Intel Celeron/Pentium/Core i3/i5/i7 - NVIDIA vs. AMD Linux Gaming Performance

[indepe]

But Xonotic and ET:Legacy are no ports And really every game shows different results with different CPUs so all of them are CPU bound. They are also GPU- bound in some times so the GPU performance does impact the result as well but it's definitely not optimal.

Yes. Also in some tests it seems that it depends less on the individual card and more on AMD vs Nvidia. I guess that could be either the lack of driver efficiency placing a higher burden on the CPU (in a way that should be optimizable), or due to the available AMD GPU commands requiring more preparation (or data) on the CPU side. (Meaning, GPUs may differ not only in their execution speed, but also in the amount of CPU attention they require. GPU memory size also being a possible factor.)

The 1050 seems to max out in about half of the tests, more or less, at some level.

[hxfhjkl]

Well now i know why my RX 470 is performing terribly under the x4 860k (which is probably worse than all the cpu's in this test)....

[oooverclocker]

indepe Yeah but the games are also games that are not very demanding so it's more likely that they're CPU bound. AMD vs. Nvidia has a smaller impact in this case perhaps - when you look at the Nouveau performance with reclocking it matches about the performance of AMD cards with RadeonSI so it mainly a driver issue.
No matter if Nvidia has really good drivers with optimizations we don't know of or really bad drivers leaving out some details and reducing the texture qualities, they provide a better performance and they have been used for the comparison with the AMD open source drivers. We shouldn't forget this point.

The 1050 shows a nice performance in these tests but yesterday I played some Talos principle with mostly 40 FPS everything maxed out, even setting the render resolution to unlimited and everything, just MSAA on 4x instead of 8x in 1440p on an RX 480@RADV.
So for these cases I highly doubt that a card with less than 1/3 of the raw performance of a 480 can reach playable FPS. As well as for Deus Ex: MK, Hitman, etc.
Though for high FPS >60 that we can often see in Benchmarks of Dota 2 and Talos here RADV seems to be kind of CPU bound as well at it's current state.

[Passso]

Thanks, but comparing to what? The only change recently made in PTS7 Git was bolding the title and shortening up the identifier height on multi-way comparisons, or is that what you are referring to?

(My 3 cents as a "standard reader with 1080p screen")

- Graphs were far too tall, (expecially compared to text) so that table results needed scrolling
- All colors were too close so it was generally hard to identify items
- small text was totally unreadable, I had to zoom in 200% just to read it, leaning my head to the screen

So, all is fixed now. Good job

[mitcoes]

Suggestion for a next "most visited post" gaming CPU + GPU combo price / performance benchmarks with AMD CPU and GPUs, plus Intel CPUs and Nvidia GPus (Intel GPUs are not for gaming). And make it for 3 price ranges and or 720p gaming, 1080p gaming and 4k gaming. But of course it will be next month or whenever new AMD CPUs will arrive.

[scorp]

Hi guys,

does anybody from you made the same (or at least some) test as they have being done here? I have an i5-3570K with RX470 and my results were always very much slower, than what Michael is getting. I always thought, that it is due to the better Xeon CPUs he is using, well now I see that I'm getting not the half of FPS, he is getting on Celeron with RX480. I know, that the GPU is slightly faster but double as fast? And I don't believe (even) the latest Celeron with two cores is so much faster, than the real 4 cores i5-3570K. I also made similar tests using my old HD7850, however the results are in no relation to the ones I see here. So, how about your results? I'd like to know what is so horribly wrong with my machine?!

[oooverclocker]

Do you use either the Padoka stable or Padoka PPA? With the old drivers being included by default the performance may be quite bad.

[xpander]

Here AMD FX-8320@4,5 ghz with GTX1070 into the mix.

in short: AMD cpu sucks when it comes to singlethreaded opengl its just a notch better than 2.9ghz celeron

Somehow Bioshock Infinite and Portal tests failed.. rest is here:

http://openbenchmarking.org/result/1702092-TA-1702083RI26

Enjoy

[gamerk2]

On Windows, ALL these titles are GPU bound.

As I've noted for a very long time: The Linux scheduler is not optimized for gaming workloads. The Linux scheduler is designed to cleanly handle lots of light threads in a way to minimize latency, but it sucks when you have one or two threads doing 99% of the work. What's happening is the main game thread and main rendering thread are getting bumped and not running, which is going to wreck performance.

For gaming workloads, per-core run-queues are a horrible idea. Stop defending a bad implementation and FIX IT.

[hxfhjkl]

X860k beats many intel cpus in windows gaming tests.

http://www.tomshardware.com/reviews/best-cpus,3986.html

Use latest drivers and a light distribution:

https://www.phoronix.com/forums/forum/software/distributions/926593-distro-for-linux-pc-in-living-room

Play windows games with wine-staging csm enabled, you beat windows benchmarks in some games.

Does not do anything for me, the games only run pretty decently on windows (which i don't use). I have tried wine-staging, overclocking, and different drivers, the cpu is just too weak.