Sign in to follow this  
Davin

Western Digital Raptor Preview

Recommended Posts

Kudos to WD for actually supplying something for testing. The results look fine. The price for a 36gig drive is not bad but with the price of 10,000 rpm scsi drives dropping to almost as low, I see no reason to even consider buying a first generation Raptor other than for the sheer novelty of owning one.

Share this post


Link to post
Share on other sites

HOLY MOLY! Look at those results of the NON-BETA Raptor! They're excellent! This is actually more on par with what the drive should be showing given its specifications. I suspected that you knew something, Eugene, when you posted your question about the cause of

300 --> 500 IO/s in High End drivemark! This can't be just a simple firmware optimization, because for example also in the transfer rate test results have come up from 37MB/s to 45MB/s and from 57MB/s to 63MB/s. I believe that the BETA Raptor was really horribly tuned and just had the nearly completed mechanics (mechanical sample), and after that WD just concentrated on getting the performance right with tuning the firmware. When you compare the FINAL results of the Raptor, I find them very plausible, since they beat every 10kRPM SCSI drive, but not quite as fast as the 15kRPM Cheetah 15k.3. If you remember the performance increase of the 7200RPM JB-series, then this is totally believable, since historically IDE drives have usually been faster in SINGLE-USER scenarios than comparable SCSI-drives.

However, evidently WD has not so much concentrated on server performance, because those results have remained almost the same. I don't mind that, because now this drive is perfectly suited for performance-oriented users and high-end desktops, yet still OK for servers.

Share this post


Link to post
Share on other sites

I still don't see what the advantage over scsi is. You still need and adapter, it's a first generation technology not only in rpm but more or less in interface as well. How could anyone recommend these in good conscience for server use? SCSI is tried and true and not significantly more expensive.

Share this post


Link to post
Share on other sites

Holy crap, that's one hell of an improvement from the Beta to the Production drive. Looks like I'll be altering my system build recommendations for high-performance gamers...

Share this post


Link to post
Share on other sites

And the impact of tuned vs. detuned firmware again amazes me. Wish I could do things with mechanics to net such differences that quickly! ;)

Share this post


Link to post
Share on other sites

Guys, we've been played. :oops: Eugene set us up brilliantly. I bet he knew the whole time we were speculating about the results of the final version of the Raptor but was concerned we would reject such a huge improvement. That's why he made sure that: (a) we thought SR's testing suite was comprehensive enough to not be rigged, and (B) we conceded that WD could improve its showing in SR's benchmarks simply by improving performance for general desktop usage with their own IPEAK like drive performance analytical tools.

Good work, Eugene. You are the puppet master. :wink:

Share this post


Link to post
Share on other sites

I certainly hope the main page is going to be updated with a news item about the new results being available, at least after noise/temp testing can be performed for a complete set of results.

I was disappointed after seeing the beta reviews here and on Anandtech, but now I'm excited again. I certainly hope these drives prove a commercial success to spur the development of further 10kRPM ATA drives. A speed increase is long overdue in the ATA arena. My next gaming rig (Prescott/Canterwood) will include a pair of these puppies unless something even better sees the light of day between now and the availability of Prescott.

Share this post


Link to post
Share on other sites
Guest Eugene
I certainly hope the main page is going to be updated with a news item about the new results being available, at least after noise/temp testing can be performed for a complete set of results.

We'll have an update pointing to the results from the front page tomorrow. A formal writeup will hit either late this week or early next.

We usually sneak in tidbits here and there in the forum a bit in advance though to encourage people to check in here ;)

Share this post


Link to post
Share on other sites
Guys, we've been played.  :oops:  Eugene set us up brilliantly. I bet he knew the whole time we were speculating about the results of the final version of the Raptor but was concerned we would reject such a huge improvement. That's why he made sure that: (a) we thought SR's testing suite was comprehensive enough to not be rigged, and (B) we conceded that WD could improve its showing in SR's benchmarks simply by improving performance for general desktop usage with their own IPEAK like drive performance analytical tools.

