Hacker News Reader

Comment by dharma1 | Hacker News Reader

Nvidia X1 already is 1 tflops GPU and the X2 (which is likely to be in the nintendo) is going to be around 1.5 tflops. Those are mobile SoC's with CUDA cores, fit OK in tablets (Pixel C has one), X1 peak power consumption is about 10 watts.

Xbox one GPU is 1.3 tflops and PS4 is 1.8 tflops. So in the same ballpark

View on HN

1. tflops is pretty meaningless, especially between different GPU architectures.

2. Both Xbox and PS4 consume around 100W. Unless the Switch has 10x the perf/watt, it seems like performance won't be comparable.

3. Both Xbox and PS4 have massive heat sinks and fans, so they'll be able to sustain higher perf for longer.

The X1/X2 are big.LITTLE setups. Between that and various other leaks, it seems that the Switch is designed to run on the low-power LITTLE CPU/GPU in mobile mode and on the beefier Watt-burning big cores when docked.

Not clear on how they're cooling it, though

That 1 TFLOPS on the Tegra is for 16-bit floats I believe while the TFLOPS on the Xbox One and PS4 are for 32-bit floats.

That is a good catch, double speed FP16, so FP32 perf should be half that on the X1. While useful for some things (like neural network inference and I guess some shaders) it's not directly comparable - NVidia style marketing

This was the same deal with how CPUs were (mis)marketed for years. Higher clockspeeds don't mean anything unless the pipelines are identical. A much smarter chip with larger cache sizes, branch prediction, SIMT, ultra-wide registers, better instruction caching, etc might have half the clockspeed of a dumber chip and outperform it by a huge margin. GPUs are no different in that texture caches, vertex buffer sizes, warp sizes, etc all play a big part and you really can't compare flops unless comparing two GPUs of the same product line.

Reply on news.ycombinator.com