Over the past few years SATA has become a standard interface on hard drives and is starting to show up in many peripheral devices. Today we’re taking a look at two similar hard drives to see if there’s a performance difference between SATA and the older Parallel IDE standard.
i dont have enough money to affort a sata mobo *and* harddisk(s) .. so i stick to parallel …
i dont have enough money to affort a sata mobo *and* harddisk(s) .. so i stick to parallel …
SATA equipment is about as cheap — sometimes cheaper — than the PATA stuff. As time goes on, SATA will become much cheaper while PATA is phased out.
Yep. I’ve had to buy a couple of mainboards for people reciently and it’s getting quite hard to find a PATA only board. Just about every single one comes with SATA now (alongside PATA) and they’re not expensive at all.
null
Yea, the server just went into a tailspin. I’ll have it figured out as soon as I get access. Sorry `bout that.
Links seems bogus….
The theoritical difference between SATA and PATA are not much, 150 mbps for the former and 133 mbps for the latter. But in my experience most specially when doing
video editing, SATA is the winner.
No no no….not true at all…
I’ve used an ATA 133 and my video editing experience is *exactly* the same as that on an SATA. At this point, it’s the subsystems BESIDES the drive that makes the difference these days.
The only way we will see significant performance increases in the storage area will be when we come up with some new technology that is cheap and easy for the masses….
Has anyone tried this? Is it actually possible for that matter? It seems to be something I’ve heard about in the past, but I’ve no idea how it might happen.
Is this what you mean:
http://www.abit-usa.com/technology/serillel_new.php
Abit gives one away with some mobos.
At least, SATA II is faster. The perfomance jump I saw when I moved from a PATA mobo to a SATA-II mobo was quite noticeable. 1.5 GB/s doubled to 3.0 GB/s.
The only time you would potentially benefit from the faster interface is when reading from the drives cache. I’m yet to find a drive that tops 100MB/s yet so the 3GB/s means nothing for the most part.
At least, SATA II is faster. The perfomance jump I saw when I moved from a PATA mobo to a SATA-II mobo was quite noticeable. 1.5 GB/s doubled to 3.0 GB/s.
I don’t think you actually get the performance you say you do — either with the SATA or PATA drives.
We might not see much difference, but at least we got new plugs and a new spec so we can dump that 20yo trash
20 year old trash? Why I oughtta..
Oy…this is your argument? Have you tried to *configure* a new SATA device? Most of the controllers I’ve seen have RAID built in-a definite nightmare for “the average user” (and yes, average users *do* buy new Hard disks occasionally). All it does is make job security for us IT people (which may not be a bad thing, now that I think about it)
Ok, so is there a mirror for this anywhere?
Hey, that 20 year old trash still gets the job done. And for OS platforms that don’t handle freaky new ports & protocols, PATA is as common as sliced bread (and good too). My friend wanted to try out hobby OSs (or even just a WinXP/Linux dual-boot) and nothing but Windows could handle the hardware. Forcing PATA drives to be primary master doesn’t override the SATA boot sequence, so the only disk that will boot is the SATA. Its nice, but honestly we couldn’t find any performance boost for all the hastle.
Hey, that 20 year old trash still gets the job done. And for OS platforms that don’t handle freaky new ports & protocols, PATA is as common as sliced bread (and good too).
I know. Platforms like Windows XP make SATA impossible for Joe Average.
My neighbor is Joe Average and imagine his surprise when he bought a new computer with an SATA drive and couldn’t install Windows on it.
The problems are 1) Windows XP doesn’t support SATA 2) the driver was on a CD and Windows can’t load drivers from a CD during install 3) he didn’t have a floppy drive in the machine.
After pulling his old floppy drive out of his old machine, hooking it up to the new machine, copying files from the CD on to an old floppy using a 3rd machine, and restarting the Windows XP install, he was finally able to get Windows installed on his SATA machine.
On the other hand, I installed Linux on my floppy-less SATA machine and guess what. It just worked.
