What can I use to stress test a nvidia chip?

[Mad_Professor]

I got a board that has the nvidia defect. Awhile back I did a HP provided board replacement and change out the passive north-bridge with an active copper one and assume everything was good. But the computer came back recently with the craptastic nvidia 630/610 chip non-boot.

Hit it with a heatgun, everything working again. The problem was the heatsink didn't pull good to the board using those shitty white push retain tabs. Now that I have changed them out with 3/4 #4 phillip round head machine screws with nuts and rubber washers, and the thing seems solid.

prime95 won't work since that's gear more toward ram and cpu stressing. I need something that will stress the onboard video/northbridge chipset?

I want to see if this thing is going to fail again, which it probably will. That way I can say "hey it this, I need to replace the board again with an aftermarket instead of an HP replacement, and do a wipe and reload."

I have the computer for 2 weeks, the client is going on a cruise. Good time to do long term stress test this week.

[mariushm]

Re: What can I use to stress test a nvidia chip?

Furmark?

http://www.ozone3d.net/benchmarks/fur/

This used to kill Radeon 4xxx series and equivalent nVidia cards when it came out so be careful. Now most new cards have power limits and drivers detect it and throttle down the cards.

Not sure if it works on integrated cards - it's OpenGL so it should...

[mockingbird]

Re: What can I use to stress test a nvidia chip?

Eventually, either the rigidity of the insufficient grade underfill or the high-lead to euctectic bonded bumps will either cause the chip to crack on the inside, or the bond to microfracture.

The best way to test it is to run rthdribl at your native resolution.

[Uranium-235]

Re: What can I use to stress test a nvidia chip?

piss on it

[Mad_Professor]

Re: What can I use to stress test a nvidia chip?

Well the thing won't run with the two matching 2GB sticks it originally came with. It will test fine in memtest86+ then when you load Windows, it randomly BSOD or stops at Black stop screen. Then when you run memtest86+ again it fails on test 5. I tested each stick individually in each dim slot too and it tested ok and even loads windows fine, but having both in causes issues.

I have tried two kingston sticks and that loaded windows fine and passed memtest86+. I also found in my collection of ddr2 sticks two matching 512MB sticks that match the same manufacturer and series as the two 2GB sticks, and even they work fine.

I don't know if it's just a failing VRM or nvidia chip can't handle 4GB or what.

I think it's safe to say this board is too flaky.

[shovenose]

Re: What can I use to stress test a nvidia chip?

Junk that mofo
the reballed motherboards are a bit unreliable and you don't know how long it's going to last, so IDK if it's even worth bothering to fix...

[brethin]

Re: What can I use to stress test a nvidia chip?

Just to recap from what I read after all the work on the board. 2 sticks of 512MB work and test fine in the board but 2 sticks of 2GB give you troubles.
KISS comes to mind here sounds like its a memory issue and not a board issue per say. I understand that the board may have had the 2 sticks of 2GB that came with it and worked orginaly but that doesn't mean they are still good or are working 100%.
Test the parts you are using for testing in a known good setup before you use them to test a problem part. And remember if they work in one part it doesnt mean they will work in all parts.

[brethin]

Re: What can I use to stress test a nvidia chip?

Ignore the lazy feks telling you to giveup/junk it. Thats what china wants you to do is throw it away and buy another one good for a year so they can sell more shit.

[brethin]

Re: What can I use to stress test a nvidia chip?

Junk that mofo
the reballed motherboards are a bit unreliable and you don't know how long it's going to last, so IDK if it's even worth bothering to fix...

It would help if you knew what you were talking about.

#1 Reflowing is a bit unreliable.

#2 Reballing is normally better than new ESP if leaded solder balls are used.

[shovenose]

Re: What can I use to stress test a nvidia chip?

I was under the impression reballing and reflowing were the same thing?

[Mad_Professor]

Re: What can I use to stress test a nvidia chip?

