• 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.

Promise Ultra66 Performance Question

L

Lex Luthor

Senior member
I am planning on possibly buying a Promise Ultra66 card and am curious as to the effect (speedwise) it might have on my system.

I have a K6-3 450, 80MB Ram, Win98SE and a 6GB Hard drive that supports UDMA66 transfers. My current motherboard only supports UDMA33.

I mainly use the PC for surfing, internet applications, some business apps and some games.

What effect would adding a promise ultra66 have on things such as boot up time, application load time, etc? Would it really be a perceptable
difference to my system?

Just to save some time, I'm not really interested in other solutions yes, I know more memory would help, etc). I'm just curious about his one solution.

Thanks in advance for your help!

Lex
 
IMHO adding more ram would boost the speed a lot more than adding a ata66 card.

Since you have an old 6 gig ata66 drive,it will not faster on the ata66 since it's not fast enough to even saturate the at33 interface.
 
A memory upgrade would result in more disk-cache and would therefore be a better upgrade, even if you think about the improved performance of a RAID 0 setup.

As the Promise software RAID steals it's RAID power from your CPU, which is not too fast it would certainly be a bad buy, all IDE RAID's that are software driven are bad choices when it comes to advances in total performance. Adaptec has an IDE hardware RAID solution, but even that one would be a bad buy in your case as the two IDE disks will use twice the CPU power compared to a single setup.

Then there is the issue of possible data loss, if you get a minor error with your RAID 0 setup, you can loose ALL of your information, actually a RAID 0 setup more than doubles the chances of data loss.

Get more RAM, a faster CPU, then if that is not enough, start looking at SCSI disks and controllers.

Patrick Palm

Am speaking for PC Resources
 


<< He wasnt even talking about a raid card,Just the Ultra66 addon ata66 card. >>



Man, i need to slow down on the reading part... ;-)

The Ultra 66 would do very little, get more mem instead... (i read your post again, but i had to give you some suggestion)

Patrick Palm

Am speaking for PC Resources
 
Not all promise products are raid .😉

If he had a brand new ata100 fast drive and he wanted to keep his bord &amp; cpu then a ata66 (100) addon card would be perfect.

But for an old 6 gig HD vs only 80megs of ram running win98se,I say more ram is needed first.
 
So, basically, with my system, there will be almost no perceptable difference if I get the promise ultra 66 card? If I can get the card for near free, is it worth even taking apart my pc and putting it in (which is a big pain for me)?

Thanks again.
Lex

 
Not with an old 6gig drive.

at best you'll get a 5% boost.

adding 64 megs of ram will make it much faster.
 
I'm curious, what's the difference if I have an old 6GB UDMA66 drive versus a new, larger UDMA66 drive? Why does that matter?

Thanks!
Lex
 
all IDE RAID's that are software driven are bad choices when it comes to advances in total performance

This is rubbish

If you run any system benchmarks, winstone, sysmark and the like, IDE Raid 0 will outperform single HDD by AT LEAST 10% in nearly all situations.

10% is at least 1 to 2 CPU speed grade if you want the boost winstone/sysmark score by upgrading just the CPU
 
with only 80MB of RAM, I'd agree that your best option is to get more RAM. Getting the promise card for your setup would be like getting Z-rated tires for a Yugo, wouldn't help at all cause your HD wouldn't really need it.
 
When I added an ATA-100 drive and a Promise ULTRA-100, I had a noticeable increase in performance over ATA-33. I installed an ATA-66 in a friends computer and he also noticed the improvement. Of coarse, I had 192 RAM to start with!!!
 


<< If you run any system benchmarks, winstone, sysmark and the like, IDE Raid 0 will outperform single HDD by AT LEAST 10% in nearly all situations. >>



I know this, but what if i don't think these benchmarks are any good? and downgrading your CPU with software RAID is not a good idea.

I have benchmarked more than 5000 systems in real world situations, and i do not use some benchmarking program because very few of my customers run only benchmarking programs on their PC's, they run real world programs and so i use those programs instead.