I know what you mean about OS platforms that don’t support those freaky new ports & protocols.
After pulling his old floppy drive out of his old machine, hooking it up to the new machine, copying files from the CD on to an old floppy using a 3rd machine, and restarting the Windows XP install, he was finally able to get Windows installed on his SATA machine.
This person is Joe Average? lol
It was an interesting comparison of performance, pitty the server seems to have died, hopefully it will be back online soon, as I found it a decent read.
*Very* sorry about that. The server is currently back online and I’m endevoring to find out what the heck just hit it.
It’s okay, just I found it interesting… no need to apologuise.
the only real test would be the same drive platters and heads with the two different interfaces. But of course any difference would really be the SATA or PATA implmentation not the protoccol. I sure there are PATA controllers that are 10-20% faster then other and the same with SATA so you get a crappy connection and chipset and the protoccol doesn’t matter. Hell I have two SATA interfaces in a W2003 server and the one gets 2 time the performance of the other with the same drive. Either way there is no way to really test these two unless you are comparing the whole system not SATA vs PATA but system Y with drive X and system X with drive W.
Isn’t one of the benefits of SATA that is supports more advanced protocol options like tagged queueing? Does the current Linux driver support this? Wouldn’t a more useful test be done on Windows, where presumably the driver support is much more mature? Determining the effectiveness of a hardware protocol based on a single Operating System and a single driver implementation seems a pretty useless test methodology.
I wonder how the results would have looked if the PATA drive would have had 8 MB of cache instead of 2 MB?
For a bunch of server I used to manage system images using G4U and an FTP server. (Think Enterprise Ghost for the cheap.) This worked great for all the systems that were using regular PATA hardware. However, drivers for SATA equipment haven’t really hit NetBSD yet (the basis for G4U) so I couldn’t make system images of the servers using SATA drives. I tried G4L (based on Linux, not NetBSD) but the keyboard/codepage was set to … russian or something. I ended up using a FC3 rescue CD and this command to do images, which is pretty straight forward:
dd if=/dev/hda
| gzip -c
| ssh backup@backuphost cat > /backup/imagename.gz
This is the dumbest benchmark I have ever seen, really.
If you take a drive that can do about 40MB/s and you put it on a 100MB/s bus and a 150MB/s bus and you test the speed of the drive on each bus, you should find THEY ARE ALMOST EXACTLY THE SAME! Oh my god! What a wonderfull thing you’ve discovered!
Come on, anyone who would think about this for 5 seconds could realise what to expect from the test. Now. Try this, get 2 high performance drives that can do 55MB/s sequential read and try them on each bus, with both drives going at the same time…simple math now:
55+55=110
We should get around 110MB/s performance out of them. On PATA we only get 100MB/s(probably less)….on SATA we get 110MB/s out of them. A 10MB/s performance boost. So what happens when we take 4 drives and do it? should we see 100MB/s out of PATA and 150MB/s out of SATA?
No. Because the 32 bit 33 mhz PCI bus generally can’t do 150MB/s in the real world.
What will make a difference inbetween SATA and PATA then? Using two drives at once, you’ll find better latencys off of SATA, because on PATA, if you have two drives on the same channel, only one drive on that channel can operate at the same time. SATA is point to point, so both drives can do operations simultaneously.
What else? Better drives are available that can push more MB/s on SATA than on PATA.
and…. on SATA our drives are hot pluggable without any kind of special controller. We don’t get that on PATA.
So this is really stupid. You could compare a SCSI vs SATA vs PATA and find that there are no differences inbetween the busses with this type of logic. Seriously, the results will look exactly like what was shown even between a PATA and SCSI drive if the drives are both low end drives.
This guy really needed to do some research.
They are comparing two _different_ drives one Hitachi with SATA and one Seagate with PATA!
It’s like comparing a red Porsche to a blue Skoda and declaring that red cars are faster.
Yeah, it’s a very random benchmark. Two different drives from different manufacturers, one with 2MB cache, one with 8MB.
