Transfer speeds of hard drives

[dchilder]

I've got a general question I've wondered about for a little while. Why is it that hard drives, rated at 100 MB/s plus, do not transfer files from one to another at anywhere near that speed? I know that the Ultra ATA xxx is a theoretical upper limit, but read this case in point:

I'm cloning one hard drive to another. In this case, I'm running them both on the same EIDE chain, which is Ultra ATA 66. I SHOULD be getting a max transfer speed of 66MB per second, or 3960MB per minute. However, the average transfer speed for this operation was more like 143 MB per MINUTE.... ridiculously slow. Does this have anything to do with the cloning process that I used (two ways, Wester Digital's data lifeguard utility, and Disk Image)? It goes into a DOS mode to run the process, and I thought that might slow it, but this rate seems to be similar to those I see in Windows when I transfer very large files from one drive or partition to another. Just wondering....

David

 

[glugglug]

There is not a single IDE drive on the market today that has a transfer rate above 70MB/s even on the fastest part of the disk (one of many reasons why those expecting an improvement from SATA are delusional), or even a SCSI drive that can get more than 80MB/s at any point.

If your drive is old enough to be ATA-66, it probably has a maximum transfer rate below 10MB/s.

Also, IDE does not support direct disk-disk transfers, so they all go to your system memory somewhere and then go back through the bus at a later time -- the 2 drives are NOT used in parallel, unless you put them on different IDE buses.

 

[Shalmanese]

The same reason 100Mbit network cards never go at 100Mbit and 12Mbit USB doesnt go at 12Mbit. 66MB/s is the rating of the CABLE. If you hooked it up to an oscilloscope and pumped a preset 66MB/s stream though, the other end would be able to read it fine. The HD's can be anything, although, obviously not more than 66MB/s.

The benifits of SATA arent from performance standpoits, its thinner, longer cables and a less antiquated protocol.

 

[TerryMathews]

It's not just that: a U/66 controller is capable of understand 66MB/second. No IDE drive on the market can sustain that transfer rate (yet), but there are plenty of drives that can peg that in burst transfers. Most of the WDxxxxJB series can.

To prove this to yourself, appropriate a fairly fast IDE drive (Like one of the JB series drives I mentioned). Hook it up to a controller you can control the DMA mode on like your motherboard's chipset. Set for U/100 and run hdtach or something similar. Write down the performance numbers. Then, reboot and set the interface for U/66. Run hdtach again and you'll get a slower transfer speed. It won't be much of a difference, but it will be there and will be repeatable. That is the beauty of burst transfers.

 

[isaacmacdonald]

this isn't really an answer. it's rather a related observation I made the other day.

having abysmal amounts of fragmentation on my download drive (a wd 1200jb) due in part to modest 4kb cluster size, I decided to convert all of my media drives to 64kb clusters. Anyway, after the first drive was reformated I transfered 80 gigs of movies from drive A (WD 80gig 5400/2mb) to drive B (WD 1200JB 7200/8mb). It took about 20 minutes to complete. After formating drive A (64kb clusters) I did the reverse and copied the 80gigs back. It took more than 40min to complete the transfer.

Simple observation, but interesting how dramatic a difference there was between the 5400 and 7200 write speed (perhaps also attributable to platter sizes).

 

[capybara]

Originally posted by: glugglug
There is not a single IDE drive on the market today that has a transfer rate above 70MB/s even on the fastest part of the disk (one of many reasons why those expecting an improvement from SATA are delusional), or even a SCSI drive that can get more than 80MB/s at any point.

If your drive is old enough to be ATA-66, it probably has a maximum transfer rate below 10MB/s.

Also, IDE does not support direct disk-disk transfers, so they all go to your system memory somewhere and then go back through the bus at a later time -- the 2 drives are NOT used in parallel, unless you put them on different IDE buses.

glugglug got it Xactly right, and yet everyone ignores his answer ???? wtf ????

