Nforce2 mobo has issues with PCI transfer rate??

[Svenn]

It should be faster theoretically because of the 256MB onboard cache, right?

 

[pillage2001]

Weird. Downloaded what Svenn asked me to but I can't seem to run the Setup. Double clicked it and nothing happened??

 

[chizow]

Originally posted by: Svenn
It should be faster theoretically because of the 256MB onboard cache, right?

And dedicated Math Coprocessor. I found that the softRAID cards would spike anywhere from 50% to 100% CPU utilization during benchmarks and apps, but remained pretty steady around 30-40%. My main reason for going hardraid, even though I just added 600mhz (that's right Pillage, MHz biatch ) of muscle to my CPU, was for video editing and encoding/decoding of streaming video. Damn, looks like the 3ware is a dud. Although the read speeds are very good at the high-end, its a very steep drop-off from 128 down and the write speeds are miserable :| If I weren't lazy and had my Comcast home page set-up I'd post my results.

Chiz

Edit: Yah, I think the WinXP softRAID results would probably be very similar to a softRAID card, but of course you don't get the benefit of having a bootable drive with its own BIOS and controller.

 

[mechBgon]

From what I've seen, the throughput is always low on the small sizes with ATTO.

 

[chizow]

Originally posted by: mechBgon
From what I've seen, the throughput is always low on the small sizes with ATTO.

Hehe well, its probably a good thing I don't have my web space set-up, I'd be too embarassed to post the low block scores. I'm talking 56k speeds at 0.5 and my cable modem probably has better download speeds at 8.0 for write speeds. Read speeds are also < 20Mbps until the 64 blocks...............

Chiz

 

[pillage2001]

Something interesting going on over at ATOT. I'll be back.

 

[ssanches]

Seagate Barracuda ATA IV, 2Mb cache, 7200RPM, Auto Acoustic Management disabled, on the KT333's built-in IDE controller, on its own IDE cable
FAT32 partition results [L=here]http://mechbgon.tripod.com. I know they're rather strange (reads slower than writes?!) but whatever, that's what I get on my KT333 board. As you can see, some of the marks are in the low 40Mb/sec range like they should be for this drive, which maxes around 45Mb/sec throughput in theory.

mechBgon, I have a Barracuda IV HDD on a VIA 686B southbridge and while my ATTO benchie's read speeds matches your read speeds, my write speeds never exceed the 30MBps mark in ATTO! In fact in all the benchmarks I've seen on the net, the Barracuda IV never exceeds the 35MBps sustained write speed. I recall very well that the Barracuda IV was critized for it's low write speeds and I'm shocked with your numbers of around 41MBps (I don't know if they're hosed). When I use HDtach 2.61 I get read speeds around 44MBps at the beginning and write speeds of around 30MBps IIRC (I can't recollect the write speeds thru HDTach very well since I haven't tested them lately; however I've verified the read speeds)

Here are my system specs:
Barracuda IV, 2MB cache, AAM disabled, SMART monitoring disabled, firmware revision 3.10 (non raid)
VT8365A with a 82C686B southbridge; PCI latency @ 64, UDMA mode 5, IDE busmaster enabled
Athlon XP 1600, 512MB PC-133 SDR 7.0ns CPU, RAM and FSB at stock)
VIA 4-in-1 ver 4.45
MS Windows XP Pro SP1, FAT 32 primary partition used for these tests.

 

[mechBgon]

I was asked if I could work today (supervisor's kids are sick and he had to stay home), so I may be delayed a bit in providing my promised experiments. Never fear, I can always come in early tomorrow. Should have at least a few results for you guys tonight, I hope.

 

[pillage2001]

Originally posted by: mechBgon
I was asked if I could work today (supervisor's kids are sick and he had to stay home), so I may be delayed a bit in providing my promised experiments. Never fear, I can always come in early tomorrow. Should have at least a few results for you guys tonight, I hope.

