https://community.intel.com/t5/Software-Tuning-Performance/Measuring-theoretical-flops-for-icelake/m-p/1351285#M8006 <P>The biggest problem with computing "peak" performance for recent processors is knowing what value to use for the frequency.</P> <P>&nbsp;</P> <P>The nominal frequency on the Xeon Platinum 8358 is 2.6 GHz. &nbsp;When running AVX512 code (required to get 32 FLOPS/cycle/core), the base frequency is 1.9 GHz and the maximum all-core Turbo frequency is 2.9 GHz. &nbsp;The actual frequency seen when running a "peak FLOPS" sort of benchmark will depend on the leakage current of the particular chip and the effectiveness of the cooling system.</P> <P>&nbsp;</P> <P>32 FLOPs/cycle/core * 32 cores * 1.9 GHz * 2 sockets = 3891.2 GFLOPS</P> <P>32 FLOPS/cycle/core * 32 cores * 2.9 GHz * 2 sockets = 5939.2 GFLOPS</P> <P>&nbsp;</P> <P>Based on experience with SKX and CLX processors, I expect you would see average frequencies for the HPL benchmark in the range of 2.2 GHz to 2.6 GHz across an ensemble of systems with this configuration. &nbsp;HPL performance will be in the neighborhood of 90% of peak based on the actual average frequency during the run. &nbsp;</P> <P>&nbsp;</P> <P>The Xeon Gold 6338 will use the same procedure. &nbsp;With SKX and CLX you had to check to see whether the processor had 1 or 2 AVX512 FMA units. &nbsp;It looks like for ICX all of the processors have 2, so 32 FLOPS/cycle/core should work for all models.</P> Thu, 13 Jan 2022 21:12:04 GMT McCalpinJohn 2022-01-13T21:12:04Z https://community.intel.com/t5/Software-Tuning-Performance/Measuring-theoretical-flops-for-icelake/m-p/1351110#M8004 <P>Hi,<BR />I have few servers each equipped with with dual icelake 8358&nbsp;processors.<BR />I would like to know that the following is correct method to measure theoretical Double Precision flops&nbsp; (RMax) -</P> <P>=&nbsp; cores/socket * sockets * frequency * ops/cycle * elements/ops * vector registers per core</P> <P>= 32 * 2 * 2.6 * 2 * ( 512 register size / 64 bits DP ) * 2</P> <P>=&nbsp;32 * 2 * 2.6 * 2 * 8 * 2</P> <P>= 2662.4 * 2<BR />= 5324.8</P> <P><BR />Also, with there be any difference apart from frequency and cores/socket if i try to calculate FLOPS for 6338 CPU Model?</P> Thu, 13 Jan 2022 12:11:03 GMT https://community.intel.com/t5/Software-Tuning-Performance/Measuring-theoretical-flops-for-icelake/m-p/1351110#M8004 psing51 2022-01-13T12:11:03Z https://community.intel.com/t5/Software-Tuning-Performance/Measuring-theoretical-flops-for-icelake/m-p/1351285#M8006 <P>The biggest problem with computing "peak" performance for recent processors is knowing what value to use for the frequency.</P> <P>&nbsp;</P> <P>The nominal frequency on the Xeon Platinum 8358 is 2.6 GHz. &nbsp;When running AVX512 code (required to get 32 FLOPS/cycle/core), the base frequency is 1.9 GHz and the maximum all-core Turbo frequency is 2.9 GHz. &nbsp;The actual frequency seen when running a "peak FLOPS" sort of benchmark will depend on the leakage current of the particular chip and the effectiveness of the cooling system.</P> <P>&nbsp;</P> <P>32 FLOPs/cycle/core * 32 cores * 1.9 GHz * 2 sockets = 3891.2 GFLOPS</P> <P>32 FLOPS/cycle/core * 32 cores * 2.9 GHz * 2 sockets = 5939.2 GFLOPS</P> <P>&nbsp;</P> <P>Based on experience with SKX and CLX processors, I expect you would see average frequencies for the HPL benchmark in the range of 2.2 GHz to 2.6 GHz across an ensemble of systems with this configuration. &nbsp;HPL performance will be in the neighborhood of 90% of peak based on the actual average frequency during the run. &nbsp;</P> <P>&nbsp;</P> <P>The Xeon Gold 6338 will use the same procedure. &nbsp;With SKX and CLX you had to check to see whether the processor had 1 or 2 AVX512 FMA units. &nbsp;It looks like for ICX all of the processors have 2, so 32 FLOPS/cycle/core should work for all models.</P> Thu, 13 Jan 2022 21:12:04 GMT https://community.intel.com/t5/Software-Tuning-Performance/Measuring-theoretical-flops-for-icelake/m-p/1351285#M8006 McCalpinJohn 2022-01-13T21:12:04Z