XFX RX 460 Slim General Benchmarks
For our general benchmarks that etst the compute capability of the card we’ve depreciated Caffe due to a lack of current OpenCL support. When MlOpen is released along with Vega in the future, we’ll reintroduce a DNN-based benchmark that utilizes a framework that can utilize both NVIDIA and AMD with their respective platforms. Until then we have a few very good compute benchmarks that are relevant and quite fun to compare.
FAHBench
Monte Carlo
In terms of overall compute capability, it’s definitely not the quickest in the shed, but it does provide a good value for the performance it can actually provide. Even more, FP16 math is fully supported and on with an equal amount of throughput as single-precision math. It’s interesting that the double precision throughput of the RX 460 is nearly on par with the GTX 980, even though Polaris is most certainly not a powerhouse by any means. The real value comes in the ability to place many of these mini-GPUs together for a minimal amount of money. Looking for an inexpensive compute rig and want to put the maximum amount of GPUs in that can run for a long time and do so without reaching enormous temperatures? The RX 460 Slim from XFX would be an excellent choice for those wanting to fill out distributed compute PCs, with a good efficiency. GCN does very well on projects that support OpenCL.
Now of course, the most important question remains to be answered.