 

[isaacmacdonald]

umm... except transfers from caches should be able to saturate the the bandwidth provided by ata-66/100. With a growing trend to boost cache sizes, this IS pertinent.

 

[glugglug]

Originally posted by: isaacmacdonald
umm... except transfers from caches should be able to saturate the the bandwidth provided by ata-66/100. With a growing trend to boost cache sizes, this IS pertinent.

If its cached in the 8MB (or more likely 2MB) HD cache its also cached by the OS in RAM anyway and doesn't go across the bus.
Only exception is read-ahead and write buffering.

Even if the cache is making a burst faster, the time to send the ENTIRE contents of HD cache to the system across ATA-100 is less than 0.1s, and even that 0.1s burst will never happen outside of a very synthetic test.

 

[Peacekeeper100]

One more thing is that both HD's are on the same bus. This means they are sharing the same 100Mb line. So in theory each drive only gets 50Mbs. Then you have to account for latency and errors, this probably reduces them down to 30Mbs even if they are new fast drives. Plus I don't think each drive can be read at the same time. This means the software has to find the data, grab the data, find where the data is going, then send the data. This all slows down the process.

 

[isaacmacdonald]

I don't know about that shared bandwidth thing. Doesn't the traditional IDE interface only allow for consecutive operations? That would mean of couse that the overall performance would be cut in half, but that wouldn't directly be a result of sharing cable bandwidth and shouldn't come into play when ascertaining the neccesity of ata-100/66.

In a related subject, don't these read ahead/write caches account for a lot of performance? Anandtech recently revised their review of the new 10k RPM SATA drive and indicated new firmware with these caches enabled were responsible for huge performance gains.

 

[foxkm]

Both of the Disk cloning utilities are DOS Based. DOS doesn't use any of it's own disk access modes.. Its Uses whats specified by the BIOS during the POST. If you have a newer AWARD or AMI Bios, there should be settings to force DMA modes and turn on 32 bit disk access. This may not be safe for all drives, but it will increase DOS performance. Since the drives are ATA66, chances are they are 20 gig and lower drives which most likely will have a burst read/write of like 25-30 megs a second.. If you divide that in half, you should be able to get close to 15 megs a second in DOS while cloning...

foxkm

 

[Hazer]

Umm, wrong on a few acounts:

Having both drives on the same channel is bad to begin with, here's why:
They dont share the bandwidth "So each HDD gets 50MB/s" that is wrong. Its worked on a master/slave system. That means only one drive can access the 32 bit path at one time. The master has precedence. But when the master drive recieves a special signal, it then replies back a signal to notify the slave drive that it can use the IDE cable.

So transfering data from one drive to another on the same IDE channel works like so: CPU receives command to transfer data from drive A to drive B. CPU issues commnd to PCI. PCI issues command to IDE bus to HDD A to write data segment to main system mem. HDD A receives command and verifies that command was issued to HDD A and not B (master/slave verification). HDD A reads platter for data and stores into buffer. Buffer contents are then sent to system mem through PCI bus. PCI bus signals CPU operation completed. CPU then issues command to PCI bus to write data to HDD B. PCI bus signals HDD B to receive data from system mem. HDD A receives the command, determines command is for HDD B, sends signal to HDD B. HDD B receives signal and responds readiness to receive data. PCI bus activates transfer of data from system mem to IDE channel. HDD B writes data physically to platter.

ATA IDE channel consists of (80 pin connector) 32 lines for transmitting data. 32 lines for addressing data. The rest are for signals to read, write, HDD boot detection, burst transfer verification, ready signals, etc etc. The address lines and the data lines are connected to both HDD sharing the channel. 1 or 2 lines are used to determine which HDD will control the channel at one time. So only one HDD can access these lines at one time.

The biggest determining factors for speed of HDDs has to do with seektime (its in ms, while transfer rates are microseconds), RPM, and disk density. SCSI drives perform better because of RPM, and seektime. Also, they perform better due to drive motor quality. Hence why they are more expensive and less in demand.

