It’s a pity that the article doesn’t even mention Flame. Since they had the system for a while to record this video and write the article they could at least run the Flame benchmark and mention it as a sidenote.
But considering the comparisons between the MacBook Pro and the Studio, I guess the frame-based render times will be around 4:40-5:00 range.
I don’t know about benchmarks - but I just loaded a project that had a timeline that had 32 shots to render. it was so fast it appeared to skip the first 20. did the whole thing in about 10- seconds. on my intel Mac Pro 2021 it would ve gone through all of them and taken about 40 - 50 seconds.
What do you mean? Flame does talk to BMD UltraStudio card on M1 configurations.
I had some time today so I ran the bench on a MacPro and my M1 Max MacBookPro
First Wario, my MacBookPro
…and Toad, the MacPro
Toad is a beefy box and substantially more expensive but for all its heft it’s only 30 seconds faster.
I’m guessing an Ultra is going to be around 14. But it would be great to see.
Oooh…just ordered my new M1max MBP yesterday, and its in stock too!
I had the exact opposite result. The benchmark took 26:15 on my 28 core MacPro with 256 GB RAM, and 25:56 on my new M1max MacBook Pro with 64GB.
hey call - downloaded the benchmark archive. I’m getting this so far - not sure if I’m doing it exactly right.
how do I get the end duration display you have
Well, I time it with a chronometer. The first frame takes a while to render, probably doing some internal initialization. So I start the render, stop it, unrender, then start again so it immediately starts to render.
I think the take/home is really that they are basically the same speed when all is said and done.
What I’m having a hard time understanding are people who have half the time you and I are seeing on that google sheet with the same hardware. I suspect they didn’t render until the end, but regardless it makes the sheet an unreliable document.
The other take home is that unsurprisingly the Ultra is nearly double in speed compared to the max. I’ll run the bench again in my i9 at work just for comparison.
@Jonhollis any way I can ask you to run the bench one more time? This time, select all, flush renders, park at the top layer, and select “render” and not “render selected”
I’m guessing you end up around or even sub 7 minutes. Either way it would be very telling.
Clearly still way behind the most bleeding edge Linux systems, but less embarrassingly so. Progress! Looking forward to the MacPro this summer.
Indeed. I am curious to see if we can narrow down what’s happening. I suspect the top times on the sheet are rendering with “render” on top layer as opposed to the “render sel” with all layers selected that I believe @Jonhollis did. If that’s the case, and it brings him down to around to 6-7 (which is my guess only) that puts an M1 Ultra only about 30%-ish off from the leaders (all threadripper/a6000rtx/3090s for the most part). If Flame goes native for the chipset (these benches are all still running under emulation), those results could become very very close. All of this on a box that is basically 6k US. An a6000rtx is 5k.
Lots of “ifs” here but that seems rather befitting of iffs…
The benchmark definitely shouldn’t be considered the be all and end all. We really should have 3 or 4 specific tasks as performance will vary depending on what you are doing.
That being said, I suspect the reason the Threadrippers perform so well has a major part to do with the PCI 4 bus as opposed to PCI 3 that all the other Linux systems have. As we see more PCI4 PCs released, the render times will greatly reduce on the HP/Dell Intel based systems.
For all Flames, the GPU is being talked to using OpenGL API and that is depreciated on MacOS (by Apple) and has never performed as well as a Linux or Windows system. M1 chips are also being used via Rosetta for Flame. Performance of Flame on the Macs is definitely going to improve in future. Macs are more versatile too so it isn’t always about performance.
Still, if I was buying a system predominantly for Flame, Id be buying a Linux Threadripper based system right now.
Great news. So almost twice as fast as the MacPro and M1 Max.
Here’s a MacPro with “render”
…here’s an M1 Max with “render”
And lastly here’s my daily driver: a Linux box, i9 and TitanRTX also with “render”
So that Ultra is faster than my day job machine by a full minute running under emulation. Now, wouldn’t it be great if it were native? Regardless thanks for taking the time to run these tests @Jonhollis.
Edit: I don’t know know why I made these snazzy little graphics, I just did.
I couldn’t compete on the snazzy graphics.