And no hard info on the state of the SATA driver in Linux, which is most likely quite a bit less mature than PATA (duh!).
This would have been interesting if:
-It was Seagate vs. Seagate with identical specs but for the interface
-It was running on an OS where the SATA quirks are known and accounted for
Otherwise it’s not worth reading.
Why wouldn’t they use a PATA and SATA drives from same manufacture and drive size. Say a WD 120GB IDE and a WD 120GB SATA 8MB for example. Instead 2 different drives, sizes, and manufactures can probably lead to a large error in their testing results.
Plus drives themselves are the limitting factor and not the interfaces no?
“Why wouldn’t they use a PATA and SATA drives from same manufacture and drive size. Say a WD 120GB IDE and a WD 120GB SATA 8MB for example. Instead 2 different drives, sizes, and manufactures can probably lead to a large error in their testing results.”
I’m more curious about that than SATA vs PATA performance too…
What’s the point of this test? To test the current implementation of SATA in Linux Kernel? To compare if SATA is a worth? Than, why not use more drives? Only now we are seeing “native SATA” harddrives; and that means that SATA and PATA models was the same besides the logic board, sometimes using bridge chips…
It’s clear that the article needs to be more objective… and a larger selection of hardware AND software would help too if the intention is to be (the) “SATA vs. PATA Performance test”…
Drivers can be a limiting factor, too.
I have two Western Digital drives that are pratically the same model, save for the interface (WD800JB (PATA) and WD800JD (SATA): 80GB, 8MB cache). So far, the PATA drive was about 40% faster on MS Windows because the driver for my SATA controller (Si3112) ain’t worth crap.
I have tried these drives on my Linux server with a better controller (Promise SATAII150 TX2plus, two SATA and one PATA ports) and while I didn’t made any extensive test, they were pretty much equal in performance.
They could at least have used the same drives, with the only difference being the connection. Sheesh – talk about dodgy results. Science for beginners – Chapter 1.
I bought a couple of those tiny SATA PATA bridge interfaces a yr ago in order to get away from parallel cables. In the end I never really saw any improvement or less hassle with the cabling, these skinny red cables are too stiff and you still have to have a separate power cable. I seem to recall higher error rates instead since my kit is open frame, which SATA isn’t designed for out of the case wiring.
To reduce par cabling I often cut a ribbon down to a short single section for 1 IDE use.
For the eventual switch to 2.5 drives it makes alot more sense to go to serial cabling though.
Did the server have SATA or PATA drives in it
Okay, so the server was thrashing a PATA drive. Turns out Apache2 KeepAlive had an odd default I didn’t catch that only came out as a problem with this article.
They’re different drives, yes. One’s a legacy PATA which you’d be likely to find in a PATA desktop computer and one’s a newer SATA in the same price range.
With the benchmarks run, the differences on straight throughput should’ve been minimized. I’d love to read an article on more similar drives and see if there’s much difference (that is, before SATA has better individual chipset support, otherwise it’s like comparing an orange to half an orange).
I’ll see if I can get Hitachi to send out a PATA drive to update it with figures from an identical-but-for-interface drive.
Could the similar results be due to misconfiguration of hdparam or unoptimized code for SATA? Just wondering. I know most OS put really safe choices for hard drives. Also, SATA is relatively new so Linux might not be optimized for it yet. Don’t really know, just a concern I have.
one would think that parallel transfer would be faster than serial, but not in this case.
One of the benefits of S-ATA is that you can transfer things to master and slave devices at the same time on the bus. On P-ATA based controllers, you can only write to a single device at a time.
EMC’s Centera storage devices are now based on S-ATA.
A hitachi SATA verse a Seagate ATA? Is that even fair in any way?
Ahh, this has to be the dumbest performance test I have ever seen! Comparing a Seagate Paralell ATA drive against a Hitachi Serial ATA drive is comparing apples to oranges. They are two *DIFFERENT* hard disk platforms and are likely to perform quite differently even if both of them were Parallel ATA.
