Radeon HD 5450 errors

[rs77063]

I'm running (or trying to) GPU mining (using cgminer), but I'm getting errors like:

Code:

GPU0: invalid nonce - HW error

According to ckolivas, this is:

Hardware errors are not uncommon at high overclocks or high temps. Just bump the clocks down a little or drop temps further. If you haven't overclocked, it may be a sign of some hardware instability but likely is harmless.

Up to that point, I never touched the clocks, so it can't be overclocking. That leaves temp. Is my temp too high? I looked on AMD's site for specs and did find them, but no mention of temperature.

I've since underclocked the GPU from 650 to 300 and the memory from 450 to 300. While mining, my temp is about 67C.

Is this high for my card? What is a good temp range? Or do I have some hardware instability, which since it is stopping me from GPU mining, isn't harmless.

[1714160848]

1714160848

[ssateneth]

occasional hw errors are nothing to worry about. If you are getting nothing but hw erorrs, then start worrying. GPUs have a tolerance of about 100C. Recommended 24/7 temperature should not exceed 85-90C.

[rs77063]

Code:

If you are getting nothing but hw erorrs, then start worrying.

Yep, time to start worrying then.  

This still puzzles me though. My desktop video, running off the same card, seems fine. I would think there would be something wrong if my GPU is messed up. Can anyone explain why I'm not seeing video problems?

Although it does occur to me I may have stumbled on to something useful. I've got tests for power supplies, cpu's, memory, and hard drives. But up until now, nothing for video cards. Hmmm....Especially if this could detect more subtle faults.

[crashoveride54902]

Code:

If you are getting nothing but hw erorrs, then start worrying.

Yep, time to start worrying then.  

This still puzzles me though. My desktop video, running off the same card, seems fine. I would think there would be something wrong if my GPU is messed up. Can anyone explain why I'm not seeing video problems?

Although it does occur to me I may have stumbled on to something useful. I've got tests for power supplies, cpu's, memory, and hard drives. But up until now, nothing for video cards. Hmmm....Especially if this could detect more subtle faults.

furmark is a heavy stress test for gpu's, it's been dubbed the power virus for your video card..i wouldn't run it thou, why stress out a video card and make no money doing it? mining is almost as hard on a card as furmark...that is why you don't see anything bad happen when you use it for your desktop...try to play a game and see if you get any artifacting...did you say your getting 67c at 300mhz core clock? if so, then i think your gpu is on the way out the door because that's really hot for a low clock...unless it's dusty in the heatsink..or the fan is worn out and not spinning as fast

[ssateneth]

I find OCCT to be a much greater stress tester. My reference 5870 cards pull about ~5-10% more amps under load compared to furmark, as well as having a built-in error checker to check for unstable oc

[-ck]

Try a different driver version.

[rs77063]

...did you say your getting 67c at 300mhz core clock? if so, then i think your gpu is on the way out the door because that's really hot for a low clock...unless it's dusty in the heatsink..or the fan is worn out and not spinning as fast...

The 5450 is passively cooled, so there's no fan to wear out. Dust isn't an issue here either. I can see the card's heatseak through my case window. It looks clean.

I did finally find some other tests that mention the 5450's temp. One test (furmark) puts it at 58C, and the other claims 43C "under load". Right now my load is about 0% and I'm getting 55.5C (room temp is 27C). Time to add another fan to the case, even though I think you're right about the gpu on the way out.

I find OCCT to be a much greater stress tester.

I'll have to check that out. Thanks for the tip.

Try a different driver version.

I have serious doubts that will fix anything, but since your advice has been dead on so far, I'll give it a shot and report back. Probably be a few days though.

[ssateneth]

I think you ignored my post. GPUs have a tolerance of about 100C. Most will throttle before that happens. Safe 24/7 operating temperature is 80-85C. Also it's just a budget card. I wouldn't be worried. Also try a different driver and OCL.

[crashoveride54902]

...did you say your getting 67c at 300mhz core clock? if so, then i think your gpu is on the way out the door because that's really hot for a low clock...unless it's dusty in the heatsink..or the fan is worn out and not spinning as fast...

