TR Forums

Anyone find the 7970 Ghz editions a little "touchy"?
I also have a 6870 and that one I can mine bitcoins and play dx11 games fine with no crashes.

For the 7970, it's kinda touchy. Running a coin miner + furmark or 3dmark doesn't crash the system. But games like crysis 3 or a few others that use dx10 or 11 crashes within 10-20 minutes. When it crashes the screen becomes unresponsive and the display goes black. It doesn't do the driver has crashed but has successfully recovered thing. I just have to restart the computer. If i don't mine coins while playing games then it doesn't seem to crash ever so maybe it's a temp issue. The GPU gets to about 82C when I have the coin miner + games running.... but my 6870 can get as high as 84C and never crashes so i guess the 7970 is just a bit more touchy?

I also noticed that the 7970 doesn't like certain things running in the background while gaming, like websites with a lot of flash/video content, playing hulu/netflix, etc either.

But underclocking to 925 mhz seems to be completely stable, aside from mining + dx10/11 gaming.

sys specs:
core i5-2500k @ 4.0 Ghz
16GB ram
windows 7 64 bit
7970 GhZ edition 3GB vRAM http://www.newegg.com/Product/Product.a ... 6814125413
650 watt corsair PSU

At the moment I'm running catalyst 13.1.

Anyone have any ideas on how to make it more stable w/o underclocking etc? Maybe I just need to get beefier fans to help keep the system cooler.... mainly using low RPM fans ~800-1200 RPM and I have a total of 5 of them on the case. 4 120MM and 1 140MM.

are you sure it's not a power issue? I don't know if your PSU has one or several 12V rails and how much power they can deliver, but if it has more than one rail it could be that you accidentally have the GPU and the CPU on the same rail, which may not be able to deliver enough current to feed both.

I doubt anything but a beefy GPU cooler will help....if you turn all your case fans off, you'll see they don't do much, at least mine don't in Antec 1100+3 fans.

Maybe the PSU is going bad. I presume you have it for several years now if you had a HD 6870 previously. The only way to test this is to use another capable PSU that is known to be stable.

In the mean time you could also test the RAM using memtest, temperatures for the whole system and voltages with OCCT while stress testing and use HD tune to see if the HDD is having any issues. A black screen without any error can be cause by many things (although if no error is displayed i'm suspecting hardware problems) that's why i'm suggesting to do a complete check. Maybe others will know better how to pin point the source of your problem. Maybe there are some records in Event Viwer that could help you find the culprit.

You could also test previous versions of the video driver that have been known to be stable.

As it doesn't crash at 925 MHz, this suggests a power or cooling issue. I've never run bit coin mining while gaming, but much over 80 degree seems a little warm, given that the Windforce triple fan stock cooler is decent.

I would reseat the cooler with new thermal paste. But since new and If worried about voiding warranty, maybe get an app where you can manually set fan speed. Test at high constant speed and monitor temps and see if issue persists. If it does and you can rule out the PSU, probably time for RMA.

(Edit: Learning how to use Swiftkey keyboard on smartphone

Oh i forgot to add something...

When I mine bitcoins, i crank up the GPU clock rate to 1150. As long I don't play games or anything, it's completely stable at this clock rate even though it gets really hot.

I am sure if i turn off the case fans, things will overheat. it feels like a little heater with as much heat its pumping out.

The PSU that I have just has a single 12V rated at 54 amps so I'm sure it's more than good enough. I can run furmark at 1000 mhz clock rate and it never crashes even though it gets to about 84-85C so I don't think power draw is the issue here.
I've previously ran memtest for hours and never reported an error, same with prime 95. I can run prime 95 + furmark (and give furmark a higher priority) and it doesn't crash.

It really seems to be game related... specifically dx10/11 type games... but only if i mine while gaming even down clocked to 925mhz, so maybe there is some kind of conflict with the mining algorithm (sha256) and the graphics engines that dx10/11 games use? I wonder if the 13.3 beta drivers will help :/

CB5000 wrote:

Maybe I just need to get beefier fans to help keep the system cooler.... mainly using low RPM fans ~800-1200 RPM and I have a total of 5 of them on the case. 4 120MM and 1 140MM.

That may be a good idea.

With my PC pushed under my desk, it heats up that whole space so that hot air is sucked back into the case when I'm gaming. I've placed a small tabletop fan to keep the air under my desk cool, which in turn allows the PC and its components to stay cool.

Using the 7970 OC TOP from Asus at 1Ghz works flawlessly no matter which game. Actually works so well that I've been having the comp on 24/7. Unfortunately, the fans doesnt stand for that, so I need to have one of those exchanged.