The only way to make a valid comparison is to get two disks from the same manufacturer that have the same specs (ie: same cache sizes, platters, heads etc etc), one running with Parallel ATA and the other running with Serial ATA. Otherwise the comparison is basically pointless.
Ignore this one for sure.
Compare different DRIVES when the article is said to be about different INTERFACES?! GREAT IDEA!
One benefit of the SATA is aerodynamics. When the cables are thinner there is less resistance for air flow and the components get cooled better. SATA RAID is also nice.
LOL…The airflow inside a computer case isn’t fast enough to be worthy of naming it aerodynamics.
Only lame marketing folks would use that term to sell their products.
Besides, you can’t use cable clips, wire/cable ties to keep the cables out of the way? Its not that hard.
According to SilentPCReview it matters whether you use thin SATA cables or wide IDE cables.
One benefit of the SATA is aerodynamics. When the cables are thinner there is less resistance for air flow and the components get cooled better
Which of course is why they went from 4 pins on the power to 15 which in the case of drives that DON’T include a molex results in your running even more wiring inside… RIGHT.
Seriously, SATA – LOVE the concept… HATE the implementation. ESPECIALLY with the rinky little garbage connectors they put on it that break in a good stiff wind and fall out of their sockets if you move the machine more than five feet. These connectors are a total joke and make USB look like a engineering masterpiece. Manufacturing thermoset resin (the type of plastic the SATA connectors are made of) thinner than a millimeter is a BAD idea…
Seriously, SATA – LOVE the concept… HATE the implementation. ESPECIALLY with the rinky little garbage connectors they put on it that break in a good stiff wind…
And molex connectors are any better? Sure, I enjoy ripping the skin on my fingertips and almost tearing my fingernails off because I can’t get pliers in the case because of the angle the drive is installed. If molex connectors died a violent death tomorrow, I would shed no tears.
At least with the SATA power connectors, when you pull them, they detach.
I’ve pulled plenty of Molex connectors over the years. You wiggle them out. Using tools like pliers is just asking for trouble. Now for those of you connecting, and disconnecting drives enough that SATA power connectors are a problem. You might consider getting a removable drive cage. Move the wear and tear somewere else.
i’ve had varios experiences with sata – from an ide->sata adaptor running an ata133 drive to a sata150 mobo and drive to a sata150 mobo and sata-ii drive.
basically all i’ve noticed is lower cpu usage with the sata drives (by up to 20%) otherwise no performance increase at all.
it’s nice not to have the master/slave crap, and the ability to have 6 sata drives instead of 4 ide.
also, afaik only maxtor do ata133 drives, and i’ve had various problems with them (stick to seagate ata100 for ide folks) no problems with any wd/seagate sata drives.
They compared a *40GB* PATA drive and a *80GB* SATA drive. This is the most unfair (and useless) benchmark I’ve ever seen.
The pata drive only has a 2mb buffer. this comparison completely sucks. I have a pata drive with a 10 mb buffer
There is no point in comparing PATA and SATA.
PATA is old, suffers of many sw and hw problems, has been patched over and over. It is time to die for PATA.
SATA is the way to go and earlier everybody switches to SATA, the best it is. SATA prices will fall fast is there is great demand.
shya… if it ain’t broke, don’t fix it. PATA just works. never had an issue with it. No reason to change over. Not everything supports SATA, but they sur as hell support PATA. The most common denominator on a PC is the HD.
shya… if it ain’t broke, don’t fix it. PATA just works. never had an issue with it. No reason to change over. Not everything supports SATA, but they sur as hell support PATA.
You naver had an issue with PATA as an user. Ask driver writers if they prefer PATA or SATA.
PATA requires more, more complex, and uglier code.
More code + many exception to the standard = more bugs and less speed.
Less code + drives that conforms to the standard = less bugs and more speed.
For you it can be just as like HTML vs XHTML. No difference, no?
