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.
MP 7,1 GPU rendering on the new Mac Pro. News, updates, chat.
- Thread starter vel0city
- Start date
- Sort by reaction score
it should work but before investing in that setup you better ask to developers*, also I would wait for the final release.
I know for sure that Otoy team was using iMacPro with 4x eGPU-Radeon VII, so that should work.
*be sure to ask specifically to developers since average forum users may or may not give you proper information.
I know for sure that Otoy team was using iMacPro with 4x eGPU-Radeon VII, so that should work.
*be sure to ask specifically to developers since average forum users may or may not give you proper information.
Last edited:
I have this configuration but with just one Razer Core external enclosure.
I use daily 3 gpus: 2xRTX2080Tis + 1XGTX1080Ti on tb3
Unfortunately the bottleneck is the tb3 bus, one GTX1080Ti plugged on the pci-x slot is much faster (+30%)
I use on bootcamp 3ds Max + Fstorm.
When the RTX3080Ti will be available probably I will replace one 2080Ti and I will move the second one into the external box. I read the power connector of the new RTX is different, also I don't know if the Mac Pro will handle 2x 320w external power (Belkin cables).
I don't think AMD will release a Navi gpu faster than the new Nvidia Ampere, also realtime raytracing is a temporary feature for AMD on Apple Macs because the new Mac Pro 7.x will not have any branded gpu.
We have to wait for a professional gpu/co-processor by Apple working on a proprietary bus, forget about pci 4 and 5, exclusively designed for realtime 3d rendering.
I use daily 3 gpus: 2xRTX2080Tis + 1XGTX1080Ti on tb3
Unfortunately the bottleneck is the tb3 bus, one GTX1080Ti plugged on the pci-x slot is much faster (+30%)
I use on bootcamp 3ds Max + Fstorm.
When the RTX3080Ti will be available probably I will replace one 2080Ti and I will move the second one into the external box. I read the power connector of the new RTX is different, also I don't know if the Mac Pro will handle 2x 320w external power (Belkin cables).
I don't think AMD will release a Navi gpu faster than the new Nvidia Ampere, also realtime raytracing is a temporary feature for AMD on Apple Macs because the new Mac Pro 7.x will not have any branded gpu.
We have to wait for a professional gpu/co-processor by Apple working on a proprietary bus, forget about pci 4 and 5, exclusively designed for realtime 3d rendering.
Attachments
Navi? No. But AMD isn't aiming Navi at Ampere. At least the computational Ampere.
Exclusive: AMD Radeon Instinct MI100 Specs, Performance, and Features
Technology Vision
adoredtv.com
This rolls in somewhat under to 2x the Pro Vega II Duo ( not the single, the Duo. If the numbers pan out in the released system: 42 TFLOPS versus 28 ). Not ray tracing only hardware, but it is a very big hammer for generalized code composed by 3rd parties.
Zero necessary need to wait for a co-processor from Apple. If Apple needs heavyweight SP Float compute 'power' all they need is a driver that looks a lot the one they already have for Vega. Will they 'fumble' getting one out? Sadly, pretty decent chance. ( The MI100 can't be coupled to some TB controllers and pump out end user graphics so decent chance Apple will ignore it. )
Don't really need PCI-e v4 for real time rendering. Just better algorithms and some hardware support. For the current Mac Pro if Apple enabled the ability to dump the finished computation data back to a Pro Vega II via Infinity Fabric, then they could get away with the "Compute" GPU not having any display output and still not run into a PCI-e v3 bottleneck. ( not particularly a huge leap from having support for two Pro Vega II and a display only hooked to one of them. )
It is a very big unjustified leap that Apple's real time ray tracing support is only going to be via Apple proprietary hardware. That is doubtful. Apple may drag their feet on extending Metal to support that until they have both theirs and some other 3rd parties to roll out with. But Apple only isn't going to buy them much of anything.
Otoy gave me a week license to test in C4D. I have a Vega II duo. Was snappy and seemed to have most everything in its place. I would not call it stable quite yet. Crashed more than a few times, but they know about the lack of stability and are on it. I look forward to Big Sur and the permeant license.
I understand this is a Mac Pro thread and the only thread where folks are discussing Octane X. I am just dipping my toes into the world of CG and was wondering if anyone has had the opportunity to try the preview version of Octane X on the new iMac with the 16GB 5700XT GPU?
I don't have the 16GB 5700 XT but I do have the 8GB version in my Mac Pro. It's way more stable than when I was using an Nvidia 980Ti and speed feels similar if not quicker (though frankly speed is of no use without stability). I'm taking delivery of an identical Sapphire 5700 XT tomorrow, so fingers crossed it'll just work and I can render faster than I have ever done before. With regards the iMac I see no reason why the 16GB version shouldn't be pretty much the same, and potentially more stable. I have had issues which were definitely memory related.
Thanks @skippermonkey !!! I might just pull the trigger on the new iMac. I have been digging through the Apple Silcon WWDC videos and the approach they have taken for the GPU/CPU architecture is interesting.
As well as the amount of time OTOY has spent with Apple getting Octane-X on to metal. Definitely interesting times ahead but I am skeptical of how the new apple architecture is going to handle the complexity of CG work.
Question to those doing cgi ~
I'm investigating cost per minute and cgi varies significantly based upon variables.
A colleague said that it probably cost $10,000 for the spaceship scene in this short film.
Is that about right? Which software would the cgi artist being using for something like this?
I'm investigating cost per minute and cgi varies significantly based upon variables.
A colleague said that it probably cost $10,000 for the spaceship scene in this short film.
Is that about right? Which software would the cgi artist being using for something like this?
