Re:Apple M2 vs. AMD Ryzen 7 Pro 6850U Performance in Nearly ...
by gbil on Friday August 19, 2022 @09:09AM
(#32520518)
at work I got a m1 pro with 16GB of RAM just because they offered the 14" managed device with 16GB only, while they offered 32GB on the 16" devices. When I complained about this, the response was "but is 16GB unified memory" as if I was disputing the performance of the RAM and how it is handled within the M1 chip while my issue was filling the 16GB. I'm happy that at least in HN people seem to be more aware on RAM sizing
[ Parent
| on HN ]
Re:Apple M2 vs. AMD Ryzen 7 Pro 6850U Performance in Nearly ...
by Damogran6 on Friday August 19, 2022 @09:58AM
(#32521005)
More ram = more performance if there's a cost in going across a bridge to keep the pipeline full. 16gb of _unified_ RAM that can be accessed by the CPU and GPU without a performance hit in going to the motherboard is a factor in some workloads. Whether you agree or not, it's an effective test comparing two architectures.
So the question is: are you spec benchmarking, or are you interested in the performance of the workload, because it's the actual work that matters, not how the bits are stored under the keyboard.
Re:Apple M2 vs. AMD Ryzen 7 Pro 6850U Performance in Nearly ...
by gbil on Friday August 19, 2022 @10:08AM
(#32521112)
I'm interested on not swapping which is a simple matter of RAM size, regardless if that is unified or not - which I believe I touched in my post
Re:Apple M2 vs. AMD Ryzen 7 Pro 6850U Performance in Nearly ...
by Damogran6 on Friday August 19, 2022 @10:30AM
(#32521425)
Only if your datasets, in aggregate, extend beyond what can be compressed into 16Gb...an what's the performance of the device you're swapping to? There's a GREAT BIG DIFFERENCE between an RLL 80 Gb drive and What you're swapping to on the motherboard today...
(In fact, you could consider the on-die RAM to be a really big Cache and the Flash storage to be REALLY REALLY big Non volatile addressable RAM.)