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.
3D Rendering on Apple Silicon, CPU&GPU
- Thread starter vel0city
- Start date
- Sort by reaction score
Macintosh IIcx
macrumors 6502a
The Studio M3 Ultra being one generation behind what is current at launch was hopefully a one-off thing. They only alluded that there might not be an Ultra for each generation, not that it would always be launched on an older M generation. So I would not rule an M5 Ultra out, although it might come a couple of months after the M5 Max to bin the chips for it.
There is also the weird state of the Mac Pro to account for, maybe that will get the M5 Ultra for WWDC 2026?
The 16” model still has some thermal headroom. It’s more a question of costs, probably. Faster GPU also requires faster everything else (more caches, faster RAM, etc.).
Tdude96
macrumors 6502a
I'm hoping we do. I was looking at release dates after someone in another thread said that Apple is on a 2-year cadence with Mac Studio, which didn't strike me as correct (it's ~15-21 months). We don't have a good enough sampling to make a remotely accurate guess.
However, what I noticed was that when there is a new Mac Studio, it follows ~5 months after the release of the same generation Max chip.
MacBook Pro M1 Max -> Mac Studio M1 Max/Ultra = 4 months, 23 days
MacBook Pro M2 Max -> Mac Studio M2 Max/Ultra = 4 months, 20 days
MacBook Pro M4 Max -> Mac Studio M4 Max/M3 Ultra = 4 months, 4 days
This makes me think that if we do see a late January release of the M5 Max MacBook Pro and if they are planning to release a M5 Mac Studio, there's a really good chance we'll see the M5 Max/Ultra Mac Studio around WWDC 2026, which would be roughly 5 months after the M5 Max MacBook Pro release and 15 months after the Studio M4/M3 release, fitting both timelines.
That's assuming, of course, Apple does plan on having M5 Studios and doesn't do something wonky with the chip generation mismatch like they did on the last Studio.
Last edited:
M5 MBP GPU-rendering benchmarked in Cinebench 2024 GPU (using the Redshift renderer).
Last edited:
With respect to stacking/packaging - assuming the tech exists (a big assumption) - would it already be feasible from a heat/power perspective in Apple's case? If an A19 Pro consumes 12W of power - all else being equal - could they stack two of them and not have heat be an issue? Would that change assuming there was available active cooling (as in a desktop)?
Last edited:
StudioMacs
macrumors 65816
If it was 96% faster than 108 seconds, it would be 4.32 seconds.
55 seconds is approximately 49.07% faster than 108 seconds.
"faster" refers to seed, and speed is proportional to the inverse of time. 1/55 is 1.96 times 1/108.
You can also say that it takes almost twice as long to finish the task on the M4. When one device takes twice the time, the other is twice faster, not 50% faster.
You can also say that it takes almost twice as long to finish the task on the M4. When one device takes twice the time, the other is twice faster, not 50% faster.
It's all relative. Yes, M5 does the job in half the time or 49.07% of M4's time but that's twice as fast as M4. M5 is 96% faster.
M5 Max 40c turned out to be about 34% faster than M4 Max 40c in Blender Opendata. It's not far from M3 Ultra 80c. It's almost as fast as a desktop 4070 Ti Super and faster than mobile 4090/5080.
If M5 Ultra will be as fast it would give around 9968 which would make it faster than RTX Pro 6000 Blackwell and desktop 5080. I think it would be even faster thanks to the new Fusion architecture and being two generations ahead of M3 Ultra. The question is if we'll get M4 Ultra or M5 Ultra. It would be weird if Apple kept using the old chip design for Ultra.
If M5 Ultra will be as fast it would give around 9968 which would make it faster than RTX Pro 6000 Blackwell and desktop 5080. I think it would be even faster thanks to the new Fusion architecture and being two generations ahead of M3 Ultra. The question is if we'll get M4 Ultra or M5 Ultra. It would be weird if Apple kept using the old chip design for Ultra.
Last edited:
I think a more reasonable extrapolation based on the observation of the data is that m3max to m3 ultra scaled about 1.8 and thus an ultra class m5 should land at about 12000 at least.
tenthousandthings
Contributor
Yes, I think if it is 2x Max, without any changes to the M1-M2-M3 UltraFusion symmetry, then that's a safe bet. TSMC goes out of its way to say that SoIC ("Fusion Architecture" in Apple's implementation) is compatible with UltraFusion ("InFO-L" in TSMC's current terminology, if I understand that correctly, they have a tendency to change their nomenclature without public comment).
