Sign in to follow this  
Davin

Western Digital Raptor Preview

Recommended Posts

Guys, you did a great job of the (p)review.

I know here at work we're waiting for the Raptor to be available. There'll be at least one, a perhaps 3 in RAID5 going into our main operations server (Exchange, custom DB, and all the other stuff).

3 Raptors and a modern board/CPU combo will go well to replace the Classic Athlon 600 machine with it's old AmiRAID and 3 IBM 36LZX.

If WD produce what should be a reliable drive, I can see the Raptor being a popular unit.

Share this post


Link to post
Share on other sites
An intellectual exercise for everyone:

If SR posts substantially better performance figures for a raptor, what would be the better explanation?

a) the beta drive's performance didn't have a lot to do with production drive performance

B) WD tuned the final drives performance to better perform on SR's tests

According to Anand's review, WD claims to already be shipping these drives to distributors. With the drive you previewed dated Feb 11, that doesn't give WD more than a few weeks for improvment.

Now the question is, if the final production drive performs much better, (which you would know better than anyone here) why on earth would WD send out such poor performing samples when productions drives are already available? That would seem an extremely stupid market blunder on WD's part.

I'd go with choice a. SR's benchmark suite is too varied to be able to tune for all of them. If it only improved in Sandra or some other synthetic benchmark I would be suspicious, but if it performs better across the board on SR's suite then the drive simply performs better.

Share this post


Link to post
Share on other sites
Well, to be perfectly honest, one of the main reasons we were hesitant to publish was the somewhat stubborn character we see from some readers. Many, for example, already seem to have passed final judgment on the Raptor and dismissed it due to this article. We feared that some readers would refuse to accept substantially improved figures as representative of the family's performance because they'd fallaciously assume that WD's final was merely a reaction to our beta figures.

If you already have the drive, it would be impossible for the drive to be a reaction to your preview unless WD was psychic and knew what you were going to post. That said, if you do have a final production that performs much better, I don't think you should have posted the preview you did. Instead, you should have posted something on the frontpage that said you have a preproduction sample that everyone else is prereviewing as well as a final version. Along with something to the effect that the final drive revision has demonstrated much better performance in early testing so you decided to not post a preproduction review. The first impression you usually the best remembered, and this previewed drive performed poorly to say the least.

Share this post


Link to post
Share on other sites
Now the question is, if the final production drive performs much better, (which you would know better than anyone here) why on earth would WD send out such poor performing samples when productions drives are already available?  That would seem an extremely stupid market blunder on WD's part.

If the production drives weren't available, then that would explain why WD sent out beta drives. Some customers like getting teabag samples, drives that do little more than read and write and have very immature code.

I agree; if production drives were ready there would be no reason to send out beta samples. It'd be counterproductive if nothing else.

Share this post


Link to post
Share on other sites
area = pi * r^2, yes

but you used pi * d^2

you need to divide your areas by 4.

heheh.. Oops. (Tells you how long it's been since I've had to do math more complex than balancing my checkbook! Also tells why I have an accountant...)

So... 3.5" disk has 9.6 square inches, 3" disk has 7.0 square inches, and a 2.5" disk has 4.9 square inches. Okay, moving from 3.5" (7200 RPM) to 2.5" (15k RPM) would produce about a 50% reduction in usable space after taking out for the spindle hole. (Anyone know how big a spindle is?) Therefore, at the same data density, a 15k RPM drive would hold about half as much. (Of course, I don't know how much of that area is actually usable, after all, we have to remove the spindle, then I doubt the heads go all the way out to the edge...)

Obviously, the anecdotal things I had heard about 3.5" to 3" being a 50% reduction probably aren't true, unless spindles are rediculously large.

Share this post


Link to post
Share on other sites

Data stroke on a 3.5" drive is ~ 1 inch, depending on model. If you assume the spindle is the same size going from 3.5" to 3" media, then the stroke should be ~ 0.75". You lose some space at the ID for parking/landing zone and tolerances for the disk clamp and other parts.

I've heard area numbers for 3" from 1/2 to 2/3 of a corresponding 3.5" drive. Since faster drives can't pack data as tightly, the actual density might be a little lower than you'd get just extrapolating from area.

Share this post


Link to post
Share on other sites
Anyone know how big a spindle is?

I have a couple of dead drives that have been taken apart. The spindle part is about 1 inch across, so that's 0.5" radius. It may vary by model, however, and I don't know if the heads write all the way to the edges (outer & inner).

Share this post


Link to post
Share on other sites

let me answer your question. I first made the fact the WDC DOES tweak their fireware, in this post. I think that your question is answered by saying that when the new one does comes out, WD aint dumb. THey know that everyone will put two and two together. You may some different but no really. YES WDC did modify the test cause they know that SR and newegg are great partners, but the review was VERY thorough, do not get me wrong. SR does the very best reviews on the net. But time will tell what the production model vs. this beta results will be. I bet the WDC is reading all of these posts and trying to find ways to make their even more "puffed up" (I mean that WDC is reliable and fast, but notice that the results are always that much more above -- hmmmm a little suspicious if you ask me)!!!...

Thx mann.

Share this post


Link to post
Share on other sites

Wouldn't it be more fair to match this drive against 10k RPM 36 GB single platter drives from IBM, Seagate, and Maxtor? As opposed to matching it against drives with 3 times the number of heads?

Just a thought for when you get a final product to review.

Share this post


Link to post
Share on other sites

Hey, if WD can tweak their Raptor to perform better on SR tests, then great! TestBed3 is all about real-world performance, so a higher score means it's actually a faster drive. What's wrong with that?

