The M1 GPU is about as powerful as a 1050TI. If the M1 Max is 4 times as powerful, that would be about equivalent to a RTX 3070, not 3080.

According to UserBenchmark, the RTX 3070 is about 394% faster than a 1050TI. And the M1 benchmarks about that level.

Note: these are desktop GPUs.

Quoteshockingly bad reporting, ignoring key assumptions and acting as if guestimates are actual benchmarks.
until to have data to show us, make sure each claims has some language like "possibly" or other qualifiers.

Agreed

Keep drinking the kool-aid.

So while Anandtech quotes Apple's figures (which are exactly the same as I estimated), they don't offer any reason to dispute those figures. Obviously testing the chips will determine the veracity of those claims.

As for the M1 GPU real-world gaming benchmarks on games like Borderlands 3, Total War: Three Kingdoms, Tomb Raider and PUBG — these are all coded for Intel chips and being run in Apple's Rosetta 2 translation layer — they are in no way reflective of games coded natively for M1 on Mac as they aren't any AAA titles ported natively for the platform yet that I am aware of (Minecraft notwithstanding).

As for Apple's performance claims, sure, pinch of salt. But rarely, if ever, does Apple overstate its claims. As an example, recently it claimed the A15 is 50 percent faster than the competition (i.e Snapdragon 888). Turns out testing showed it is actually 62 percent faster than the S888. Given the performance of the M1, I would be surprised if Apple's performance claims are overstated.

So...once again Apple releases a product roughly 2 years behind current technology for the sheeple to line up for.

They're just looking at Tflops, which are only a paper calculation of ALUs * clock speed * 2 operations per ALU per clock. I think Apple groups 128 ALUs into each core. You can work this out mathematically on any GPU and see it's the case.

This isn't to downplay it - I'm confident the M1 GPU is MORE efficient per Tflop on paper than RDNA2, and so should the M1 Max be, so it might end up more performant than that.

This is akin to spreading the decades old MHz myth that clock speed is a CPU performance number.

shockingly bad reporting, ignoring key assumptions and acting as if guestimates are actual benchmarks.
until to have data to show us, make sure each claims has some language like "possibly" or other qualifiers. manufacturers have a long history of claiming their product to be "better" than another, but it turns out it's only on a fringe benchmark and not for the things people actually do.

M1 is on par GTX 1050 desktop, M1 pro is like GTX 1060 6GB
M1 Max will be around GTX 1660ti perf, the bandwith is limited by LPDDR5 around 200GB/s even is 52000MT/s or higher 4000GB/s. maybe is more near to GTX 1080, but RTX 2080 never, PS5 APU is trash, old 1950X with 24 vega CU's @180W with good SSD is near to GTX 1660Ti aka GTX 1070.

Always better than AMD APU trash

The fact is, Anandtech is going from claims themselves, nothing to back up their claims other than apples claims. Notebook check have an array of benchmark results to look at. If Apples 4x claim turned out to be true, why don't you go and check this websites benchmarks across all the games tested on the M1 using the highest visual settings (this will take any potential CPU bottleneck out of the equation) and then multiply the results by 4 and you will find it still falls substantially short compared to the 3070 laptop GPU let alone the 3080. When the hardware is available for testing and you show us it competing in the real world, then I will believe you, my bet is, it will gain just over 3x the performance in those games tested compared to the M1

Thanks RobH - yes. Fixed.

Do you mean " based on the assumption that Apple has NOT clocked the new GPU cores higher in the M1 Pro and M1 Max chips or made any other changes. " ?

This Anandtech article since published backs up everything I had to say regarding the figures and performance written in this article: https://www.anandtech.com/show/17019/apple-announced-m1-pro-m1-max-giant-new-socs-with-allout-performance

Does it mean the cheapest base model with 14 core GPU is as good as GTX 1060 6GB card ? (which I think is at 4.4 Teraflops)

Quote from: Steve Smith on October 19, 2021, 09:36:21
"Apple's M1 Max GPU is as powerful as an Nvidia RTX 2080 desktop GPU"

🤣🤣🤣🤣🤣🤣 This site just gets worse... So according to this guy, apples mobile GPU is now faster than nvidia best efforts in the mobile space, the RTX 3080 Mobile, as the 2080 desktop is slightly faster than the 3080 mobile. 1st of all, performance doesn't scale linearly, so adding 4x the amount of GPU cores doesn't necessarily mean you're getting 4x the GPU power, even if if you did, 4x would still fall short of the 3080 let alone the 2080 desktop, by this sites own benchmarks.

I'd love to know what happened to this website, it used to be a place for reliable tech info, now it's turned into a clickbait site, I suggest you change the authors.

I, too, am skeptical of Apple's claims, but I'm yet to see any responses laying out any concrete details as to how this analysis is wrong.

It would be good if more knowledgeable people could share this.

"Apple's M1 Max GPU is as powerful as an Nvidia RTX 2080 desktop GPU"

🤣🤣🤣🤣🤣🤣 This site just gets worse... So according to this guy, apples mobile GPU is now faster than nvidia best efforts in the mobile space, the RTX 3080 Mobile, as the 2080 desktop is slightly faster than the 3080 mobile. 1st of all, performance doesn't scale linearly, so adding 4x the amount of GPU cores doesn't necessarily mean you're getting 4x the GPU power, even if if you did, 4x would still fall short of the 3080 let alone the 2080 desktop, by this sites own benchmarks.

I'd love to know what happened to this website, it used to be a place for reliable tech info, now it's turned into a clickbait site, I suggest you change the authors.