The NVIDIA GPUs Delivering The Best Performance Per Watt & Per Dollar For Linux Gamers

[phoronix]

Phoronix: The NVIDIA GPUs Delivering The Best Performance Per Watt & Per Dollar For Linux Gamers

Following last week's NVIDIA GeForce GTX 950 launch I took the current complete NVIDIA desktop line-up of Maxwell GPUs and ran a second set of Linux OpenGL gaming tests on each of them while this time looking closely at the performance-per-dollar and performance-per-Watt performance. Here's the look at these NVIDIA Linux results if you're wanting to find the graphics processor delivering the best value as a Linux gamer.

http://www.phoronix.com/vr.php?view=22065

[darkbasic]

960 is the best one

[Michael]

Here's another discussion item I had been thinking about as well - https://github.com/phoronix-test-sui...uite/issues/77

[schmidtbag]

Makes me wonder how the results may scale if there were 2 960s in SLI

[GreatEmerald]

Yea, clearly GTX 960 wins here. Not that surprising, but good to have it confirmed that it's midrange, latest generation cards that take the cake. Though it would be even more interesting to see how radeonsi performs in a test like that; I'm thinking it would be even more geared towards middle/low-end.

[Michael]

Yea, clearly GTX 960 wins here. Not that surprising, but good to have it confirmed that it's midrange, latest generation cards that take the cake. Though it would be even more interesting to see how radeonsi performs in a test like that; I'm thinking it would be even more geared towards middle/low-end.

As mentioned in the article, a open-source perf-per- test will likely come in future... But then again probably won't be too interesting given I only have a handful of modern AMD cards and the un-relocked Nouveau cards will be a wash.

[Michael]

Makes me wonder how the results may scale if there were 2 960s in SLI

Likely not too interesting... I was thinking of buying a mid-range replica of one of my Maxwell cards to do some modern SLI Linux testing, but was told it would be a waste of money. Of course, that makes me sort of want to test it even more, but don't have the extra funds anyhow :/

[biggerfisch]

Why wasn't at 780 Ti included in this comparison? In recent other posts, it's punched far above it's age and given that it's now previous generation, I'd expect it to score highly on the per $ comparison.

[Mike Frett]

Why wasn't at 780 Ti included in this comparison? In recent other posts, it's punched far above it's age and given that it's now previous generation, I'd expect it to score highly on the per $ comparison.

It's been my experience that the older a GPU, the more expensive it becomes. To the point where buying a Modern GPU becomes less expensive.

[mitcoes]

Thanks a lot, again, this price/performance ad watts/ performance tests are a great deal and i have been asking for them here a lot.

Also a metatest to give to any tested hardware an overall score so twice the thanks.

My suggestion as you make the meta test is to give the 100 (or 1000) score to one popular device (As it can be a Mac Mini or even better for a Linux benchmarking web the most popular future Steam Machine, perhaps the Alien) and at the meta test, score benchmarking vs that known piece of hardware, that, of course can be changed as times goes by, perhaps each year being with the time the "Phoronix metascore 20xx result" a tag vendors would like to have at their web pages and product cases.

And this time 4k and 1080p (almost 2k) results are similar as all where Nvidia Maxwell GPUs, but may be in the future, a chip can be better in 4k price/performance and worse in 2k (1080p) price/performance or even better in 4k and worse at 2k fps (rare case but can happen between close cards from different brands), and a 2k meta score and a 4k meta score (as I think it will take long to be relevant the 4k gaming) will help a lot the next Steam Machines buyers and some Linux user gamer.