VOGONS


First post, by acl

User metadata
Rank Oldbie
Rank
Oldbie

Hello

During my (soon ending) vacations, I wanted to troubleshoot an issue I've been experiencing for quite a while with my 2005 system.
It's one of my "permanent" gaming system with the following configuration:

  • A8N-SLI Premium
  • Opteron 154 (~= Fx57) @3.0Ghz
  • 1GB Kingmax DDR 433 (PC3500)
  • 2x GeForce 7800 GTX 256 in SLI
  • Sound Blaster X-Fi Platinum
  • 2x WD Raptor 36G RAID0 (NForce4)
  • Win XP SP3

The issue is frequent system freeze during games.
More often with F.E.A.R / Oblivion. Less crashes with UT2004 / Doom3.
The image freeze (not black) and the system is not responsive, requiring a reset. Sometimes the first reboot also get stuck before windows desktop, requiring another restart.

I'm using games from GOG (Except Doom3 from DVD)
I've tried several Nvidia drivers : 175.19 and the 3xx (last avail. from Nvidia website) with the same result.
It seems to occure only with SLI. (Or when using the 7959GX2 which uses SLI internally)

I also tried multiple alternative parts :

  • Other CPUs (Opteron 180, A64 3200+)
  • Other memory (Corsair XMS)
  • Other mainboard (A8N-SLI Deluxe)
  • Other GPU (7950 GX2, 7800GTX 512 SLI before one died)
  • New PSU (650W 80+)
  • Re-installing windows

Things i did not tried to change : Disk RAID, Sound card.

I also checked the temperature (by running games windowed with GPUZ + Core-temp visible) and it's probably not temperature related. The CPU runs super cool and the GPUs never goes above 80 C° (for GTX 7800) or ~70 (for 7950 GX2).
The chipset seems really hot, but I have no reading of it's temperature.

My next attempt will be to remove the sound card, but have very little hope with this solution.
I will also (hopefully) receive a shipping with other 7800 GTX 512 MB.

Do you have any other ideas to fix the issue ?
Thank you very much!

Last edited by acl on 2023-01-24, 16:42. Edited 1 time in total.

"Hello, my friend. Stay awhile and listen..."
My collection (not up to date)

Reply 2 of 9, by acl

User metadata
Rank Oldbie
Rank
Oldbie
agent_x007 wrote on 2023-01-21, 11:31:

What's your airflow inside case ?

Currently it's one fan in the lower front side, blowing cool air on the HDDs + GPU
A tower cooler + fan on the CPU, blowing toward the back.
One fan in the upper back extracting air (in front of the CPU fan)
PSU extracting air in the upper back.

But I tried several setups.
I added up to two other case fans (pumping fresh air)
I also changed the CPU cooler.
And the two motherboards I tested have different chipset cooling system (Deluxe has a fan. Premium has a heatpipe heading to a radiator near the CPU fan)

My two GTX 7800 used to have some cooling issues because of the new pads I added on the memory. The cooler barely touched the GPU die and temperatures sometimes raised to 100C° but I solved that by adding a metal shim between to fill the gap. And now they max out around 80C°

The CPU runs quite cool (even with OC) and I still have the issue with several non OC CPUs I tried.

I initially really focussed my attention on temperature monitoring, but I'm not sure it is what cause this problem.

It seems that the problem does not occur on single GPU config...

I'm currently thinking about :
- lower the graphics settings
- remove the sound card
- remove the raid
- try single GPU more extensively.

But still not sure.

"Hello, my friend. Stay awhile and listen..."
My collection (not up to date)

Reply 3 of 9, by chinny22

User metadata
Rank l33t++
Rank
l33t++

I think your on the right track.
Clean install of windows with just 1 graphics card installed and test.
If all ok keep adding hardware 1 part at a time and test again

Reply 4 of 9, by acl

User metadata
Rank Oldbie
Rank
Oldbie

Ok, I think I finally found the issue.
I ran 3DMark06 multiple hours, left Doom3 running alone 2 hours more and played F E.A.R yesterday evening... no crash at all.

The issue was... : Instability with 4 ram sticks.

I all my previous tests I used 4 sticks. Either 4x Kingmax 256MB DDR433 or 4x Corsair XMS 512 DDR400 CL2.

I started to suspect a memory issue after a strange memtest report. No errors initially, buy some errors appearing after a few runs. I started to test with only two slots populated. The system stopped crashing either with 2x Kingman or 2x Corsair ram sticks.