You could gain a marginal, and probably unmeasurable in realtime, difference if you transfered the data by having each HDD have its own IDE channel. This would release the need to verify master/slave conditions. But its a drop in the ocean. Burst rate is a moot point also. So what if the buffer can burst the data on the PCI channel directly to memory if it takes the same amount of time to fill the buffer? Its like filling a bucket with water. You can fill it faster if you fill up a balloon first, then pop the baloon over the bucket. But it takes the same amount of time to fill the baloon beforehand. Burst transfer works best when look-ahead reading is used and the data is sequential and small enough to be less than the buffer size. It doesnt mean alot when the filesize you are transferring is 100's of MB large. Also, transfering small files means the HDD has to find each file and wastes time because of the seektime rating. Each file will take (for IDE) 9 ms to find.

Whats 9ms to transfer speed? ATA133 means transfering 133MB/s. The PCI bus is 33MHz and 32 bit. 33 million cycles x 32 bits divided by 8bits/byte = 133MB/s. Or 133 kb/ms. 9 ms is wasted trying to find a 64kB file that can transfer in 0.5ms (IDE ATA bursted).

SATA is gonna be a joke. At least for single drives. It wont have much in the way of better tech for the platters, motor, and magnetic head. Its all about the interface. It will have a higher capacity to transfer data by using a serial signal instead of a parralel signal. But the true limitations of a HDD that can be seen by the naked human eye will not be improved. Realworld operations will hardly be increased. They will be increased, the interface will have more headroom for inovation, but the increase will likely be about 10%. Considering the average transfer rate for your everyday IDE HDD is around 45MB/s in good conditions, getting to 50MB/s doesnt really impress me. The true limitations need to be addressed more than the interface.

 

[drag]

Sata is not a joke. It's just plain better technology, so what if even the newest HD won't fill up current IDE ribbon cables? Much of the complexety of modern HD's is just to deal with the crap that goes on in those ribbons. They have no sheilding to speak of and they are fragile, and bulky. Not only that you can get away from the horrid master/slave BS that plagues us with all this legacy hardware... It's important to introduce new technology now, because in the near future we are going to see drastic changes in PC technology, they need to get as far from that 20+ old parallel cabling technology inorder to introduce new invovations.

Could you imagine what will happen now if they were introduce new super fast ide HD that actually came close to filling up the 100Mb per second limit? Massive data corruption, that's what. Who would want these new super fast HD's? who would be willing to pay the exorbiant fees of the newest technology even faced with the uncertainty of any new tech breakthoughs? Well the PC hobbiest that's who. Who would get blamed if you had a overclocked motherboard that worked fine until you plugged in a new HD? now you got massive data corruption, windows XP gets wiped out in a week... would they actually do reasearch and learn to blame the neon lights and 19-24 inch ide cables wrapped in black tape (or there 12 dollar metal braded "sheilded" cable? ) Are they bother going to run tests compare copied files to find the source of bit rot is the hyper active IDE controllers on the overclocked bus?

Nope they would conclude: The HD is obviously running on the ragged edge of it's materials..(a reflection of overclocking various motherbords) flaky technology thats what it is already sent out pre-overclocked/stressed from the factory, obviously can't be trusted, probably gonna wear out in a year, released it to soon. They should of know better than to inflict buggy hardware on US! So you boys in marketing better stick with those SCSI's drives, you don't want what happened to my windows XP happen to your finacial spreadsheets....

And what about all those years of crappy case design bottom dollar, twisted, folded and smashed cabling? Sure they can get away with it now, but do people realy want to pay Compaq or Dell a extra 50 bucks for a better case design so they can spend a extra 150 bucks on a high performance/lowercapaticy harddrive with a spotty reputation? who ever would spend money to develope the technology to get that fast would be almost suicidal waste of money with today's motherboards...

