[8RGM2i] Unstable in 3d?

[Itachi]

Ok, this is the first time I start a thread here... I must say I like this forum for so far.

Ontopic:
My situation:
For always I've wanted a small PC for HT PC and LANs, it's much easier than a bigtower. So, I bought an EPoX 8rgm2i and an Athlon XP 2600+. When I plugged it all in, it worked! Even the overclock functions are great on this bord. But, when I started to test things like 3dMark2001 I encountered a few major problems.

The problems:
When I enter a 3d game, or program the system freezes or gives a BSOD. The BSOD says nv4_disp.dll is the cause of this problem. For that I've tried to install older and newer Detonator (and Forceware for the newer version) drivers. Still crashes. :/
Ok, to test the drivers I inserted a GeForce2MX GPU, ripped out my brother's computer. It worked! I started Need For Speed: Underground and I could even drive! Normally, with the intergrated 4MX GPU, it would crash as soon as I get to the menu's. This is with the same Detonator driver!

And another problem. When I test my computer with PCMark, it crashes (PCmark) on test 4; audio comversion. I tried with PCmark 2004 and 2002, it both crashes at the same point.

So it looks like it's a GPU problem...

I checked for IRQ conflicts; the GPU doesn't share its IRQ.
I tried to underclock(!) the CPU to 100FSBx11,5.
I've got all the Windows updates.
I tried to update the C-MEDIA AC97 driver.
I tried to install the latest nForce driver (3.13?)
I tried to use less hardware, so I could check the PSU. No difference. The voltages are O.K.
I tried to stress the CPU in Windows, so I could check if it is a temperatur problem. It's rock stable in 2D.
I tried to run some Direct3d tests... strangely enough it ran fine.

Specifications:
XP 2600+, 333FSB (Barton)
EPoX 8rgm2i
2x Corsair value 256MB PC3200 cas 2,5 memory (dual channel)
AOpen PSU
MaXtor 40GB 5400RPM

I thank you in advance.

[flytek]

try using one stick of memory or run in single channel mode. try cooling the northbride. try different memory. the onboard graphics adds plenty heat to the northbridge (especially at 333fsb) and is notoriously fussy about memory

[Itachi]

Hmm, the 8RGM2i is officially designed to work on 166FSB (and 133)... is it really impossible to run stable with the passive cooled northbirdge?
I'm afraid that if I'm going to mod a fan on the heatsink my supplier isn't going to like it, if you catch my drift.

I've checked my memory with Memtest86, after 4 passes without a single error I had enough.

I'll try to run with one stick this evening. Thanks for your reply.

[flytek]

i've never had heat issues at 166fsb but it certainly couldn't hurt. did all your benchies work with the agp card inserted? the memory incompatability usually only shows up under 3d.

[Itachi]

I am at the moment not able to run the PCMark benchmark with the AGP card. Tommorow I'm able to use a low-profile GF4MX card to test it. I'll post the results tomorrow.

Tomorrow I'll be able to test with one bar of 256MB Kingston valueram too.

[flytek]

hmmmm. most likely memory then. with the agp inserted dual channel is not nearly as important. have you tried to manually lower memory timings to 7-3-3-2.5 at 100%of fsb?(ie: underclock it to ddr333) try the memory in the slots closest to the cpu or try some other memory. also try just one stick of memory. try the fsb jumper in different positions or off altogether. try cpu interface aggressive in the bios. maybe even a new bios is available.

[Itachi]

flytek... I love you. :P

I just inserted some of Kingston's valueram (256MB CAS3 PC3200) and I ran it at the exact same speed as the Corsair. That is 170FSB/320DDR CAS 2-2-3-7.

3dMark ran fluently.

What is it with the GPU that it's so obstinate about the memory? There's nothing wrong with Corsair, that's for sure.

Does someone have a list of compatible memory with this board? Should I just buy some Kingston memory?

[flytek]

glad to help. there is a compatability list somewhere. maybe somebody will drop by with a link.

[Áedán]

Originally Posted by Itachi

What is it with the GPU that it's so obstinate about the memory?

It's the integration that causes the problems. With non-integrated graphics, most of the RAM access is done by the CPU. When you move to an integrated solution, both the CPU and the graphics access the RAM. That's a lot of extra loading on the RAM, significantly above what the CPU could manage on it's own!

Hence, if the RAM has problems or is close to the edge of stability with just the CPU alone, throwing integrated graphics in there will easily push it over the edge.

[EPoX Tech]

http://www.nvidia.com/content/nForce...patibility.asp

A list that exists for a reason!

[kfl]

MB: Epox 8RGM2i
CPU: Duron 750 jumper ok, FSB 7,5/100, 1,75V; Temp 44C
RAM:1x PQI 256MB DDR 400MHz; optimal in BIOS
PSU: 300w JNC (new, tested ok)
HDD:WD800BB new
CD: ok tested
OS: W2k pro sp4 clear install
BIOS: default

The Problem: self restart in Win
EST

[EPoX Tech]

Please check with http://www.nvidia.com/content/nForce...patibility.asp and note support of IGP with DDR333 memories

[hakrist]

I have the (almost) same system and am to experiencing major problems.

With only one memory dimm I experience major artifacts on the graphics, with 2 inserted no artifacts but unstability when running 3dmark.

System:
XP 2500+, 333FSB (Barton)
EPoX 8rgm2i
2x Corsair value 256MB PC2700 cas 2,5 memory (dual channel)
460W PSU (checked ok)

I returned the mobo for a replacement MSI nforce-2 mobo. Same problems!!!

Maybe nforce2-IGP doesn't like corsair memory? I also tried a pair of twinmos winbond PC3200 memory dimms. With these the system wouldn't post at all.

Damn annoying.

[kfl]

Originally Posted by EPoX Tech

Please check with http://www.nvidia.com/content/nForce...patibility.asp and note support of IGP with DDR333 memories

Support
?????????????????
http://www.nvidia.com/content/nForce...ufacturer=EPOX
?Adata
?Apacer
?Corsair
?Hynix
?Kingston
?PQI
?Samsung
?Spectek
?TwinMOS
?Viking

[EPoX Tech]

Just check the generic list from the link provided by myself, not the link which you have suggested.