3 ms·
Given that the (serial) compute task (only CPU/RAM) that I started before I went shopping finished in 4073 seconds on my iMac Pro, and in 2648 seconds on my Mac
by auggierose 6y ago
Given that the (serial) compute task (only CPU/RAM) that I started before I went shopping finished in 4073 seconds on my iMac Pro, and in 2648 seconds on my Mac mini, the conclusion is not that weird (even if it is not much of a conclusion).
- klelatti 6y agoInteresting. What are the iMac Pro specs?
- auggierose 6y agoIt's the 18-core one.
- klelatti 6y agoThanks - and wow!
- em500 6y agoDepending on the workload, that's not as impressive as it sounds at first. Single threaded M1 performance is around 50-60% better than the 18-core Xeon in the iMac (which is impressive). But high-core CPUs are typically slower per core. I have plenty of programs that run faster on my dual core laptop than on the 16 and 24 core Xeon servers at work.
- klelatti 6y agoFair comment - I guess I'd assumed that someone with an 18 core Xeon would be running multithreaded workloads!
- rfoo 6y agoOP explicitly said serial program tho.
- ben-schaaf 6y ago> But high-core CPUs are typically slower per core. While this is true for intel CPUs, this hasn't been the case for AMD Threadripper. The 2990wx has marginally higher single core performance than the 2700x (https://www.cpu-monkey.com/en/compare_cpu-amd_ryzen_threadripper_2990wx-886-vs-amd_ryzen_7_2700x-876 https://www.cpu-monkey.com/en/compare_cpu-amd_ryzen_threadri...), while the 3990x has only slightly lower single core performance than the 3700x (https://www.cpu-monkey.com/en/compare_cpu-amd_ryzen_threadripper_3990x-977-vs-amd_ryzen_9_3950x-932 https://www.cpu-monkey.com/en/compare_cpu-amd_ryzen_threadri...).
- bsagdiyev 6y agoI think that’s the “typically” portion, and the comparison was against Intel anyways which does seem to follow that.