https://community.intel.com/t5/Intel-oneAPI-Math-Kernel-Library/MKL-on-SGI-Altix-hardware/m-p/974246#M16962 I think you will have to test your own application. Any generalization such as you request will be dead wrong. If your application depends on the library knowing about the various latencies on the Altix system, you may have difficulties. The SGI ads I've seen emphasized running OpenMP jobs on groups of 12 CPUs maximum, but they undersold the ability of Altix to take advantage of 32 or more CPU's in properly parallelized applications. It might be worth your while to look up posted Altix benchmarks which might be relevant to you. <DIV></DIV> Sun, 18 Jan 2004 13:32:47 GMT TimP 2004-01-18T13:32:47Z https://community.intel.com/t5/Intel-oneAPI-Math-Kernel-Library/MKL-on-SGI-Altix-hardware/m-p/974245#M16961 <DIV>Hello,</DIV> <DIV></DIV> <DIV>Can you tell me how effective MKL is on SGI Altix CC-NUMA machines when run on a large number of CPUs (up to 64)?</DIV> <DIV></DIV> <DIV>Thanks in advance</DIV> <DIV>Giga</DIV> Sat, 17 Jan 2004 20:16:43 GMT https://community.intel.com/t5/Intel-oneAPI-Math-Kernel-Library/MKL-on-SGI-Altix-hardware/m-p/974245#M16961 Intel_C_Intel 2004-01-17T20:16:43Z https://community.intel.com/t5/Intel-oneAPI-Math-Kernel-Library/MKL-on-SGI-Altix-hardware/m-p/974246#M16962 I think you will have to test your own application. Any generalization such as you request will be dead wrong. If your application depends on the library knowing about the various latencies on the Altix system, you may have difficulties. The SGI ads I've seen emphasized running OpenMP jobs on groups of 12 CPUs maximum, but they undersold the ability of Altix to take advantage of 32 or more CPU's in properly parallelized applications. It might be worth your while to look up posted Altix benchmarks which might be relevant to you. <DIV></DIV> Sun, 18 Jan 2004 13:32:47 GMT https://community.intel.com/t5/Intel-oneAPI-Math-Kernel-Library/MKL-on-SGI-Altix-hardware/m-p/974246#M16962 TimP 2004-01-18T13:32:47Z