Share this post


Link to post
Share on other sites
Hey, if WD can tweak their Raptor to perform better on SR tests, then great!  TestBed3 is all about real-world performance, so a higher score means it's actually a faster drive.  What's wrong with that?

Note, of course, that neither they nor any other manufacturer have SR's desktop tests...

Share this post


Link to post
Share on other sites

They don't need to have SR's desktop tests. All they have to do is capture typical desktop usage. Optimize for that and you will likely perform better on SR's tests... that is, unless you have wacko usage patterns, Eugene :)

Share this post


Link to post
Share on other sites
SR's benchmark suite is too varied to be able to tune for all of them.  If it only improved in Sandra or some other synthetic benchmark I would be suspicious, but if it performs better across the board on SR's suite then the drive simply performs better.

This brings up another point. I doubt WD would be able to improve results across the board with SR's test suite unless they made a mechanical change to the drive or unless the firmware were extremely immature because of the conflicting requirements of desktop and server performance.

Assuming no mechanical changes and firmware that is already close to maturity, WD can only really tweak the drive in one direction -- make it perform better in server situations or make it perform better in desktop situations. You can't have it both ways because cache segmentation is optimized for a specific randomness and locality. Change it and you will change what you are optimizing for. I will be very suprised if we see significant improvements in both desktop and server performance... it is possible that we may even see sacrificing performance in one area to give to another.

-----------

So what's it going to be? Improvement in desktop performance or server? Place your bets. I'll say +0% in desktop performance, +5% in server performance.

Share this post


Link to post
Share on other sites
The Raptor fared a bit worse in his tests than ours. He used an SI 3112 -based controller, one that I believe has some issues when paired with these beta Raptors. That's why we went ahead with the Promise controller despite strong pushes from Seagate and HyperMicro to standardize TB3's SATA extensions on the SI chipset.

Tom's Hardware recently posted a review comparing the SI controller to others, including the Promise model you used. The SI did do somewhat better. I think your review should at least have mentioned that you may not have used an optimal controller.

Share this post


Link to post
Share on other sites
So what's it going to be? Improvement in desktop performance or server? Place your bets. I'll say +0% in desktop performance, +5% in server performance.

Hmm. I'm going to guess we'll see an improvement on the server performance numbers rather than the desktop, since this drive *is* supposedly targeted at low-end servers and the like, not desktop stations. Figure on 10%? Code tweaks can work wonders from what I've seen.

Share this post


Link to post
Share on other sites
So what's it going to be? Improvement in desktop performance or server? Place your bets. I'll say +0% in desktop performance, +5% in server performance.
I'll say: Don't underestimate Western Digital. But then, I'm cheating. :-P

Share this post


Link to post
Share on other sites

IBM? They don't even make hard drives anymore.

Share this post


Link to post
Share on other sites
Maybe performace will be greatly improved with a native SATA

controller chip like the silicon image 3112 which is available on some mother boards. See Tom's hardware guide review.http://www.tomshardware.com/storage/20030307/index.html

Nicolaim and Sebm,

As we stated earlier in the thread, we tried this Raptor with no less than three different SI3112-based boards. The numbers were so bad that its likely there was some kind of unexpected extraneous reaction between the SIs and this particular sample of the Raptor.

Anand used an SI3112 board for his tests, hence the drive's relatively worse showing compared to our results.

Share this post


Link to post
Share on other sites

Well, thanks Eugene...I think.

I had been leaning toward a RAID5 array of 4 raptors on an Intel SRCS14L SATA controller for my new worskstation.

(http://www.intel.com/design/servers/buildi...CS14L/index.htm)

The SRCS14L uses: "Two Silicon Image* Sil3112A* SATA controllers".

And the LSI SATA RAID controller also uses the SI3112.

http://www.lsilogic.com/products/stor_prod.../sata.html#1506

These are exactly the types of products that the Raptor is aimed at. They can't ignore this market segment...or they're in real trouble. I find it hard to believe that they'd be that stupid in the long run.

We'll have to wait for the final product Review. I'm guessing that SR will be revisiting the SI3112 issue in that review?

Encyclia

Share this post


Link to post
Share on other sites
They don't need to have SR's desktop tests. All they have to do is capture typical desktop usage. Optimize for that and you will likely perform better on SR's tests... that is, unless you have wacko usage patterns, Eugene :)

LOL, not wacko, no... this is the point I'm trying to drive home to folks, though its better to lead someone else to say it rather than say it myself. Thanks ;)

So what's it going to be? Improvement in desktop performance or server? Place your bets. I'll say +0% in desktop performance, +5% in server performance.

Speculate no longer. Results for a WD supplied final version on the Promise controller have been placed in our performance database.

This particular sample doesn't seem to have the adverse reaction to the SI controller that the first sample did... but results that I have so far on the SI simply don't live up to the promise. I'm not done with Raptor+SI tests, though. We'll see.

Regards,

Eugene

Share this post


Link to post
Share on other sites
Speculate no longer. Results for a WD supplied final version on the Promise controller have been placed in our performance database.

This particular sample doesn't seem to have the adverse reaction to the SI controller that the first sample did... but results that I have so far on the SI simply don't live up to the promise. I'm not done with Raptor+SI tests, though. We'll see.

OMG! the final version is alot faster with most applications than the preview.

And some more performance in the server benchmarks.

Btw eugene: is that whining gone with the final version? or does it still make the same amount of noice as the preview?

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.

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