Jump to content

SCSI drives running freakishly slow...


Recommended Posts

Topic says it all.

 

Working off a fresh system build, new DFI Infinity nF4 SLI, opteron 146... yada yada (see sig for specs). My drives run freakishly slow. A SiSoft Sandra test will put all of my drives around 20-25MB/s for every benchmark. Anytime my computer, for whatever reason, has to hit the drives with a vigor, everything grinds to a crawl... even the mouse cursor. This rears it's head in games like BF2, FEAR, etc. If my hd is being hit hard (say, verifying client data in BF2), everything gets real chunky, my Teamspeak starts stuttering horribly, etc. I've defragged on a nightly basis, I've reinstalled...

 

This problem started from day one. since then I replaced my Adaptec 2940UW with the 29160... I've replaced the SCA adapters on my Seagate and Compaq drives... I've bought new cables, replaced my terminators. Nothing is fixing the problem.

 

In my previous system (ABit AI-7, p4 2ghz, with the adaptec 2940UW), heavy filesystem usage never caused the rest of the machine to slow. The major litmus test being that my teamspeak never stuttered when verifying client data in BF2. The seagate drive was purchased for this system to replace a really slow single platter Compaq ATA/66 drive. With the compaq at/66 drive, verifying client data was guaranteed to cause stuttering and was viciously slow. Upgrading to the seagate drive and using the adaptec 2940UW pulled from a linux server, everything was peachy. Now with my new machine and what should a staggeringly fast drive setup... i'm suffering the same problems that caused me to upgrade to the better drives in the first place.

 

Any help would be GREATLY appreciated.

 

Thanks,

 

-m

Share this post


Link to post
Share on other sites

I hate to self bump... but...

 

Bump.

 

I searched the forums pretty vigorously for anything related to SCSI performance and didn't find anything. It seems like everyone is sticking to SATA. Hopefully there's someone out there with experience with DFI boards and SCSI drives/controllers and can send some suggestions my way.

 

Thanks, as always,

 

-m

Share this post


Link to post
Share on other sites

Are you sure you did install the HD's to the LVD channel in you SCSI card?

 

The Seagate Ultra320 drive is connected via a terminated 5 drive Ultra 160 LVD cable to the LVD/SE connector. My 29160 is a single channel controller with a 68-pin LVD/SE connector, a 68-pin SE connector, a 50-pin SE connector, and a 68-pin lvd/se external connector. It's an SCA drive with an SCA-to-Ultra320 adapter. However, instead of posting at 160 in the SCSI bios, it posts at 80.

 

My other drives are connected to the 68-pin SE connector because I only have one SCA-to-Ultra160 (or better) adapter. Although in hindsight, the WDE 18130 is Ultra2 LVD and shoud probably get along just fine sharing the cable and connector with the Seagate.

 

I'm planning on ordering a few more SCA adapters when i get my next paycheck. I'm going to get a couple SCA-to-Ultra160 and a couple SCA-to-Ultra320 and mix and match to see if I can get the drives to post at their maximum bus speeds.

 

Which version of the 29160 do you have?

 

I have the vanilla Adaptec 29160. Manufacturer link here: Adaptec site

 

Did you install the latest Adaptec drivers?

 

I did. I upgraded to the most recent 29160 family drivers and repeated the Sandra benchmark... still performing at between 19-29MB/s on all of the tests. The slowest test is buffered writes.

 

 

Thanks for responding... hopefully with some extra brainpower applied to the problem I should be able to this worked out.

 

-m

Share this post


Link to post
Share on other sites

Please use the tools in the begining of this thread and post your results would love to know how SCSI can do.

 

I am sure everyone would love to help you out. If nothing else it will give a clearer idea at how well they really are performing it may be something other than the drive itself. Many people wonder about SCSI performance it would be great to give them hard NFORCE4 specific data.

 

 

 

 

 

 

http://www.dfi-street.com/forum/showthread.php?t=4177

Share this post


Link to post
Share on other sites

Please use the tools in the begining of this thread and post your results would love to know how SCSI can do.

 

I am sure everyone would love to help you out. If nothing else it will give a clearer idea at how well they really are performing it may be something other than the drive itself. Many people wonder about SCSI performance it would be great to give them hard NFORCE4 specific data.

 

 

 

 

 

 

http://www.dfi-street.com/forum/showthread.php?t=4177

 

I'll be glad to when I get home from work today.