The 5450 is passively cooled, so there's no fan to wear out. Dust isn't an issue here either. I can see the card's heatseak through my case window. It looks clean.

I did finally find some other tests that mention the 5450's temp. One test (furmark) puts it at 58C, and the other claims 43C "under load". Right now my load is about 0% and I'm getting 55.5C (room temp is 27C). Time to add another fan to the case, even though I think you're right about the gpu on the way out.

I find OCCT to be a much greater stress tester.

I'll have to check that out. Thanks for the tip.

Try a different driver version.

I have serious doubts that will fix anything, but since your advice has been dead on so far, I'll give it a shot and report back. Probably be a few days though.

ahh didn't know it was passive...but i probably should have caught that at the low 650 clock to start with...could always check your vrm temps with gpu-z but if its a budget card they probably didn't put vrm temp diods in...i got a hw error on my 5870 and scared me cause i've been pounding it with mining...no biggie thou if it goes out, only payed 35bucks for it...well good luck to ya

[GenTarkin]

HW errors such as those are a sign of bad RAM or cold solder or components on the PCB being burned out but barely limping along...
Eventually you will start seeing these HW errors being thrown more and more over time, to the point windows/linux will no longer be stable & just have to remove the card.
I have had this happen on 3 different 5970's now on one GPU...
Im convinced these cards arent just meant to handle this type of 24/7 load =(

[rs77063]

Try a different driver version.

OK, I downgraded from amd-driver-installer-catalyst-13.1 to amd-driver-installer-catalyst-12.10. Same error.

Thinking it's really a bad card, I bought a new one. This time a Radeon HD 6570. It's fast for mining anyway. And still I'm getting the same errors. But now I'm getting them faster. OK, what are the odds of two bad cards?

So for kicks, I tried another miner, phoenix 2. Surprise, it mines!   It's also running cooler, at about 53C than my 5450. Admittedly, the hash rate with phoenix 2 is lower by a significant amount than with cgminer, but that could be because phoenix 2's docs are almost non-existent and I haven't tuned it properly yet.

So this looks like a bug in cgminer to me. What can I do to help you, ckolivas, find it?

Update: After some tuning, phoenix is now slightly faster that cgminer.

[crashoveride54902]

really? what kind of tweaks?

Try a different driver version.

OK, I downgraded from amd-driver-installer-catalyst-13.1 to amd-driver-installer-catalyst-12.10. Same error.

Thinking it's really a bad card, I bought a new one. This time a Radeon HD 6570. It's fast for mining anyway. And still I'm getting the same errors. But now I'm getting them faster. OK, what are the odds of two bad cards?

So for kicks, I tried another miner, phoenix 2. Surprise, it mines!   It's also running cooler, at about 53C than my 5450. Admittedly, the hash rate with phoenix 2 is lower by a significant amount than with cgminer, but that could be because phoenix 2's docs are almost non-existent and I haven't tuned it properly yet.

So this looks like a bug in cgminer to me. What can I do to help you, ckolivas, find it?

Update: After some tuning, phoenix is now slightly faster that cgminer.

[rs77063]

really? what kind of tweaks?

Really only two things. Keep in mind this is an AMD/ATI card. I changed two lines to the following:

Code:

    kernel = phatk2 #The kernel to use for the device
    worksize = 128 #Work group size, tweaking this option may improve performance

I'm still playing with aggression. At aggression = 3, there is occassional desktop lag, but not too bad. When I'm away, I set it back to 6.

[vdragon]

On the hardware side, check the thermal grease under the heatsink, it sometimes looses its specifications, simply turns into a brick, then you need to remove it, and put new one.

[crazyates]

Not all 5450s are passively cooled. Some do have a fan.

We recently got a cheap Asus 5450 with a heatsink held on by 2 pins, and it was so flimsy, the heatsink was flopping all over the place. I couldn't imagine it cooling that thing under load, especially mining.

[FreedomCoin]

one of my 6950s has started getting HW errors, about 5 every 12 hours. Funny thing is, i still get less rejects than my card with no HW errors :-/

yeah check the paste thats what im doing next. And dont forgot to lube those fan bearings if you have a fan.