8800GTX + Hi-perf RAM = Trouble?

[Deinonych]

Hi all,

I posted this over on NVIDIA's forums, but I thought I'd run it by the users here as well.

Like many others, I've had problems with my EVGA 8800GTX. I recently RMA'd my card due to graphical glitches, freezes and BSODs. The new card has been running fine for about a month. Last week, I got my first BSOD - at the desktop, right after rebooting. It kept crashing, so I reinstalled the drivers (158.22).

Last night, I was attempting to play some C&C3 and the game kept locking at the main menu, with lots of blocks and graphical glitches. I checked the temperature of the card, and it was at ~65C - well within operating range. Subsequent reboots and driver installations produced the same results - lock at the main menu. I tried Counter-strike, and it locked up during the first few seconds of the video stress test.

I finally got the card working again by completely disconnecting all power (including the PCIe power connectors) and reconnecting power. I also updated the drivers to the latest 162.18 betas. While searching for a possible solution, I came across this post:

http://forums.techpowerup.com/showthread.php?t=31266

The interesting part of this post is where he mentions high-performance RAM. I have Crucial Ballistix DDR2 in my machine, which supposedly (according to this post) doesn't "play nice" with the 8800GTX. I'm thinking about picking up a pair of 1.8V 1GB sticks to replace my Crucial sticks to see if my problems go away. Any thoughts on this theory?

For the record, I've ruled out just about everything with my system:
Heat: the card idles at 63-65C depending on whether the chassis door is open or closed (Antec P180). Proc idles between 39-42C.
Motherboard/Proc: System is rock-solid during Prime95, as well as other processor-intensive tasks like video editing and ripping FLACs
Memory: I've run Memtest86+ overnight with no errors
Power: I have power coming from two separate PCIe-specific modular connectors on the PSU, plus the PSU is rated for more than enough juice for this card. I haven't experienced any power-related instability in the system.

 

[aka1nas]

Can you give us more detailed hardware specs? I'd bet your PSU isn't able to supply enough amperage while the video card and the rest of the system is under load. A better test would be to run Large FFT Orthos in the background while running 3dmark or some other 3d benchmark to try to stress everything at once. Memtest isn't going to cut it if the memory is not bad, just not receiving enough power when the GPU is under load.

 

[Deinonych]

Sure. System specs are in my sig. I'm running a Seasonic M12-700 PSU. It has two PCIe-specific modular plugs. Each power connector on the card is connected to a dedicated plug on the PSU. The PSU is rated to run SLI, so I don't really think it's the problem (but I could be wrong).

I'll try running Orthos + 3DMark this evening.

 

[Slugbait]

I see you have a P5N32-E. It has the notorious 680i. Have you flashed to the latest BIOS? We already know eVGA 8800 GTX problems were solved this way on a 650i.

 

[Deinonych]

I have the P5N32-E SLI Plus, which uses a 650i. I'm running the most recent (non-beta) BIOS for the board.

[edit] Thanks for referencing that thread. I'm going to try to get my hands on a Kill-A-Watt to see what the total system draw is.

 

[Cookie Monster]

From what i can tell Asus P5N32-E SLI Plus is infact a 680i. They had a massive talk over at xtremesystems trying to find out the exact chipsets that was being used on that board. It was something along the lines of 680i+570Sli(south bridge).

 

[Deinonych]

Well, according to AnandTech, it's a 650i northbridge with a 570 southbridge. Either way, I'm running the latest BIOS and chipset drivers.

 

[Slugbait]

Originally posted by: Cookie Monster
From what i can tell Asus P5N32-E SLI Plus is infact a 680i. They had a massive talk over at xtremesystems trying to find out the exact chipsets that was being used on that board. It was something along the lines of 680i+570Sli(south bridge).

Nope, we're both wrong, it's a 650i

 

[aka1nas]

Originally posted by: Slugbait

Originally posted by: Cookie Monster
From what i can tell Asus P5N32-E SLI Plus is infact a 680i. They had a massive talk over at xtremesystems trying to find out the exact chipsets that was being used on that board. It was something along the lines of 680i+570Sli(south bridge).

Nope, we're both wrong, it's a 650i

The conclusion from the XS thread IIRC was that we are all forgetting that it's a chipSET. I.E., both the 680i and 650i use the same/similar NB chip with the 680i possibly being a better binned version. The extra PCI-E lanes are provided by the SB, which is where the 2 chipsets differ. The 650i uses a newer SB that provides no extra PCI-E lanes, while the 680i and the Plus uses a 500 series NB to provide enough lanes for x16/x16 SLI and the 3rd PCI-E slot.

 

[Deinonych]

Back on topic: I've run Orthos for 2 hours now, with a couple runs of 3DMark06 and a few CS:S video stress tests (while Orthos was running). No lockups, hangs or other problems.

