I recently got a GTX 780 from @mutation666 and wanna compare it to my current Radeon R9 270X TOXIC. I have no idea how to do so fairly.
I am looking at both synth and real tests, gaming and non. On both Windows and Linux.
I don’t wanna go overboard tho, I want to know just the gist.
I am thinking of benching GTA V, Dirt Rally, Doom (2016), Quake Champions, and CS:GO. While on the topic of games, are 3 runs of the same bench enough? Should I just use in game benches where available? How do collect data, Fraps on Windows, what about LinuX?
As for synth I was thinking Blender Benchmark, but GCN1 cards ain’t supported as OpenCL compute for Cycles.
Anything else I should think about?
In any case, thanks!
I don’t think results would change much between runs unless you’re dealing with an unstable driver, so 3 runs should be more than enough.
Unigine has a couple of benchmarks, they run on both Windows and Linux
Most benchmarks should have built-in data collection, but i think glxosd has that functionality, but has been recently deprecated.
If you want to go crazy, you can always use PTS
Yeah already did unigine. Thanks!
Forgot about phoronix, but i just want to get the general gist of perf diff, not to torture test it
almost done benching 270X
it’s 5:40 AM I am slow as fuk lol
Are you still sleeping? WTB updates
For that age of cards, here’s the weird ass benchmarks that I would do.
#1: WOW quality switch. How long does it take to redraw?
#2: Source Game Frame Shudder. Source games are easy to lag with just moving your mouse. So, see what one does less lag.
#3: Planetside 2. Its just a good benchmark game.
#4: GTAV / IV quality swap. Again, redraw time.
Do these all in actual working games, not in benchmark modes.
Metro last light benchmark is a good one if you have it.
Check against other (older) benchmarks as sanity check.
Doom 2016 on Linux doesn’t work in Vulkan OpenGL works and got impressive FPS.
Vulkan just has black textures. I got angry and stopped benching lol. Drank beer and went to bed.
P.S. this is how to bench, not bench result thread. I’ll make a new one for those.