Safe to overclock 7950?

[VirtualLarry]

I have an HIS 7950 3GB card, with a single fan in the middle of the card. Nothing fancy as far as the cooling goes. The stock clocks for this card are 800 / 1250.

Just wondering what would be a safe overclock?

I just tried using CCC Overdrive, and boosted the core clock from 800 to 900. I've seen cards with factory OCs of 1000 or 1050 core clock, so I didn't think that was too excessive.

I do distributed computing on the card, and on my G3258 CPU, so I'm putting a really hard load on the card. I don't want to shorten the lifespan of my card in any significant way.

Temps are showing as 70C for the GPU in GPU-Z 0.8.2, with 99% GPU load, and fan speed is at 56% 3440 RPM. Before OCing the GPU by 100Mhz, temp was 67C, fanspeed was 52-54%, 3000RPM.

 

[schmuckley]

yikes
Too hot,IMHO.

 

[Shmee]

The OC boost shouldnt hurt it, though heat over time can. Have you tried decreasing the vcore to see if it is still stable at lower?

 

[Essence_of_War]

I don't think 70 is that hot...and you haven't tweaked the voltage at all, right?

What gpu z say about the vrm temps?

 

[Erenhardt]

Ice-cold for a gpu.
anything up to high 80's is ok.
I run my 7950 at 1200MHz@1.3V or something close to this. Custom fan profile to keep the fans quiet and temps in the 80-85'C.

Count for yourself how much performance you can get.
If you are afraid of boosting voltage, going as high with core freq as it is stable is fine.

 

[VirtualLarry]

GPU-Z shows VDDC as just under 1.000v, and VRM temp 1 and 2 are 25-27C.

 

[Chicken76]

I wouldn't overclock a card that's doing distributed computing. If you don't care if the card still works 6 months from now, sure, go ahead and push it as far as it will go. I've had bad experience with this. It shortens the life of the card significantly, sometimes to just a few months.

Instead, I would look to see how low it can go on the voltages and still remain stable.

 

[Chicken76]

GPU-Z shows VDDC as just under 1.000v, and VRM temp 1 and 2 are 25-27C.

Those VRM's are either unused or the sensor is bad. You don't expect something that has tens of amps going through it to be that cold a couple of inches away from something that's sitting at 70 Celsius. (not to mention that it's connected to it with metal traces. Really, it's doubtful that anything on that card is that cold (except maybe for the corners of the PCB).

 

[VirtualLarry]

I wouldn't overclock a card that's doing distributed computing. If you don't care if the card still works 6 months from now, sure, go ahead and push it as far as it will go. I've had bad experience with this. It shortens the life of the card significantly, sometimes to just a few months.

Instead, I would look to see how low it can go on the voltages and still remain stable.

It's the same card as an R9 280, and most of those ship with a 950-1000Mhz core clock. My card shipped with only an 800Mhz core clock. So I figured up to 900-950Mhz would probably be safe.

 

[Techhog]

yikes
Too hot,IMHO.

You must either stick with low-end cards or undervolt everything if 70C makes you say "yikes."

 

[schmuckley]

GPU-Z shows VDDC as just under 1.000v, and VRM temp 1 and 2 are 25-27C.

Scroll down to the other VRM temps

 

[Techhog]

It's the same card as an R9 280, and most of those ship with a 950-1000Mhz core clock. My card shipped with only an 800Mhz core clock. So I figured up to 900-950Mhz would probably be safe.

My 7950 can't even get to 1GHz without raising voltage, and then if I do that I'll have issues with driver crashes if I try to play a game after waking from sleep without resetting the clocks. Don't make assumptions.

 

[Essence_of_War]

My 7950 can't even get to 1GHz without raising voltage, and then if I do that I'll have issues with driver crashes if I try to play a game after waking from sleep without resetting the clocks. Don't make assumptions.

Yeah, agreed. It may be the "same" GPU, but there is still binning that comes into play.

Instead, I would look to see how low it can go on the voltages and still remain stable.

If you're planning on using for DC 24/7, I think this is your best option. Although 70C gpu temp isn't really that hot for intense gaming, no one expects to be doing intense gaming 24/7. Also, as chicken76 pointed out, those VRM temps seem verrrrry low.

 

[tential]

I went straight to a 1 ghz OC with 0 issues.
Obviously not every card is good like that and you shouldn't do that probably.

I have the Sapphire Tri-X design I believe.

 

[Yuriman]

I believe many 280's ship with higher stock voltages, too.

Personally, I'd back off maybe 50mhz from your highest stable overclock without a voltage increase. 70-75c isn't concerning for Tahiti, and I've never known a non-overvolt OC to kill a card.

 

[Geforce man]

I'd agree with the previous. Stick it at 900Mhz and you're probably good to go. If you are comfortable with doing any kind of work on the card, you could replace the stock thermal compound with some AS Ceramique or similar, and probably shave off 2-3 degrees celsius (possibly more) from your temperatures as well.

 

[Headfoot]

yikes
Too hot,IMHO.

Wrong.

Larry,
You can go up to 80-85c on 7xxx series radeons. They overclock like a dream. I got 1200 core with only the smallest uptick in voltage on my last 7950. I was able to get 1000 core at -50mv from stock. Play around with it and see what happens. The only way to see how good your chip is, is to try.

 

[VirtualLarry]

Not sure if I smelled something "electrical". Been having power glitches, like every day. Thought I smelled insulation burning in the wall some months ago. Told maintenance about it, they said it had to be my equipment. Of course, I carefully smelled all of my equipment, and couldn't smell anything that seemed burned. All of it was working.

So, just to be safe, I removed the 100MHz overclock on my video card. It raised the temps of my video card from 70C to 72C, which I didn't think was significant, and was still within the allowable range of temps, but whatever, I don't want to be blamed, if there is aluminum wiring in the wall and it catches fire.

For the record, the PC on the left is drawing 273W max, according to my UPS software, and the PC on the right is drawing 240W (monitor connected to PC on right takes less power).

This apt has 20A circuit wires.

 

[Headfoot]

if something burns out in your computer, you'll know. I've had it happen and its usually pretty catastrophic and smelly