 

[dclive]

Originally posted by: Deinonych
Back on topic: I've run Orthos for 2 hours now, with a couple runs of 3DMark06 and a few CS:S video stress tests (while Orthos was running). No lockups, hangs or other problems.

I'd be happy to look at the minidump files if you'd like to see if anything jumps out at me. If successful, that would show what's causing the dumps - but typically it isn't as useful for hardware issues...

 

[Deinonych]

Originally posted by: dclive
I'd be happy to look at the minidump files if you'd like to see if anything jumps out at me. If successful, that would show what's causing the dumps - but typically it isn't as useful for hardware issues...

Thanks! PM Sent.

 

[Deinonych]

Looks like there's a new BIOS (0801) that just showed up on Asus' site a couple days ago. I don't see anything that specifically references problems with the 8800GTX, but who knows. Maybe it will solve my problems. *crosses fingers*

 

[chizow]

The 3 times I thought my GTS was the source of my reboot/BSOD/CTD/instability problems all turned out to be memory problems, and ultimately, motherboard problems. I had a P5N-E SLI, one of the first boards from early January. Basically it killed 3 kits of RAM, including 2 Ballistix PC-8000 and 1 Buffalo PC-6400. I looked into it and it does seem to be somewhat of a problem with Asus/650i/680i boards. Thankfully Crucial RMA rocks (upgraded me to some PC-8500 after the 2nd RMA) and Buffalo RMA is also very good (although a bit clunkier process).

My guess is that the boards have problems regulating power to the dimm slots and ultimately spike and kill the memory modules. Not sure exactly what the problem is with the boards, I think I read somewhere most of the Asus boards (that cost less than $200) are 2-phase power vs. the higher-end boards which run 7-phase. I purchased and upgraded to an MSI P6N SLI Platinum about a month and a half ago and haven't had any problems since (except the BIOS, make sure to flash to 1.3 if you want to run 4GB!).

 

[Deinonych]

Originally posted by: chizow
The 3 times I thought my GTS was the source of my reboot/BSOD/CTD/instability problems all turned out to be memory problems, and ultimately, motherboard problems. I had a P5N-E SLI, one of the first boards from early January. Basically it killed 3 kits of RAM, including 2 Ballistix PC-8000 and 1 Buffalo PC-6400. I looked into it and it does seem to be somewhat of a problem with Asus/650i/680i boards. Thankfully Crucial RMA rocks (upgraded me to some PC-8500 after the 2nd RMA) and Buffalo RMA is also very good (although a bit clunkier process).

My guess is that the boards have problems regulating power to the dimm slots and ultimately spike and kill the memory modules. Not sure exactly what the problem is with the boards, I think I read somewhere most of the Asus boards (that cost less than $200) are 2-phase power vs. the higher-end boards which run 7-phase. I purchased and upgraded to an MSI P6N SLI Platinum about a month and a half ago and haven't had any problems since (except the BIOS, make sure to flash to 1.3 if you want to run 4GB!).

Interesting. According to the specs of my motherboard, it runs 8-phase power. Fortunately, I haven't had any problems with RAM so far. But, I'll be sure to keep that in mind. I have Crucial sticks as well, so it's good to hear their RMA process is customer-friendly.

There's got to be something wonky going on with one of my components. It could be the motherboard - hopefully BIOS updates will eventually work things out. I have to say that after many many years of using NVIDIA cards, this is the first problem I've ever had. Odd.

 

[dclive]

IMAGE_NAME: nv4_disp.dll

DEBUG_FLR_IMAGE_TIMESTAMP: 45df38b5

MODULE_NAME: nv4_disp

FAULTING_MODULE: bf9d5000 nv4_disp

DEFAULT_BUCKET_ID: GRAPHICS_DRIVER_FAULT

CUSTOMER_CRASH_COUNT: 1

BUGCHECK_STR: 0xEA

PROCESS_NAME: hl2.exe

LAST_CONTROL_TRANSFER: from 00000000 to bfa7fc55

STACK_TEXT:
a2762bb8 00000000 bfb43d93 a2762ce4 e3f5cef0 nv4_disp+0xaac55


STACK_COMMAND: .thread 0xffffffff8a414020 ; kb

FOLLOWUP_IP:
nv4_disp+aac55
bfa7fc55 ?? ???

SYMBOL_STACK_INDEX: 0

SYMBOL_NAME: nv4_disp+aac55

FOLLOWUP_NAME: MachineOwner

FAILURE_BUCKET_ID: 0xEA_IMAGE_nv4_disp.dll_DATE_2007_02_23

BUCKET_ID: 0xEA_IMAGE_nv4_disp.dll_DATE_2007_02_23

Followup: MachineOwner

__

Is there a newer nVidia driver release you can try?

