The NVIDIA/AMD Cards On Linux With The Best Value For 2015 Holiday Shopping

[phoronix]

Phoronix: The NVIDIA/AMD Cards On Linux With The Best Value For 2015 Holiday Shopping

If you are wanting to buy an AMD Radeon or NVIDIA GeForce graphics card this holiday season, here is a fresh round-up of thirteen different graphics cards using the latest AMD/NVIDIA drivers. Beyond just running several Linux OpenGL game tests -- including some Steam tests -- these results also have the performance-per-dollar benchmark results computed too for finding the best value for 1080p Linux gaming this season.

http://www.phoronix.com/vr.php?view=22495

[xpander]

what are your metro settings? seems to be not the highest settings, or does the CPU make so much difference.. im getting(metro 2033 redux) ~85 FPS and (metro LL redux) ~105FPSon GTX 970 with AMD FX8320@4,2ghz, thats with very high settings. the problem is that you need to start up the metro to change your settings. by default i think its on medium settings when you haven't changed anything

[dungeon]

It is all about CPU cap somewhere... i even laughed now on those multithousands max fps in Xonotic, that happens only with Intel drivers and nVidia drivers

That high load to the sky spikes never ever happened with any AMD driver either opensource drivers or fglrx... guessing that it has something to do with some CPU cache maybe or some instruction missues or broken for AMD drivers.

Or maybe if someone can better explain what CPU instruction gets mad there? SSE2 i guess? Read from something tackle sweet G-Spot

Does someone know maybe if nvidia driver uses intel compiler to make their blob?

With Unigine Valley is one of the rare cases where the AMD proprietary driver performance can compete with the NVIDIA binary driver.

Of course, when CPU is not there all is fine

[Kano]

@Michael

I always wonder why you only use the i7-5960x at stock speed - i7-4790k or i7-6700k would be certainly faster at stock speed for single threaded games. If you OC the CPU to 4.4/4.5 GHz it would be fine. If Intel does not send an i7-6700k then you could use the i7-4790k, but I did not see any benchmarks with that the last months. The Turbo speed of the Haswell Refresh is 900 MHz higher compared to the Haswell-E - thats 25% difference! Then you could see more scaling in CPU limited tests - otherwise it is lost time if every card gets the same result.

[whitecat]

The R7 370 series does deliver good open-source value too

No, it's a shitty card. The R7 370 and more globally PITCAIRN/CURACAO chipset must be ban. Buy it only if you want troubles or having a underclocked card.
In fact Michael you're lucky that your card works.

https://bugs.freedesktop.org/show_bug.cgi?id=76490

[vmicho]

Well, just bought 750 Ti KalmX, for 130eur here in Vienna 2 months ago. Cannot complain at all. And it's completely fan-less ^.^
And thanks for the nice tests.

[WOLF308]

@Michael

I always wonder why you only use the i7-5960x at stock speed - i7-4790k or i7-6700k would be certainly faster at stock speed for single threaded games. If you OC the CPU to 4.4/4.5 GHz it would be fine. If Intel does not send an i7-6700k then you could use the i7-4790k, but I did not see any benchmarks with that the last months. The Turbo speed of the Haswell Refresh is 900 MHz higher compared to the Haswell-E - thats 25% difference! Then you could see more scaling in CPU limited tests - otherwise it is lost time if every card gets the same result.

Agreed, need to dump that 5690 for gpu benchmarking. Honestly even the i7's are overkill. The i5 4690K is pretty much the go to chip in gaming circles, though it will slowly be replaced with the 6600K Skylake part obviously. Give it a decent overclock to 4.4-4.8 depending on your luck with the silicon lottery and your good to go. The extra ~$100 you save on the i5 vs the i7 can be put towards upgrading the gpu.

[schmidtbag]

Agreed, need to dump that 5690 for gpu benchmarking. Honestly even the i7's are overkill. The i5 4690K is pretty much the go to chip in gaming circles, though it will slowly be replaced with the 6600K Skylake part obviously. Give it a decent overclock to 4.4-4.8 depending on your luck with the silicon lottery and your good to go. The extra ~$100 you save on the i5 vs the i7 can be put towards upgrading the gpu.

An i7 isn't "overkill" for gaming, it's just simply unnecessary. HT isn't that useful for gaming, and in some cases will actually lower performance.

As for Skylake, the only reason to choose it over Haswell is because of the IGP. Gamers don't want the IGP, so Skylake isn't worth getting.

[Slartifartblast]

An i7 isn't "overkill" for gaming, it's just simply unnecessary. HT isn't that useful for gaming, and in some cases will actually lower performance.

As for Skylake, the only reason to choose it over Haswell is because of the IGP. Gamers don't want the IGP, so Skylake isn't worth getting.

It's a pity they didn't make a Skylake with IRIS pro L4 cache eDRAM like the i7-5775C and i5-5675C, it would have been interesting to see what that would have brought to Skylake.

[efikkan]

Agreed, need to dump that 5690 for gpu benchmarking. Honestly even the i7's are overkill. The i5 4690K is pretty much the go to chip in gaming circles, though it will slowly be replaced with the 6600K Skylake part obviously. Give it a decent overclock to 4.4-4.8 depending on your luck with the silicon lottery and your good to go. The extra ~$100 you save on the i5 vs the i7 can be put towards upgrading the gpu.

Some overclocking might be in order, but downgrading to a quad core is rather silly for benchmarking. There are already games utilizing 4+ cores, and there will be more to come.
For most optimized games a CPU well below 4 GHz should be enough and the GPU be the main bottleneck (unless it's multi-GPU).