Not really. It didn't vastly improve across the board. You're forgetting who WD is trying to sell these drives to. The drive improved 9% in the fileserver and only 1% in the webserver benchmark. WD just built the best gaming drive on the planet. Who cares? How many people are going to buy a drive to load Quake3 faster? What happened to targetting the lowend server market? It's still getting trashed in the server benchmarks. WD did a phenomal job optimizing this drive, unfortunately they did it for the wrong applications unless they decided to change their market to the gamer.

One thing this optimization makes you think about, is how well a 15k SCSI drive would perform if someone decided to optimize it for "workstation" applications.

Share this post


Link to post
Share on other sites
... It didn't vastly improve across the board.  ... improved 9% in the fileserver and only 1% in the webserver benchmark.

9% in file serving is not too shabby considering all they did was tweak the firmware (server performance is more dependent on drive mechanics than on firmware so I guess if you think about it a while, it is not surprising that it is a smaller improvement compared to what a similar tweak would do for desktop performance). The improvement in server performance is especially significant if you consider that the small improvement server performance is accompanied by significantly improved desktop performance (15-20%), especially in high-end apps (75%). I don't know about you, but I expected a significant improvement in desktop performance to come only at the expense of server performance, or vice-versa. To attain such improvement without sacrificing in any area is noteworthy, IMO.

WD just built the best gaming drive on the planet.  Who cares?  How many people are going to buy a drive to load Quake3 faster?

I agree with you personally, but you would be surprised how many people would do just that. I have read countless posts by gamers on several boards who ask how they can improve their disk performance so that they can load levels faster. Many of them ask if IDE RAID 0 or a U160 controller and a 15k.3 will load their levels faster. Some ask why their RAID setups aren't benching well. If the Raptor were pushed by an editor or a leader on the boards of a gamers website, you bet all these gamers "with nothing better to do but blow their money on high end gaming rigs" will be all over it.

What happened to targetting the lowend server market?  It's still getting trashed in the server benchmarks.  WD did a phenomal job optimizing this drive, unfortunately they did it for the wrong applications unless they decided to change their market to the gamer.

I feel the same way, but maybe that's all they could muster without taking a big hit in desktop performance? -- i.e., it wasn't worth the sacrifice

One thing this optimization makes you think about, is how well a 15k SCSI drive would perform if someone decided to optimize it for "workstation" applications.

<homer> drool </homer>

Share this post


Link to post
Share on other sites

I'm a hardcore gamer with a home network. 2 of my 4 mobo's have serial ATA ports and I plan to use them to the max with the Raptor. The only problem is getting one right now! Where the hell can I order the Raptor?

Share this post


Link to post
Share on other sites

Eugene -

I think it might be worth including the 15K.3 (and/or Atlas 15K?) in the review charts, just to demonstrate the difference between the best ATA drive and the best SCSI drive for those that don't check the database.

I know it's not exactly apples to apples, but since it's close to or beating 10K SCSI in several tests, you could forestall some of the inevitable "Raptor vs. 15K.3" questions.

Spod

P.S. No pressure, but when can we expect the Atlas 15K review? Any idea of it's release date?

Share this post


Link to post
Share on other sites
Guest Eugene
I think it might be worth including the 15K.3 (and/or Atlas 15K?) in the review charts, just to demonstrate the difference between the best ATA drive and the best SCSI drive for those that don't check the database.

I know it's not exactly apples to apples, but since it's close to or beating 10K SCSI in several tests, you could forestall some of the inevitable "Raptor vs. 15K.3" questions.

The drives we include in review graphs are those that target the same market as the review drive. 10k RPM SCSI drives and high-end 7200 RPM ATA drives address the entry-level enterprise segment, hence the comparison. 15k drives just don't fit in anywhere aside from a tiny overlap in the market that represents crazy users who'll buy anything that happens to be the fastest thing around.

P.S. No pressure, but when can we expect the Atlas 15K review? Any idea of it's release date?

I don't have a formal release date, no. Maxtor claims to have shipped two batches to distributor representing an unnamed customer, however.

Our review sample was an 80-pin SCA unit; we were forced to use a converter to give it a spin in testbed3. Though unlikely, Maxtor is concerned that the converter may have had an effect on (what they view as) the poor Atlas 15k server showing.

