There is no moving of goalposts by anyone other than you. We were discussing the performance when using redshift as compared to the previous model. You mentioned the M1, we were discussing the 5600m.
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
It's going to be interesting for sure. I have a 2009 Mac Pro with a 1080ti that I use solely for Redshift render. I am very familiar with that machine and setup and am eager to see how my new M1 Max 64GB compares when it arrives. Metal vs CUDA. In terms of responsiveness, transferring the scene from memory to the render, render bucket speed, I have got my eye on everything.
Even saving out a hi-res EXR to disk from Redshift should be vastly improved with the write speeds and SSD in the M1 Max, yes?
Even saving out a hi-res EXR to disk from Redshift should be vastly improved with the write speeds and SSD in the M1 Max, yes?
Thread is moving fast, not sure if this was posted: https://www.macrumors.com/2021/10/21/new-macbook-pros-high-power-mode/
"New 16-Inch MacBook Pro With M1 Max to Feature High Power Mode for Intensive Workloads"
"Your Mac will optimize performance to better support resource-intensive tasks. This may result in louder fan noise." The new mode is not likely to be used in typical work cases, but instead when users may be rendering larger files or graphically intensive tasks that require an added boost of performance."
Interesting and only avaialble in the 16" model.
"New 16-Inch MacBook Pro With M1 Max to Feature High Power Mode for Intensive Workloads"
"Your Mac will optimize performance to better support resource-intensive tasks. This may result in louder fan noise." The new mode is not likely to be used in typical work cases, but instead when users may be rendering larger files or graphically intensive tasks that require an added boost of performance."
Interesting and only avaialble in the 16" model.
Yeah, I wrote it before other scores were available. At this point my hypothesis is that the Max scores are 14“ machines that don’t have enough thermal headroom to run the GPU at full power. But it could also be Geekbench submitting the work in a weird way that fails to fill up the GPU properly.
Please do. Actually, consider writing a blog post or something since this is valuable information that is easily lost on these forums.
I think if we’re saying going from 16 to 32 scales by 1.6, we need to explain why the same cores scale perfectly from 8 to 16 but then suddenly change.
Probably, but there's this:
So still important to wait.
Such scaling issues are not uncommon when clock speed are lowered. But that should've effected everything, including Apple's benchmarks in their presentation ... unless benchmarks were very carefully chosen. Not the first vendor to do that, but even so ... What's weird is all of the benchmarks on Apple's website show imperfect GPU scaling, less than 2x between Pro and Max, some more understandable than others given what they are, but ...
It was, or maybe in another similar thread. Difficult to keep track. But yes, could be part of it.
Given what we now know, I'm doubtful. We might see a few higher scores that break 70k, but I'm fairly certain this is a 32 core now given Redshifts own material on the matter. The numbers line up far better with theirs.
Scaling has never been perfect at the upper end and memory bottleneck are the common previously known issues.
Yea, there's no question we need to wait and see what happens, but I think it's fairly safe to say that this is the 32 core after Redshift gave its numbers and its matching really quite closely to what Geekbench is getting. It's entirely possible geekbench is optimized properly here for the worklord or something. I just find it super strange that the 8 to 16 core is literally double, while the 16 to 32 isn't, without something else going on like having cores spread out over the SoC and/or different clock speeds.
If this is accurate, three reasons I can think of off the top of my head:
1) Clock speeds are lowered in the Max model to compensate for more cores. Pretty standard to keep heat down.
2) Andrei mentioned the Max GPU model actually looks like two separate GPUs on the same die with some shared logic. Could also affect things depending on interconnect.
3) Memory bottlenecks, keeping big GPUs fed is hard. However, they did double bandwidth for the bigger GPUs with double the cores relative to the smaller ones. So this should be less of an issue than the others. Latency probably went up though with double the bandwidth, so not entirely implausible that it had an effect.
Last edited:
I would argue memory bottleneck is probably less likely since they also doubled bandwidth from 16 to 32, but it's possible. GPUs even at the high end scale reasonably well ... until power and as you say memory become an issue. Latency probably went up though with double the bandwidth, so not entirely implausible that it had an effect.
Which, from what it seems, they've mostly solved. This according to Apple. I'm really eager to see the results now come Monday. This is gonna be a really long weekend XD
Given Mark Gurman's leaks and the fact they've been very accurate so far, I think the Max has no interconnect. Its the main high end SoC, and the Pro is the "chop" version. The news of "High Performance Mode" is interesting. I wonder if it could explain some of these results.
What Redshift material? Did I miss a part of the discussion? Can you point me to it?
The M1 Pro/Max reveal was pretty exciting, and there are so many things we all want to know. I'm on the edge of my seat, captivated by the leaps Apple is making in their silicon design, in spite of the fact that I never intend to buy these versions.
Oh I'll ABSOLUTELY be getting one. I've been waiting for these for years (namely miniLED and ProMotion). The new Apple Silicon is the icing on the cake, what with getting all the ports back and a better keyboard etc.
Sorry bad choice of words, not an interconnect like between different dies, but whatever the shared logic is and on die speed of transfer between the “two” GPUs is if the reading on the die shots are accurate and they are two on die GPUs and not one.
Surely this image shows that the 14 is slower than the 16 even if both have 32 cores? It's 4x the 5600m on the 16" and 4.1x the Iris on the 14.Here you go
Check what they are comparing it against. It's the old iris graphics. Speeds should be the same. We will find out Monday if there is any real speed difference, but I do doubt it (aside from the potential of thermal limitations, which I doubt will be significant).
The test on the left is different (test 14) and is testing real-time 3D performance, the redshift rendering test (test 15) is different.
It's really confusing because Apple gonna Apple and I have no idea why they ran test 14 on a 14" and test 15 on a 16"
The Iris must be much slower than the 5600m. How can the same gpu be 4x both? The 14" must be clocked lower.
EDIT: I didn't read!
Ok I see, yes I misread it. I thought they tested the same thing on both. Apologies.