+- hashcat Forum (https://hashcat.net/forum)
+-- Forum: Misc (https://hashcat.net/forum/forum-15.html)
+--- Forum: Hardware (https://hashcat.net/forum/forum-13.html)
+--- Thread: Ryzen OpenCL benchmarks? (/thread-6357.html)
RE: Ryzen OpenCL benchmarks? - epixoip - 05-11-2017
Password cracking is an embarrassingly-parallel problem and acceleration comes from parallelism. AVX2 with 256-bit registers enables us to operate on eight 32-bit words or four 64-bit words simultaneously, versus only four 32-bit words or two 64-bit words with SSE2. For most all hash algorithms, a lower-clocked CPU with AVX2 will run circles around a higher-clocked CPU with only SSE2. So yes, AVX2 is critical for a modern password hashing rig, especially now that high-end Intel CPUs can add an additional e.g. 1 GH/s to NTLM.
RE: Ryzen OpenCL benchmarks? - NikosD - 05-13-2017
For mainstream desktop CPUs Intel offers 4C/8T and AMD offers 8C/16T
For HEDT (High end desktop) Intel offers 12C/24T and AMD offers 16C/32T
Leveraging AVX2 is far more difficult than multi-core.
That's why more cores are better.
RE: Ryzen OpenCL benchmarks? - epixoip - 05-13-2017
8 threads * 8-way SIMD = 64 32-bit ops in parallel
16 threads * 4-way SIMD = 64 32-bit ops in parallel
32 threads * 4-way SIMD = 128 32-bit ops in parallel
24 threads * 8-way SIMD = 192 32-bit ops in parallel
That's why 256-bit registers are better than more cores. And leveraging AVX2 is trivial, the OpenCL compiler does it for you even.
RE: Ryzen OpenCL benchmarks? - NikosD - 05-13-2017
Not true.
Embarrassingly parallel integer workloads like video decoding/encoding haven't managed to leverage AVX2 that much even using sophisticated hand-written ASM e.g x264 manages a 5% speed-up using AVX2 and x265 ~20% using medium preset and 4K.
If you manage to find out an embarrassingly parallel Integer workload that using AVX2 a 4C/8T Intel Kabylake has the same performance with a 8C/16 AMD RyZen CPU at the same clock speed, feel free to mention it.
Or I can save you from the trouble and tell you that there is no such thing.
RyZen 8C/16T is always faster in embarrassingly parallel workloads at the same clock speed compared to Intel 4C/8T even using the most AVX2 optimized code.
RE: Ryzen OpenCL benchmarks? - undeath - 05-14-2017
You do realize this is the hashcat forum and epixoip is talking about hashcat kernels, right? Nobody here cares about audio/video performance.
RE: Ryzen OpenCL benchmarks? - epixoip - 05-14-2017
Yep, you can argue other use cases all you want, it's just going to fall on deaf ears. These are the hashcat forums, all we give a fuck about here is password cracking performance, and Ryzen's fake AVX2 has no place here.
RE: Ryzen OpenCL benchmarks? - NikosD - 05-14-2017
OK.
Do you have benchmarks of hashcat and Core i7 7700K vs RyZen 7 at the same clock speed ?
Or even not the same clock speed, we all can do the math.
It would be useful to see various algos performing on those CPUs