They way I see it now the current standard of IDE contolers/cabling and harddrives technology has fullfilled it's potential. The move from 40 to 80 wires was just a stop gap. Those extra 40 wires are grounded... spaced evenly between active wires, just to keep the super fast electronic signals from unsheilded wires from interfeiring with the others. (in other words reduce crosstalk) It's got the same 40 wires to carry info as its older brethern which in turn in a stopgap solution for a 20 active/20 grounded wire solution from the wayyyyy back when wire technology was REALY crappy. It's fragile, error prone, relatively slow and has a maxamum of 4 devices, where only 2 of the devices will be able to reach their potential IN A IDEAL SITUATION.

At least case modders will have something better to spend their money on then buying spagetti cabling smashed together (oh ya weaved together like ethernet...) in a stiff plastic tube that will slowly rot their data to mishmash.

To say SATA is a joke is like saying that parrallel printer cable are all that are needed... Why would anyone want USB, there is now reason why someone would fill up the bandwith in those wire with just todays printers? Hell they got fireware and usb 2 now too!!! I think its just a scam to get people to buy the new motherboards... I sticken with my slot-7 pentium! and that's that...

 

[Hazer]

OK. Misrepresented. Yes SATA is a better interface.

So lets say the next big thing is is to use opto-electronic cabling for signal transfer. But, what does that mean if the average transfer rate (due to the physical limitations) are still around 50-80 MB/s?

They still havent been able to make a HDD that can fill up the ATA100 standard in sustained transfer rate. And your saying that soon HDD tech will have drives that can reach this. But thats the problem. They are not. They are pushing the physical limits of ferroceramic media. They have stopped trying to squeeze what little they can, and have started using 'tricks' like read-ahead and burst transfer rate. These tricks do nothing to improve the sustained transfer rate limits. Either they have to change media, or the way they look at how platters are used. Increasing the interface when the transfer rates are still limited means no performance increase. Thats what the joke is.

 

[glugglug]

Originally posted by: drag
Sata is not a joke. It's just plain better technology, so what if even the newest HD won't fill up current IDE ribbon cables? Much of the complexety of modern HD's is just to deal with the crap that goes on in those ribbons. They have no sheilding to speak of and they are fragile, and bulky. Not only that you can get away from the horrid master/slave BS that plagues us with all this legacy hardware...

The only ways in which SATA is "better" are:
1) reduced cost from lower pin count <- these first 2 are the ONLY ones which manufacturers could give a rat's ass about
2) increased hype/marketing factor to dump new mobos/controllers onto the market
3) command queueing (its actually in the protocol spec that drives can reorder the operations they are told to do if the firmware determines it will improve overall speed. AFAIK no SATA drive does this yet but a few IBM PATA drives support their own proprietary equivalent).
4) Varied pin lengths designed for hot-swapping

It's important to introduce new technology now, because in the near future we are going to see drastic changes in PC technology, they need to get as far from that 20+ old parallel cabling technology inorder to introduce new invovations.

No. It isn't. Most of the negative attributes of PATA were fixed with the ATA/33 spec. The inherent scalability limits before they would need to change it are 10 times higher transfer rates than SATA BTW due to the parallel nature.

Could you imagine what will happen now if they were introduce new super fast ide HD that actually came close to filling up the 100Mb per second limit? Massive data corruption, that's what. Who would want these new super fast HD's? who would be willing to pay the exorbiant fees of the newest technology even faced with the uncertainty of any new tech breakthoughs? Well the PC hobbiest that's who. Who would get blamed if you had a overclocked motherboard that worked fine until you plugged in a new HD? now you got massive data corruption, windows XP gets wiped out in a week... would they actually do reasearch and learn to blame the neon lights and 19-24 inch ide cables wrapped in black tape (or there 12 dollar metal braded "sheilded" cable? ) Are they bother going to run tests compare copied files to find the source of bit rot is the hyper active IDE controllers on the overclocked bus?

