GPU voltage limit, "force constant voltage" confusion. I could use some help.

[bmanzzs]

Long story short, I have two R9 280x's in crossfire. Both cards work just fine on their own individually, but overheat when installed in crossfire.

In Crysis 3, cards core reaches 80-85C, and VRM temps hit 90-95C after about half an hour at max settings. Eventually the computer will either: 1. driver crash 2. BSOD. 3 freeze

 

Now, this wouldn't be TOO much of an issue because I can undervolt the core from the original 1.2v to about 1.155v by enabling "force constant voltage" in AB. this way I get around max 75C core and 85C vrm. awesome.

HOWEVER, I don't like doing this because it means my GPUs will be constantly sucking up power and unable to switch into 2d mode, meaning my GPU's will also idle at around 20C higher than what they normally would when at a low-power state.

 

So my question is:
Is it possible to "limit" the amount of voltage to my cards, so they never go above 1.155V? The description in AB says it should do just this, but my VDDC readings always jump from their low power voltage state (0.850v), to (1.2V), seemingly completely ignoring my voltage setting and eventually crashing. Because of this, I have to use "force constant voltage".

Please, brothers. HALP?

specs:
i5 2500k

asrock extreme3 gen3

8gb g-skill

2x r9 280x

evga 1000W G2 Gold

[Fulgrim]

You can just lower your core voltage, but you'd need to lower your clocks too.

[bmanzzs]

That's what I mean, I can't. I set the voltage in AB to 1.155v, but the cards use the default voltages (1.2v) while in 3d mode unless I use force constant voltage.

[-rascal-]

Your motherboard's PCI-E slots are spaced out enough so there shouldn't be any overheating issues...

I need you to answer a few questions before I can continue

• What computer case are you using?
• What is the fan / cooling configuration?
• What is the make and model of your R9-280X's?
• What version drivers are you using?

I'm running two HD 7970's (aka R9-280X) in my system, and I am not getting anywhere close to your 80*C+ temperatures.

[bmanzzs]

Thank you for your response. It is very unfortunate that these cards reach those temperatures, but they do. In fact after about 30 minutes of Metro 2033 Redux my GPU1 VRM temps hit 100C, and the GPU2 VRM temps were hovering around 90.

1. Cooler Master CM 690II Steel. love this case
2. Stock fans, nothing custom except AS5 thermal paste (which actually did help temps about 2-3C)

3. Both cards are XFX R9 280X TDFD

4. I'm currently using 13.7, but the exact same things were happening in 13.4 and 13.1

The cards appear to have enough room, but while in crossfire they both come up 10-15C higher. I also have a case out take fan directly next to them on the door to dissipate some heat, but that doesn't seem to help that much.
I'm gonna assume your cards aren't XFX, because apparently XFX is infamous for incredibly poor VRM cooling... hence my 100+C temps.

[-rascal-]

@bmanzzs (damn quote button is broken again...)

 

Yes, you are correct, XFX's "Ghost Thermal" technology cooling has been an known issue.

It was a problem when they first used it on the HD 7000 series...but they continued with it on the R-200 cards...

Frick, my XFX HD 5850 from back in the day ran amazingly cool.

 

I don't know which particular version of the CM 690 II you have, but I assume the one with the option to mount 120mm / 140mm fans to the side panel. Have you tried setting up the fans on the side panel as intake, blowing cool air to the graphics cards?

 

If you can, try the following:

• See if mounting a 120mm intake fan on the HDD cage spot helps in any way.
• Add one or two 120mm fans to the bottom of the case as intake, to feed cool air to the GPUs.

Basically, ALL the cool fresh air onto the GPUs.

[bmanzzs]

Thank you again.

Unfortunately I pretty much have tried that fan configuration already. Using case fans by themselves isn't sufficient enough to cool them, I need a desk fan blowing straight into my open case for VRM temps to stay at 90. But even then, I still get crashes and BSODs. Frustrating as hell. 

here's a screenshot of results after running the Valley benchmark for 10 minutes.

[pdv789]

Try replacing he thermal compound and pads on the card. I often flip the HD 7000 series after replacing those and getting 70c oced at max temp.

[bmanzzs]

Already tried replacing the thermal paste as stated above. It helped a tiny bit with temperatures but did not solve the problem. Not sure if replacing the pads would help much.

Thanks for the response though.

[bmanzzs]

Good news(ish?)!

It looks like I was wrong about something. I decided to remove the first card entirely and place the second card in the top PCI-e slot. after about 15 minutes in Metro Redux, I crashed. Core was at 75C and VRM maxed at 86C when it froze. So I tried the card on my other PC, and guess what? Same thing, freezing in Valley with okay temps.

So this leads me to believe that the second card has been the culprit this entire time... XFX sent me a replacement fan because the previous fan failed on this particular card. Seems this card is toast! Which I guess is a good thing?   I'm going to send XFX an e-mail to RMA this thing. If they can't recreate the multiple freezings and BSODs I've encountered, then I know they are lying!