Yeah, for simple tests like Geekbench gpus should scale pretty much linearly. That's what they're good at. No one is suggesting it'll be the same for Photoshop or FCP etc.
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
So M1 Max with 32 Core GPU matches RTX 3080 Mobile.
- Thread starter iBug2
- Start date
- Sort by reaction score
With the amount of hype, I expect it to be more evenly split between good/bad surprises.
There will be goal post movers. There will be told ya so's. There will be a lot of happy new MBP owners.
The point is not to be good at Geekbench, it's to understand what's going on. There is no reason for it to scale perfectly from 8 to 16 and then to...collapse from 16 to 32. That doesn't make sense. Furthermore, it matters because the results for this gpu will have consequences for how any future iMac Pro or Mac pro will do.
Well, given you can go on their website and see certain benchmarks they used, but then broadly claimed the 32 core was on par with a 3080 overall, I'd disagree with you there. In any case, we will see what happens come Monday. I'm dramatically tempering my expectations now.
Geez. This is the same company who tried to compare its XDR display to the Sony broadcast mastering monitor. It's not even in the same league. Not even close. But it also only costs 20% of the Sony display, even with that stand.
You should know by now that if you want to compare performance, you should have a specific set of criteria in mind, and compare those, instead of just relying on press releases and product unveilings.
Now, if your criteria is Geekbench 5 Metal performance or whatever, then fine, but people don't Geekbench all day long for their workflow.
GB is more garbage that I thought. Should only compare GB scores if it's the exact same minor version so 5.4.1 with 5.4.1 and not 5.3.1 or even 5.4.0.
I think you're misunderstanding my position. I'm disappointed in the data performance we've gotten thus far (just not GFXBench and Geekbench, mind you), so I'm trying to set my expectations to something I feel like is more realistic. Given what Apple has given us and the leaked benchmarks we've gotten, I am fearing that the 3080 claim that Apple broadly used to compare the 32 core against, is likely not going to be the case across the board. Or in other words, it's going to only complete with a 3080 in very select spaces, which is rather disappointing if it turns out to be true. I am really not sure why you're lambasting me or other users of the forum for tempering our expectations haha
The issue is they also compared to their own chips and said 4x scaling relative to the M1 and I can’t remember if people calculated it or they said it but that’s a 10TFLOP level of performance. As @JimmyjamesEU said, if it fails to achieve that in compute scenarios then it simply fails to do that. Between the pro and 32-core Max, for double the cores, double the memory bandwidth, and same clocks, you should get double the compute performance, not 1.6x. Since we know the core count and the bandwidth, that leaves the clocks. We also know that there’s a high power mode specifically for the 16” max GPU that will run it higher than its normal clocks, which mode are which benchmarks on? Does it matter? Maybe not. We’ll see.
It’s more than there are confusing discrepancies in the claims of performance with different benchmarks that make it difficult to figure out what is going on. Yes, it’s a very good GPU regardless, but even so … it isn’t clear exactly what’s going on with even amongst their own performance claims.
New 16-Inch MacBook Pro With M1 Max to Feature High Power Mode for Intensive Workloads
The new 16-inch MacBook Pro with the M1 Max Apple Silicon chip will feature a new High Power Mode for intensive, sustained workloads, according to...
www.macrumors.com
This might explain the discrepancy in benchmark performance.
Of course it's not going to beat or tie 3080 across the board. They're completely different designs after all. Plus, that's Apple's advertising practice and it hasn't changed. But Apple's GPU will most likely do very, very well for many workloads.
I'm making those statements because for past 20 pages or whatever you guys have been trying to come up with excuses as to why it isn't performing better. Hell, in retrospect, even Geekbench is looking like it's more accurate that some might have believed, despite the fact it's not a good test for real world workloads.
Thank you. Someone can explain my own thoughts better than I can! ?
Absolutely. It’s a great gpu, and I’d be happy if they’d said this was what to expect. Given they undersold the A15 performance, and these figures line up with what a 24 core gpu would produce, it’s … weird and confusing.
Yeah, no reason to speculate too much about it when people soon will have the hardware on hand to test
I personally wouldn't be surprised if the 32-core GPU is clocked lower. That could explain the non-linear performance increase but we will know soon.
Yeah, Monday morning is going to be a wild morning!Yeah, no reason to speculate too much about it when people soon will have the hardware on hand to test
I personally wouldn't be surprised if the 32-core GPU is clocked lower. That could explain the non-linear performance increase but we will know soon.
The m1 max performs about the same (some higher, some lower) as the RTX3080 in GFXBench 5.0. GFXBench concentrates on graphics rendering
GFXBench - Unified cross-platform 3D graphics benchmark database
The first unified cross-platform 3D graphics benchmark database for comparing Android, iOS, Windows 8, Windows Phone 8 and Windows RT capable devices based on graphics processing power.
Here's the issue though. Look at the on-screen FPS, not off-screen. You'll never play game with stuff off the screen in that manner.The m1 max performs about the same as the RTX3080 in GFXBench 5.0
GFXBench - Unified cross-platform 3D graphics benchmark database
The first unified cross-platform 3D graphics benchmark database for comparing Android, iOS, Windows 8, Windows Phone 8 and Windows RT capable devices based on graphics processing power.gfxbench.com
Looking at that, it's closer to a 3060, which is oof.
I think in gfxbench the offscreen is more important due variations in screen refresh rates etc. Offscreen is pure gpu measurement.
Oh is it? Today I learned. I thought the on screen was more important.
So then does that mean when it comes to gaming, the M1 Max is actually better than the 3080 laptop???? ?????
EDIT: I'm guessing the thumbs up that crazy dave gave you indicates you're correct
The comparison uses OGL which is the slowest of the available API's. DX performance is better, and for some stupid reason DX12 and Vulkan results are missing.
Oh is it? Today I learned. I thought the on screen was more important.
So then does that mean when it comes to gaming, the M1 Max is actually better than the 3080 laptop???? ?????
EDIT: I'm guessing the thumbs up that crazy dave gave you indicates you're correct
It depends truthfully: on screen measurements introduces new variables, so when measuring “pure GPU power” (heavy scare quotes) in gaming off screen is more pertinent. But yes you do game on screen obviously
. Also, the newer API results are missing for DX and Vulkan in that test. So … yeah …New 16-Inch MacBook Pro With M1 Max to Feature High Power Mode for Intensive Workloads
The new 16-inch MacBook Pro with the M1 Max Apple Silicon chip will feature a new High Power Mode for intensive, sustained workloads, according to...
This might explain the discrepancy in benchmark performance.
What I am thinking, the M1 Max is clocked for the 14” MBP.
However the 16” MBP has the ability to go “unleashed” with it’s better thermals.
Guess someone on Monday has to confirm this high-performance mode on the 16”.