Is there any definitive answer about what is actually different about these tailored Xeon chips in iMac Pros and why they are underclocked compared to stock Xeons?
I remember during initial press coverage during iMac Pro launch in December, I heard someone quoting Apple that these custom-made Xeon chips are optimised for AVX-512 performance so that they can perform AVX-512 optimised code without applying negative AVX offset. Unlike the standard chips that are optimised for general instruction performance, but have to throttle down when too dense AVX-512 code is executed.
However, I can't source this information anymore. I was wondering if someone has more information about this.
I remember during initial press coverage during iMac Pro launch in December, I heard someone quoting Apple that these custom-made Xeon chips are optimised for AVX-512 performance so that they can perform AVX-512 optimised code without applying negative AVX offset. Unlike the standard chips that are optimised for general instruction performance, but have to throttle down when too dense AVX-512 code is executed.
However, I can't source this information anymore. I was wondering if someone has more information about this.