SicTransitGloriaMundi
macrumors regular
that's a lot of big words, coming from the user who claimed it cannot be an m1max, "because 24 cores".
MacBook Pro's M1 Max GPU is Over 3x Faster Than M1 in First Metal Benchmark
- Thread starter MacRumors
- Start date
- Sort by reaction score
Weird since I took that number from Geekbench as well. How can two scores differ that much?
Anyway, if that's true then M1 Max 32 Core cannot have these numbers, or Geekbench is not measuring it correctly. Because I don't believe for one second that M1 Max 32 is slower than 5700XT. It will be substantially faster if we go by the comparisons on the M1 Max page.
Last edited:
One thing this benchmark article settles is whether the M1 Max is based on the A14 or the A15 design. The A14 single core benchmark was only around 1500’ish while the A15 was in the mid-1700’s. With the M1 Max clocking in at 1742, this tells us that the cores are A15 cores, not A14, which the original M1 is based on. The number following the M means generation of Apple Silicon Mac, not generation of processor, apparently, and is not due to which A-series processor it shares cores with. The old assumption was that M2 would refer to an A15 design, but that won’t necessarily be true. Likely a MacBook Air in 2022 using an M2 would be called M2 because it is a second generation AS Air and will probably still be A15 based like the M1 Max. But a MBP in late 2022 would use M2 Max despite being based on the A16.
I think in this case Tflops is misleading since 3000 series have double the Tflops of previous generation but the Laptop Apple chose in their graphs clearly has 3080M and Apple claims the GPU matches that one with 100Watt less power.
Nope. The M1 CPU core is already around 1700. The M1 Pro/Max cores are the exact same cores as those in the M1.
Probably it’s more irrational to make wild assumptions about who has what machine instead of looking at the available data.
PsykX
macrumors 68040
It's not this linear though, but I still suspect this score is attributable to the 24-core GPU.
BenRacicot
macrumors member
Yup, that’s why it’s an “M1” series naming convention.
They gave us LPDDR5 RAM and the bandwidth to actually use it ALL across 32 cores. That’s the main difference with this chip over the M1 from last year.
Another thing…
When gaming with 32 cores we have 8 incredible CPU cores and a neural engine sitting around with access to the same cache data. I wonder if I’m the future those will able to be leveraged by high end games?
keifer.street
macrumors regular
Yes even the regular M1 is more than strong enough to render 4K video much faster than your existing 2018 machine…
This makes me wonder if the M chips will be updated every 2 (maybe 1.5) years. About 1 year to scale up the M# chips.
adamjackson
macrumors 68020
faster than the iMac Pro Vega as well. My vega 48 iMac is definitely slower. Amazing.
JippaLippa
macrumors 68000
Does the quantity of Ram (specifically 32 or 64GB) change the GPU performance?
I really couldn't afford the 500$ extra (with my countrie's exchange rate) for the 64GB but I'm tortuing myself it might end up halving the performance of the 32-core GPU I have gotten.
I don't understand how this unified memory thing works...thanks!
I really couldn't afford the 500$ extra (with my countrie's exchange rate) for the 64GB but I'm tortuing myself it might end up halving the performance of the 32-core GPU I have gotten.
I don't understand how this unified memory thing works...thanks!
This is the score of my current 15" 2016 MBP:
Single: 4393
Multiple: 14077
This is the score of the 14" 2021 M1 Pro I just bought
Single: 1743
Multiple: 12353
Doesn't look like an improvement, or is Geekbench v5 that different from Geekbench v4?
Single: 4393
Multiple: 14077
MacBook Pro (15-inch Late 2016) - Geekbench
Benchmark results for a MacBook Pro (15-inch Late 2016) with an Intel Core i7-6920HQ processor.
browser.geekbench.com
This is the score of the 14" 2021 M1 Pro I just bought
Single: 1743
Multiple: 12353
MacBookPro18,3 - Geekbench
Benchmark results for a MacBookPro18,3 with an Apple M1 Pro processor.
browser.geekbench.com
Doesn't look like an improvement, or is Geekbench v5 that different from Geekbench v4?
Zdigital2015
macrumors 601
My Core i9 Radeon Pro Vega 48 scores 57717
My 2016 i7 Radeon Pro 460 scores 14469
My 2012 i7 NVIDIA GeForce GT 650M scores 1962
I’m not sure why anyone here is fussing about this score for the Pro Max, 32 cores or not? At this point, no units have shipped and embargoes are still in place. This may be a preproduction unit, it may be 24 cores or 32 cores, we’ll certainly know a lot more next week. The fact is that if that number holds true, you now have a laptop with desktop GPUs speeds along with dedicated ProRes encode and decode engines. Comparisons are cute and all, Apple does them and I wonder if they should. Me, I could give two sh*ts what NVIDIA PoS it compares to anyways. How it performs with Apple Silicon native apps and if that gets more done than Windows is all that matters.