Just to recap from what I read after all the work on the board. 2 sticks of 512MB work and test fine in the board but 2 sticks of 2GB give you troubles.
KISS comes to mind here sounds like its a memory issue and not a board issue per say. I understand that the board may have had the 2 sticks of 2GB that came with it and worked orginaly but that doesn't mean they are still good or are working 100%.
Test the parts you are using for testing in a known good setup before you use them to test a problem part. And remember if they work in one part it doesnt mean they will work in all parts.

two 256MB works ok, two 512MB works ok, two 2GB sticks does not.

I test the two 2GB in my rig and it tested ok then I install windows 7 on a 10GB ide drive, that went fine with no issues. I took my corsiar memory two 2Gb dual channel kit from my rig stuck it in this craptastic hp nvidia chip board, and memtest passed, boot windows and got random BSODs, failed memtest after that.

This is also a board I recapped since I recapped the failed board that this board replaced figure might as well, since it's not cover by warranty and paid 89$ for it.


Edit: Anyone know what year/series the nvidia defect affects? I'm looking at a replacement board that has nvidia geforce 9300 northbridge/onboard video.
This board is done for.

Ignore the lazy feks telling you to giveup/junk it. Thats what china wants you to do is throw it away and buy another one good for a year so they can sell more shit.

I would, but I got to keep the clients well being in mind. If I send this computer back to him only to fail, it would make me look bad or I don't know what I'm doing. The fact that it will not run 4GB of memory is suggesting something else failed on the chip or board.

[mockingbird]

Re: What can I use to stress test a nvidia chip?

Edit: Anyone know what year/series the nvidia defect affects? I'm looking at a replacement board that has nvidia geforce 9300 northbridge/onboard video.

ALL nVidia chipsets after Geforce 5/FX (But not including GeForce 5/FX), and ALL GPUs after Geforce FX up until Fermi...

And according to this article, also the Geforce GTX480.

In case you were wondering, nVidia doesn't make chipsets anymore... Also, they nearly put HP's PC division out of commission.

[Mad_Professor]

Re: What can I use to stress test a nvidia chip?

ALL nVidia chipsets after Geforce 5/FX (But not including GeForce 5/FX), and ALL GPUs after Geforce FX up until Fermi...

And according to this article, also the Geforce GTX480.

In case you were wondering, nVidia doesn't make chipsets anymore... Also, they nearly put HP's PC division out of commission.

Thanks for that article, it explains why my p5n-d northbridge and gtx260 runs hot.

so I guess an intel board is the way to go.

[mockingbird]

Re: What can I use to stress test a nvidia chip?

No, GTX260 has eutectic bumpers and eutectic pads, as well as the more appropriate underfill, the GTX 2xx cards become defective only because of RoHS soldering, not because of the actual chip.

Reflowing GTX 2xx cards should fix them. But yea, the P5N-D might stop working at some point.

Hypothetically, reflowing a motherboard with a defective nVidia chipset may actually cause the bump inside the chip to re-bond with the pad (Don't quote me on this), but once the die is actually destroyed because of too much pressure from the rigid underfill, there is nothing to do.

[b700029]

Re: What can I use to stress test a nvidia chip?

two 256MB works ok, two 512MB works ok, two 2GB sticks does not.

How long did you run memtest for and what type of errors were they? (Repeatable, random, # of bits in error?)

[Mad_Professor]

Re: What can I use to stress test a nvidia chip?

How long did you run memtest for and what type of errors were they? (Repeatable, random, # of bits in error?)

The first test it would pass fine no problems let it run for 3 more passes which took like 2 hours. But the minute you try to load windows it instantly does a random BSOD. Like NTFS.sys, or IRQL_NOT EQUAL or something, memory management error. I get this one commonly SYSTEM SERVICE EXCEPTION, then if you reboot and reload memtest it fails half way through test 5.

It kept saying 0.1-2048MB failing address random numbers.

Something driver side triggers a hardware failure, it use to run both sticks fine 3 months ago when I gave it back to the client.

FUCK IT! I ordered a new board from newegg along with a new psu to get rid of this propitiatory miniature 20 pin atx PSU. I do a data backups of the driver and prepare my phone for a call to M$ to activate Vishit.