[SOLVED] 5970 (GPU0 isn't putting out ~400 mhash/s but GPU1 is)

[someone703]

Had it working after getting help in my other thread, but then just a day or two ago things started acting up in GUIMiner as far as "Speed" and "Accepted" statuses not updating or just sitting on "connecting..." instead of showing hash speed even though it was mining.  I haven't made any hardware/software changes.

Tried switching over to cgminer to see if it'd be better since there are threads about GUIMiner not working for others as it did before.

Now it would seem that GPU0 is only putting out half the mhash it should be.

Also, will the overclock settings done in cgminer reset to defaults once cgminer is closed?  Just curious but my settings are stable for mining/browsing but if I decide to game those same settings will cause my computer to lock up.

Screenshots of GPU0 & GPU1 (I've also tried changing the engine/memory/voltage settings for GPU1 in cgminer and it says that it's successful but it still displays 400/900 and 1v):

[ssateneth]

Had it working after getting help in my other thread, but then just a day or two ago things started acting up in GUIMiner as far as "Speed" and "Accepted" statuses not updating or just sitting on "connecting..." instead of showing hash speed even though it was mining.  I haven't made any hardware/software changes.

Tried switching over to cgminer to see if it'd be better since there are threads about GUIMiner not working for others as it did before.

Now it would seem that GPU0 is only putting out half the mhash it should be.

Also, will the overclock settings done in cgminer reset to defaults once cgminer is closed?  Just curious but my settings are stable for mining/browsing but if I decide to game those same settings will cause my computer to lock up.

Screenshots of GPU0 & GPU1 (I've also tried changing the engine/memory/voltage settings for GPU1 in cgminer and it says that it's successful but it still displays 400/900 and 1v):

yeah..... your "slow" gpu is definitely in a different performance level. it needs to be in perf level 2 to run full speed, that looks like perf level 1, which is used for basic hardware accelleration, such as youtube videos. Close all your browsers and any videos that may be using hardware acceleration, or turn off hardware acceleration of videos.

[someone703]

Thanks, turns out it was just a youtube tab I had open that wasn't actually playing anything...

Sent you a small tip (it's not much but I've only just started into mining and it's gonna be like 2 weeks for my other GPU RMA; assuming they don't take longer).

[ssateneth]

Thanks, turns out it was just a youtube tab I had open that wasn't actually playing anything...

Sent you a small tip (it's not much but I've only just started into mining and it's gonna be like 2 weeks for my other GPU RMA; assuming they don't take longer).

NP, thanks for the buck. Yeah you're better off turning off hardware acceleration for any sort of video playback. you can right click a youtube video to turn off video acceleration for all flash applets. DXVA is also hardware acceleration used for video playback such as bluray and certain video codecs in your video player of choice and sets the GPU to perf level 1 (overrides perf level 2). Switching to software decoding shouldn't be a problem as long as you got a half decent CPU for decoding 1080p videos. Anything less can be decoded by a 1.6ghz single core celeron, current generation of course.

[Ed]

also,
keep in mind what 5970 cores is not full identical, due to different cooling efficiency

GPU core near fan has lower temperature and as result higher OC capability

[ssateneth]

also,
keep in mind what 5970 cores is not full identical, due to different cooling efficiency

GPU core near fan has lower temperature and as result higher OC capability

But really had nothing to do with the OP's original problem. /shoo

[BTC guy]

I wish my 5970 gave me 400 hash. Im only getting 325 per core.

[someone703]

What kind of clocks are you running it at?

At 850/150 I'm pulling 375-380 mhash each with 1.05v.

If I up it to 900 I pull ~400 each but that was with 1.1v.

[jjshabadoo]

850/150 on air?

seriously?

Just one card or what? what brand?

[someone703]

Yup on air.  Fan speed set to 50% and I get around 75-80C.

Just 1 card.  Diamond 5970.

I'm sure my temps will jump up a bit more when I get my other 2 5970s (one should be here Tuesday; the other 1 more week if Athlon Micro sticks to their 1-2 week timeframe for RMAs on my Sapphire).  And the other one I think is probably a Diamond.

The person on Craigslist just said it's an AMD Radeon 5970 from Newegg and I didn't really bother to ask him what brand it was at the time.

When the other cards come I'll probably either turn the fan speeds up or put an external fan of some sort on them.

[someone703]

Well, 2nd card arrived and works just fine.

It's able to mine overnight stable at 825/150 as well.

Temps jumped a bit - ~80-85C on 65% fan speed; also got a small external fan going on them too.

But now one GPU won't hit 375 mhash/s even though they're on the same clocks and intensity.

GPUs 0/1/3 were doing 375-380 mash/s and GPU 2 was doing like 250ish.

Closed down all other programs and browser too.

[Dyaheon]

Well, 2nd card arrived and works just fine.

