I think the imac will ise the same m1 pro and m1 max, like the 24” did with the m1
The mac pro is said to use 24c and up to 128 gpu cores
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
The mac pro is said to use 24c and up to 128 gpu cores
The rumors for the Mac Pro lineup revolve around Jade 2C & Jade 4C...
I believe these are both maxed-out M1 Max SoCs (two for Jade 2C & four for Jade 4C) lashed together...?
M1 Max
- 10-core CPU (8P/2E)
- 32-core GPU
- 16-core Neural Engine
- 64GB RAM
Jade 2C
- 20-core CPU (16P/4E)
- 64-core GPU
- 32-core Neural Engine
- 128GB RAM
Jade 4C
- 40-core CPU (32P/8E)
- 128-core GPU
- 64-core Neural Engine
- 256GB RAM
Rumor also has a "smaller Mac Pro", with a nod towards the G4 Cube...? I would LOVE to see a Jade 2C powered Mac Cube...!!!
Unsure if there will be a new iMac Pro...? I feel if they make the larger mid-level iMac stay with the 27" display, then there would be room for a Jade 2C 30"+ iMac Pro... (30" or 32"...?)
It is impressive, too bad the RT performance isn't that great (extrapolating performance from M1 to M1 Max with existing Metal RT API).
Ray Tracing Performance Comparison - Tellusim Technologies Inc.
Tellusim Technologies Inc. Ray Tracing Performance Comparison
tellusim.com
Well to be fair, M1 dosn’t have ray tracing cores at all. So it’s expected. Perhaps the M2 with new GPU architecture.It is impressive, too bad the RT performance isn't that great (extrapolating performance from M1 to M1 Max with existing Metal RT API).
Ray Tracing Performance Comparison - Tellusim Technologies Inc.
Tellusim Technologies Inc. Ray Tracing Performance Comparison
but then remember that the M1 is kind of slow.
M1 is equivalent to a 1050M Ti~
M1 Pro is equivalent to a 3050Ti mobile
M1 max is equivalent to a 3080m!!! With an i9 cpu
Last edited:
Unless the AI cores and Machine Learning cores can be used for ray tracing?It is impressive, too bad the RT performance isn't that great (extrapolating performance from M1 to M1 Max with existing Metal RT API).
Ray Tracing Performance Comparison - Tellusim Technologies Inc.
Tellusim Technologies Inc. Ray Tracing Performance Comparison
Are we sure the RQ CS score is going to impove enough to match those cards though?
I dunno, @leman what do you think?
Thats the mac pro and yes that will probably see in it. But i was talking about the imac 30” pro or not pro
But i guess now with these chips marketing names the bigger imac will be called imac pro
Dont think we will see another 27”
Next year even the macbook air is supposed to get redesign
So yes, next year probably the transition will be over and we still have to see the new mac mini, the 30” imac , the new redesign macbook air and of course , the “big one” the mac pro
They're not designed for that. AMD and nVidia GPUs have parts whose sole purpose is to find intersections between rays and triangles, for instance.
That might be true. Still could be used for it as any image processing they do now.
man’s likely better than brute forcing it.
never said it will be equal as it’s always better to have purpose made silicone.
well if their sliders are to be believed they already are equivalent with a vastly lower power draw in laptops.
M1 max have 32-64Gb LDDR5 memory with 409GBps bandwidth 512-bit bus
RTX 3080m have 16GB or 8GB of GDDR6 on a 256-bit bus, and 384GBps
And DDR5 have vastly better memory latency than GDDR6, while GDDR6 have much faster bandwidth.
But it seems apple have solved that gap with some kind of non standard solution.
Weren’t the performance comparisons talking about Cinema 4D?
I feel like I saw that somewhere now I cannot find it.
It would make sense for the A-series to come first, then the Mx-series, then the MxX-series; this way the largest user base gets the lowest power (electrical & compute) SoCs, then each smaller group gets a more refined & powerful product...?
If that makes sense...
If that makes sense...
More like a 1650ti as predicted. Still potentially decent power per watt except for cost.
OpenCL Score: 60,167
OpenCL Score: 60,167
MacBookPro18,2 - Geekbench
Benchmark results for a MacBookPro18,2 with an Apple M1 Max processor.
browser.geekbench.com
More like a 1650ti as predicted. Still potentially decent power per watt except for cost.