 

[Deinonych]

Originally posted by: dclive
IMAGE_NAME: nv4_disp.dll

DEBUG_FLR_IMAGE_TIMESTAMP: 45df38b5

MODULE_NAME: nv4_disp

FAULTING_MODULE: bf9d5000 nv4_disp

DEFAULT_BUCKET_ID: GRAPHICS_DRIVER_FAULT

CUSTOMER_CRASH_COUNT: 1

BUGCHECK_STR: 0xEA

PROCESS_NAME: hl2.exe

LAST_CONTROL_TRANSFER: from 00000000 to bfa7fc55

STACK_TEXT:
a2762bb8 00000000 bfb43d93 a2762ce4 e3f5cef0 nv4_disp+0xaac55


STACK_COMMAND: .thread 0xffffffff8a414020 ; kb

FOLLOWUP_IP:
nv4_disp+aac55
bfa7fc55 ?? ???

SYMBOL_STACK_INDEX: 0

SYMBOL_NAME: nv4_disp+aac55

FOLLOWUP_NAME: MachineOwner

FAILURE_BUCKET_ID: 0xEA_IMAGE_nv4_disp.dll_DATE_2007_02_23

BUCKET_ID: 0xEA_IMAGE_nv4_disp.dll_DATE_2007_02_23

Followup: MachineOwner

__

Is there a newer nVidia driver release you can try?

Thanks for the analysis! Yes, I've since moved to the latest beta build (162.18). So, hopefully with the new drivers + the newly released mobo BIOS, the problem will go away. Time will tell, I suppose.

I'm going to test the PSU with a Kill-a-Watt this weekend just to be sure.

 

[Deinonych]

Well, I've tested the total power draw from the system (sans monitor). Highest power draw with Orthos and a 3D app running was 335W. I tried 3DMark06, CS:S Stress Test and a Supreme Commander replay while Orthos was running. 335W was surprisingly low, but that's what the Kill-a-Watt display read. So, based on this info, my PSU isn't even breaking a sweat with my components at load. I guess I'll have to hope that the new BIOS fixed it.

 

[Acanthus]

Originally posted by: Deinonych
Well, I've tested the total power draw from the system (sans monitor). Highest power draw with Orthos and a 3D app running was 335W. I tried 3DMark06, CS:S Stress Test and a Supreme Commander replay while Orthos was running. 335W was surprisingly low, but that's what the Kill-a-Watt display read. So, based on this info, my PSU isn't even breaking a sweat with my components at load. I guess I'll have to hope that the new BIOS fixed it.

This doesnt tell you how much load is on each individual rail though, the wattage of the PSU is a combined rating of all rails.

You could be very low on your 5v rail, and over the limits of your PSU on the 12v rail. (where the GPU draws from mainly)

 

[Deinonych]

That's true. However, I don't have the tools to test any of the 12V rails. From what I can tell from the specs, my PSU has two dedicated 12V rails @ 18A each for the PCIe connectors. Since each power connector on the video card is connected to its own PCIe power connector on the PSU, I'm quite skeptical that there's a power problem.

Again, I could be wrong, but it seems highly unlikely that power is the problem.

 

[Heinrich]

I get an occasional BSOD (once a week average) that I think is related to logitech drivers, because it seems to happen while clicking quickly. I found a 'tutorial' of sorts on how to analyze the dump files that are generating. I downloaded processexplorer. Then I got lost and could not get much further...permissions errors and I was given commands to run that made no sense. Can one of you smart guys help - this is Vista 64

 

[Deinonych]

That's interesting. I'm using a Logitech mouse (MX518) with Logitech's drivers. If I get another BSOD, I'll try using the standard Windows driver for my mouse.

 

[dclive]

Originally posted by: Heinrich

I get an occasional BSOD (once a week average) that I think is related to logitech drivers, because it seems to happen while clicking quickly. I found a 'tutorial' of sorts on how to analyze the dump files that are generating. I downloaded processexplorer. Then I got lost and could not get much further...permissions errors and I was given commands to run that made no sense. Can one of you smart guys help - this is Vista 64

Put the dumpfile(s) on a website and post the URL, and we'll look.

 

[customcoms]

Anyway to test the card in a buddies system? I HIGHLY doubt its PSU related, but that is a common component. The other common component is the motherboard. RAM killing a gpu=utter bs in my opinion. ram killing a motherboard, or much more common, a motherboard killing ram, oh yeah. Especially high performance ram that requires STABLE volts to perform. However, memtest86 is giving no errors, your cpu's fine, everything points to the graphics card. Have you tried a different slot on the board for the card? With SLI mobo's, you can usually get away with running the gpu in the lower slot and telling the pc your in sli mode (in bios/drivers). just a thought...

 

[dev0lution]

What voltages is your board running for the memory, HT (MCP and SPP), and PCI-E? Don't know if it matters, but by default on my eVga 680i + XFX 8800GTX these were reading higher than what nv monitor considers "safe" - and this is without an OC. I believe it's related to the board detecting the LinkBoost feature. NV Monitor might be overly cautious, but it might be worth looking into...