You can build a scene like that on most renderer(both CPU and GPU based). Personally I would ask more than 10K since water simulations requires more work. There are about 2 minutes of CG animation(3600frames) that alone will cost quite a bit to render on a farm.
You will need a modeling software(Maya, Cinema4, Blender, ecc), a texturing software(Mari, Substance, ecc) and a simulation software(Houdini, ecc). Then as said you can use any commercial renderer(well most of them) for the final frames, like Vray, Arnold, Renderman, Octane, Redshift, ecc.
You will need a modeling software(Maya, Cinema4, Blender, ecc), a texturing software(Mari, Substance, ecc) and a simulation software(Houdini, ecc). Then as said you can use any commercial renderer(well most of them) for the final frames, like Vray, Arnold, Renderman, Octane, Redshift, ecc.
Last edited:
That gets you to completed assets, but don't forget the time / effort / software to properly composite those assets into the shot(s)...!
To expand on what I posted earlier: I think you will want to be looking for a generalist who approaches a shot with the end result in mind, willing to fake whatever can be done on the cheap and who knows enough common VFX techniques and most aspects of their tool of choice to not trip over technical hurdles.
Check out this talk around the 4:00 mark for a few minutes to see what I mean:
The other way would be to build proper models, texture, light, simulate, render and compose them, that's probably work for a small team of more specialized people with a whole bunch of software and concerns like how to simulate or render all the necessary bits within the time limit. Obviously this comes with a lot more budget attached.
Check out this talk around the 4:00 mark for a few minutes to see what I mean:
The other way would be to build proper models, texture, light, simulate, render and compose them, that's probably work for a small team of more specialized people with a whole bunch of software and concerns like how to simulate or render all the necessary bits within the time limit. Obviously this comes with a lot more budget attached.
We are just now getting proper GPU rendering support, and I was super excited what the possibilities of Apple Silicon would mean for 3D production and rendering on macOS, but this spat with Epic has me worried about the future of content creation on our platform. UE5 looks super interesting and loosing it on macOS would be a serious blow to the platform for our work.
Apple’s response to Epic threatens the future of VR, AR, TV, and filmsExhibit B in the lawsuit between Epic Games and Apple could easily become Exhibit A in a U.S. antitrust case, clouding the maker of iPhones and AR software.venturebeat.com
Apple’s response to Epic threatens the future of VR, AR, TV, and filmsExhibit B in the lawsuit between Epic Games and Apple could easily become Exhibit A in a U.S. antitrust case, clouding the maker of iPhones and AR software.venturebeat.com
Yeah the gaming benchmarks look promising. Nothing beating the Nvidia options but parody is nice! It'll be interesting to see if Metal API can tighten the gap. I'm really excited for the Redshift release and it sounds like it's ready for the Big Sur release. Also interested to see if these chips come to Mac Pro in the MPX form factor. If so maybe we'll get more VRAM, 16 is nice for rendering but simulating on the GPU could benefit from the extra memory, or even dual chip cards like the Pro Duo.
If the latter happens that'd be stellar since the 3090 is pushing the limit when 4 are in a standard workstation.
If the latter happens that'd be stellar since the 3090 is pushing the limit when 4 are in a standard workstation.
Well, unfortunately you won't get an answer, because both developers insist on an NDA and no information can be published until it is allowed. We have to wait for an Public Release on that.
AMD 'juiced' the 6900 benchmarks presented a bit. They used the "Smart Memory Access" feature that only works with the newest Ryzen 5000 series CPU packages and the newest , bleeding edge logic boards (and BIOS/UEFI). Also had "Rage Mode" on ( which is heuristic overclocking). Extremely doubtful either one of those show up in the Mac drivers ( on Intel ). (if the "Smart Memory Access" is something proprietarily quirky over PCI-e ( something past setting up shared MMU mappings and cache coherence. ) then probably not going to see it on the Apple Silicon either).
Those two are around 4-6% worth of boost so would slide backwards a bit with an Intel CPU where the driver stack didn't overlocking. ( usually not lots of 'whiz bang' nobs on Mac graphics drivers. ) . Of the three cards presented the 6900 is likely to be the one that is hiding behind a fig leaf the most once the independent reviews expose how much 'cherry picking results' the non-objective benchmarks represent.
.
However, it looks to be reasonably close in classic rendering. They are probably behind on a wider set of workloads but the gap isn't as large. They didn't have anything "super duper" to show with raytracing though. Likely better than zero hardware support, but doubtful they have leapfrogged Nvidia with their first iteration. ( I suspect there is some cache contention with trying to run Ray Trace at same time something else want to heavily lean on the internal cache ( "Infinity Cache"). The cache is a 'hot rod' boost up until have a highly active working set that is more than 128MB of data to cache. That probably contributes to why they aren't sounding the trumpets early on this feature.
The lower level driver stack gaps (and software libraries layered on top) that are probably be a bigger differentiator than the ''raw' hardware here. If Apple is working highly synergistically with AMD to make the AMD hardware shine then fine. If Apple has a "our GPU first" and the others are "copious spare time" effort then this hardware isn't going to close the gap. It doesn't out 'brute force' the Nvidia offering. It should make Apple 'happier' when measuring with their favorite performance/power yardstick.
Now it’s a question of Redshift vs Octane. Performance, price, stability. I’m running Octane at the moment, but could quite happily switch if Redshift is more performant and even slightly more stable, especially as I use C4D anyway.
Rats. No demo, just the beta for existing license holders.
Rats. No demo, just the beta for existing license holders.
Last edited:
I started a thread on the AS forum about 3D rendering on the new Macs: https://forums.macrumors.com/threads/3d-rendering-on-apple-silicon-cpu-gpu.2269416/