Patrick Palm

Am speaking for PC Resources
 
What are you talking about

Winstone and Sysmark dont run real world programs? It is WinBENCH that dont run real world programs.
It is THESE benchmarks, WINSTONE and SYSMARK, that shows 10% improvement by using IDE RAID0. These are SYSTEM benchmark that stress CPU, Cache, RAM and HDD. By changing from non-raid to RAID0, at least 10% performance is gained on the same setup. To achieve the same 10% score by upgrading the CPU, you must go from a P3/700E to a P3/800E or more likely a 850E.

The CPU Utilisation is INSIGNIFICANT. Using High Point 370 with 2x IBM 75GXP drive only shows around 12% CPU usage on a Celeron 850 in HDTach. I dont see that is a huge burden on a CPU.
 
all IDE RAID's that are software driven are bad choices when it comes to advances in total performance.

Again, total rubbish.
Using even software RAID solution, on SYSTEM benchmarks shows at least 10% improvement. How is that a 'bad choice' to increase total performance. you sound like as if RAID 0 will decrease overall computing speed., which is TOTALLY wrong.
 


<< It is THESE benchmarks that shows 10% improvement by using IDE RAID0. The CPU Utilisation is INSIGNIFICANT. Using High Point 370 with 2x IBM 75GXP drive only shows around 12% CPU usage on a Celeron 850 in HDTach. I dont see that is a huge burden on a CPU. >>



THOSE benchmarks are INSIGNIFICANT. very few people use benchmarks for anything else than benchmarking, i could use sysmark to show you how much better a Celeron is compared to a Duron, but that doesn't mean that it is true.

There are two major flaws with IDE RAID 0, the first one is related to data loss, the risk more than doubles, and then there is CPU utilization.

12% is nowhere near true, in real world tests (even if a utility uses real programs to test with, that doesn't mean that those programs are used) it is closer to 20-25% during high loads.

That would be like having a 650 T-bird instead of a 1 Ghz T-bird, and you say the performance hit is insignificant????????? Even with 12% the utilization is HUGE.

So you improve your disk troughput with 10%, increase the seek time with the double (20%) and downgrade your CPU performance with 12% (20-25), not to mention you increase the risk of data loss (and remember with RAID 0, ALL data is lost) with more than 100%. This is your idea of a good improvement to your system??

Patrick Palm

Am speaking for PC Resources
 
So you improve your disk troughput with 10%, increase the seek time with the double (20%) and downgrade your CPU performance with 12% (20-25), not to mention you increase the risk of data loss (and remember with RAID 0, ALL data is lost) with more than 100%. This is your idea of a good improvement to your system??

You seriously dont understand what I am talking about.
EVEN WITH ALL THAT CPU UTILISATION, it is STILL faster at least 10% overall.
- Transfer is faster up to 100% (usually about 50-60%)
- Seek time DOES NOT increase much at all, 20% is exagerrated.
- During peak load on even a Celeron 850 is only 12%. Less if you have a Tbird or Duron I can imagine.

Anyway, SO WHAT about the 12% extra utilistaion when your HDD read is done in a LOT less the time. Let me give you an example. Let say a program file is 10MB on the HDD and CPU processing takes 1sec on a certain CPU. Lets suppose a SINGLe HDD can read at 5MB/s sustained. Due to RAID 0 is not 100% effective at transfering (also seek time doesnt improve). I estimate it reads at 7.5MB/s sustained.

Calcuations of Single HDD time load
- 10MB / 5MB/s + 1 second. = 3seconds to load this program

Calculations of RAID time load considering 12% cpu utilisation
- 10MB / 7.5MB/s + 1sec * (1+0.12) = 2.45sec...

If you think 12% is Way too low then I will also do a 25% cpu utilisation calculation.
- 10MB / 7.5MB/s + 1sec * (1+0.25) = 2.583sec

Hell, even if you say it has a 50% cpu utilistaion, time is still only
- 10MB / 7.5MB/s + 1sec * (1+0.5) = 2.833sec

I didnt even include the 5% cpu utilisation even on single drive. If i include the 5%, then the relative time of the Signle HDD is
- 10MB / 5MB/s + 1sec * (1+0.05) = 3.05sec

Well you see, in this situation, RAID 0 theritically will give you a 2.45sec load time, and Non-RAid will give you a 3.05sec load time. A 24.5% speed increase. This is like upgrading a Tbird 800 to a Tbird 1000. Surely it is theoritical, but overall there is still a ~10% speed increase.