Few understand that one XHTML parser requires half the number of lines of code than an HTML parser that need to parse all the HTML s**t that can be found on internet.
They kill off the floppy disk, fine, now they’re planning on deep sixing the IDE ATA hard drive… Fine…
Want to try to install Windows XP on a machine with an SATA only motherboard with no provision for a floppy drive? Oh… Feck you can’t (easily).
Thats the problem, its all well and good saying, “lets dump XYZ – its old and difficult to program for” but the thing is, people know what the issues are with ATA, there are variances within the spec, but they are known quantities and can and are accounted for.
I can install windows on any motherboard I want with an ATA hard disk and a cd rom. To do the same on an SATA board, I also need (at least at some point) a floppy drive hooked up to do the same, not to mention the fact that in using something other then ATA, I’m reducing what I can install OS wise onto that motherboard.
Ultamately this has to be about the users, and quite frankly there are very few benefits for the user in SATA over ATA, Raid is nice yes, but you can do that with ATA, and really, if you want raid, you’ll probably be better off running it from a third party PCI card (if your motherboard goes after a couple of years, does the new model have the same controller chip etc).
Besides, I don’t as a rule trust small and fragile looking connectors to last all that long, and if you’re as ham fisted as me, they’ll probably not appeal much to you either. I smell a nice sideline in SATA replacement centres popping up to repair boards comming soon.
I got a huge performace boost when I used a SATA drive. Its an SATA150 Samsung 160GB drive. I installed Windows in 20 minutes vs. 40 on a PATA drive. It is faster.
A Windows XP SP2 disk installs on a SATA disk without requiring additional drivers. But there are much fewer CDs shipped with SP2, of course.
SATA is register-compatible with PATA, so SATA controllers can be presented as old-style drives. If you have SATA ports on your board, you can install XP without the proper drivers. or Linux. Or DOS.
For the manufacturing side, there are much fewer wires to route. 7 wires is much easier to route than 40.
For cramped cases, SATA cables are sooo much easier to hook up. Also, they only fit on one way, so no putting the cable on incorrectly because your ribbon cable isn’t keyed and you don’t know the red wire goes near the power connector. SATA also allows longer cable lengths, which is nice for those who need it.
Also, on the SATA connector, the first wires to make contact are the ground wires, which is VERY important for reliable and safe hot-plugging. Of course, your controller still needs to support it.
[quote]SATA is register-compatible with PATA, so SATA controllers can be presented as old-style drives. If you have SATA ports on your board, you can install XP without the proper drivers. or Linux. Or DOS.[/quote]
I haven’t tried it, but AFAIK you would have to change a setting in the BIOS to make SATA work in PATA compatible mode. And then find some way to update the driver to SATA-capable.
…update the driver to SATA-capable following the Windows XP installation, that is.
>> SATA is register-compatible with PATA, so SATA controllers can be presented as old-style drives. If you have SATA ports on your board, you can install XP without the proper drivers. or Linux. Or DOS.
Try it sometime… Legacy ATA support in SATA chipsets was one of the FIRST things vendors realized they could save money by cutting that corner. Hell, the linux kernel crashes with the wonderful “AIEE!!!” error on pre SATA kernels, and on some builds of the ones that are SUPPOSED to support it… Have yet to see a nForce4 board pull it off successfully, including the much vaunted ASUS A8N SLI chokes in that department… Even the Linux support isn’t a ‘complete’ implementation on most chipsets, falling back to the older ‘fakeraid’ instead of the newer ‘dmraid’.
Also:
>>Also, on the SATA connector, the first wires to make contact are the ground wires, which is VERY important for reliable and safe hot-plugging. Of course, your controller still needs to support it.
and you’ve apparantly never seen the inside of a SATA connector – all those pins make contact en-mass… There are no “First wires to make contact”, if they wanted that route, they should have put it in a nice stable metal hood (like say, a good old 9 pin din) instead of the rinky little piece of crap they did end up with.