Nope they would conclude: The HD is obviously running on the ragged edge of it's materials..(a reflection of overclocking various motherbords) flaky technology thats what it is already sent out pre-overclocked/stressed from the factory, obviously can't be trusted, probably gonna wear out in a year, released it to soon. They should of know better than to inflict buggy hardware on US! So you boys in marketing better stick with those SCSI's drives, you don't want what happened to my windows XP happen to your finacial spreadsheets....

That's what WILL happen with SATA. In order to achieve the same transfer rate as PATA it must be clocked 10 times higher. This puts the current SATA/150 only slightly lower frequency than your cellphone & microwave, and overclocking it will put it squarely into the same band. The cost-centric nature of the whole thing encourages the cables to be poorly or completely unshielded.

They way I see it now the current standard of IDE contolers/cabling and harddrives technology has fullfilled it's potential. The move from 40 to 80 wires was just a stop gap. Those extra 40 wires are grounded... spaced evenly between active wires, just to keep the super fast electronic signals from unsheilded wires from interfeiring with the others. (in other words reduce crosstalk) It's got the same 40 wires to carry info as its older brethern which in turn in a stopgap solution for a 20 active/20 grounded wire solution from the wayyyyy back when wire technology was REALY crappy. It's fragile, error prone, relatively slow and has a maxamum of 4 devices, where only 2 of the devices will be able to reach their potential IN A IDEAL SITUATION.

IDE was never 20 pins or 20 signals ever. Most of the base 40 pins have always been used. 20 pins with an 8 bit data bus and a few pins for control would get you at best an addressing space of 128K on the drive (assuming standard 512K block sizes). I'd put good money on SATA proving more "error prone". The only real point you've made here is that only 2 of the 4 devices on 2 IDE buses can reach full potential -- the master/slave system IS a back hack, but can be ignored.... just put one drive per IDE chain, everything is master, no significant disadvantage vs. SATA.

To say SATA is a joke is like saying that parrallel printer cable are all that are needed... Why would anyone want USB, there is now reason why someone would fill up the bandwith in those wire with just todays printers?

USB1.1 was a sick joke, and the old Parallel port IS/b] FASTER AND MORE RELIABLE!!. BTW there were plenty of parallel port devices that saturated it over a decade before USB.
Sure USB2.0 and firewire are faster than a parallel port but what do you expect from a technology that comes 20 years after the one its being compared to was last revised?

 

[drag]

I don't know about all that.... have you ever tried hooking up a keyboard and mouse and a printer and a pen drive and a digital camera to a parallell port? I sure that'll be much better/faster/stable... Compared to newer technology the parallel printer port is hopelessly outclassed. But I am sure that in your world all you should be able to do with extra ports on the back of your computer should only be used for printers and joysticks... and maybe a scanner if your lucky... You kinda missed my point (my little exitable freind) which in turn kinda proves the point in a perverse way... Just because the HD's for right now aren't going to be made to go any faster right now doesn't mean that it is not worth adopting... What if somebody came out with a affordable 3-d storage medium? or came out with harddrives that used some sort of magnetic feild create 'air' bearings in a vaccum that would allow platters to spin at a 1,000,000 rpms or something equally freakish... Or what if they created a affordable way to have massive RAM storagage... or a secondary RAM stash of cheap 3 gigs of very cheap high latency sdram to augment your main DDR RAM? kinda like a L1 and L2 caches? Or howabout using the high speed IDE bus to set up parallel proccessing motherboards? I know they can network thru scsi cables, we use it at work for dealing with a freakish government post office computer...

I am sure when they designed the 8086 computers that they would know that we would be able to plug a camcorder into the back of one of it's desendants and push play to send the video stream into space and back down again so that anybody around the world to could watch that stream.. with all this happening in a fraction of a second.... and from a teenager's basement bedroom...

