Am I right, that first quarter of next year seems like a good time to build a new PC? Intel introduces new CPU architecture (Kaby Lake), AMD introduces new GPU architecture (Vega), and new CPU architecture (Zen).
the generation after that will be faster
Neither Zen nor Vega will be out in Q1 2017.
Try Q4
Kaby is a very small improvement over Skylake.
>AMD
lol
>>57370613
Both are supposed to launch early 2017.
>>57370618
AMD started this year with low-mid end GPU market, and Vega should introduce a high-end one (with 4096 ALUs, HBM2 memory and 12 theoretical TFLOPS).
>>57370715
how do they measure the TFLOPS? Just running OpenCL benchmarks?
Also what would be better for OpenCL computation? nVidia or AMD?
>>57371172
>how do they measure the TFLOPS? Just running OpenCL benchmarks?
Theoretical FLOPS is a bullshit figure GPU manufacturers use, basically it's amount of FLOPS you could achieve if all ALUs were constantly doing FMA.
>Also what would be better for OpenCL computation? nVidia or AMD?
From what I've found so far, AMD is a clear winner here.
>>57371202
why would AMD come out above nVidia? Its seemed to me that nVidia has been more worked into aspects of Computerland (namely CUDA Acceleration)
Im getting a 780TI for $185 to hopefully pit against my R9 290X in order to compare not only the gaming performance but also the computational power (HASH and otherwise)
What could you recommend as a good OpenCL program to run? Or where to learn how to program in OpenCL
>>57371256
Nvidia was a pioneer in GPGPU and CUDA existed before OpenCL, but nowadays and since a long time, AMD beats NVIDIA at GPGPU, at least when it comes to consumer hardware (so not NVIDIA TESLA line).
>What could you recommend as a good OpenCL program to run?
Depends what you want to do?
E.g. with https://github.com/meriken/merikens-tripcode-engine-v3 you can mine a nice trip.
>Or where to learn how to program in OpenCL
I have no idea, perhaps https://developer.nvidia.com/opencl and http://developer.amd.com/tools-and-sdks/opencl-zone/