• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

Nforce2 mobo has issues with PCI transfer rate??

Page 2 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
Okay, with the tuner off, here's my scores (starting with 128 block, as scores didn't change much from before and drop off precipitously)

Block Size___Read___Write

128............. 13624.....45567
256............. 14463.....58283
512............. 16249.....81382
1024............21173.....71991

Running RAID 0 on a 3ware Escalade 7000-2 64KB stripe size. Read speeds are about where they should be, but the write speeds are very low, still higher than when running my tuner though.

Makes me wish I had saved the results from my Sil680 (RMA'd). I still have my Adaptec 1200A (HPT370) but its slower than the Sil680 and I don't wanna reformat again. :|

Chiz
 
I'm doing what I can. I got school at 9am EST so I don't wanna be up TOO late... Maybe just a test with single drive on IDE controller and then single drive on RAID card... How would I know if there's a problem anyway?
 
Originally posted by: pillage2001
Correct me if I'm wrong but I thought Read speeds are always higher than write??
Click to expand...

They are, but these speeds are lower than my single 80GB 8MB Ultra, I'll double-check and post its scores. Also, it FEELS much slower writing. I installed the card on Friday, and noticed its write speeds to be much more sluggish than the Sil680, which was a softRAID card. These results just confirm it.

Chiz
 
Originally posted by: Svenn
I'm doing what I can. I got school at 9am EST so I don't wanna be up TOO late... Maybe just a test with single drive on IDE controller and then single drive on RAID card... How would I know if there's a problem anyway?
Click to expand...

Low read/write speeds. In that link up there, they said the transfer speed was limited to 21Mb/s. So you should be getting scores around that.
 
Scores around that on the PCI slot? And higher on the IDE channel? With my RAID setup I had up to 40k write and 80k read as I said before... that was PCI?
 
RAID?? You're running a PRomise RAID card aren't you?? The RAID card uses the PCI interface to transfer data. It's part of a PCI device although it's onboard. The SB's IDE is independent of the PCI's transfer's speed.

edit: Corrected N with SB. Thanks MechBgon. 🙂
 
The IDE controllers are built into the SB but you've got the main point correct: the built-in IDE controllers are not on the PCI bus, they're on the Hypertransport bus. I do have a Promise Ultra66 card in my office somewhere (I think) that I can stick in there, it's fast enough for my Barracuda ATA IV. Another thing to test. Dang, I better get to bed so I can get up early enough to DO all this stuff! 😕
 
Svenn, if your write speeds are capped at 45MB/s, but your read speeds are capped at 80MB/s, then it's very unlikely the PCI bus is causing the low performance. If that was the case then the read results should be capped at the same amount.

I hope this possibility is not the case. I am going to be building an AMD based system for my girlfriend soon, and I wanted to use an nForce2 based motherboard.
 
Yeah, that's what I was talking about. Sorry, I'm tired, eheh. If I was getting ATTO scores of 40k and 80k on my RAID card, does that disprove the theory? That's higher than the 20MB/S stated, but still low scores for a RAID setup...
 
Originally posted by: Svenn
Yeah, that's what I was talking about. Sorry, I'm tired, eheh. If I was getting ATTO scores of 40k and 80k on my RAID card, does that disprove the theory? That's higher than the 20MB/S stated, but still low scores for a RAID setup...
Click to expand...

Your scores were constant at around 40k right?? That's pretty darn low for a pair of RAID drive. Not too sure if that's the cause of the PCI bus. Have to run a single drive to make sure. 🙁
 
Bleh, just did some research and it looks like the 3ware 7XXX series cards suffer from poor write speeds across the board :| Yah, its a PCI card, but it looks like the PCI bus isn't holding it back, its the card. I was talking about the TV tuner using the PCI bus through the SB and then the hypertransport bus to the NB using video overlay.

Hmmm, Newegg isn't gonna like me much if I ask for another RMA on a RAID card lol.

Chiz
 
I'm trying a single drive on the IDE controller now. Just got windows installed, installing nForce2 drivers and things, what all drivers do I need to install?
 
Originally posted by: Svenn
I'm trying a single drive on the IDE controller now. Just got windows installed, installing nForce2 drivers and things, what all drivers do I need to install?
Click to expand...

Try the SW drivers first 🙂 Then use the standard MCP-T IDE drivers.

Chiz
 
Originally posted by: Svenn
Yeah, that's what I was talking about. Sorry, I'm tired, eheh. If I was getting ATTO scores of 40k and 80k on my RAID card, does that disprove the theory? That's higher than the 20MB/S stated, but still low scores for a RAID setup...
Click to expand...
It sounds like it does disprove the theory, or at least it proves that it does not apply in all situations. I guess my question, which I will pursue tomorrow, is whether nVidia's "Performance" IDE drivers boost the performance of the southbridge's non-PCI IDE controllers at the expense of the PCI-bus performance.

Let's say nVidia wanted to boost their Content Creation 2002 score. Let's say they found that giving the native IDE controller top priority boosted those scores, but reduced the performance of the PCI bus. So what, they might say. There's nothing on the PCI bus anyway, let's do it and get ourselves some good reviews at the big review sites.

If this were the case, then switching to these "performance" drivers should make my PCI SCSI card and my PCI Promise IDE card slower.

So how about this: I do an initial set of benchmarks using WHQL drivers. My SCSI card should hit ~120Mb/sec like before, and my Promise Ultra66 IDE card should allow my IDE drive to do its usual ~42000 in ATTO. The IDE drive should also do about ~42000 on the southbridge's non-PCI IDE controller, because it simply doesn't go any faster than that.

Then I switch to the "Performance" IDE drivers. If the SCSI throughput and the Promise Ultra66 IDE card ATTO benchies drop significantly, then that may indicate that nVidia has opted to improve the native IDE controller's performance at the expense of the PCI bus (and perhaps the other devices that share the Hypertransport link?)

The suspense builds... 😀 FedEx tracking on my 8RDA+ :Q😀😎🙂
 
Svenn, Quick ?, are you using ECC or unregistered PC133 with that SX4000? I have tons of unregistered PC133 laying around, but I don't want to buy ECC if I don't have to. I've read reports you can use either, and that you can even use 512MB as long as the RAM is dual-bank.

TIA,
Chiz

Edit: E-mailed 3ware Tech Support, said this was basically unacceptable performance out of a $120 hardware RAID card and that if there wasn't a fix I was gonna get something else :| Looks like either RocketRAID 133 or the SX4000.

 
You know, my impression was that just striping the drives in Windows' Disk Management actually was capable of pretty good performance. With as much CPU horsepower as we've got on hand these days... anyone game to try it, please fill us in on the results 😀
 
Originally posted by: mechBgon
You know, my impression was that just striping the drives in Windows' Disk Management actually was capable of pretty good performance. With as much CPU horsepower as we've got on hand these days... anyone game to try it, please fill us in on the results 😀
Click to expand...

Software RAID you mean??
 
Downloads, Applications/Tools/Utilities, ATTO ExpressPCI Host Adapter Utilities, Windows SCSI Utilities Version 1.63
 
Originally posted by: pillage2001
Originally posted by: mechBgon
You know, my impression was that just striping the drives in Windows' Disk Management actually was capable of pretty good performance. With as much CPU horsepower as we've got on hand these days... anyone game to try it, please fill us in on the results 😀
Click to expand...

Software RAID you mean??
Click to expand...

Yes, exactly.
 
You must log in or register to reply here.

TRENDING THREADS

Back
Top