So what if the SATA spec sesze that the clock for the electrical current is 10 times faster then a PATA? it's that superior design that allows this to happen. Its not like the electricity has to move ten times faster or anything. Its the very design of the PATA devices that make them so suspectable to interference, the electrical waves and limits the scalability of them.. And just because something is cheaper to manufacture doesn't make it inferior. AND I AM SURE, that making stuff as cheap as possible is NOT the manufacture's main priority... I'd hate so see what happens to a business that follows that flawed logic... They want to give a reason for people to actually BUY THEIR PRODUCT. you know with new inovations, faster more inexpense harddrives... more efficent and userfreindly etc etc... If ford thought the way your accusing the HD/motherboard manufactures.. we all be driving model T cars that were inflatable with 2 horsepower motors powered by cow farts and beans...

do you suppose that since the SATA spefications expect scalability up to about 1.5Gb of bandwidth that due to the inherint nature of the wonderous PATA that runs at a frequency 1/10 of that of SATA that the PATA will be able to do 150 Gbs of bandwidth with the same cost?

edit: oh and before I forget, even though the SATA must use 10 times high frequency (which I would like to do more research about this if you would be so kind as to provide a link...) it also does it at a fraction of the voltage of PATA.. But I suppose that doesn't matter...

 

[neologan]

I suspect that though SATA is being firmly touted as a desktop replacement for 2003, in 2004 you'll start to see the real reason behind SATA. There are already demo's of huge SATA raid controllers that will probably be aiming for SCSI RAID in the next 2-3 years and be far cheaper. I find it odd that desktop users are actually getting the technolgy before the server market, that's got to be a first. SATA are hot swappable and as someone said above "command queueing" is now implemented which has until now has been a SCSI feature. I definetly think that a 600m/bs SATA controller with 20 or more SATA 10,000rpm drives will kick the sheeze out of any SCSI config. in terms of price/performance.

I'm no expert, if I'm wrong please say so

neologan

 

[drag]

Ya I think that it is normal now that new pc hardware is going to be inflicted on the PC crowd before everyone else, because for lots of people having the newest and flashiest stuff is more important then the money. THat way the manufactures can get SATA's teeth cut on a group of people that tend to be very enthusiastic and very abusive towards there hardware. So once the obvious bugs get worked out they can appeal to the more high-end crowd, which tends to like to error on the side of being overcautious.

Could you image RAID support being a option like onboard NIC when choosing a new motherboard?

 

[glugglug]

Originally posted by: drag AND I AM SURE, that making stuff as cheap as possible is NOT the manufacture's main priority... I'd hate so see what happens to a business that follows that flawed logic... They want to give a reason for people to actually BUY THEIR PRODUCT. you know with new inovations, faster more inexpense harddrives... more efficent and userfreindly etc etc... If ford thought the way your accusing the HD/motherboard manufactures.. we all be driving model T cars that were inflatable with 2 horsepower motors powered by cow farts and beans...

That's got to be the funniest thing I've read in awhile. If you really are naive enough to believe that senseless penny pinching and cost cutting is not the main priority of pretty much every company out there, then you obviously haven't been out in the real world yet. BTW, Ford and other car manufacturers DO think that way - if there was something that could double the life of your car but would increase the cost by $10, you could be sure as hell it would never be done. Why do you think so few cars have mirrors integrated into the visors when most people end up getting aftermarket ones? Answer: it would increase manufacturing costs by like $0.25

 

[Hazer]

OK. Obviously no one gets my 'joke'. Why not ditch the SATA serial wire commands and just stepup to optical? Its way faster, and has no magnetic interferance. You can have the cable cross the room without signal integrity degredation. Easily get 10GB/s transfer rate. Crap, in Europe, the had optical going about 1TByte/Km !

But where are the HDD that can push out that bandwidth? Fastest IDE drive is 50MB/s? And only becuase IDE is the cheaper drive, They can put the motors and stuff from a SCSI drive (quadruple manufacturing costs) and make it work just as fast as a sigle SCSI drive.

