R9 390 + OCCT question/problem

[SentinelAeon]

Hello, i have ASUS STRIX R9 390 DirectCU III OC, 1050/1500 default clocks. I am trying to undervolt the card to make it more efficient and quiet. Its all good but i am having problems testing my undervolt (and maybe even a future underclock).

I am using the latest OCCT. Graphic card at default (1050 gpu, 1500 mem, 1.275V on gpu, 0% power limit). I run a gpu test with error testing, everything at default in the program. It passes like 1 hour of stress testing. But when you look at the actual clock, you can see it is nowhere near 1050Mhz on gpu. I know this is normal since power limit is set, but do you see why this is a problem ? How am i to know that the gpu is stable at 1050Mhz, if it never reaches that frequency ? Just to be objective, i tried setting the power limit to +25% and then the clock really reaches the 1050Mhz on gpu, but thermals are insane in this case, my vrm goes over 100C until the OCCT reports errors after a few minutes.

So my question is this: How am i to stress test whether my undervolt of 1.250V (-0.025V) at 1050Mhz is stable, if the OCCT test never goes even over 1000Mhz ? The only other option in OCCT is to limit the GPU usage to like 90 or 95% in which case clocks will go to 1050Mhz, but then again, i wont have the full stress of gpu at that frequency.

 

[SentinelAeon]

Or maybe you can tell me how you test stability of gpu

 

[EarthDog]

I know this is normal since power limit is set,

I'd try looping Unigine Superposition or something like that to test.

 

[SentinelAeon]

I was playing with power limit a bit, since my first goal is to lower the heat and noise when i play games. I currently play Dragon age: inquisition in 1080p, most things maxed. I am happy with 40-50 fps average, right now i have far more and it is a waste. So i simply set the power limit to -20%. I am getting some wierdness about power limit ... sometimes drivers use that power limit, sometimes they simply ignore it and i have to restart pc for it to take effect. I did some testing of it, i was interested to see what the curve of power vs fps was, trying to find the optimal point for best performance per watt. As you can see it seems to be 275W on killawatt (complete pc draw). Why i say 275W on killawatt and not -20% on power ? Well .. because this power thing is acting strange. For instance setting power to -5%, i get 313W instead of 388W, which simply isnt possible, thats not -5%, thats - a lot more. And then i will set the settings to default and apply -5% again and suddenly im getting 360W ... So yes, very strange.

 

[RJARRRPCGP]

I'd try looping Unigine Superposition or something like that to test.

With my Zotac GeForce GTX 1660 Super, I get power throttling that lowers the boost clock some, even in Unigine Superposition at 1080p high and PC Halo MCC. That's at the default 125W TDP limit and it only lets me go to 130W TDP for the max-override, according to MSI Afterburner.

 

[EarthDog]

I sit around that too. It's where you want to be.

 

[SentinelAeon]

As i mentioned, right now i don't need the full power of GPU. I could use AMD fps limit, but i read online that this introduces mini stutter and ever since i saw that, when i play with FPS limit, i simply feel stutter even if it is not there.
So i decided i will test what is better: lowering GPU clock or lowering power. At first glance lowering power for 30% sounds better, but i am getting some strange results. I set limit to -30%, run the same test ... sometimes max power draw is 80W, sometimes 150W and sometimes 200W. I have yet to come to the conclusion what is the reason. On the other hand, i can set lower max clock and also undervolt. I will compare both, write down the average fps and average power draw. I downloaded the latest 3dmark and i might use OCCT also but i might also not.