We don't really know the transistor count of Apple GPU. Area-wise, an Apple GPU core and an Nvidia Ampere SM are roughly the same. And while it's true that Apple GPU cores lack some functionality compared to Nvidia, the opposite is true as well. These GPU simply do not implement the same feature set.
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
We don't really know the transistor count of Apple GPU. Area-wise, an Apple GPU core and an Nvidia Ampere SM are roughly the same. And while it's true that Apple GPU cores lack some functionality compared to Nvidia, the opposite is true as well. These GPU simply do not implement the same feature set.
Do you count the transistors/area of the neural engine and the multimedia engine to do an apples-to-apples comparison with the Nvidia GPU?
What functionality does the Nvidia GPU lack that the Apple GPU has?
Any kind of comparison will be flawed. There is simply no apples-to-apples with these devices and how they are organised. I am comparing an Apple core to Nvidia SM because they are functionally equivalent from the perspective of GPU compute (both have 128 ALUs and a private memory pool that is shared among the threads executing on the device). But of course they are not the same, as the hardware is organised differently.
Well, for starters there is TBDR, which probably has a non-zero transistor cost on its own. Only half of Nvidia's ALUs can do an INT operation while all ALUs on Apple are INT capable. M2 seems to support more complex data routing from registers with their shuffle and fill functionality. There is on-the-fly framebuffer compression. In addition, the hardware is organised differently. On Nvidia TMUs and rasterisers are shared between a block of SMs, on Apple, each core has its own TMUs and rasterisers. Cache hierarchy is most likely different as well, although I have no idea how.
To be clear, none of this means that Apple GPUs are superior to Nvidias. Not at all. It's just an illustration that comparisons are hard, because different hardware makes different tradeoffs.
Last edited:
New Renderman, but nothing new on the AS front from the sounds of it (other than the new denoiser doesn't work even under Rosetta). The macOS related notes from the release docs are:
Still have the installer error (at least for Renderman 24) in Ventura; it's been 10 years Pixar and you can't fix an installer
Speaking of denoisers, the updated version in Renderman looks impressive (although not GPU based and can't denoies Renderman xPU renders yet).
Still have the installer error (at least for Renderman 24) in Ventura; it's been 10 years Pixar and you can't fix an installer
Speaking of denoisers, the updated version in Renderman looks impressive (although not GPU based and can't denoies Renderman xPU renders yet).
Last edited:
From Redshift blog
Open Beta of AMD support on Windows coming to RS 3.5.15
''For anyone wondering about the supported AMD GPUs. This is the current list.Radeon PRO W7900, W7800, W6800, W6600, VII and W5700
Radeon RX 7950XTX, RX7950XT, RX7900, RX6950, RX6700XT, RX6600 and RX5700XT
In addition the following GPUs should be supported (but are not actively tested):
Radeon Vega Frontier Edition, Radeon RX Vega 56, Radeon RX Vega 64 (gfx900)
Radeon RX 5500 (gfx1012)
Radeon RX 6800, Radeon RX 6900 (gfx1030)
Radeon RX 6700 (gfx1031)
Radeon RX 7800 (gfx1101)
Radeon RX 7600, Radeon 7700 (gfx1102)''
I think is quite important and relevant, everything that helps to break Nvidia monopoly is good news and everyone will benefits directly or indirectly from this.
It also show how good Maxon is to adapt to every platform.
It also show how good Maxon is to adapt to every platform.
Could be a long bow, but with this in place, their attention might also turn to optimising for Apple Silicon now. That is, if their resources are adaptable like that? Logically, pushing AMD support out for Windows would take priority over A/Si at this point in Apple's transition, particularly without the A/Si Mac Pro.
Anyway, as many would know, this is what Maxon (Panos) had to say in the Redshift forums last year (15 Apr.) . . .
Hey guys,
I'm afraid it will be very hard (or, rather, impossible) to comment on such deep technical details due to a number of NDAs.
What I can say (and this is not specifically about Apple or the M1 Ultra but a general comment about our dev philosophy) is that we want Redshift to run as efficiently as possible on all the GPUs it supports today - or the ones it will support in the future. For this reason, we try to keep close communications with the hardware vendors. In some cases those communications result in discussions about possible optimizations that are fairly easy/straightforward - so we go ahead and do them immediately. In some other cases, the optimizations might require more work and are pushed further down the line. And, in yet other cases, the optimizations are deemed unrealistic given the existing architecture of Redshift.