Share this post


Link to post
Share on other sites

OK, now the IDE boys have got up to 10k RPM I think it's time for the SCSI men to get something more exciting that 15k :D

Seriously though, the reliability, sound and heat levels of current 15k drives are really impressive - a 20k drive could surely be no more hot and noisy that 1st generation 10k units? I suppose the performance delta for 15 to 20k is not large enough to justify the effort just now, but I just want something exciting to covet.

Dave

Share this post


Link to post
Share on other sites
OK, now the IDE boys have got up to 10k RPM I think it's time for the SCSI men to get something more exciting that 15k  :D 

Seriously though, the reliability, sound and heat levels of current 15k drives are really impressive - a 20k drive could surely be no more hot and noisy that 1st generation 10k units?  I suppose the performance delta for 15 to 20k is not large enough to justify the effort just now, but I just want something exciting to covet.

Dave

Yeah, 20k is probably possible (wasn't there a rumor a few months ago about somebody making an 18k drive?) but they'd most likely have to shrink the platter again. Pretty soon, we'll be up to a 100,000 RPM drive that can only hold ten megabytes. But, HOLY COW will you be able to read those ten megabytes fast!!! :P (And with that small of a stroke length, seek times should be amazing!)

Share this post


Link to post
Share on other sites

It's an official Review from anandtech:

http://www.anandtech.com/storage/showdoc.html?i=1799

And since he has the release ver drive now, the results are much more favorable, showing the same sort of dramatic improvement that's in the SR database for desktop benchmarks. I'll wait for SR's review before making any decisions, though. His server benchmarks will be in a part 2.

He does have some details from WD about why the beta performed so differently from the release. Apparently write caching was disabled on the beta, in addition to there being firmware changes in the final release.

Maybe he should have read the SR preview before publishing his?

Share this post


Link to post
Share on other sites
Btw eugene: is that whining gone with the final version? or does it still make the same amount of noice as the preview?

I haven't had the chance to perform formal heat and noise measurements on the drive yet (have to set up an environment, etc ;))... but subjectively, the whine seems lessened. With the previous sample, for example, I found the whine bothersome while I worked in the same room.

This is something I unfortunately attribute to the erratic results one might get from one sample to the next when it comes to that type of noise.

Eugene.

Out of curiousity - this whine which annoyed you was it mounted in a case in the same room or on a benchtop?

Share this post


Link to post
Share on other sites

Hello, beside the impressive new benchmark figures I am currently to build up a PC which has a good performance and is as quite as possible. I am very keen on getting information about the subjective noise level of this hard disk.

What is your impression is it noisier compared to 10k SCSI drives ?

As I also don't use any additonal system fan (because I want to avoid any additonal noise source) the heat generation is also important for me, I assume in this point the drive is much better than SCSI drives, is this correct?

Thank You

Share this post


Link to post
Share on other sites
Hello, beside the impressive new benchmark figures I am currently to build up a PC which has a good performance and is as quite as possible. I am very keen on getting information about the subjective noise level of this hard disk.

What is your impression is it noisier compared to 10k SCSI drives ?

As I also don't use any additonal system fan (because I want to avoid any additonal noise source) the heat generation is also important for me, I assume in this point the drive is much better than SCSI drives, is this correct?

Thank You

mpaulu: it's hard getting an answer, I think Eugene is busy OR he only replies to regulars! :)

Either way this drive - even if it DOES have fluid bearings OR even if it's in a no vibes III noise restrictor the "whine" of 10k MIGHT come out of the hard disk - this will be really hard to test :(

Those "cases" that hold the entire drive in there would work, but they are for 5400 rpm ONLY :)

Share this post


Link to post
Share on other sites
Guest Eugene
What is your impression is it noisier compared to 10k SCSI drives ?

All of the current-generation 4-platter SCSI drives bother me less than both raptor samples... this includes the MAS3735 (which, incidentally, looks like it's going to be the fastest drive of its generation for both desktop and server tasks... more on this soon) who's idle noise weighs in at a whopping 50.5 db/A. Yet more evidence that everyone must be careful with objective measurements.