Work comes first dude.

 

[Dug]

There was a problem with scsi write speeds and WinXP/2000 awhile back. Could XP be a culprit? Or possibly ATTO not giving accurate results.

 

[mechBgon]

Ok, just a heads-up: did some various tests with the nForce 220D (classic nForce) as a baseline. Suffice it to say that the evidence points to a healthy PCI bus on the nForce220D. With the Promise UltraATA 66 controller AND the 19160 running SCSIBench at the same time, both on the PCI bus, their combined same-sector-reads come out to about the same as when just the 19160 is running it: ~120Mb/sec. Running ATTO on one and SCSIBench on the other had some impact on the 19160 and a lot of impact on the Ultra66, but their combined throughput was still way up there.

I also compressed the contents of UT Disc 1 at Maximum compression in WinZip 8.1 (took 3:20) and then transferred the resulting 380Mb Zip file from the Barracuda ATA IV to the Cheetah (across the PCI bus, both drives were on PCI-based controllers here). Copying the 380Mb file took 10 seconds, which is 38Mb/second... considering the Barracuda's maximum read speeds, that's pretty good, definitely much higher transfer than watching a DVD or copying a CD would demand.

The nForce2 board is lying here waiting for an installation. Let's see if it can whup its older sibling. Back in a couple hours or so... Oh, and I did take some screenshots of my benchies but I'll need to get home before I can post them, probably.

 

[SpideyCU]

Just to jump in here, if I may make a suggestion. If you're noticing oddities in your scores, it's not necessarily the PCI bus's fault. There have been funky issues with SCSI and RAID drives under Windows XP, and AFAIK, these haven't fully been resolved yet. People doing ATTO testing sometimes showed their single IDE hard drive out-reading their RAID setup. I don't have the link offhand, but, I ran across this on storagereview's forums several months ago when I was looking to get myself a SCSI hard drive (which eventually went into my Linux box ). I'll try to find references to what I'm yammering about so I'm not just the "crazy cousin" that stumbles into a party uninvited.

Edit: OK, so it's not technically a "WinXP issue", and may or may not be relevent (I hadn't been keeping up-to-date on this issue). Anyway, what I was ranting about was explained HERE. "The only reason it occurs on SCSI-class devices (SCSI & IDE RAID) and not on standard IDE, is that IDE does not have a WRITE_THROUGH flag in its command set." ...and later on... "As far as I can tell, all IO operations on XP will suffer some performance loss due to the synchronous flushing of meta data. I still have some digging to do, but it appears that the NTFS log file gets FUAed to disk, even while your own data can pend in the write buffer."

 

[mechBgon]

I'm starting to think what you just said, SpideyCU. I got my new board in place and what do you know... Win2000 was all, "hey, some kind of nForce board like that last one" and aside from having to use a PS/2 keyboard to get into Windows so it recognized my USB keyboard, it looks like no reinstall is needed.

Here's the kicker: benchmarks are essentially coming out the same as before, to make a long story short. There is no PCI problem showing up. Now remember: I'm using Win2000. No SCSI issues. And when I reinstalled the nForce UDP, I was not offered the option to use a Performance IDE driver. In Device Manager, the IDE controllers are showing up as standard IDE controllers, not SCSI, and I bet that's where WinXP is different... it'll offer you the Performance driver, it will make it look like SCSI, and the WinXP SCSI Monster rears its ugly head and makes everything work like junk. Am I getting warm?

On my little UT/WinZip benchie, it performs the same as before. I think the 1700+ is the bottleneck... hmm, oughta get me one of those Tbred-B's The board does seem to re-launch my slowest-launching app from RAM a bit faster than it used to, although that's purely a seat-of-the-pants impression.