The specific problem discussed here (which is "why doesn't the M1 Ultra render exactly twice as fast as the M1 Max") does not belong to the unrealistic category. I.e. there might indeed exist workarounds. But these won't be simple modifications either!
And in a follow up response . . .
Hopefully some of those NDAs will result in us seeing optimisations debut with the A/Si Mac Pro announcement.
"Full Metal" Blender is getting closer and closer.
devtalk.blender.org
2023-05-08 Eevee/Viewport Module Meeting
Practical Info This is a weekly video chat meeting for planning and discussion of Blender Eevee/viewport module development. Any contributor (developer, UI/UX designer, writer, …) working on Eevee/viewport in Blender is welcome to join. For users and other interested parties, we ask to read...
Lightwave3D has been purchased away from Virtz. So far it looks as though things are going to get real good, although it is going to take a while. Support for native Apple Silicon is on the list of things.
I worked with Andy Bishop (the new owner) back in the day. Really nice guy and genuinely passionate about Lightwave. Tbh couldn't imagine a better person to be in charge of Lightwave's development; hoping he can turn things things around there.
After years of following this thread I finally got an Apple silicone computer and did some C4D/Octane work. The responsiveness is very nice however I have to ask anyone else with a M1/2 MBP - when rendering does your laptop get super loud?
I know it's a very demanding task for any computer to do, but all I heard leading up to my purchase was that these computers stay cool and quiet under heavy workloads.
For reference I have an M2 Max 14in and I've been using the latest software versions: C4D 2023.2.1 & Octane 2022.1.1(r4) and macOS 13.3.1
I know it's a very demanding task for any computer to do, but all I heard leading up to my purchase was that these computers stay cool and quiet under heavy workloads.
For reference I have an M2 Max 14in and I've been using the latest software versions: C4D 2023.2.1 & Octane 2022.1.1(r4) and macOS 13.3.1
Might be these are not Apple silicon optimized variants of the software, but are running thru Rosetta 2...?
I believe Octane X is ASi optimized, and Maxon should be working on an ASi optimized variant of Cinema 4D as well...?
I got to thinking the 14in might be the issue unfortunately but turning the render priority down to low reduces noise to a nice level - but now it has the same performance as an M1 haha
And yes this version of octane is Apple Silicone only - running native.
And yes this version of octane is Apple Silicone only - running native.
C4D was native since the first iteration of AS, actually one of the first native software in general.
Have a 14" M1 MBP and it gets loud when hitting GPU (or in the worst case CPU and GPU) at the same time. CPU is pretty close to silent or very quiet even when rendering with all cores. It seems mainly the GPU that causes the fan to kick in.
Still quieter than my old iMac Pro, which got frustratingly loud very quickly (or 2016 15" MacBook Pro for that matter).
I have Studio Max and Ultra also at the at the our studio we have got Studio Max 32GB, Studio Max 64 GB a couple of 16 MBP's a couple of PC's ext. I have a lot insight about AS devices.
For average 3D workflows Studio Max with 32 GB Ram is not enough. With 64 GB ram could be but still there is CPU core counts issues for simulations ext.
We never use MBP's for 3D rendering. Because they are still certain degrees of CPU/GPU temperatures and fan noise even with Davinci Resolve while 4K/6K workflows. I have no experience with M2 MBP, 96 GB of RAM could be + but it will still reach certain degrees of CPU/GPU temperatures and fan noise while 3D rendering sessions. I think fan noise is fine at MBP's, but I am not sure for body components which could effect from that heat ( connect ports could be very hot ).
On Studio Ultra I have no issues while working on 3D or overnight 3D rendering sessions. It is the most reliable device I ever used. There is no fan noise but still it could be warm on the chassis which is perfectly normal.
Other thing is about Octane. Octane uses much more Vram/Ram than Redshift. Arnold is using possible small size of RAM. Vray is similar with Redshift about Ram usage. If its possible I would suggest you try Redshift or Vray. It will make a difference based on my experiments.
No 16 inch not quite under heavy workloads. You could hear fan noise very clear & loud. Also 16 inch is reaches maximum degrees quicker than 14 inch MBP's.