No one is coughing up $3899 for a gaming laptop, regardless. If you can fool witless rubes here, good for you. But unless you have disposable cubic bucks, you’re not buying this for games that don’t exist, you’re buying this to get work done. Based on what an upgrade this is from the 5600M, the non-Prosers are going to be happy clams. The rest of you who can’t afford it or have no use for it, but want it are just going to spin your wheels for no good reason.
My 2016 i7 Radeon Pro 460 scores 14469
My 2012 i7 NVIDIA GeForce GT 650M scores 1962
I’m not sure why anyone here is fussing about this score for the Pro Max, 32 cores or not? At this point, no units have shipped and embargoes are still in place. This may be a preproduction unit, it may be 24 cores or 32 cores, we’ll certainly know a lot more next week. The fact is that if that number holds true, you now have a laptop with desktop GPUs speeds along with dedicated ProRes encode and decode engines. Comparisons are cute and all, Apple does them and I wonder if they should. Me, I could give two sh*ts what NVIDIA PoS it compares to anyways. How it performs with Apple Silicon native apps and if that gets more done than Windows is all that matters.
No one is coughing up $3899 for a gaming laptop, regardless. If you can fool witless rubes here, good for you. But unless you have disposable cubic bucks, you’re not buying this for games that don’t exist, you’re buying this to get work done. Based on what an upgrade this is from the 5600M, the non-Prosers are going to be happy clams. The rest of you who can’t afford it or have no use for it, but want it are just going to spin your wheels for no good reason.
mjtomlin71
macrumors 6502
It is using the same cores as the M1, but they've also been sped up a bit. The M1 GPU cores run at ~1.25W each, these according to Apple's vague charts and a little math are running at ~1.75W. We can assume the above results are in fact from a 24-core GPU. We should be seeing numbers closer 90,000 if Apple's "x4" performance holds true, which we have no reason to doubt, since the CPU is in fact ~70% faster than the M1 as claimed by Apple. ...and let's face it, if there was single benchmark they would base those claims on, it would be Metal.
Might be better to wait for more results? Although, what are Internet forums if not a place for knee-jerk reactions? We just went through this with that original CPU result, which proved to be the low-end of those results.
Good news though, we already know that the next gen GPU cores "bigger" and much more efficient! So the M2 should see larger gains in GPU performance!
Last edited:
YES, vastly different numbers between versions.
hagjohn
macrumors 68000
I think your expectations may be a big high.I hope this is not the 32-core gpu otherwise this would be a bit disappointing
jlc1978
macrumors 603
It would be nice, but I see several roadblocks that performance doesn't overcome; the primary one being the size of the Mac game market vs the cost of development. If a company wants to d a proper port it will require a lot of development to take advantage of Apple's tech and tools, whose cost my be high enough that the margin isn't appealing. If the do a port like the old "take the Windows version, put a wrapper around it and don't bother making it Mac like," gamers will be vocal abut how the port sucks.
Apple, on the other hand, seems to be all in on Apple Arcade and handheld gaming; and not poisitioning the Mac as a game machine.
So in the end, it's not lack of hardware capability that is keeping the Mac from being a game machine.
ignaciobarrena
macrumors 6502
Of course it's different.
MacBookPro18,3 vs MacBook Pro (15-inch Late 2016) - Geekbench
browser.geekbench.com
The M1 Pro you just bought is x2,3 faster in single-score and x3,6 faster in multi-core
Appledoesnotlisten
macrumors 6502a
I see now! So I overpaid $200 for the cores I'll probably never need)Dude, I have literally ordered M1 Max with 24-core GPU, it was $200 cheaper than the 32-core version.
what are you talking about
As a game developer, this will do nothing to encourage me to port my game. More macOS market share will. It won’t matter if the $700 Mac mini has the equivalent of the desktop RTX 3090, the market share is just not there. That’s why my game is being made in windows only.
Zdigital2015
macrumors 601
The scores for Geekbench 5 were completely recalibrated. The 2016 15” 6920HQ I own now scores 908 SC and 3681 MC. Run GB5 on your 2016 and then post your score. The M1 blows away the 2016 MacBook Pro.
darngooddesign
macrumors Core
@MacRumors.com What a lot of us want from your reviews are benchmarks as they relate to battery life so we can decide if the performance is worth the battery hit.
Geekbench 4 is extremely different from Geekbench 5
Your old MacBook should come close to 3400 in "multiple"
MacBook Pro (15-inch Late 2016) Benchmarks - Geekbench
browser.geekbench.com