AMD 290 still throttling even with low temps?

thilanliyan

Lifer
Jun 21, 2005
11,848
2,051
126
I have 2x290 video cards mining right now and the temps are kept <65C. One of the cards sticks to the 950MHz clocks and 1.15v, while the other one keeps throttling to between 915-950MHz and voltage fluctuates as well. Powertune is at 30% for both cards (I think).

Any ideas why it could be throttling?
 
Feb 19, 2009
10,457
10
76
It will throttle naturally unless its pegged to the max constant. Mining has ups and downs.
Try powertune 50% and see as well, tho I dont think its that.

Also try force --gpu-engine 950 in cgminer.
 

Stuka87

Diamond Member
Dec 10, 2010
6,240
2,559
136
Load up MSI Afterburner and watch the GPU % along with the clock speeds. I am betting you are not hitting 100% utilization, which means its going to declock on its own to conserve power/heat. The voltage should fluctuate as well.
 

Gloomy

Golden Member
Oct 12, 2010
1,469
21
81
The powertune power limit doesn't change the voltage. It governs the amount of power that the card can draw from the PSU. There is no reason not to set it at +50%, especially when you're mining... unless of course your PSU can't deal with both cards. I just set mine to +50% and let it draw as much as it needs.

Push it to 50% and see if that changes anything, but like they've said, it probably isn't that.
 

(sic)Klown12

Senior member
Nov 27, 2010
572
0
76
Load up MSI Afterburner and watch the GPU % along with the clock speeds. I am betting you are not hitting 100% utilization, which means its going to declock on its own to conserve power/heat. The voltage should fluctuate as well.

GPU usage doesn't work correctly with Hawaii. It'll continuously bounce from 0% to 100% no matter how much of a load you're putting on it. According to Unwinder it's something that AMD will have to fix within their drivers.

Guys, in different forums I've seen claims that AB reports wrong GPU usage on 290 cards because it is not supporting new AMD GPUs yet. That's incorrect statement, new versions of AB won't fix it, GPU usage is reported as it is calculated by AMD driver and it is not anyhow calculated internally in AB. So it is not a case of unknown GPU, it is as correct as AMD driver team implemented it. Currently GPU usage sensor implementation inside AMD driver is broken and it may return unreliable data if other sensors were polled immediately before GPU usage. So the amount of wrong data can be minimized by adding artificial delays between polling the sensors, it may be minimized by putting GPU load sensors to the first positions in the list, but the only true fix can be expected from AMD drivers update.

Silverforce11's answer is most likely the correct one. If the load drops any the GPU clock will drop accordingly as Hawaii clocks down rather than waste cycles.
 

FalseChristian

Diamond Member
Jan 7, 2002
3,322
0
71
I wouldn't worry about it. When you fire a graphics-intensive game like Crysis in XFire (Yes, I said Crysis) both cards will be utiliized 100% and they won't fluctuate (my experience with nVidia GPUs, anyways).
 

thilanliyan

Lifer
Jun 21, 2005
11,848
2,051
126
Yes it looks like the GPU load is fluctuating on the 2nd card, causing the voltage and core to change as well.

That's odd though, as my 7950s were not doing that before. HOWEVER, with the 13.11 beta driver, even my 7950s are doing it, and also only on the 2nd card (only 2 of my HIS cards are mining right now).

I will try forcing the clocks on the 290. The clocks on the 2nd 7950 are constant but the GPU load fluctuates, and like I said it wasn't doing that before.
 
Feb 19, 2009
10,457
10
76
Mining load will vary, sometimes lots of people hammer the pools. Which I suspect is happening due to he massive +50% extra workers in the past few days. Just let your card do its thing on auto, set the temp target, up powertune, let it go nuts with the fanspeed activating jetmode.
 

thilanliyan

Lifer
Jun 21, 2005
11,848
2,051
126
Lol thankfully it's winter here in Canada and I can just let this rig run outside where I don't have to hear it :D... -15C today, but the cards are still running at 70C for the bottom card, and 85C for the top card at 50% and 60% fan respectively. I don't have any fans blowing into the system though because my PSU is running at about 15C, which is about as low as I want it to go.

One thing I thought of...maybe the device IDs in cgminer are reversed meaning it's actually the main card that is throttling (which would make sense to me because it has to render windows as well), but then Afterburner shows it as "GPU2" that is throttling...not sure.
 
Feb 19, 2009
10,457
10
76
Afterburner doesnt read CF R290 load properly btw, known bug.

Im using GUIminer and Hawaii 1-1 = top card (the one sucking hotter air from the bottom card's PCB, also the one driving the monitor). I originally thought it was 1-0 for the top card.
 

thilanliyan

Lifer
Jun 21, 2005
11,848
2,051
126
Afterburner doesnt read CF R290 load properly btw, known bug.

Im using GUIminer and Hawaii 1-1 = top card (the one sucking hotter air from the bottom card's PCB, also the one driving the monitor). I originally thought it was 1-0 for the top card.

I have XFire disabled. Would it still read the GPU load incorrectly?