The GeForce GTX 1060 Founders Edition & ASUS Strix GTX 1060 Review
by Ryan Smith on August 5, 2016 2:00 PM ESTCompute
Shifting gears, let’s take a look at compute performance on GTX 1060.
As we already had the chance to categorize the Pascal architecture’s compute performance in our GTX 1080 review, there shouldn’t be any surprises here. But it will be interesting to see whether the GTX 1060’s higher ratio of memory bandwidth per FLOP materially impacts overall compute performance.
Starting us off for our look at compute is LuxMark3.1, the latest version of the official benchmark of LuxRender. LuxRender’s GPU-accelerated rendering mode is an OpenCL based ray tracer that forms a part of the larger LuxRender suite. Ray tracing has become a stronghold for GPUs in recent years as ray tracing maps well to GPU pipelines, allowing artists to render scenes much more quickly than with CPUs alone.
While GTX 1060 could hang with GTX 980 in gaming benchmarks, we don’t start off the same way with compute benchmarks, with the last-generation flagship holding about 17% ahead. Unfortunately for NVIDIA, this is about where GTX 1060 needed to be to best RX 480; instead it ends up trailing the AMD competition. Otherwise the performance gain versus the GTX 960 stands at 65%.
For our second set of compute benchmarks we have CompuBench 1.5, the successor to CLBenchmark. CompuBench offers a wide array of different practical compute workloads, and we’ve decided to focus on face detection, optical flow modeling, and particle simulations.
Like with GTX 1080, relative performance is all over the place. GTX 1060 wins with face detection, loses at optical flow, and wins again at particle simulation. Even the gains versus GTX 960 are a bit more uneven, though at the end of the day GTX 1060 ends up being significantly faster than its predecessor with all 3 sub-benchmarks.
Moving on, our 3rd compute benchmark is the next generation release of FAHBench, the official Folding @ Home benchmark. Folding @ Home is the popular Stanford-backed research and distributed computing initiative that has work distributed to millions of volunteer computers over the internet, each of which is responsible for a tiny slice of a protein folding simulation. FAHBench can test both single precision and double precision floating point performance, with single precision being the most useful metric for most consumer cards due to their low double precision performance. Each precision has two modes, explicit and implicit, the difference being whether water atoms are included in the simulation, which adds quite a bit of work and overhead. This is another OpenCL test, utilizing the OpenCL path for FAHCore 21.
Finally, in Folding@Home, we see the usual split between single precision and double precision performance. GTX 1060 is solidly in the lead when using FP32, but NVIDIA’s poor FP64 rate means that if double precision is needed, RX 480 will pull ahead.
Facebook
Twitter
RSS
189 Comments
View All Comments
anandreader106 - Friday, August 5, 2016 - link
First thought: Still no Doom benchmarks being factored in?Ryan,
You are my favorite GPU reviewer. Period. However I do think I need clarity on your Final Words.
It's my opinion that DirectX 11 performance is "good enough" from Nvidia and AMD thus far in this new generation. So I'm left wondering, why aren't you going more in-depth with DirectX 12 and Vulcan titles/performance? Wouldn't that give us the best indication of what to expect going forward?
cknobman - Friday, August 5, 2016 - link
The best indication you will get is that when reviewing Nvidia cards none of these things will be addressed?Why, because Nvidia is not doing so hot at them and it would not make their cards look better than AMD's.
Look at the other 1060 benchmarks and comparisons and you will see that:
A. Nvidia is behind on dx12 and the 480 => 1060
B. @1080p the 1060 is overkill and a $200 480 4gb (or even a $180 470) is all you need
C. Because of Nvidia's "founders edition" price gouge model most 3rd parties are trying to get away with charging more than $250. Reality is most 1060's are >= $270 which makes the AMD 480 the better buy.
StrangerGuy - Friday, August 5, 2016 - link
It's funny the AMD fanboys always harp about the evil $300 1060 and never mentions how their favorite $200 480 is essentially vaporware and 8GB versions price gouged to death.Ryan Smith - Friday, August 5, 2016 - link
"It's my opinion that DirectX 11 performance is "good enough" from Nvidia and AMD thus far in this new generation. So I'm left wondering, why aren't you going more in-depth with DirectX 12 and Vulcan titles/performance? Wouldn't that give us the best indication of what to expect going forward?"The benchmark suite only gets updated periodically. It's a lot of effort to design and validate a testing sequence, and then run (and possibly re-run) 30 some-odd cards through it. So adding games has the net effect of slowing things down even further.
At this point we're updating the testbed to Broadwell next month, at which point we'll refresh the games list as necessary.
Though I will note that there's a reason we run so many (9) games: one game is too small of a sample size. Right now Doom is the only Vulkan game on the market,* so while it's a very interesting first look at Vulkan, it's not something that's going to be representative of Vulkan as a whole.
* We'll ignore DOTA 2 since it's not meaningfully GPU limited on these fast cards
CHADBOGA - Friday, August 5, 2016 - link
Doom is one of those few games out there that will inspire people to go one way or the other and should be included in your benchmark suite.Scali - Saturday, August 6, 2016 - link
Aside from that, the Vulkan implementation in DOOM is not yet complete.As you can read in the DOOM FAQ, they use AMD shader intrinsics extensions, but no equivalent for nVidia. Likewise, on AMD hardware, async compute is enabled, on nVidia it is not yet. The FAQ says they're still working on optimizing the code with nVidia.
While it may be interesting to benchmark DOOM's Vulkan implementation to get an idea of where we currently stand, I don't think it is mature enough at this point to say anything about performance in Vulkan games in general, or how AMD and nVidia stack up, since you're comparing apples to oranges at this point.
rhysiam - Friday, August 5, 2016 - link
I too am curious as to why the whole DX11 vs 12 comparison wasn't even raised. DX12 does not appear once in the conclusion page. The 1060 is the better DX11 card, no question. It's early days for DX12, but what we're seeing so far is enough to suggest things may well be quite different. The three DX12 titles in the review (Hitman, RoTR & AoS) are the three strongest games for the 480 by far. Add Doom via Vulkan into the mix and you have 4 NextGen API titles that put the 480 at or above 1060 performance. Of course we can't make hard and fast recommendations based on a few titles like this, but surely it's worth mentioning at least, if not exploring in detail?This might be a minor point except for the fact that you dismiss the 4GB 480 based on speculation/extrapolation that its VRAM won't be enough to keep it competitive future demanding titles. Surely those demanding titles will increasingly be (or at least offer) DX12 though? So if you're advocating a 1060 over the 480 4GB based on longevity and future performance, the DX12 question has to be raised doesn't it?
rj030485 - Friday, August 5, 2016 - link
Think Ryan needs to work on his math. He says the 1060 is 17% faster than the 480 in GTA V when the difference more like 30%.Ryan Smith - Friday, August 5, 2016 - link
Oh geeze. This is what happens when you read the wrong column in a spreadsheet. Thanks!onemoar@gmail.com - Friday, August 5, 2016 - link
I don't know why anands witcher 3 scores are so lowI am pushing 80FPS in places with everything turned up to ultra and post effects on with no hairworks