The hope, as I see it, without any expertise other than having good reading-comprehension skills and a bit of time on my hands, is that Apple could introduce an Ultra-only secondary SoIC chip (to be paired with a standard Max) that replaces the CPU die with a second GPU die. To my mind, that fits Apple's criteria -- it doesn't require much additional R&D other than with regard to UltraFusion's local ("L") silicon interconnect, which is presumably already part of the Ultra budget.
This theory assumes that the CPU and GPU dies are similar in size, which I don't think is something we know at this point. I don't recall seeing any representations of Fusion Architecture -- the last time Apple did the CGI imagery was M3 (press release) -- the M4 (press release) doesn't use CGI and doesn't show relative sizes (at least not directly), but it does show graphical representations of the layouts, so an improvement overall, IMHO. But unless I've missed something, the M5 (press release) has none of that, neither CGI nor layouts, which makes me think more is to come when the Ultra launches. Perhaps there's something there they don't want to reveal at this point?
[On the other hand, the M3 Ultra (press release) was the same, so maybe I'm reading too much into the absence of any graphics.]
Last edited:
The m5 ultra class if release soon will be a true beast. basically a CPU on par with a 32 core threadripper pro and a GPU mostly on par with a 4090 but with unified mem up to 512 GB. Then of course the x86 does not stand still so who knows what will be the correct thing to compare to.
I really hope that the ultra does not get released at end of year or even worse, we get a m4 ultra since that will lack som much of the massive improvements form the m5 arch.
Then there is alos the fantasy of something even more extreme but tbh, the only thing I really would like is better networking than 10GBe and the ability to have a ssd extension inside the chassis.We'll see.
I really hope that the ultra does not get released at end of year or even worse, we get a m4 ultra since that will lack som much of the massive improvements form the m5 arch.
Then there is alos the fantasy of something even more extreme but tbh, the only thing I really would like is better networking than 10GBe and the ability to have a ssd extension inside the chassis.We'll see.
Boil
macrumors 68040
Apple is currently limiting the M3 Ultra Mac Studio to 256GB of RAM, so who knows what they will do with any new releases of the Mx Ultra Mac Studio...?
Introducing the all-new Apple Mac Pro Cube featuring:
- Mx Extreme SoC
- (up to) 1TB UMA RAM
- (up to) 32TB system SSD (4@8TB NAND blades in RAID configuration)
- Four available 2280 M.2 NVMe PCI6 x4 slots for standard third-party expansion
- Two 100GB/s Ethernet ports
- Assorted TB5 & USBx ports
AI video applications for advanced rendering run up to a dozen times faster on the RTX 5000/RTX Pro 6000 than on Mac silicon. Provided, of course, that you’re using real-world software rather than Geekbench benchmarks. Or perhaps the whole world is mistaken about Apple and is actually using Nvidia hardware 😃
Last edited:
Same here. Still using an M1 Max Macbook pro as my main machine, but I no longer have the need for portability but I do need additional rendering capability. Very excited about the M5 specifically for C4D and Redshift. I think this is going to be the year I get a new desktop Mac and a couple of Apple displays.
I do not use that kind of software myself, can you give some examples? I do however work with developing AI models for images though and training on even a low end nvidia card is often 10x the speed of on the best Macs. Basically a no-go. Even if the neural accelerators in the m5 is 4x current state it will still lag massively. But very few people do actual training compared to inference so I just hope the inference speeds improved enough. Would love to test the m5 max for these purposes but I will not buy one just for that. I guess (and hope) that once the ultra comes out we have more comprehensive data to go on. It really has the potential to be game changer.
On another note, today nvidia will present new stuff at GTC. Expecting a new DGX Spark version with massive improvements with a new nvidia SoC. Nvidia is not standing still at all but improving faster than ever.
Pressure
macrumors 603
Are you implying AMD NAVI 21 is better than modern Apple GPUs in 3D rendering?
I mean, have you seen the Blender or Maya performance for that AMD GPU compared to Apple?
In Blender the AMD GPU gets around 2500 points. That's less than the M4 Pro GPU. Let's not talk about the performance to watt ratio 😅
https://browser.geekbench.com/metal-benchmarks 😃
Five years have passed, and Apple's GPU hasn't evolved.
Or maybe it's starting to catch up to the RX 6900 XT )
In professional applications, power consumption isn't an issue.
Suffice it to say that you might have to pay $100 for an hour of cloud computing.
And if you have a powerful graphics card and can do it on your own computer,
just calculate how much you're saving.
Five years have passed, and Apple's GPU hasn't evolved.
Or maybe it's starting to catch up to the RX 6900 XT )
In professional applications, power consumption isn't an issue.
Suffice it to say that you might have to pay $100 for an hour of cloud computing.
And if you have a powerful graphics card and can do it on your own computer,
just calculate how much you're saving.