Just yesterday I imaged over to a Raptor- though I feared the slight whine of the production sample would get to me over time, it's been fine so far.... the drive is quite speedy compared to what I've used in the past month (WD800JB, DM+8, DM+9, Cuda V) ;)

As I also don't use any additonal system fan (because I want to avoid any additonal noise source) the heat generation is also important for me, I assume in this point the drive is much better than SCSI drives, is this correct?

Much better than 4-platter SCSI drives yes. Single-platter units may be just as heat-competitive though.

Share this post


Link to post
Share on other sites
Guest Eugene
Much better than 4-platter SCSI drives yes. Single-platter units may be just as heat-competitive though.

But, as e_dawg would be rightfully rapid to point out, the Raptor is faster than all single-platter 10k SCSI units for non-server use.

Share this post


Link to post
Share on other sites

Hi everyone! :)

First of all, great article and great website!

Now on to my question ... I would have made another thread for it, but seeing that it is very much related to WD's Raptor drives posting it in here made more sense.

Currently my best system in the house is a P4 1.6A with 256RD Ram (PC800) on an Asus P4T-E mobo with (ready?) 5400RPM drives :oops: ... I've lived fine with them for quite a while and I DO have a 7200RPM WD (120 Gig non-SE) in my server just because it's accessible from the whole house that way at all times.

I'm finally getting sick of the slow loading times in some recent games and have decided to do 2 things - double my ram and get new HDDs at last. Right when I was about to place my order on 3Ware's Escalade 7000-2 ATA 133 RAID Controller Card and 2 WD "SPECIAL EDITION" 80GB 7200RPM EIDE hard drives in hopes to get some great RAID 0 performance WD HAD to announce the 10K drives, so I've been waiting and doing research at the same time.

The new plan involves 2 WD Raptors in a 10K RPM SATA RAID0 setup where the controller is a mistery ... I hear great things about Promise in general ... 3ware's SATA controller's simply not for me at the kind of price they are selling it.

Narrowed it down to 2 options:

[*]Promise SATA150 TX2

[*]HighPoint RocketRAID 1520

There is 1 thing that bothers me about each controller

in the case of HighPoint RocketRAID 1520 it is:

This guy's troubles combined with the fact that Tom's Hardware mentioned a bunch of companies WD is working with, for great SATA support, but HighPoint was not listed anywhere while 3ware, Promise and a bunch of other more popular ones were RIGHT THERE. :(

in the case of Promise SATA150 TX2 it is:

Originally posted on Tom's Hardware Guide

The SATA150 TX2 achieved good performance figures in the Winbench 99 2.0 applications benchmark and in the H2benchw application test. It was marked down because of the interface transfer rate with motherboards using Intel chipsets, where it cannot achieve any more than 63 MBytes/sec. Comparative measurements with an Athlon system using a KT400 chip and VT8235-Southbridge showed a healthy 82 MBytes/sec.

*sigh*

According to Mark1Computers the drives will be selling on Monday ... and I'm still pretty clueless as to what controller to pick ...

This is getting much longer than I anticipated it to be, but here are the last 2 questions...

[*]Just how high pitched of a whine is it that WD Raptor HDDs give out?

[*]Is it a bad idea to have your WHOLE main OS (Windows XP Pro most likely) on a controller card, because some people say you only achieve the highest performance by enabling dynamic disks under windows and I'm thinking that would be kinda impossible if you are trying to enable it for the disk that your OS is on.

Well...That's it. *phew*

Sorry for the long post, I hope I didn't bore anyone too much. :oops:

Any advice on any of the questions I asked would be greatly appriciated!

Thanks in advance guys, and keep up the great work on the website...now that I know about it I'm definitly not forgetting it when friends ask me about harddrive upgrade advice :)

Nk2k over and out.

Share this post


Link to post
Share on other sites
But, as e_dawg would be rightfully rapid to point out, the Raptor is faster than all single-platter 10k SCSI units for non-server use.

Absolutely, Eugene. The superb showing of the production Raptor in the desktop benchmarks despite its much smaller capacity is particularly impressive.

