Single page Print

Sizing 'em up
Take some clock speed information and some other numbers about per-clock capacity from the latest crop of high-end graphics cards, and you get this neat table:

  Peak pixel
fill rate
(Gpixels/s)
Peak
bilinear
filtering
int8/fp16
(Gtexels/s)
Peak
rasterization
rate
(Gtris/s)

Peak
shader
arithmetic
rate
(tflops)

Memory
bandwidth
(GB/s)
Asus R9 290X 67 185/92 4.2 5.9 346
Radeon R9 295 X2 130 358/179 8.1 11.3 640
Radeon R9 Fury X 67 269/134 4.2 8.6 512
GeForce GTX 780 Ti 37 223/223 4.6 5.3 336
Gigabyte GTX 980 85 170/170 5.3 5.4 224
GeForce GTX 980 Ti 95 189/189 6.5 6.1 336
GeForce Titan X 103 206/206 6.5 6.6 336
GeForce GTX 1080 111 277/277 6.9 8.9 320
GeForce GTX 1080 Ti 139 354/354 9.5 11.3 484
GeForce Titan X (Pascal) 147 343/343 9.2 11.0 480

We won't be testing every card in the table above, but we think it's useful to see how far we've come from some popular graphics cards of the past. The GTX 1080 Ti and the Titan X Pascal both offer more of, well, everything than any other card in this table save memory bandwidth. As we'll soon see, however, theoretical peaks alone don't tell the whole story.

To see how these theoretical numbers play out, we turn to our trusty Beyond3D test suite.

In the first of our synthetic Beyond3D tests, the GTX 1080 delivers what will soon become a familiar sight throughout this review. That 130 GPixels-per-second figure likely falls short of our theoretical peak because of the complex way that Nvidia's cuts to GP102's resources likely affect its real-world performance.

Thanks to Pascal's delta-color-compression mojo and the fact that GP102 is just a whole lotta chip, the GTX 1080 Ti excels at moving a whole lotta data with a 100% compressible black texture. Its incompressible-texture performance approaches the Fury X's compressed performance. Good grief.

Likely thanks to GPU Boost doing its thing, the 1080 Ti happily beats its theoretical peak texture-filtering rates. It also happily beats up on every other card here.

Here's a benchmark that's posed a long-running mystery for us: how do Nvidia's Maxwell-and-newer chips outstrip their theoretical polygon throughput? Well, we can finally explain what's going on. TR has long surmised that Nvidia's Maxwell cards have used a form of tiled rendering to achieve this counterintuitive result, and friend-of-the-site David Kanter proved it a while back. Nvidia admits as much now, and the technique explains the Pascal cards' performance here.

The GTX 1080 Ti continues to excel in these tests of pure number-crunching ability. It's hard to say too much about these numbers with my jaw on the floor. Let's see how these startling benchmarks translate into real-world performance for the GTX 1080 Ti now.