6 ms·

What is the maximum performance % difference between optimizing for perf/$ and perf/watt? Sure, there are wafer scale chips now, but the TDP for a phone is stil

by ece 4y ago

What is the maximum performance % difference between optimizing for perf/$ and perf/watt? Sure, there are wafer scale chips now, but the TDP for a phone is still ~5W watts, average laptops have gone from ~15 to ~30W, and desktops from ~300-600W+. I suppose with Zen 4, there might actually be an apples to apples comparison baring ISA and uncore differences. If ADL is anything to go by I imagine performance will be within ~15% of each other, but with a ~30% price difference if you care about a more efficient and cooler running chip. Sure the efficiency gains add up, but so do the performance gains on the other side.

wmf 4y ago
What is the maximum performance % difference between optimizing for perf/$ and perf/watt? Alder Lake and M1 Pro are good demonstrations of those two approaches.
- ece 4y ago
  For a crude comparison, the new framework with Alder Lake is ~20-30% less than the 14" MBPs with M1 Pro. If the uncore becomes more important, maybe one approach will take precedence. It's too bad Apple Silicon isn't as well documented (ie. AMX) as Intel/AMD chips are.
tlb 4y ago
It can be a lot. Speculation requires executing operations before you're sure they'll be needed, which can double or triple power draw in order to increase instruction-level parallelism. And all the machinery needed to enable speculation, like branch predictors, draws power too. This graph shows a factor of 100 between the highest-performing and most-efficient systems: https://en.wikipedia.org/wiki/Performance_per_watt#Examples https://en.wikipedia.org/wiki/Performance_per_watt#Examples
- ece 4y ago
  For general purpose compute relying more on ILP than DLP, I would think the range is going to be narrower. Narrow enough to be a meaningful choice for consumers using battery powered devices and wanting more efficient or higher performance laptops and desktops. For servers, it seems like Intel and AMD are choosing optimizing more for perf/$ over perf/watt in their roadmaps. HPC roadmaps might be different still.

by ece 4y ago

wmf 4y ago

What is the maximum performance % difference between optimizing for perf/$ and perf/watt? Alder Lake and M1 Pro are good demonstrations of those two approaches.

ece 4y ago
For a crude comparison, the new framework with Alder Lake is ~20-30% less than the 14" MBPs with M1 Pro. If the uncore becomes more important, maybe one approach will take precedence. It's too bad Apple Silicon isn't as well documented (ie. AMX) as Intel/AMD chips are.

tlb 4y ago

It can be a lot. Speculation requires executing operations before you're sure they'll be needed, which can double or triple power draw in order to increase instruction-level parallelism. And all the machinery needed to enable speculation, like branch predictors, draws power too. This graph shows a factor of 100 between the highest-performing and most-efficient systems: https://en.wikipedia.org/wiki/Performance_per_watt#Examples https://en.wikipedia.org/wiki/Performance_per_watt#Examples

ece 4y ago
For general purpose compute relying more on ILP than DLP, I would think the range is going to be narrower. Narrow enough to be a meaningful choice for consumers using battery powered devices and wanting more efficient or higher performance laptops and desktops. For servers, it seems like Intel and AMD are choosing optimizing more for perf/$ over perf/watt in their roadmaps. HPC roadmaps might be different still.