AMD RX 480 Compute-Focused Benchmarks
No review article would be complete without including some much needed general purpose compute benchmarks to compare just how much improvement has been made in compute. They have full operational FP16 capability which should translate into far better DNN capability. Here we’re testing Caffe, which isn’t quite FP16 yet, though is still a very viable test of compute capability. But just how much better is this going to be in comparison to it predecessors, or even with its rivals. Caffe DNN can easily be run by anyone on a *nix distribution. Simply download the applicable binaries, compile it with either cuDNN or use AMD’s open source OpenCL version. You can then test the benchmark that’s inherent in the Caffe framework.
Realistically we see a very large jump in efficiency with Polaris 10. This workload is easily handled and even bests the heavily compute focused 390X. Hawaii and Tahiti are still among some of the best performers in some workloads precisely because of how much raw double-precision throughput they have. But this is not that, and it the RX 480 does exceptionally well. It makes us wonder just how the larger variant, Vega 10, would do in such a test. Next up we’ll test the familiar, and updated, Folding@Home benchmark which is a test of single and double precision work.
It’s no surprise to see the Titan X in the lead in single precision. What’s surprising, and in fact quite fascinating, is that the RX 480 is able to do so well despite having much less raw throughput. Double precision, however, is another story. It was able to do much better than anticipated, beating the Titan X and coming very close to the Fury X. Not bad considering the price and segment this card is being marketed towards. It may be a very efficient wolf in sheep’s clothing for compute.
Next we’ll test what we’ve all been waiting for; games.