Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Probably for 99% of use cases yes. 5600M is about 11.6 TFLOPs at half precision and ~5.8 at single. The current M1 is about 2.6 single precision, with 4X the number of cores in the 32 core version you'd expect almost linear scaling to about 10.4. If they took the GPU from the A15 which is significantly faster (20 to 40%) you'd be looking at up to ~14 TFLOPs of FP32 performance putting it above the 3060M from Nvidia and the 6800M from AMD.

One area were I don't expect they be able to compete is on pure memory bandwidth. They will have a wider 256bit bus due to the increase in CPU cores from 4 to 8 but they'll likely still be using LD-DDR4X which is slower but much less power hungry compared to GDDR6.
Exactly my concern as well. If there are updates to this chip’s GPU cache, bus or RAM I’m wondering if a 64GB config could be worth it? Thoughts?
 
Have you used Final Cut on an M1? As a video editor myself, It’s extremely capable and holds it’s own against some very decent desktop rigs I’ve used. That being said, I’m eagerly waiting for the M1X announcement on Monday. We are all aware the M1 is Apples low-end consumer chip and though it is very impressive for what it is, does have some limits. Let’s see what they can do with a bit more power at hand :cool:
Yes I am running FCPX on my M1 Mac Mini. And I am more than happy with it. Editing multiple tracks of 4k h265 with grading, noise removal, etc in high quality mode and without proxies. It plays fluid and is a pleasure to work with.
And as soon as it is possible I will order a M1X macbook pro as M1 has shown me that it is has a great future. Especially since all my essential software as well as the FCPX plugins I use are all available as M1 native apps.
 
Last edited:
  • Like
Reactions: Roode and Jorbanead
Exactly my concern as well. If there are updates to this chip’s GPU cache, bus or RAM I’m wondering if a 64GB config could be worth it? Thoughts?
Apple is in my view is unlikely to have a 64GB model at launch if they are using LPDDR4X just due to there being no 16GB LPDDR4X chips being manufactured. The max is 12GB chips and if they do 4 of them you'd get 48GB. Bandwidth wise you are looking at ~136gb/s which is about on par with the 2019 Mac Pro with 6 memory channels which is a lot of performance.

If Apple pulls a fast one and launches with LPDDR5 at 6400MT (highly unlikely) you are looking about 205gb/s which puts it against 8 channel memory systems in todays fastest server systems. Given the GPU and CPU can perform operations on the same data as they have unified memory they will take the performance crown.
 
Apple is in my view is unlikely to have a 64GB model at launch if they are using LPDDR4X just due to there being no 16GB LPDDR4X chips being manufactured. The max is 12GB chips and if they do 4 of them you'd get 48GB. Bandwidth wise you are looking at ~136gb/s which is about on par with the 2019 Mac Pro with 6 memory channels which is a lot of performance.

If Apple pulls a fast one and launches with LPDDR5 at 6400MT (highly unlikely) you are looking about 205gb/s which puts it against 8 channel memory systems in todays fastest server systems. Given the GPU and CPU can perform operations on the same data as they have unified memory they will take the performance crown.
Samsung already using LPDDR5 for a while. I dont think it's not unlikely.
 
Apple is in my view is unlikely to have a 64GB model at launch if they are using LPDDR4X just due to there being no 16GB LPDDR4X chips being manufactured. The max is 12GB chips and if they do 4 of them you'd get 48GB. Bandwidth wise you are looking at ~136gb/s which is about on par with the 2019 Mac Pro with 6 memory channels which is a lot of performance.
Heh heh ... Apple could surround the SoC with memory modules ... two on each sides would bring the max to 64GB for LPDDR4X modules. The traces would be a nightmare tho.
 
If its coming true, then the 128cores gpu from the next Mac Pro should be able to stand against RTX A6000 and RX 6900XT or QUADRO RTX 6000
 
  • Like
Reactions: richinaus
They have patents for that. If I remember correctly one patent shows over 24 RAM modules
Patent should be on how they stack up the RAM modules then I presume. 24 RAM modules on the same plane as the SoC ... I cannot imaging how those will be placed ... :p
 
Patent should be on how they stack up the RAM modules then I presume. 24 RAM modules on the same plane as the SoC ... I cannot imaging how those will be placed ... :p

No, it wasn't stacking. The patent directly depicted a large PCB littered with RAM modules. Of course, it has to be just a schematic depiction, I cannot imagine this being even remotely practical.

But mounting RAM chips on both sides of a package is a thing, and I am sure we will see it in some way.
 
