Sapphire Radeon HD 4850 giving me grief

[gregsteg]

I live in Canada and I do not know whether I should pay the $50 fee (and postage) to send in my Sapphire ATI Radeon HD4850 1G DDR3 dual dvi video card for RMA (paid $210 March 2009, can they really charge me? I still have a warranty).

So, I tried to build my own computer for the first time almost a year ago (my warranties are coming up soon). I have always had trouble with what I assume is the video card, I thought it was overheating, because:

-The computer just restarts after playing 3D games for a while (recently I reformatted with Windows 7 and I get a blue screen for a few seconds before the computer restarts)
-It usually only cuts out when I am playing 3D games, or other 3D rendering programs
-I can increase the amount of play time by opening up my case and turning the fans to max but it still eventually cuts out
-I can extend play time by turning down in-game graphics settings
-3D shooters cut our early, Company of Heroes last longer if I use the standard view, whereas Civ4 on lower resolution goes for hours.
-Most recently I have been using this weekends free Shattered Horizon which will give me a blue screen inside 10 minutes

I have tried any number of driver combinations:

All windows drivers still lead to problems
-Windows 7 64-bit, CCC and driver versions 9.4 to 10.1
-Windows XP, the original drivers from the manufacturer's CD, to 10.1
-After my reformat I get error messages during the "drag and drop transcoding" install but the drivers and CCC still install

-Linux ATI drivers seem fine but I don't have any 3D intensive programs on Ubuntu (.avi movies flicker)

I recently replaced the thermal paste which does keep the temperature down but the card still cuts out.
During 'Shattered Horizon' crashes the, last temp I saw on the CCC overdrive read-out was 76degrees C so I think I can rule out overheating.

Some of the blue screen errors (scribbled down in the few seconds before reboot, can I collect the full report somehow?):
Either something about a "Page fault" or "driver IRQ not less than"
Files mentioned:
srt.....sys
dxgmms1_sys
and others.


Has anyone seen anything similar to this before?
Can I be confident this is my video card giving me trouble? Is there a way to rule out all my other hardware?
Should I bother with the RMA? Any recommendations for a cheap temporary replacement while I wait for the RMA?
I thought this was a good card, I do not want to sink significant money into another card that would drop performance (if I had any to start with)
Should I buy nVidia next time? I am limited by my power supply (500W and only one power connection for PCI-E card)


Screen Resolution: 1680 x 1050

Video Card read-out from cpuid:
Graphics
-------------------------------------------------------------------------

Number of adapters 1

Graphic APIs
-------------------------------------------------------------------------

API ATI OD5
API ATI I/O

Display Adapters
-------------------------------------------------------------------------

Display adapter 0
Manuf. API index 0
Display name \\.\DISPLAY1
Name Radeon HD 4850
Codename RV770
Technology 55 nm
Memory size 1024 MB
Memory bus width 256 bits
GPU ref clock 100000
PCI device bus 5 (0x5), device 0 (0x0), function 0 (0x0)
Vendor ID 0x1002 (0x174B)
Model ID 0x9442 (0xE810)
Performance Level 0
Core clock 500.0 MHz
Memory clock 993.0 MHz


Software
-------------------------------------------------------------------------

Windows Version Microsoft Windows 7 (6.1) Ultimate Edition (Build 7600)
DirectX Version 11.0

System:

Intel(R) Core(TM)2 Duo CPU E8400 @ 3.00GHz
2x2Gb ocz pc2 6400 DDR2 @ 400Mhz 5-5-5-15
Mainboard Model P45 Neo3 (MS-7514) (0x0000043D - 0x1272BD20)
Earthwatts 500W max


Thank you for any input.

 

[cybot_x1024]

what are your gpu temps? (and cpu temps too?)
you replaced the thermal paste on the gpu or cpu?
4800 series have a very high thermal threshold for heat therefore you could be easily overlooking a cpu cooling problem. Is the cpu cooler secured on the socket firmly? And what thermal grease are you using? Ive had such a previous ordeal and it turned out that it was the cpu cooler that was loose.

 

[gregsteg]

I replaced the thermal paste on the GPU with Arctic MX-3 thermal compound.

The CPU does not seem to heat up much when the crashes happen. The CPU temperature did not exceed 40 degrees C during the last in-game crash I had. I have an after market cooler and no OC for the CPU.

Some of these crashes cause the system to reboot several times in a row before everything seems to get worked out. I had my bios reset once as well, I'm starting to think there is a more serious underlying problem. I only seem get system crash/reboots from 3D gaming though (and folding@home).