Whats the fastest single SCSI drive? 80 MB/s?

It took 8 years to go from 20MB/s to 80 MB/s for HDD transfer rate. It took the same time to go from 200MHz CPU to 3GHz CPU. Go talk to some hardrive manufacturers RP people, and ask them when they will release their 500MB/s hardrives. They will s**t thier pants. I have some investment in this, and Im finding out that ferroceramic technology is reaching the very limits of what it can do. They have refined platter technology quite a bit to go from just 40MB/s to 80MB/s. They are not looking to jump from 80MB/s to 200MB/s anytime soon.

So the joke is, creating an interface that can transfer 1.5GB/s for a <100MB/s device is kinda useless. Yes, they will eventually reach the 150MB/s limit, but to do so anytime soon will cost the enduser quite the bit of cash. Forget about reaching 1.5GB/s.

BTW: In order for SATA II to be viable, the PCI bridge is gonna have to change its standard also. Thats the big secret they wont say. Its a 32 bit bus that runs at 33MHz for most systems. There is 66MHz systems. Even so, fastest the PCI bridge can push 32 bits to the main system memory is 133MB/s for 33MHz, and 266MB/s for 66MHz.

 

[conehead433]

SATA is great because it leads the masses to buy something they think is better.

 

[neologan]

BTW: In order for SATA II to be viable, the PCI bridge is gonna have to change its standard also. Thats the big secret they wont say. Its a 32 bit bus that runs at 33MHz for most systems. There is 66MHz systems. Even so, fastest the PCI bridge can push 32 bits to the main system memory is 133MB/s for 33MHz, and 266MB/s for 66MHz.

Eh, what are you on about? At present SATA conrollers are not integrated onto any chipsets Southbridge so yes you are confined by the PCI bridge, but in the next few months (with the introduction of Springdale i think) SATA will be integrated and wil no longer be held back by the PCI bridge.

Doh doh doh to you m8

neologan

 

[glugglug]

BTW, yes, PATA cabling is crappy and being used past its design intentions, however, going serial is not a solution IMO. The proper move would be to use twisted pair cabling, like SCSI, rather than just tying the extra 40 wires to ground.

 

[neologan]

I don't see why everyone hates SATA so much. Most people are pissed at the low performance of hardrives but are taking it out on SATA. SATA is a controller standard and can't all of a sudden make hard drives faster. SATA is capable and will be supporting new technolgies in the future because of the scope for high bandwidth. PATA and SCSI will be phased out in the next 5 years, I'm sure of it. SATA will be around long after this. As I've said in an earlier post, SATA as it is right now is a joke but it's not really meant to be any more at this stage because hard drive technolgy is a joke too. The next big jump for SATA will be with high end RAID and will be far cheaper than SCSI.

So I ask you, what do you really have against SATA considering the fact that hard drives at present are slow, what is it you really need that it doesnt give you?

neologan

 

[Hazer]

What I have against SATA is that its being marleted as "the new thing to replace PATA". They cant even fill-up the 'old' PATA system, so they design a new one? First SATA drive out costs $50 more than its PATA coutnerpart, and is shown to go 10% slower. Great. I dont really have anything against SATA. Its just that its not needed. Its like putting PC3500 into a motherbaord with an unclocked Duron.

Improve the drives first. And one thing that IS getting me is this: Ferroceramic drives are reaching their limits! They have already dualled the magnetic coat. They have created special film to increase the head fly. They have gone as dense as they can go with this medium. Its not the same as with CMOS technology. You make it sound like HDD are gonna reach like 200MB/s at some point in the near future. Im telling you right now, they wont. Not without creating a whole new machanism. And if they do that, SCSI will look cheap compared to the new medium. HDD tech has been at a standstill for the past 8 years. They have introduced rekatively small improvements and tricks to squeeze out just a little more.

The big joke is that why create a 1.5GB/s transfer path if there is no (and wont be for decades) device that uses that speed?