So as a conclusion, either :

  1. My two DDR kits have defects. But I was lucky enough to pick two good sticks among 4 for the 2 kits in my 2 consecutive tests.
  2. Dual Channel with 4 sticks is not super stable by design on Athlon64 (Not sure if MB has something to do with stability because of the CPU integrated memory controller)

It must be on of these two answers.
I can only validate on the long run, but with 4 sticks, I was not able to play for more than ~20 minutes without a crash. So for now, I consider the issue as fixed.

"Hello, my friend. Stay awhile and listen..."
My collection (not up to date)

Reply 5 of 9, by Grem Five

User metadata
Rank Member
Rank
Member

I'm not sure if this applies to your system but my old 1366 system was unstable when I mixed RAM from different manufacturers with identical specs. Even though the ram sticks all had the same specs they didn't play nice with each other.

If I only used all the sticks from one manufacturer or the other manufacturer then it was fine. Odd thing was it only had probs in 2 of my games where it would crash the whole system. In my animation software which seemed to hit the ram way harder when rendering it was fine but load up one of those games and about 5 mins in it would take my whole system down.

My system could render for hours using about 100% of those mixed sticks fine. Took me awhile to figure it out.

Reply 6 of 9, by acl

User metadata
Rank Oldbie
Rank
Oldbie
Grem Five wrote on 2023-01-24, 20:10:

I'm not sure if this applies to your system but my old 1366 system was unstable when I mixed RAM from different manufacturers with identical specs. Even though the ram sticks all had the same specs they didn't play nice with each other.

If I only used all the sticks from one manufacturer or the other manufacturer then it was fine. Odd thing was it only had probs in 2 of my games where it would crash the whole system. In my animation software which seemed to hit the ram way harder when rendering it was fine but load up one of those games and about 5 mins in it would take my whole system down.

My system could render for hours using about 100% of those mixed sticks fine. Took me awhile to figure it out.

1366 is not that old 😀. I stopped using mine as daily gaming machine last year. But yes, with six memory slots for triple channel, im sure that it's pretty sensitive. I use 6 identical 4GB sticks on my 1366 system.

And yes I experienced the issue you're describing with a socket 775 and DDR2. I had 4 stick of ram, visually identical, but one was of a slightly different model with different specs. The system even had issues to boot (but I was overclocking)

For my s939 problem, I never mixed ram types.
I have two sets of four identical modules. And I have issue with both of the 4 modules sets.
In both cases, the issue vanish as soon as I remove 2 sticks....

"Hello, my friend. Stay awhile and listen..."
My collection (not up to date)

Reply 7 of 9, by pentiumspeed

User metadata
Rank l33t
Rank
l33t

Socket 1366 uses Nehalem microarchitecture which are Bloomfield, Gulftown, Jasper Forest, Lynnfield, Gainestown and Westmere-EP. I do have notebook T410 based on Arrandale, also both Arrandale and Clarksfield processor can be plugged in high end notebooks like T510 for example.

Reason I didn't look at Nehalem stuff is this:
However! Performance is not as great as C2D (I went from E8600 with 16GB) and C2Q socket 775 vs socket 1366 processors and I didn't bother going with this. I hopped from C2D to Ivy Bridge computer when I upgraded. Later on I got donated two Sandy Bridge computers and they are slightly faster than the Nehalem. But the greatest difference when you go to Ivy Bridge both on socket 1155 if one overclock the memory or go for socket 2011 with 4 dimms.

Anyway, Nehalem is interesting to look at and to play with and I don't have space for one at this moment.

Cheers,

Great Northern aka Canada.

Reply 8 of 9, by The Serpent Rider

User metadata
Rank l33t++
Rank
l33t++

You need to set Command Rate to 2T for stable operation with 4 sticks.

I must be some kind of standard: the anonymous gangbanger of the 21st century.

Reply 9 of 9, by acl

User metadata
Rank Oldbie
Rank
Oldbie
The Serpent Rider wrote on 2023-01-25, 03:41:

You need to set Command Rate to 2T for stable operation with 4 sticks.

I don't know why, but I've never been able to set T1. With 2 or 4 sticks. The memory is probably not performant enough.

For the Corsair XMS, 2 sticks out of 4 are from a kit I bought in 2005 and I bought two others a few years ago to have four. And even in 2005 with a K8N-Neo2 platinum (Nforce3 ultra), i remember not being able to set T1. (But even without T1, i stayed with this kit anyway because it handled 240+ MHz )

"Hello, my friend. Stay awhile and listen..."
My collection (not up to date)