Mac: M3 - *Hardware accelerated RT (Part 1)

[claus]

I’m dissappointed with @cekuhnen and @anon55679826 's behaviour in this thread. You have both repeatedly stated that the Max performance competes with “the 3080”. @Bullit is trying to nuance the fact that there is no such thing, and it needs to be specifified what type of 3080 is being used in performance comparisons since there is objectively an enormous difference between the types of 3080’s out there. It has no meaning and it is misleading to do a general comparison with “a 3080” without specifiying which one exactly. Please concede that fact, without getting angry about it and shouting that you don’t care. If you don’t care, you shouldn’t be comparing performance.

You guys oddly took that as an attack in general on the latest macbook pro, and almost seem to take it personally. There is no point in argumenting other benefits the macbook pro has in a discussion that is purely about performance. Stay on subject. @Bullit doesn’t hate the macbook pro, so don’t respond as if he does. Bullit, what they want to hear from you is acknowledgement that the Max is an impressive achievement when taking in all factors.

That said, I’m eagerly awaiting Metal Blender so we can actually see what it’s capable of, instead of endlessly speculating. I’d love to see less personal bickering and more updates on the topic of this thread here.

[Tiwaz (Steve)]

In rasterization it is somewhat close however depending on the task we in compute the M1 Max seems to have no chance in is about 2,68 times slower than a 3080.

example from redshift forum

If anybody is interested in up to date 3080 laptop specs, since updating Redshift, all drivers of my Legion 7 + jumping to Windows 11, the benchmark is now taking 23 seconds less than before! (Make sure you update AMD chipset drivers on Win11)

Redshift 3.0.59 (Windows)
CPU: 16 threads, 3.29 GHz, 27.86 GB
GPU(s): [NVIDIA GeForce RTX 3080 Laptop GPU 16 GB 0.008ms (RTX ON)]
Time: 00h:03m:45s

aslo from the redshift forum:
I did the benchmark with M1 Max 32 core, 64 go and I get 10min03 sec

Also unfortunately the M1 Max does not have RT cores.

Redshift 3.0.58 (Windows)
CPU: 32 threads, 3.40 GHz, 31.92 GB
GPU(s): [GeForce RTX 3060 12 GB 0.008ms (RTX ON)]
Time: 00h:04m:44s

Even the 3060 is quite fast in that sense.

So I would not expect to much for metal gpu rendering in blender.
Now viewport metal could help a lot more

Also I mentioned this before it all lines up as the M1 Max has close to 1/3 of the TFlops of the 3080

If we can to beat the 3080 for that task you well need an M1 quadro (or whatever they might call it) with 128 core GPU
Actually if they make that it will probably beat the 3090 even.

Also if I want the close to perfect match for the M1Max

I’m on macOS 12.0.1:

Redshift 3.0.59 (macOS)
CPU: 8 threads, 2.70 GHz, 16.00 GB
GPU(s): [AMD Radeon RX 5700 XT 8 GB 0.073ms]
Time: 00h:10m:21s

I think that is still very amazing for integrated graphics.

Sorry it got a bit long.

[anon55679826]

@claus

I’m dissappointed with @cekuhnen and @anon55679826 's behaviour in this thread. You have both repeatedly stated that the Max performance competes with “the 3080” @Bullit is trying to nuance the fact that there is no such thing, and it needs to be specifified what type of 3080 is being used.

We’re not flat out randomly stating those claims past just sharing articles and videos with benchmarks showing that the M1M is competing with the 3080. Each video and article states which machines is used and which card. Bullit starting the obvious after each and every video has become old.
As much as you want him to be the victim, the replies of “…manipulated by Apple propaganda” and “…another Bull $h1t video” is exactly what it’s meant to be, troll material.

@claus

It has no meaning and it is misleading to do a general comparison with “a 3080” without specifiying which one exactly. Please concede that fact,

Every article and video does this, but if not, we are all big boys and girls here, and know which cards are being used, and it’s suuuuuuper obvious to the folks in this thread. It’s also incredible amazing that the M1M is doing this.

Should I just scroll past his reply after the twentieth time? Yes and I will now, but it makes it harder when he’s quoting my reply to alert me and let me know, again.

And I’m not the only one calling him out on this.

[anon55679826]

@claus I think you owe cekuhnen a public apology for publicly calling him out.

I just want to defend @cekuhnen cause he probably wont do it himself, being that he’s was being civil with his replies, and could recognize being trolled.

@cekuhnen

Yes there are different 3080. Also there are probably differently performing 165W 3080.
We all know that.

3080 is used as a ballpark.

@Bullit

So two very different quality values are now a ballpark?

If the ballpark here is just the tag 3080 then any propaganda can manipulate you.

@cekuhnen

I am not manipulated by propaganda - simply because I know this are rough estimates - ballpark expressions. And maybe you should refrain from making such personal accusations here.

@Bullit

Not only you are manipulated a more precise word would be by branding, but now by your answers it is not unconscious it is with full knowledge.

[Metin_Seven]

@cekuhnen

Getting my cardio stress tested

Test faster nurse !!!

Hmm, what’s that nurse doing?

[dcrosby]

Save me a slice!

[cekuhnen]

@claus

I’m dissappointed with @cekuhnen

Maybe it is time to take a deep breath here.

I think it is more than obvious that with a new piece of hardware for rendering like the m1 max one has to be careful when compared to other dedizier GPUs.

Currently the m1 Max is mainly used in a labtop so thermal development is a limiting factor.