OpenCL Score: 60,167
MacBookPro18,2 - Geekbench
Benchmark results for a MacBookPro18,2 with an Apple M1 Max processor.browser.geekbench.com
*1660Ti, not 1650Ti
But yes, this is a bit disappointing. I was hoping for RTX 3060 Laptop GPU level (c. 80-85,000 OpenCL). The alleged close performance to 3070 and 3080 laptop GPUs (from Apple presentation graphs) seems to be very optimistic unless they cherry-picked a specific workload.
It is only a sample of one though, and real-world results may be better given the high memory bandwidth of the M1 Max, unified memory etc.
I'll withhold judgment until I see comparisons for tasks that I'm actually likely to want to do (mostly photo / video editing)
I wouldn't be basing anything on one single benchmark, especially when we have no idea about how it works internally.
How about 3Dmark Wildlife?
This performance is perfectly in-line with what the M1 gets. Around 3-4x higher scores than the M1.
How about 3Dmark Wildlife?
This performance is perfectly in-line with what the M1 gets. Around 3-4x higher scores than the M1.
And you can add +15% to the openCL score to estimate the Metal score. openCL run on emulation on the M1.
This is only a compute task, I'm more interested in rendering tasks.
Anyway, that openCL score seems too low. There is no reason why it shouldn't be 4x the M1 score, so >70000.
There are sub-test where the M1 Max is about 4x faster, and some others where it's just a bit more than 2x faster. Not sure why that is, but I'rather see Metal scores. Who uses openCL on the Mac these days?
This is only a compute task, I'm more interested in rendering tasks.
Anyway, that openCL score seems too low. There is no reason why it shouldn't be 4x the M1 score, so >70000.
There are sub-test where the M1 Max is about 4x faster, and some others where it's just a bit more than 2x faster. Not sure why that is, but I'rather see Metal scores. Who uses openCL on the Mac these days?
Mac mini (Late 2020) vs MacBookPro18,2 - Geekbench
browser.geekbench.com
Last edited:
OpenCL isn’t emulated, but it is a deprecated API. Apparently it’s buggy and may not be all that well optimized beyond “it works … mostly”.
OpenCL is deprecated in MacOS.More like a 1650ti as predicted. Still potentially decent power per watt except for cost.
OpenCL Score: 60,167
MacBookPro18,2 - Geekbench
Benchmark results for a MacBookPro18,2 with an Apple M1 Max processor.browser.geekbench.com
Wait for Metal scores.
Metal scores will be around 70k, so still way short of the high-end mobile Ampere.
This is a bit disappointing, but then again not unexpected. M1's Geekbench compute scores are not too hot either, but the GPU clearly punches above it's weight in real-world tasks.
I am wondering however why Apple GPUs don't perform too well in Geekbench. Is it something about the benchmark itself, is it how Geekbench measures performance, or is G13 indeed poor at utilizing its ALUs? I though that Geekbench results might be biased towards GPUs with higher memory bandwidth but it's clearly not an issue with the M1 Max...
Might as well base it on marketing then since Metal is proprietary single-platform so you can't compare performance across cross-platform.
Fresh GFXBench results OMG look at that High Tier offscreen score 275.9 fps!
3.45 X faster than the M1. As fast as a RTX 3070 laptop or about 15-20% less than the RTX 3080 laptop or RX 6800M. That's crazy to achieve that at only 60 watts, Apple truly created a monster!
3.45 X faster than the M1. As fast as a RTX 3070 laptop or about 15-20% less than the RTX 3080 laptop or RX 6800M. That's crazy to achieve that at only 60 watts, Apple truly created a monster!
That’s true for DX as well and for GPUs “cross platform” APIs still depend on how well the drivers were implemented and the match between the API and the underlying hardware. Same with any cross platform benchmark that uses different APIs - you’ll never know how well it was written for each and even it’s a theoretical since you’ll never know how well any given game or production software will port their engine.
You can compare to metal scores for AMD chips and extrapolate but I’ll admit that’s far from ideal as well. The only way we’ll get true cross platform is if/when Asahi Linux writes graphics drivers but you still run into the first problem.
Comparing GPUs is harder than comparing CPUs and doubly hard when they were designed with different APIs in mind and different OSes.
Yes .. they will reuse the chips where they can to reduce cost. e.g. M1 in the iPAD Pro
Also note that the higher end rumoured designs reuse the same chips, I assume people are thinking of chips on a package like this: https://www.pcmag.com/news/new-amd-ryzen-threadripper-cpus-unveiled-32-core-chip-is-1800