Well Unigine Heaven is more of a stability tester than a FPS indicator for one. Secondly, you're using a tool developed in the exact code of Warzone, C++, very CPU intensive. You're running a 3600, not the hottest CPU with a B450 board also not optimal for AAA gaming. Running a serious game or test means all the hardware needs to be factored as a whole when looking at results and such. It would be like putting that card into a AM3+ FX 8150 system and expecting 200 FPS in any game. Not happening because the card exceeds the power/speed of the CPU, BUS and RAM. You're missing the benefit of PCIe 4.0 which that card takes advantage of.
Try running 3D Mark TimeSpy, Super Position other tests. Try other games. Are you running the test with back round apps or unnecessary Windows services on? Is "enhanced sync" on? Have you tried at least moving the minimum GPU clock to within 100Mhz of the max GPU clock and enabling "fast" VRAM? What monitor specs are we talking about? DP or HDMI connection, refresh rate, GTG response all play a part in the smoothness or perceived FPS drop. Remember these tests weigh heavily on the CPU as well as the GPU, especially single threaded functions that AMD 3000 series CPU's fall behind in.
I've read the Power Color cards, especially the Red Devil's aren't living up to expectations as well. So at the end it could be brand. Very hard to say being the hardware attached. The hardware of the build is just "ok" not top of the line like an X570 and other matching components where diagnosing a performance issue could really be pin pointed to the card or something else.
I've also red posts dating back over a year with Nvidia having FPS drops using that test, so definitely try others. Bottom line if you drop in a high performance GPU expecting miracles using parts that don't totally match, you're in for disappointment. There's a science to building a gaming PC and just slapping high priced parts in doesn't mean the rig will fly. I can virtually guarantee that there's nothing wrong with the card itself in this build. My feeling is many of the complaints have the same in common of using lower end CPU's and boards, even RAM and expecting 300 FPS because the card cost about $1k.
My build is a X570 board with a 3600X/PBO enabled/10XScalar/motherboard voltage, 16GB DDR4 3733 custom tuned to 3600, 1800 IF, Gigabyte WindForce RX 6800 Gaming OC, Corsair RM750x, Corsair H110i. So far no issues with FPS and test scores are very nice. I say this to demonstrate parts matching parts to achieve a desired result, not to brag.
Now it's working
And you can also see the graph of my core clock
I now have this results, with this settings, core clock still having drops and I can't even change the VRAM, or the bench just crash....
And I was wondering, when you see those 2 graphs, the GPU usage goes down as the power consumption goes down aswell. I got a power supply 750w from Aerocool. I didn't know when I bought it but those power supply have very bad reputation. Could this be the problem ?
Old question, but in any case, the graphs show a drop at a particular timeframe. Does that correspond with a frame drop situation or is it an actual "movie" part of the game where frames often go to ~25-50 FPS?
Another thing is the Tjmax seems high at 80c+. My non-x only hits ~60c in games running on max graphics and OC'ed to the hilt. Might want to change the fan profile or get some more airflow there. Yes, in spec, too high? Absolutely in my book. AMD cards always seem to have an issue around the 77c mark on the VRAM. The "hotspot" seems to be the same way.
Make sure you use 2 separate, 8 pin power connectors, not the split. In earlier posts (see pic below) , my settings are for a non-xt Gigabyte RX6800. The "XT" uses a tad bit more power but not enough to call for 84c. In Radeon settings under "performance", enable all adjustments then: select "manual", select "under-volt GPU" and write that number down, same screen, enable "fan tuning". Make sure to turn off any "zero RPM" option and make your own fan curve, similar to mine for coolest results. Not the least noisy but far from loud. Cooler is better no matter how you achieve that goal, to a point.