 

[tomvertommen]

I don't see why you are so convinced it's the gpu. It's not overheating, you cleared that out. I don't know what else it could be...Maybe you could try it with an old gpu and see if it keeps on crashing ?

 

[gregsteg]

I've taken to running as many stress tests as I can find recommended in various forums.

I ran Prime95 which successfully crashed the system but not before giving me this:

[Sat Feb 20 03:08:32 2010]
FATAL ERROR: Rounding was 0.5, expected less than 0.4
Hardware failure detected, consult stress.txt file.
FATAL ERROR: Rounding was 0.5, expected less than 0.4
Hardware failure detected, consult stress.txt file.

Any ideas?

 

[gregsteg]

The rounding an error in my computers effort to find the next digit in pi?

 

[kikireeki]

messing with the GPU thermal paste was not a wise decision as it will void your warranty! the PSU is fine but if it is faulty it might cause such crashes so Try another PSU.

 

[gregsteg]

Thanks for all the advice, I have indeed started to look into other causes than the video card. I ran some more stress tests and found that my computer cannot calculate Pi accurately when I use a RAM intensive setup. After my computer crashes it fails the windows memory diagnostic which it normally passes. I am going to run with this and buy some different RAM tomorrow.

If it is the RAM, is the RAM damaged or defective or incompatible? Is there a way to know for sure what will be compatible with the mobo? I thought I had figured this out when I first bought the stuff last year.

I unfortunately do not have spare parts lying around which I can swap in and out to pinpoint the problems.

 

[kokin]

messing with the GPU thermal paste was not a wise decision as it will void your warranty! the PSU is fine but if it is faulty it might cause such crashes so Try another PSU.

Replacing the TIM on CPUs/GPUs/chipsets(north/southbridges) does not void the warranty. If you cause physical damage because of changing the TIM, THAT will definitely void your warranty, but otherwise it's relatively simple and straightforward. I just did it twice to my GPU and my northbridge/southbridge.

@OP, try to see if your RAM is running at it's rated voltage in your BIOS, if it is and is still failing the tests, then I would look into new RAM.

 

[gregsteg]

Should I start again in a new forum for a memory issue?

I seem to consistently run into trouble when I run the Window's 7 memory diagnostic tool, it leads to a crash/reboot and I never get the see the results. It also often leads to a string of reboots and resets the BIOS.

I tried switching the RAM to it's rated voltage and tried recommended timings but that also led to the same problems and I had to manual jump/reset the BIOS once when I set up bad timings.

Mainboard Model P45 Neo3 (MS-7514)
BIOS version 1.A 9/04/2009
Ram, 2x2gb OCZ2F10664GK
running as PC2 6400 (400Mhz) 5-5-5-15

Tried 6-7-7-15 and 6-7-7-20 @ 2V, and 5-5-5-15 at 2V
I tried to play with the FSB/DRAM ratios to get 1066Mhz as well.

I think I will buy some proper PC2-6400 ram because the mobo will default to that anyway.

 

[gregsteg]

Looking at this:
ADATA DDR2 800 240pin 4GB Kit (2x2GB) Unbuffered-DIMM Non-ECC Memory (w/ Heatsink)

Apparently I need non-ECC and unbuffered from what the web tells me.

 

[gregsteg]

The new RAM fails, I re-posted in the memory forums:

http://www.tomshardware.co.uk/forum/270666-12-computer-computer-accurately-fails-memtest

 

[gregsteg]

Best answer selected by gregsteg.

 

[Nerraw99]

I'm getting a similar problem with an HD 4850. If I up the fan speed or lower the clock on my GPU, it goes away. I guess I need a new heatsink.

It just reboots my computer 5-10 minutes into an intensive game. Lowering the resolution or graphics settings seems to do the trick too. I'm sure it's the vid card.

 

[gregsteg]

I'm getting a similar problem with an HD 4850. If I up the fan speed or lower the clock on my GPU, it goes away. I guess I need a new heatsink.

It just reboots my computer 5-10 minutes into an intensive game. Lowering the resolution or graphics settings seems to do the trick too. I'm sure it's the vid card.

My problem turned out to be my motherboard. It couldn't use the RAM properly (not sure how). I don't know if opening my case up and turning down the graphics settings could have helped the motherboard issue but it seemed to at the time.

I did redo my thermal paste on the HD4850 and it brought the temperature down significantly (>20 degrees).