PDA

View Full Version : Video card crashes on game or benchmark start up



Cowtown_Raider
06-20-2015, 10:57 AM
Hi guys,
Anyone here good at trouble-shooting video cards?

I picked up a used R9 290X to set up crossfire. It's basically the identical card to what I'm currently running. It came with a water block on it instead of the stock fan cooler. I don't have hoses, pump or rad for the water cooling so I removed the block and installed the stock cooler back on. I noticed that the fan clip was totally crushed, but I was able to bend the wires up and hook up the fan. I installed the card in my system and hooked up the power. Booted up fine and Catalyst recognized the card and enabled crossfire. All seemed fine until I started the Heaven benchmark to make sure crossfire was enabled. As soon as the benchmark started, the screen goes black and the fan goes to 100%, which is really damn loud. PC reboots and fan settles down after a Windows loads. I thought maybe the power was insufficient so I put the original power connectors on the new card...same thing. I put just the one card in my PC on the original slot and connectors and same thing. Finally, I pulled the card out and put my original card back in, system works fine again, games work fine.

Definitely a problem with the card. I'm thinking of pulling the stock cooler off again, cleaning the GPU heatsink because it was a mess with thermal compound and trying again. He put way too much on. I might put the water block back on it and see if it'll even start the Heaven benchmark. It'll heat up pretty fast without a fan but that might rule out a short with the fan connector or something.

It's entirely possible that I bought a brick too so I could just be SOL. The guy who sold me the card won't help me try to get the card working even though he has all the water cooling gear installed on his PC.

I welcome any suggestions from you seasoned gamers. Thanks

UndrgroundRider
06-20-2015, 11:24 AM
I'm confused. You say you were the one to put the stock cooler back on, but "he" was the one who put too much thermal compound on. Does that mean you didn't reapply new compound?

Cowtown_Raider
06-20-2015, 12:21 PM
Yeah, I didn't reapply compound, which in hindsight wasn't a good idea. There was excess gooped around the heatsink so I removed the excess, leaving a sufficient amount on the heatsink and re-installed the stock cooler.

Still, if it was an overheating issue, the benchmark should at least run briefly to build up heat, and then fail, but it fails instantly, while GPU temps are near idle.

I plan to pick up some new Arctic Silver 5, clean everything off, and try it again, but I just have a feeling it's not an overheating issue.

revelations
06-20-2015, 12:40 PM
Isolate.

Run the new card by itself, benchmark etc.

Use a good card and then test all the slots individually, (benchmark).

Once the slots are known good and you have a good GPU card, add the other. If if fails then I would change out the PSU - I assume you have a at least an 800W or greater unit already??

UndrgroundRider
06-20-2015, 12:44 PM
Originally posted by Cowtown_Raider
Yeah, I didn't reapply compound, which in hindsight wasn't a good idea.

Yea, that's a huge no-no especially when switching heat sinks. It could be to blame or it might be totally unrelated. The big problem with reusing old compound is that it hardens over time and instead of filling in surface imperfections it actually prevents a flat mating of the surfaces and creates air gaps. That's a really big issue because you can get parts of the die heating up while the rest is cool. This will trick the thermal protection sensors and create a great deal of internal stress on the die, which is really fragile to begin with. If this was the case you could've damaged the die on the first run, and now it crashes instantly. Who knows, GPUs do funny things when damaged.

Again, I'm not entirely sure if that's the issue, but it's a definite possibility. I wouldn't recommend running the card until you clean off the old compound and apply some new arctic silver.

Another thing I would look into is the power supply. You say it's a fairly instant crash, which could be caused by the power draw ramping up under load. I would remove your original card and try to get the new card to work on its own. That should rule out the power issues and any crossfire issues.

Cowtown_Raider
06-20-2015, 12:58 PM
I haven't run the card since. I'm going to disassemble it and clean off all thermal compound, then reapply the correct amounts.

I did run the card isolated in the same slot and cables as the good card and it failed instantly.

I have a Corsair 1200 watt PSU so not a problem with power. Hopefully I haven't damaged the die, shit.

Thanks for your input, fellas.

Cowtown_Raider
06-21-2015, 01:24 PM
I cleaned everything off and applied AS5. Good news is the card handled the Heaven benchmark, but it was running crazy high temps. I pulled the card and re-applied AS5. Second time, temps were a bit better, but still about 10 degrees higher than my other 290X. Using MSI Afterburner's active fan control, I can get through a full benchmark without crashing now. Temps at full load are around 88 degrees.

Seems to be a lot of different advice on how to apply AS5. Pea sized dab in the middle, a line, and X, spreading with your finger, etc. I went with a line down the middle about the thickness of a grain of rice. What have you guys found works best?

AE92_TreunoSC
06-21-2015, 02:32 PM
I always used the pea size and spread out with an old business card. Really its hard to screw up, some guys just spread it on in a line and smear the cooler on.

Are you running the correct video card rails from your psu? 290x's use quite a bit of draw and if on the wrong circuit can cause a crash.

Cowtown_Raider
06-21-2015, 04:17 PM
I wasn't able to find an 8-pin and 6-pin male to male connector to hook up to the PSU so I used dual molex connectors for the 6 and 8-pin power. I've been playing Witcher 3 for a few hours and monitoring temps. Existing card runs about 81, new card runs about 87. The fans are running about 50-60%, which is pretty loud. Damn, this game looks good with everything on Ultra.

If the rail that the card is on has insufficient power, would that cause it to overheat? It's not crashing anymore.

I don't think my case has the best cooling too. Might need to figure out more airflow.

AE92_TreunoSC
06-21-2015, 04:28 PM
Molex adapters are less than ideal, especially for high draw cards. You have a decent power supply and it should have 2 6+2 connectors. Poor connections inside the adapters can cause voltage drops which will cause crashes etc.

Personally I wouldn't run xfire or sli without proper power feed. You'll just have more problems later on.

UndrgroundRider
06-21-2015, 07:15 PM
Originally posted by Cowtown_Raider
I cleaned everything off and applied AS5. Good news is the card handled the Heaven benchmark, but it was running crazy high temps. I pulled the card and re-applied AS5. Second time, temps were a bit better, but still about 10 degrees higher than my other 290X. Using MSI Afterburner's active fan control, I can get through a full benchmark without crashing now. Temps at full load are around 88 degrees.

Seems to be a lot of different advice on how to apply AS5. Pea sized dab in the middle, a line, and X, spreading with your finger, etc. I went with a line down the middle about the thickness of a grain of rice. What have you guys found works best?

Good to hear it was a heat issue and sounds like no permanent damage. Pretty lucky there. As far as spreading the thermal compound, you want to put a dab in the centre of the proc and spread it with a flat edge such as a business card. You want thin even coverage of the whole chip surface. Don't worry about getting it perfect in one pass, just work it until it's even and fully covered.

When you attach the heat sink try and put in on straight down with no twisting or side to side movement. Cross-torque the screws if it's that kind of mount.

In terms of temp. discrepancies between cards, that's totally normal. I ran a huge bitcoin farm for a long time, even cards of the exact same make and model can run differently. There's a whole ton of reasons for that, between manufacturing defects, different heat sinks and different clock speeds (even among the same line of card).