 

I would like to clarify... I'm 99% certain my problems are not representative of proper scsi performance. Whether that's the fault of the DFI board, the NF4 chipset, or my configuration is yet to be determined. But for comparison, similar drives (10krpm 16mb cache ultra320 sca scsi) on a similar controller (adaptec 39160, which is the dual channel version of the card I have) in a similar system (opteron 148 of the same stepping as my 146 and similar spec'd memory using the same TCCD chips) perform MUCH MUCH faster. The only difference between the two system that I think is particularly salient is that my machine has a DFI infinity nf4 sli and the other machine has the ASrock agp/pcie uli motherboard (I don't remember the exact model number, but it's the socket 939 board that has both an agp slot and a pci express slot). That gives a SiSoft Sandra benchmark aggregate of >55MB/s. Some of the individual tests peak in the high 60s. Effectively, the ASrock-based system is performing 2-3x (depending on which specific test) faster than my DFI-based system with roughly equivalent hardware on every other component. This is really apparant in loading maps in BF2... The Asrock system takes <45 seconds to load the map and verify client data and my machine takes 2-3 minutes.

 

It's definitely the consensus opinion between my brother (he's got the ASrock-based system) and I that my filesystem should be performing better. He provided me with the 29160 thinking that the bottleneck for my filesystem was the bus. Well, with tbe bus improved, I'm still seeing the same poor level of performance.

 

I wish I had benchmarked the filesystem before I rebuilt the system. My only point of comparison is now strewn around my office in anti static bags.

 

-m

Share this post


Link to post
Share on other sites

Clarification... I stated earlier that the Seagate drive (which is my primary drive and should be the fastest...) is an Ultra 320 80pin drive... it's an Ultra160 80pin drive.

 

http://www.seagate.com/support/disc/specs/...st173404lc.html

 

Officially, my drive is the S*X*173404LC, not the S*T*173404LC... I think the difference in model numbers dilineates oem versus retail and I have the oem version.

 

-m

Share this post


Link to post
Share on other sites

Alright, just ran the benchies...

 

Just wanted to add another clarification. I earlier stated that the Seagate was on the LVD cable and connector and the other two were on the SE connector with another ultra2 cable... This is wrong. The Seagate and WD use the LVD cable/connector. Both post 80 in the 29160 bios. The Compaq is on the SE connector and posts 40. Only one of my available SCA adapters is rated for LVD (it's listed as ultra320 spec). That adapter is on the Seagate drive. If I put the compaq (the other SCA drive) on the LVD cable with the Seagate and WD, it forces all of the drives on the chain to go down to 40 in the bios if I use an SCA-to-Ultra2 adapter on it That's why I have it split off on the other cable and connector, so that the WD and Seagate can do 80 (although the seagate should be doing 160... still haven't figured that one out).

 

Anyhow, here are the HD Tach screens. ATTO screens following in the next post.

Share this post


Link to post
Share on other sites

And hopefully this works to display them in the thread as inline images:

 

Seagate: 32MB HD Tach

attachment.php?attachmentid=3438

 

Seagate: 8MB HD Tach

attachment.php?attachmentid=3439

 

Compaq: 8MB HD Tach

attachment.php?attachmentid=3440

 

WD: 8MB HD Tach

attachment.php?attachmentid=3441

 

----

 

Seagate: ATTO

attachment.php?attachmentid=3443

 

Compaq: ATTO

attachment.php?attachmentid=3444

 

WD: ATTO

attachment.php?attachmentid=3445

 

 

----

 

The ATTO benchmarks look particularly interesting... each drive is definitely hitting a wall on the reads... and there's a big disparity between reads and writes, much more so than what was posted in the other drive benchmarking thread.

 

My brother, who has similar drives, controller, processor and memory (only major difference is the motherboard, which is an ASrock), will be posting benchmarks in this thread for comparison. As I mentioned earlier, his sandra scores were about twice mine and his load times in BF2 are about 1/3rd. Hopefully there'll be here tonight.

 

As always, thanks for the help. Hopefully we can get this resolved so we can see how scsi performs for the other DFI users... and so I don't end up having to plunk down money on some Raptors (grumble). And as a long time scsi true believer, I really want to see my file system fly like I know it should.

 

-m

Share this post


Link to post
Share on other sites

Here is your problem.

I've replaced the SCA adapters on my Seagate and Compaq drives... I've bought new cables, replaced my terminators.
The SCA adaptors will limit the performance of the drives considerably. Adaptec and several do not recommend using SCA drives on 68 pin Ultra160 or Ultra320 conotrollers as they have a tendancy to cause signal distortion.

 

Your best bet would be to get a set of native 68 pin drives.

Share this post


Link to post
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

×
×
  • Create New...