Would it be possible to 3D stack RAM chips, around and above/below the SOC, and thus obtain more space that way?
that will make the device thicker..for now, along side the package is the right way
Surface latpop studio could do that with that design, but they didnt
 
that will make the device thicker..for now, along side the package is the right way
Surface latpop studio could do that with that design, but they didnt
Apple could order custom RAM modules with the RAM dies stacked ... Oct 18 couldn't come fast enough.
 
  • Like
Reactions: Serban55
Apple could order custom RAM modules with the RAM dies stacked ... Oct 18 couldn't come fast enough.
Ofc, Apple could do anything they want...they control everything now on their products
But i still think the Ram dies stacked on top of each other can be an desktop thing like mac mini, mac pro
Apple wants every mm if they can, to be erased and the laptop to be as thin as it can be...maybe im wrong nowadays since little Jony Ive is gone
Lets keep our hands together the time will pass quicker till Oct 18 :D
 
Last edited:
  • Haha
Reactions: quarkysg
It should be 12-15 Tflops single precision at 32 core GPU.

Apple's current top of the line mobile GPU is rated at 5.5 Tflops. They will at least double that. No way around it.
 
It should be 12-15 Tflops single precision at 32 core GPU.

Apple's current top of the line mobile GPU is rated at 5.5 Tflops. They will at least double that. No way around it.
, the top of the line mobile Gpu is not the 5600M HBM2 at about 11.6 TFLOPs?
What are you talking about? Maybe i understood wrong

The top of the line mobile GPU that Apple has, is in the 27" iMac and its called 5700 XT and it has 19.51 TFLOPs

So i hope this M1x 32 gpu core it will be the standard for the bigger imac with an additional gpu cores as option to beat or at least be on par with that 20 TFLOP
 
Higher Tflops does NOT mean faster. AMD GPU has faster Tflops and yet, Nvidia GPU performs better than AMD GPU so it's up to something else.
 
Higher Tflops does NOT mean faster. AMD GPU has faster Tflops and yet, Nvidia GPU performs better than AMD GPU so it's up to something else.
the 5700 XT could top the 32 gpu cores ..so yes, Apple needs to step up the gpu power for the next imac even further than the upcoming m1x with 32 gpu cores
 
  • Like
Reactions: richinaus
the 5700 XT could top the 32 gpu cores ..so yes, Apple needs to step up the gpu power for the next imac even further than the upcoming m1x with 32 gpu cores
I think the most interesting thing is how they step all this up, from iMac to Mac Pro.

I am 100% confident however, that they know how to do it all, as they would have been crazy to shift architectures without knowing how they would do it.
 
, the top of the line mobile Gpu is not the 5600M HBM2 at about 11.6 TFLOPs?
What are you talking about? Maybe i understood wrong

The top of the line mobile GPU that Apple has, is in the 27" iMac and its called 5700 XT and it has 19.51 TFLOPs

So i hope this M1x 32 gpu core it will be the standard for the bigger imac with an additional gpu cores as option to beat or at least be on par with that 20 TFLOP
I think you are confusing half precision (FP16) vs single precision (FP32). The 5700 XT is actually a desktop class GPU and is only ~9.75 TFLOPs at single precision which would put a 32 GPU M1X/M2 mobile above AMD's desktop performance. If the rumors about a 128 GPU Mac Pro are true, well the numbers are going to be incredible.
 
In fact, JEDEC has already (July 28th) updated the standard to include LPDDR5x (up to 8533Mt/s) with manufacturers and IP-suppliers on board.

LPDDR5 is only new for Apple products.
I'd it is possible but its highly unlikely just because of the global shortage of chips, most manufactures have just started production (Micron looks to be the furthest along). But I'd happily accept it if they do!
 
I personally don't see a need for more than 32 GB on a laptop but that's because I'd put 64 GB or 128 GB on a desktop. I'm glad that I don't have to only have one device.
 
  • Like
Reactions: richinaus
I personally don't see a need for more than 32 GB on a laptop but that's because I'd put 64 GB or 128 GB on a desktop. I'm glad that I don't have to only have one device.
I wish I could just have one device, but that will only happen if I ever retire……
 
I think you are confusing half precision (FP16) vs single precision (FP32). The 5700 XT is actually a desktop class GPU and is only ~9.75 TFLOPs at single precision which would put a 32 GPU M1X/M2 mobile above AMD's desktop performance. If the rumors about a 128 GPU Mac Pro are true, well the numbers are going to be incredible.
The Radeon Pro 5700 XT used in the 2020 iMac is effectively a downclocked Radeon RX 5700 XT with twice as much memory. It's single-precision performance is ~7.7 TFLOPS, which is roughly equivalent to 24 M1 GPU cores. The consumer version has a TDP of 225 W, which is way too much for the iMac case, while the "pro" version uses 130 W.
 
  • Like
Reactions: leman
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.