However, I do not feel that the SR desktop performance suite is 100% representative of the drive's performance... at least for my usage. I would gladly trade off some desktop performance for server performance and access time. No matter what the benchmarks say, there is no way the WD800JB feels as snappy as my Atlas 10k III, let alone the Atlas 10k. Access time must play a role in "snappiness". This is not to say that "snappiness" is the ultimate measure of desktop performance -- as unscientific and unencompassing as it is -- but that is what I notice as a user. Some of it could be psychological as you indicated in the DM+8/9 thread, but I do try to be careful in this regard. (objectivity is very important to me as well)

Share this post


Link to post
Share on other sites

24 hours and no reply...

Then again I do tend to get hooked on forums and doublecheck for replys every 15-30 minutes. :D

Sadly, before I ever get any questions answered I've already come up with another question about "hardware RAID" vs "software RAID" and while I understand what each means, I thought the only way to get software RAID going was trough the OS ... now sure I shouldn't believe everything people say but here's something 2 people on www.newegg.com posted:

Originally posted by Chizow about 3ware's 7000-2

This is a great hardware RAID card for those looking at RAID 0 or RAID 1 for a high performance gaming/workstation or small server, respectively. 

Its faster than any softRAID card (Highpoint RocktRAID 133 being the fastest), and CPU utilization is lower as well due to the onboard math co-processor and cache. My ATTO scores are extremely high and surpass expensive SCSI disks. Write and Read speeds are in the 70Mbps-90Mbps ranges beginning with the 8Mb block. This card also performed better than 2 other softRAID cards I tried in the last month.

If you are less than thrilled with your onboad RAID controller's performance, or you've tried the softRAID cards and weren't particularly pleased with the results, this is the card for you. Its meant for the power-user that does lots of multimedia work and to a lesser extent hardcore gamers. If you need increased performance and reliability, you'll want to look at the 7500-4 (for RAID 10 or RAID 5) or a softRAID card that supports 0+1.

One IMPORTANT note; Win2K and WinXP users will need to upgrade their "basic" disks to "dynamic" disks in Windows to fully unleash the capabilities of this drive. Its a painless procedure that requires you to change the setting in computer management/disk management, and is effective upon reboot. Performance will be poor until the disk is converted.

that last part caused my question in the previous post...and then

Originally posted Kman about the same card

...If you're going to get RAID, get real hardware RAID. You could go with cheaper BIOS hacks like the promise and highpoint controllers, and that might be "good enough..." but why not get the real deal? Go with 3ware and you won't be disappointed (and no, I don't work for them... I'm just a very satisfied customer).

obviously I shouldn't believe everything I read online, but these guys got me thinking about actually saving up for 3Ware's 4 port SATA card despite its insane price ... even IF it'd be a "stupid" investment as SATA's just barely starting to evolve now.

btw as a side question: how come we can't edit/delete posts? I had decided to move my dilema to its separate own thread after all but then found out I had no such ability on here :( Well it is related to selecting the proper controler to match WD's Raptors but still...

Share this post


Link to post
Share on other sites

Interesting review.

If the cost is relatively low (any word?) then this could be a nice drive in highly parallel environments. A problematic trend in data centers is that as capacity has skyrocketed, the number of physical drives have gone down and with them the ability to load-balance across many different spindles. If I have, say, a half-terabyte data warehouse to build, I'd conceivably have gotten better performance out of 64 or 72 first or second gen 9GB 10K drives (from four or five years ago) than I would out of four of five of the modern 100+ GB 10/15K drives. This is humbling, and it is a real problem.

The fact that the flagship enterprise-class Serial ATA drive is launching with a single-platter 36GB design could be an attempt to address this problem. How well do serial ATA controllers & drives operate in RAID environments? I wonder. Yes, the beta unit's performance isn't anything to write home about, but what about four of these units in a RAID-5? Single platter = low noise, low heat, low vibration, mechanical simplicity. Perfect drive to populate a SAN or array chassis. Times four and I'd have 108GB available after RAID with four separate spindles to read from. I'd love it.

Thoughts?

Share this post


Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
Sign in to follow this