So I hope that's of some help to you guys. I don't have a DVD player to get the audio and AGP card into the action, or I'd give that a test as well, and I can't opt for the "Performance" IDE driver to test that. Microsoft will be supporting Win2000 Professional through the end of 2008 IIRC, so if you're not pleased with your XP performance, I believe you do have "downgrade rights" to run 2000 instead... call up Microsoft and ask for your Win2000 downgrade, maybe?

 

[chizow]

Just finished upgrading my drives from basic to dynamic in WinXP...what a HUGE difference

My RAID scores are pretty silly now, much better than I expected, and my write scores in ATTO are actually HIGHER than my read scores across the board

I'm getting writes in the 70MBps range starting with the 16 block and peaking at 80MBps and change. My read scores are now hitting 50MBps in the 8 block and climbing quickly into the 70MBps range at the 32 block.

I'm happy

I didn't do any further testing though, been busy gettin this thing straightened out. Had to reformat b/c my initial Dynamic Disk upgrade failed and Windows refused to read from it.

Those of you with SCSI problems and WinXP should definitely look into converting your basic drives into dynamic drives, there's no way I woulda thought performance would increase this much!!!!

Chiz

 

[mechBgon]

SWEET

 

[Svenn]

How do you convert basic drives to dynamic?

 

[chizow]

Originally posted by: Svenn
How do you convert basic drives to dynamic?

I had no clue about it either Luckily I e-mailed 3ware tech support before I dipped into the Newegg RMA route again. The difference is amazing, I'm talking like 1000% percentage points in the very low blocks and easily 400-600% in disk writes. Interestingly though, reads went down slightly.

Anyways, right click on my computer --> manage ---> disk management.

It'll bring up a window with drives listed in a top and bottom split-screen. In the bottom portion, not the graphical area, but to the left of each graph that blends with the window background tan color (default) it should list the drive name, type, size and status:

Drive 0
Basic
228GB
Online

Its really unintuitive (blended), as you need to right click this to bring up the option to "convert to dynamic disks".

MAKE SURE you close EVERYTHING. XP tends to require 2 restarts if you have stuff open, and this pretty much killed my conversion the first time. If it doesn't shut down and restart by itself, its reformat/reinstall time (I learned the hard way).

Hope that helps, and improves your scores as much as they did mine!!! (although yours were much better to begin with.

Chiz

Edit: I'm not just happy my benchies increased, I noticeably feel the RAID performance difference again (like with the SoftRAID cards)..benchies just quantified it for me

 

[pillage2001]

So, what does converting the drives to dynamic really do?? I'm ready to convert but not sure if I would wanna take the risk.

 

[SpideyCU]

Originally posted by: pillage2001
So, what does converting the drives to dynamic really do?? I'm ready to convert but not sure if I would wanna take the risk.

I'm not sure what it "really" does (again, I just installed my SCSI drive on Linux and let it be ), but a lot of people have used it as a workaround for months and months now to improve performance in WinXP, and I haven't heard any bad things about it. If you really want to be sure, I'd suggest checking the forums on storagereview.com or some similar site - those people are hardcore about their storage needs.

so if you're not pleased with your XP performance, I believe you do have "downgrade rights" to run 2000 instead... call up Microsoft and ask for your Win2000 downgrade, maybe?

Yup, people with WinXP do indeed have downgrade rights.

In Device Manager, the IDE controllers are showing up as standard IDE controllers, not SCSI, and I bet that's where WinXP is different... it'll offer you the Performance driver, it will make it look like SCSI, and the WinXP SCSI Monster rears its ugly head and makes everything work like junk. Am I getting warm?

Warmer than most, mech. If there's one thing that I don't like, it's a mis-diagnosis. I like to know what the REAL culprit is, and I seriously doubt the NF or NF2's PCI bus has any bandwidth limitations that we're going to run into with standard devices.

 

[Bovinicus]

The help file claims that dynamic disks simply allows more than 4 volumes per disk to be used. I was thinking of doing it myself, but apparently it is only useful for RAID arrays. I looked up this information on MS's website.