I'd still suggest starting by manually setting the Windforce gpu cooler fans near 100% and monitoring temps and seeing if instability persists. This should tell you right away if an overheating issue (which it seems to be - HD 7970s don't seem to like temps much above 80 deg C when overclocking at all in my experience - I have an older version of the same Gigabyte HD 7970)

Yeah heat might be the issue... I'm pegging the fan down to ~60% because I can't stand the annoying whirring sound above 65%-70%, which is about 2700 rpm. I'll try to get a better case fan or two because these low RPM ones barely move air. Maybe the 2000 rpm ish fans would work better than the 800-1200 rpm ones that I currently have.

CB5000 wrote:

Yeah heat might be the issue... I'm pegging the fan down to ~60% because I can't stand the annoying whirring sound above 65%-70%, which is about 2700 rpm. I'll try to get a better case fan or two because these low RPM ones barely move air. Maybe the 2000 rpm ish fans would work better than the 800-1200 rpm ones that I currently have.

Manually jacking up the GPU fans is just for diagnostic purposes. If this fixes the issue, you could then decide whether you wanted to try reseating/reinstalling thermal paste or getting a better aftermarket GPU cooler.

Checked the VRM temps on your GPU? I don't think it's system cooling...

I have a 7870 and I'm unable to OC it to any margin because the third party cooler Gigabyte uses doesn't actually make contact with the VRMs. That's something I noticed, my card crashes due to VRMs overheating. I'm sure it's similar for other manufacturers too (This is why I try to buy cards with a stock cooler). They don't seem to think they need to cool VRMs, running OCCPT mine hit over 110c and sometimes crashes, sometimes does not, but after noticing the temps for prolonged run sessions I stopped trying to OC it. I thought about putting VRM coolers on it, but decided it wasn't worth it.

I would suggest pulling off the heatsink and making sure it makes contact with your VRMs. GPUZ can also monitors VRM temperatures as well as the GPU, it's a good way to check to see if they're overheating.

If it's not a VRM issue, I would suggest moving the power slider to +20% in the catalyst panel as an attempted fix.

Ah, NM guys. I think it's just the bitcoin miner. underclocked to 900 mhz, mem underlocked to 1000mhz and it still crashes with temps at 60C when I have the miner plus a game running. Overclocked to 1150 doesn't crash when mining alone. stock speeds while gaming also doesn't crash at all either.

....... so it seems to be just mining + gaming just doesn't mix with my setup now.

My GPUz isn't showing vRAM temps :/

Just in case anyone else was encountering this problem... downgrading to catalyst 12.8 drivers solved the problem for me. Was not a hardware related issue but a software one.

I'm not sure if - I personally - would consider rolling back all the way to an August of 2012 release (8 months old) of a driver as much of a fix.

I have almost the exact same system down to the overclock on my my CPU, just different wattage on the PS/Win8/Latest Catalyst. I run games just fine. It might just be a bad card. It's quite loud now that the weather has changed and it's a bit hotter in the house.

Ryu Connor wrote:

I'm not sure if - I personally - would consider rolling back all the way to an August of 2012 release (8 months old) of a driver as much of a fix.

Yeah but I don't play any of the AAA titles out right now that demand the most up to date drivers. In my case it looks like the AMD stream SDK 2.8 can cause a hard lock due to some driver incompatibilities with catalyst 13.1. In my experience with graphics drivers, it seems like the latest isn't always the best. It seems like that holds especially true for litecoin mining. Some people are still running versions as old as 11.2 just because it's stable and gives the best hash rates for them.

In any case, the 7970's driver quality doesn't seem that great at the moment. Even during crashes from overclocking, it locks the entire computer and the screen goes blank instead of being able to "recover", and lurking on various forums, it seems like a lot of other people have similar issues. So rolling back to a version of the driver that works a little more stable prolly isn't a bad idea.

In comparison the 6870 on another computer is running 13.1 and can run games while mining concurrently without any problems so... And whenever I crash it from too much of an overclock it can recover like 99% of the time.

Oh just out of curiosity I looked at my 7970's ASIC quality on GPU-Z.... it was 70.5%.

If your OC is too high it'll simply hard lock the system. Usually the only time the driver reset keeps the system from hard locking is when the OC is borderline stable. That really goes without saying for anything. Like when you OC a processor too much and it hard locks, but if it's borderline stable it'll throw out errors in prime95 or whatnot instead of locking the entire system up. If your driver is reseting it's not stable to begin with though.