is for the value-minded. . . How can someone who's never driven a Ferrari say it's not worth it?" -- Radboy
"Just because you may be a student on a tight budget or you're tighter than bark on a stump with your money, don't condemn SCSI or write it off as a waste. I believe most, not all, but most of the negativity toward SCSI spanks of sour grapes" -- Toolman
"Sounds like sour grapes from someone who either can't afford it, or isn't willing to spend that type of money on their computer." -- Pariah
"If you worship the gods of price/performance so much, go back to the K62, get a motherboard with everything integrated, and a nice cheap winmodem. . . And you don't have to be rich to afford SCSI." -- Sir Fredrick
"SCSI is not for everyone, It's not even for most people .. and it's certainly not for the person who can't afford it. But many of the ppl who hangs here seem to be able to afford the finer things in life" -- Radboy
"BTW, this poll does not help to show which interface is superior at all, it just helps to show what most people are able to afford." -- Sir Fredrick
Modus
that's a monster post.
<< Oh brother. Don't you see how stupid that is? It's not a real world test if you limit Photoshop's RAM to 10 megs! >>
granted you would never encounter a situation with a 10 meg limitation in photoshop, but open up a larger image, or multiple images, with 256 megs of ram which is a common amount these days, and you quickly will run into a situation very similar, as i do with 450 megs from time to time. i would take the result of this test over many of the synthetic ones though to demonstrate access times, and video capture to demonstrate str since it is more real world than wb99 will ever be.
<< It is when it comes at a 300% price hike. >>
you seem to never take into account the price premium certain motherboards that come implemented with eide raid have over ones that don't. take for example the abit KT7-RAID which lists around 141 bucks on pricewatch. now, for those of us that could care less about this, why should we have to pay for this useless feature for a motherboard. another price premium towards the eide camp. there are motherboards that are 50 dollars less than ones with raid implemented on them for the same platform. if one was going for a scsi setup, that 50 dollars is saved.
also, you seem to think that i think that str's are totally pointless. this is not true, but to an extent, str's cannot be the sole factor to judge hard disk performance, which we both agree upon. however, when would hard disk performance become most notable? during swap file operations, which are almost purely based on seek times and hardly have anything to do with str's because programs in memory that are being swapped to the pagefile aren't in contiguous memory locations, so the read head is jumping around like a rabbit on crack. memory that does get written into the page file gets extremely fragmented at times, even though the pagefile does not, and this even places more strain on seek times more than str. this is really where scsi will shine imo. for those of us that do max out our ram on occasions, the added speed is worth it, which i frequently do sometimes. of course i could add more ram, but my dimm slots are full, and i cannot buy a 256 meg stick without paying a huge price premium for it compared to other sdram because i have a bx chipset and require 16Mx8/16Mx4 modules. and then i'd have to throw out the dimm i'm replacing or sell it at lower price i paid for, a lose lose situation, when i would prefer more disk storage. if i had to just unload some media like movies and mp3's from my main disk and put them on a cheap eide hard drive for phat storage, then that would be more ideal since i don't need speed from it.
<< No, it only tells you how fast streams of data are read off your drive. That can't mean much. >>
again modus, this is sequential. hardly any reads or writes to the disk are sequential. why does sdram outperform rdram when rdram has a HIGHER bandwidth rating? latency, or in this case, access times.
<< Oh brother. Don't you see how stupid that is? It's not a real world test if you limit Photoshop's RAM to 10 megs! >>
granted you would never encounter a situation with a 10 meg limitation in photoshop, but open up a larger image, or multiple images, with 256 megs of ram which is a common amount these days, and you quickly will run into a situation very similar, as i do with 450 megs from time to time. i would take the result of this test over many of the synthetic ones though to demonstrate access times, and video capture to demonstrate str since it is more real world than wb99 will ever be.
<< It is when it comes at a 300% price hike. >>
you seem to never take into account the price premium certain motherboards that come implemented with eide raid have over ones that don't. take for example the abit KT7-RAID which lists around 141 bucks on pricewatch. now, for those of us that could care less about this, why should we have to pay for this useless feature for a motherboard. another price premium towards the eide camp. there are motherboards that are 50 dollars less than ones with raid implemented on them for the same platform. if one was going for a scsi setup, that 50 dollars is saved.
also, you seem to think that i think that str's are totally pointless. this is not true, but to an extent, str's cannot be the sole factor to judge hard disk performance, which we both agree upon. however, when would hard disk performance become most notable? during swap file operations, which are almost purely based on seek times and hardly have anything to do with str's because programs in memory that are being swapped to the pagefile aren't in contiguous memory locations, so the read head is jumping around like a rabbit on crack. memory that does get written into the page file gets extremely fragmented at times, even though the pagefile does not, and this even places more strain on seek times more than str. this is really where scsi will shine imo. for those of us that do max out our ram on occasions, the added speed is worth it, which i frequently do sometimes. of course i could add more ram, but my dimm slots are full, and i cannot buy a 256 meg stick without paying a huge price premium for it compared to other sdram because i have a bx chipset and require 16Mx8/16Mx4 modules. and then i'd have to throw out the dimm i'm replacing or sell it at lower price i paid for, a lose lose situation, when i would prefer more disk storage. if i had to just unload some media like movies and mp3's from my main disk and put them on a cheap eide hard drive for phat storage, then that would be more ideal since i don't need speed from it.
<< No, it only tells you how fast streams of data are read off your drive. That can't mean much. >>
again modus, this is sequential. hardly any reads or writes to the disk are sequential. why does sdram outperform rdram when rdram has a HIGHER bandwidth rating? latency, or in this case, access times.
TRENDING THREADS
-
Discussion Zen 5 Speculation (EPYC Turin and Strix Point/Granite Ridge - Ryzen 9000)- Started by DisEnchantment
- Replies: 25K
-
Discussion Intel Meteor, Arrow, Lunar & Panther Lakes + WCL Discussion Threads
- Started by Tigerick
- Replies: 25K
-
Discussion Intel current and future Lakes & Rapids thread- Started by TheF34RChannel
- Replies: 24K
-
-
