Video card not stable at stock clocks?

T

TopGun

Gawd
Joined
Jan 22, 2005
Messages
766
I've been having issues with my card, trying to find out what was causing it for the last couple of weeks. Finally decided that I should underclock it to 500/700 from 550/800.

A half an hour before I underclocked I tried playing cod4 but it froze with "DirectX Has Encountered An Unrecoverable Error". Once I underclocked it I played through two maps of a 48 person deathmatch server.

I'm guessing bad graphics memory maybe?
Should I just go straight for an rma or is there possibly another cause?

Powersupply is an Antec Earthwatts 380w with 12v @ 27a for all you wonderers, Powersupply calculator says that I only need 180w.

Thanks
-TopGun
 
I'm quite surprised that the PSU calculator says 180W. Maybe I'm not familiar enough with the GS type (more the GT) but I would have thought that was way too low. What does your video card manual say about recommended PSU, my 8800GT said min 400W, but I had problems even with that wattage and had to upgrade to more to be stable.

Make sure your GPU temps are Ok, If you have control over GPU fan speed, try 100% and see if that helps.

I found a good video memory tester:
http://mikelab.kiev.ua/index_en.php
 
Have it as spectator on cod4 and played a map at stock clocks and it didn't freeze or crash.
I'm really frustrated with this, completely random and out of the blue crashing / freezing in games. Temps are fine, Psu should be fine...idk but it's frustrating.
 
How old is install of windows? Sounds like it might be driver issue, could try new install on extra HD if you got spare with latest driver and see if that works.
 
The install is fresh, I've also tried with different installs & drivers.
 
"Temps are fine"

What temps ? you using riviatuner or something to give you GPU temps and if so what are they ?

I too am of the opinion the power supply is marginal (and I am not an advocate of people running out and buying big supplies ).

Fact is the manufacturers have gotten very good at binning GPU chips and especially the new ones like the G92 where the "better" chips all go to the factory OCed cards and very few, if any, chips used in the mainline cards have much overclocking or thermal headroom at all. My factory OCed card has very little "extra" headroom when playing games even after water cooling the core and installing ram sinks with a PCI slot fan blowing directly on the entire card. I can get very high clocks in tests like ATITool etc, but a game will crash it quickly with only a little bump over factory OC. The memory system on all 8800s is regarded as fairly crappy by all due to mediocre memory chips used and the supporting memory interface circuitry is fairly el-cheapo too.

Without more detailed information i would suggest proceeding in the following trouble shooting manner.

1) Determine exactly what the GPU temps are under full load by using rivatuner or some other utility to record the GPU temps under load to a log file. It would be great if it would crash while logging. Then take the side panel off, or get more fans /whatever to reduce the GPU temp by at least 5C under the exact same game/test conditions and see if it still crashes. If it does not crash, well, there you are, its heat.

If it does still crash

2) Get a Corsair VX550 or similar and say a little prayer that that does the trick. You are going to need to move up in a power supply sooner or later and unless you go total top of the line multi graphics card setup etc. on the next big upgrade a 500-600 watt good quality unit bought now should still be usable even if you go quad + 1 video card later.

3) It could just be your game and drivers although you have done about all you can for that. So back to item 1 we go.

Good luck.
 
Computer seems fine "for now", Maybe reverting back to the first bios for my board helped :confused:
I'll probably end up getting the VX550 once I get some extra money.

By the way the card would freeze in the 50's, It's currently idling at 31*
 
I was just playing cod4 when my game froze and a few seconds later bsod, didn't have time to read what it said.

I realise that a lot of this may point to my power supply but how can
http://www.ocforums.com/showthread.php?t=512807 that person run all that stuff on the same one? Isn't his system more demanding?

The event viewer says:
Code: 
Error code 1000007e, parameter1 c0000005, parameter2 bfadf12d, parameter3 b3732c48, parameter4 b3732944.

I really want to figure this out, games run great...when they run without freezing :(
 
Yea i think its not enough power in your PSU try getting a 550W one
 
I know a lot of people are always dead set on blaming the power supply.
But how can I run the hdr demo and prime 95 at the same time and not bsod instantly
if it's power issues? There's no way that cod4 puts a greater demand on the power supply.
 
Looped 3dmark06 for 7hrs overnight and it was still going when I just checked in.
I'm starting to think it's punkbuster, which friggin sucks because a ton of games use it :(
 
Whenever DirectX and my game crashes after I close the error and the game it takes a while for the icons on my desktop to load. The system seems sluggish yet I had music playing and it was still playing fine throughout all of this.

Do I rma it or what? I really am not in the position to waste money
on a powersupply that I probably wouldn't need.
 
You have more than enough power. that rig if at stock speeds full tilt i doubt your pushing 200w. My oc'd to all hell sli'd 4 gig system with 2 hd's doesn't pull more than 240 at full tilt.

That measuring from the wall with a power meter. I was having random game crashes and bsod's with my 7900 gt and the issue was video ram. Just search around for a good video ram tester and see if it artifacts/ throws error codes. It should be obvious and quick if it is a problem. My card also wouldn't run stable at stock speeds or underclocked but would run oc'd for a little bit then bsod.

If cpu loads run fine, more than likely its not cpu related. If ram passes memtest, then it's either software or video card.
 
OP, or anyone with more info on this, OP, you said reverting to older BIOS fixed the issue?

I've got a SN25P Shuttle (yea I know 350W PSU)

Playing Single Player COD4, no problem, I've had 1 crash there(running max settings, no AA @ 1680x1050 on HD3870OC, once I fire up MultiPlayer, all hell breaks loose. I've been running 1280x720 on multi just go have a bit of an edge on getting smooth FPS, so I'm thinking I'm stressing the system less by doing that(v-sync on). I've updated Punkbuster to latest client, 2.04? No help there. COD4 crashed with ver1.0, 1.4 and 1.5, nothing fixed there.

I'm thinking PSU, I might actually try running a second PSU just to the card and see what happens ... I might try flashing BIOS, any other ideas?
 
You must log in or register to reply here.
Back
Top