>In applications that actually use Metal natively, Apple Silicon GPUs do compare favorably to Nvidia laptop GPUs while using drastically less power.
Is that really the case?
I'm not being facetious here, I'd really like to see more useful datapoints.
There aren't many benchmark comparisons out there that strive for actual useful comparison, especially outside synthetic stuff with questionable applicability like 3dMark.
And for the "drastically less power" claim...it really doesn't help that most benchmarks are with decked-out "Gamer" machines using the highest available TDP configuration, despite most GPUs having their sweet spot significantly below - especially Ada Lovelace seem to scale down really well (from what I've gathered, still 60-70% performance at 60W compared to 150W with 4080 Mobile, for example).
>Even if you look at the raw technical specs of the M2 Max GPU, it's comparable to Nvidia laptop GPUs - with the exception of ray tracing.
The specs put it roughly between GA106-GA104/AD107-AD106 respectively, and I'd expect it to land there in the general, adequately optimized case.
> Is that really the case? I'm not being facetious here, I'd really like to see more useful datapoints. There aren't many benchmark comparisons out there that strive for actual useful comparison, especially outside synthetic stuff with questionable applicability like 3dMark.
Yeah, on synthetic benchmarks a higher-end MacBook Pro GPU compares favorably with recent Nvidia laptop cards (say, 4070 or so). But in games that drops off dramatically ... more like a 3050 or 3060 at best.
Is that really the case? I'm not being facetious here, I'd really like to see more useful datapoints. There aren't many benchmark comparisons out there that strive for actual useful comparison, especially outside synthetic stuff with questionable applicability like 3dMark.
And for the "drastically less power" claim...it really doesn't help that most benchmarks are with decked-out "Gamer" machines using the highest available TDP configuration, despite most GPUs having their sweet spot significantly below - especially Ada Lovelace seem to scale down really well (from what I've gathered, still 60-70% performance at 60W compared to 150W with 4080 Mobile, for example).
>Even if you look at the raw technical specs of the M2 Max GPU, it's comparable to Nvidia laptop GPUs - with the exception of ray tracing.
The specs put it roughly between GA106-GA104/AD107-AD106 respectively, and I'd expect it to land there in the general, adequately optimized case.