WD1200JB perform way below par in Sisoft. Any suggestions??

[pillage2001]

The title says it. I benchmarked my drive and it gave me a score of 26mb/s. The other drives were all well above 30mb/s.

Anybody know where I can download HD tach??

 

[Rand]

HDTach 2.61 Download
HDTach will give you a good idea of sustained and peak transfer rates. It's access time tests are pretty unreliable though, and ithas no way of measuring application disk uage performance but it's decent for what it is.


Frankly I'd ignore SiSoft's HDD benchmarks entirely. It's great for analysis, but worthless for HDD benchmarks. It's unreliable, inconsistent, and does not test in a manner even remotely similar to real world desktop HDD usage.
I've seen SiSoft claim 5400RPM drives are faster then 10K SCSI drives, and 7200RPM drives are slower then 4200RPM Mobile drives.... the next reboot the scores may vary by a good 4-5MB/s.
Pick a number at random and it means about as much as SiSoft's tests will.

 

[pillage2001]

I've downloaded HDtach 2.61 but there's a problem. I'm on win2k.

I was just wondering how come my ratings were so slow. Does the PCI IDE latency patch do anything to remedy it?

 

[Viper96720]

I use atto disk benchmark from atto express pro tools. I was getting low scores from sandra also 17000-18000

 

[SUOrangeman]

Pill-

Are you having problems getting HDtach to run on Win2K? Just run it as Administrator. I don't think I've ever had any problems with the app itself (WD1200JB + Tyan Tiger MPX-4M).

-SUO

 

[TenaciousPee]

Administrator or no, HD tach doesn't work on NT/2k unless you buy the retail version.



It could be alot of things, but my guess is that it's NTFS, which I assume you're using. Oh, and I presume that you have all the correct: drivers, cables, service packs, registry tweaks, service settings etc. etc.


I won't go into great detail because I just spent 45 minutes writing up a lengthy response to this thread, only to have it deleted just prior to posting it (stupid Java Scripts!). But in a nutshell, NTFS has to do several write transactions for every whole file it reads from and then writes to. Those write transactions are to the filesystem transaction log and to each of the multiple, redundant MBRs spaced throughout the hard drive itself. Every one of those backup writes uses the cached file location information in the HD cache itself and doesn't cost the disk system all that much in terms of transfer speed. However, random reads and writes, or reads/writes to small, individual clusters tend to not make use of that cached information and therefore suffer an enormous performance hit when those backup writes are sent to the MBR & file system.

That, combined with all the performance monitors (diskperf, perfproc, perfos etc. etc.) hog up system resources as well, although I think the majority of the blame can be leveled on NTFS' behavior itself.


I have a Seagate Barricuda IV that, under Win98SE and Sandra, benchmarked out to 38MB-42MB/s. Under Windows 2k SP3, I'm down to around 17-20 MB/s... less than half the performance. Here's a recent breakdown of the individual performance results on Sandra:


Benchmark Breakdown
Buffered Read : 78 MB/s
Sequential Read : 28 MB/s
Random Read : 9 MB/s
Buffered Write : 60 MB/s
Sequential Write : 25 MB/s
Random Write : 8 MB/s
Average Access Time : 4 ms (estimated) (<----- 4ms access time? Hahahaha! I didn't know I had a SCSI drive...)



The 'buffered read' and 'buffered write' benchmarks are essentially worthless as you never get that in the real world except under certain specialized circumstances. The sequential read/write are the most realistic results and are a good indicator of real world performance as they mimic normal-sized files. The 'Random Read' & 'Random Write' benchmark results tell the story, though. Both of them are horribly low. If memory serves me correctly, I got 17 and 18 MB/s for both those benchmarks under Win98SE Fat 32.


And that is what I think the problem is, NTFS quirks slowing down two of the six benchmarks and dragging the overall score down.






Edited for clarity, content