Tweaking disk speed

[LiamC]

O.K. What I'm trying to do is optimise disk speed under Windows 2000 Professional

I'm using Wbench 99 to "measure" any improvements.

With the 2MB cache drives, I could get similar scores to SR, even though it was a different platform.

With the 8MB cache drives, I can't get close, and in fact, the drives score little better than the 2MB jobbies. Is there some tweak I'm missing?

What I'm using
Fresh install, video, chipset, DX 8.1, NTFS. Make sure UDMA is set, defrag registry, swap file, hives etc. (all contiguous)

W2K , SP2 and SP3 default IDE drivers (no difference)

VIA 4.43 -> 13% faster in Bus Disk. Same in HE Disk

ArsTechnica Reg Tweaks
http://arstechnica.com/tweak/win2k/index.html

(with SP3 default driver) No difference in Bus Disk, 10% faster in HE

UDMA mode is enabled in BIOS.

Tried setting PCI latency timer to 32 (default) 64 & 96. No discernable difference

According to the Disk Transfer graph, there are no troughs so that looks fine.

According to the WD data lifeguard tools, The drive is operating in ATA100 mode.

Hardware
Athlon 2000+
KT266A
512MB PC2100
WD 1000BB-SE
WD 400JB

Now I know that some might say that SR uses an 80GB drive, but the transfer rates are the same for the 40 and 80GB models, and as the software only takes up about 2GB total and is at the start of the drive, then I don't see this mattering too much.

SR scores for Testbed two and testbed three score a good 15~20% better than I can do.

Comparison drives are a 60GXP and D740X.

I will try a Promise Ultra 100 and see what I come up with. Am I missing something?

FWIW, I am seeing about

7600 for Bus and 29~30 000 for H-E from the 400JB

 

[Tea]

I believe that the Promize controllerz deliver a zignificantly higher tranzfer rate than VIA onez. I zuzpect that you are on the right track already.

 

[Cliptin]

This is not the answer to yor problem.

But wouldn't you want to decrease a latency timer?

 

[Tannin]

Unlike most of her silliness tonight, Tea's suggestion is good. (Well, her support for your own suggestion, I should say.) I remain agnostic as to the question of whether or not the benchmarked performance gain is real, or indeed as to the question of whether Winbench is a particularly useful measure. On the whole, these days I incline to the view that, failing a compelling case for any other drive benchmark, you might as well use Winbench.

 

[Bozo]

If you note, SR uses Intel motherboards. In my limited experiance with VIA chipsets, they suck when compared to Intel. No matter how many 4-n-1 drivers and upgrades I've tried, VIA cannot equal the the performance of an Intel chipset, especially the IDE buss. (nor the stability)

The Promise card should help.


Bozo