>You heard it, right folks, AMD’s 2019 CPU family is designed to tackle the Intel 10nm Ice Lake Xeons favorably and things are looking really good for AMD as their Rome CPU family will only be competing against Intel’s 14nm++ server refreshed family, aka Cascade Lake-SP.
More evidence that AMD may soon pull ahead in the performance race as they take good advantage of Intel's 10nm stumble.
We already saw that the Threadripper 2990 is starved for bandwidth between Zen packages due to the infinity fabric design.
So you get incredible cinebench performance (and some other 'embarrassingly parallel' workloads), but it doesn't scale to other kinds of programs that have more interthread/interprocess communication.
I guess it remains to be seen what happens with EPYC, but my initial guess is that it will have a similar infinity fabric design to hit 64 cores/128 hyperthreads.
BTW has anyone done an R mclapply benchmark on Threadripper 2990? I was set to buy one until I saw the uneven benchmark scores.
Yes, infinity fabric has limitations.
Threadripper suffers in memory intensive workloads because only half of the CCXs have a dedicated memory controller thus some cores need to talk to another CCX to access memory.
On EPYC each CCX has its own memory controller so it's less of an issue.
Twice the number of cores compared to first generation EPYC with its main competitor misfiring. So do they go for profit and double the price or go for market share and keep it the same? As both a consumer and a stock holder, I'm hoping for closer to the latter.
Anyone here familiar with overall performance / server costs to give an insight if AMD might actually become competitive considering with all the hype around AMD these days, I have still yet to see any of the 3 major providers offer an AMD chip?
More evidence that AMD may soon pull ahead in the performance race as they take good advantage of Intel's 10nm stumble.