Risk is yes 100% higher, however that is STILL VERY LOW. For the last few years, I have had like 10HDD and NONE of them actually broke down.
 


<< Well you see, in this situation, RAID 0 theritically will give you a 2.45sec load time, and Non-RAid will give you a 3.05sec load time. A 24.5% speed increase. This is like upgrading a Tbird 800 to a Tbird 1000. Surely it is theoritical, but overall there is still a ~10% speed increase. >>



Even if your theoretical calculations were true, they only tell you how long it will take for a program to load, the difference is too low to be mentioned.

There are programs out there which make your CPU run at a constant load of 100% when used, if you were to use RAID on these, what would be the outcome?

I would agree, if you spend most of your time loading and closing programs RAID 0 would be a great solution, if you actually use high-end cpu intense software it is not.

As for the safety... Well, the risk is MORE than 100%, not only that, you do not have to crash a HDD for the whole RAID info to be lost, all you have to do is lose a little piece of info, or get an error in the RAID info table.

I will end my post with the last centence of many of my posts: No information is worthless enough to deserver IDE RAID 0.

Patrick Palm

Am speaking for PC Resources
 
Talk about CPU utilisation of RAID under Heavy CPU loading situation, the RAID solution still wins...

Remember you 0% CPU is used when the RAID ISNT loading. Considering the Pure speed of RAID is 50% faster than Single Drive, and the utilisation of RAID is NEVER as high as 50%.

RAID - 112% * 50% = 0.6272 relative time of READING
Single - 105% * 100% = 1.05 relative time of READING

if the hdd isnt loading, there isnt any CPU utilisation difference between raid and not raid. Therefore the argument is that IDE RAID 0 is takes away your cpu time is NOT a valid point, as the REAL HDD loading time is done at only half the time.

I have to talk about the RISK.
It isnt over 100% compared to Single HDD. It is just on 100%. If you can corrupt anything on a RAID partition, you can do it HALF AS EASY as you can do it on a single drive. But honestly that is still low. It is like overclocking. There is the risk that you have to take for the ADDED speed, such as overheating and shorter life span of the product. I found doing this ok as long as the chance is double, but STILL very low. Another thing, if you have got 2 HDD but not in raid mode storing your data, EVEN if you have 1 died on you, HALF your data are gone. MOSt of the time, things are useless if you only have half of it there anyway. It is unfair to compare a FAULTLESS single single drive and compare it to a faulty RAID dual drive. You must compare a Faulty Dual drive NON-RAID also.
 
OK, bickering about RAID aside...

I rather doubt than an old 6 GB drive, ATA66 compliant or not, could saturate even an ATA33 controller. More mem, or a CPU/mobo upgrade are your best bets.
 
I think PCResources has a serious problem against IDE raid.

Any chance he gets,even if the topic is not regarding IDE raid he always barges in tries to put down ide raid and preaches that SCSI is GOD.

Not everybody can afford those nifty scsi cards &amp; 1500rpm drives.

I see ide raid as a very nice solution for somebody who needs a nice little speed boost and is on a budget.

now i am going off topic too . 😉
 
Lex, if it is &quot;close to free&quot; -- I think it's worth it. For one thing, it's just cool. For another, you can potentially free up an IRQ. Promise cards use 1 IRQ for both channels, where as your M/B uses 2 IRQs for the same task.

As to the diff b/t a 6GB U/66 drive and a larger U/66 drive, it's simply a matter of the transfer speed of the drive. Assume a 60 GB U/66 drive with otherwise similar characteristics. The 60GB drive would (theoretically) have 10x the data density, so 10x as much info is traveling under the head in a given period of time, so the data transfer rate is potentially 10x as great. Where a 6GB drive will never saturate even U/33 (maybe not even PIO4), a 60GB drive certainly could (in a burst).
 
You must log in or register to reply here.

TRENDING THREADS

Back
Top