Further software is not even fully updated and optimized.

Even in pc tests u can take the same GPU and get different results based on well a lot of bottleneck factors.

In addition when you see test charts that just say 3080 with out any further information your basic high school teaching (analyzing data) should be enough to make you cautious. The same applies for all the YouTube videos people made - check the data.

Everybody here knows that btw.

I have a super busy life - family - baby - toddler - 75 university students I mentor plus my professional
client work and research.

I worked this night till 2:31 am.

This is the last time I respond to this topic (m1 vs what xyz 3080 ) here in this thread.

There are seriously more important topics to discuss here than exercising an exhausting discussion about proper bran counting.

[Renzatic (I Love You, Dinosaur Tree)]

…too late!

[Remyglass (Remy)]

I did a couple gaming tests comparing the M1 Pro GPU (16 core) to the RTX 3070 in my g15 laptop. In metal optimized games it outperformed the rtx 3070 (90w) by quite a bit. For instance, using exact same settings in Dirt Rally benchmark, the m1 pro averaged 96 fps with a low of 84fps, the rtx 3070 averaged 76 fps with a low of 51fps. I know this doesnt directly translate to blender when its running metal, but it does backup apples claims. Keep in mind this is the m1 pro I was testing, not even the max.
(waiting for Bullitt to tell me how this test is unfair… The windows laptop was on turbo mode which is a slight overclock and was pulling all the power it could from my wall)

M1 Pro vs RTX 3070 Gaming Test - Dirt Rally

[Remyglass (Remy)]

As someone who had the 14" but then decided to get the 16" because you get so much more for the money…well I am moving back to a 14". Its still fast as hell and honestly the 16 is MASSIVE…and heavy. I cant imagine trying to open this on an airplane. Even sitting in bed with it is sorta strange…the 16 is best used on a table.

[hatchli (Wade)]

Yah I’m a cafe and travelling warrior, so the 16” just seems like a non-starter despite the lovely real estate.

[anon80315389]

What’s more, the Mac rendered 10% more pixels.

Unfortunately the comparsion is unfair. I think you should set the 3070 TDP to 165W first to make it fair

[anon55679826]

Wait! You’ve been doing these test with the Pro model and not the Max? F-ing thing is a beast!
So ready to dump my Windows laptop!!!

[anon80315389]

@Remyglass

I am moving back to a 14"

You’re not alone:

Why I RETURNED my 16" M1 Max MacBook after Extensive Tests..

[amartinezch (Alej M-C)]

@jamesbstudio

I completely underestimated how these tests were going to take… but, all done now. Have run quite a few tests on the M1 Max 16” MBP now. It was fun, and eye-opening, to see what this machine can do. Hope it is as useful to you, as it was to me.

Blender on the M1 Max 16” MacBook Pro - An Animator’s Perspective

Thanks a lot for these. Really nice tests.
Would be great to have a comparison point of some of those with the likes of the newest Surface Studio Laptop that has been mentioned lately, i.e, the max amount of duplicated/instanced geometry that can be handled.

On the Surface Studio Laptop:
@Renzatic I don’t know if you have seen it, and excuse me for always referring the Max Tech channel (they often have a direct comparison point to what we are discussing), there’s exactly one test of a M1 Pro 14” binned version vs a bit more expensive latest Surface Laptop with RTX (to honour the discussions here, fine with the whatever the low wattage is that explains the crippling performance, what was it around 80W?):

• The video is called: 14” MacBook Pro vs RTX Surface Laptop Studio
• A bit of the usual, Geekbench RTX has double the scores of the M1 Pro but the Wild Life Extreme ‘real game test’ performs considerably worse
• They do mention the inclined tablet mode (for using the pencil and the keyboard) or full tablet mode (with the whole screen flipped converting the whole keyboard). This is indeed very very nice. Don’t know if this laptop comes with the pen or it needs to be bought.
• Blender Party Tug renders are still faster on the RTX, ~6s vs 9s.
• Everything else seems to be considerably less performant (file copying, battery life, etc)

That said, I think what you mention of letting it sit for a couple of months isn’t a bad idea, the latest Alder Lake benchmarks appearing here and there seem to be quite the performant powerhouses, provided that they do handle thermals and everything else properly, it might be ideal for those that would like to stick to Windows if that’s their favorite or familiar OS.

[Remyglass (Remy)]

In apples presentation and on their website they list the EXACT systems they compared them to. If you guys actually read the claims from Apples own website you wouldnt be on here wasting your time trying to explain that a leafblower is more powerful than the m1 max.

[Remyglass (Remy)]

yes all of my testing has been on the pro. but I do have a 14in max arriving next week. Getting rid of this 16" massive unit.

[Tiwaz (Steve)]

I am still a bit thorn about that too, but loving how quiet the air is I will probably stick with the 16”.

[Tiwaz (Steve)]

Let me know how that goes and sounds

Quite curious what you think considering in some workloads I have seen it is like 4k (4300 if I remember right) something rpms vs 1700k rpms.

14” pro?

[Remyglass (Remy)]

I had the m1 pro 14" before getting the m1 pro 16" which im currently using. The fans even when on full speed are not very loud on the 14in. I have seen a couple youtubers say the fans at equal rpm are actually louder on the 16 but they rarely spin up like they would on the 14. And compared to my windows laptop the fans are completely tolerable…the g15 has a high pitch fan that drives me crazy. At least apple made sure their fans arent annoyingly high pitched…more of a woosh sound.