Using the default auto fan-curve I get 67°C with open side-panel and 69°C with closed side-panel. The GPU fans spin at around 50% at this point. GPU clock at constant 1995 MHz.
Curiously most of the heat seems to point upwards instead of outwards, despite the heat-sink's fin orientation.
Keep in mind that at this low CPU/VRM load my case fans only spin below 500 rpm (4x push/pull on intake CPU AIO radiator + 1x rear).
Temps looks really good. Also your case cooling is quite good as it gets only 2 degrees warmer.
Originally Posted by lightsout
Apparently my temps are similar to yours. Which I agree are fine. But with max power at 2025mhz I got up to 69c with Superstition.
The air-flow might be just equally bad, because at higher GPU fan-speeds the difference between panel on and off is bigger. There even is insulation material on the side panels (bitumen + fabric) to keep noise down. It helps to use an AIO for the CPU, though, even with the radiator heat blowing into the case. Running Prime 95 Small FFT along-side does not make any difference, even with the radiator/case-fans spinning up towards 1000-1300 rpm then. Overall there is plenty of empty space in there, which likely matters most.
At 100% fan-speed (3300 rpm) and open panel the temps stay at 53°C with constant 2025 MHz and 39-40 fps. If more of the smoke is placed in the view then clocks dip down to 2010. 80% (2650 rpm) can keep the clocks and fps mostly the same, but temps increase towards 57°C. Even with auto-fan curve and temps rising towards 69°C with a closed side-panel frame-rates stay around 39 fps, so it all doesn't matter too much in practice.
I tested this all at a position with 740 materials, 23 light-sources, around 9.1 - 9.3 m triangles and some of the smoke being visible.
On a side note: Removing the 280 mm fan-window from the side-panel (or the whole panel) has much less of an impact than it had before with my GTX 780 GHz Edition. With the latter card removing the window would easily lower temps by 10°C. That's likely because the old card blew more hot air towards the side-panel and then pulled the same air back in again, while the Asus Dual blows more hot air upwards and thus away from its own fan intakes.
Last edited by Timur Born; 03-21-2019 at 02:56 AM.
I have some very interesting results! I have done some more testing with the TDP of the EVGA RTX 2060 XC with a higher OC. Superposition is quite TDP hungry benchmark, so it is very important to have TDP of your card at max levels. Or is it?
This is manual +125 core +1000 mem 113% TDP:
This is manual +125 core +1000 mem 95% TDP:
And this is stock:
At 113% TDP and max OC, this small card struggles to dissipate more heat that it is designed to do, it gets at max 78C during benchmark.
Meanwhile with 95% TDP and same OC, this card scores very close as with 113% TDP while running at 72C max.
At stocks speeds with all stock setting it scores way lower and runs max 73C.
So you can increase card's performance by OC while reducing TDP. This gives some extremely efficient results from what I saw in my tests.
And this should be the max performance reduction by lowering TDP as Superposition is very TDP heavy. I am quite surprised by this!
Wow not bad. Is that 1000mhz mem oc +250 on the slider? That always confuses me.
+1000 on the silder, from 7000Mhz to 8000Mhz.
Originally Posted by Timur Born
120% power limit, +145 core, +925 mem: 165 - 175W, Score 11413 (11544 at 100% fan-speed)
100% power limit, +145 core, +925 mem: 156- 161W, Score 11322 (11396 at 100% fan-speed)
95% power limit, +145 core, +925 mem: 148 - 153W, Score 11186 (11259 at 100% fan-speed)
No curve, no undervolting, just changing the power limit slider.
And what's your stock score? Because it's a very important baseline.
Yes, I came back to using the manual slider for the core, no oc curve and undervolt. It wasn't stable in other benchmarks and too much messing around. So now I use +117 core, +1000 mem, 95% TDP, stock fan, max temp 71C (22c room), Score 11187. Stable in 10+ games+benchmarks. I will stick with this for now.
I ran 1080 "Extreme" at 100% vs 120% power limit using the same overclock values. Peak power draw was 12% higher at 120% limit, GPU load was 99%. Score difference was only 2% at 100% fan-speed and open side-panel. So realistically there seems to be only little benefit in upping the power limit.