It's able to mine overnight stable at 825/150 as well.

Temps jumped a bit - ~80-85C on 65% fan speed; also got a small external fan going on them too.

But now one GPU won't hit 375 mhash/s even though they're on the same clocks and intensity.

GPUs 0/1/3 were doing 375-380 mash/s and GPU 2 was doing like 250ish.

Closed down all other programs and browser too.

Did you check the actual clocks? It could be thermal throttling due to high VRM temps, 5970s tend to do that thanks to inadequate VRM cooling.

Try turning up the fan to see if that helps.

[DeathAndTaxes]

Likely throttling.  That was why people were surprised when you said 850 core.

Kinda hard to keep 2 (even harder w/ 3 or 4) cool enough to do 850.

In GPU-Z (sensor tab) or Afterburner take a look at the GPU load % on the "slow" core.  My guess is it is not at 100% or it is changing (100% then 80% then 100% then 80%).  That is a sign of thermal throttling.  Same time in GPU-Z look at the VRM temps.  They can be 20 to 30C higher than the highest core temp.  If they hit 110C the card will start to throttle to avoid an overcurrent situation (hot VRM = less efficient = more current = hotter VRM = less efficient = more current ...).

[someone703]

Thanks for the replies all.

CGMiner shows the temps as being good.  I also tried removing the temp settings from the cgminer.conf but don't know if that actually removes any throttling or such.

Kinda weird, I did restart my comp and started it up and it was doing fine speedwise and then later it seems like maybe it gets dropped to a different performance level but doesn't kick it back up after I stop doing whatever is lowering it.

CGMiner showed an average hash rate of 1400 mhash or so rather than 1500 mash/s.

Guess maybe I should try underclocking and undervolting for efficiency until the prices recover a bit.

[wabber]

Thanks for the replies all.

CGMiner shows the temps as being good.  I also tried removing the temp settings from the cgminer.conf but don't know if that actually removes any throttling or such.

Kinda weird, I did restart my comp and started it up and it was doing fine speedwise and then later it seems like maybe it gets dropped to a different performance level but doesn't kick it back up after I stop doing whatever is lowering it.

CGMiner showed an average hash rate of 1400 mhash or so rather than 1500 mash/s.

Guess maybe I should try underclocking and undervolting for efficiency until the prices recover a bit.

I don't think it's CGMiner clocking them down it's probably the driver that's why ppl said you should take an other program to check the GPU load.

First post outside the newbie section WOW

[someone703]

Alright thanks guys, figured it out sorta but it led me to a different question.

GPUs 0/1/2 are well withing acceptable VDDC temps at 825/150 @ 1.049v.  This is the one (GPU3) that was hitting high VDDC temps and I downclocked that one in particular and lowered the voltage so that the temps are more acceptable as shown here (610/150 @ 0.899v):



Using GPU-Z and changing up the core clocks I determined that it was GPU3 in CGMiner that was getting the high temps and had the GPU Load bouncing around and it's more stable now; however GPU3 was always stable at ~375 mhash/s from testing overnight the past couple days.  It was GPU2 (as shown below) which was giving the less than desirable mhash rates:



So my question is: I downclocked GPU3 so shouldn't it be giving ~250 mhash/s instead of GPU2?

[DeathAndTaxes]

AMD driver library is inconsistent with the way it returns order of GPUs.
GPU #3 in GPU-Z may not match GPU #3 in cgminer.

This problem only exists in Windows. 

The best way to test it is to set each GPU to a different clock speed (500 Mhz, 600 Mhz, 700 Mhz, etc) then check the GPU speed reported by GPU-Z for each GPU (in sensors tab).

[-ck]

Alright thanks guys, figured it out sorta but it led me to a different question.

GPUs 0/1/2 are well withing acceptable VDDC temps at 825/150 @ 1.049v.  This is the one (GPU3) that was hitting high VDDC temps and I downclocked that one in particular and lowered the voltage so that the temps are more acceptable as shown here (610/150 @ 0.899v):

So my question is: I downclocked GPU3 so shouldn't it be giving ~250 mhash/s instead of GPU2?

GPU order as returned by the display library doesn't always correlate with the device order and cgminer can't tell so it simply reports what it's told by the ATI display driver. You can try the --gpu-reorder switch with cgminer which makes cgminer try to correlate them based on the PCI BUS ID.

[someone703]

That's just the thing though.

With GPU-Z open on the side, I've gone into CGMiner and altered the clocks.

When I changed GPU3 (in CGMiner) to 610, I saw the change in GPU-Z (which was the one with the higher VDDC temps).

It's GPU2 though which is giving me the lower hash rates.

Will look into how to use the --gpu-reorder and give that a try.

Edit:  If I were to put it in the cgminer.conf file would I just have a line with "gpu-reorder"?