Turns out the M1 pro gpu (16 core) is faster than my rtx 3070 laptop when using a metal-optimized game

[Ceed]

The sarcasm was supposed to imply 1: in the end, all anyone does is waste their time, and 2: the most successful people are the ones who can afford to sit around and do nothing but entertain themselves (and if you find joy in learning a language, or you find joy in helping other people, or you find joy in creating things, you are on equal ground with the person finding joy in movies and video games, because we're all on the clock and nobody and nothing matters aside from happiness). Sorry if that didn't come across.

Getting a little too deep here for a thread about M1 gaming, apologies. Carry on.

 

[dgdosen]

And yet you are wasting time arguing with strangers on a crazy tech forum... Sorry, but that's just funny

I'd think people on this forum could benefit from reading a Cal Newport book or two...

 

[jeanlain]

Well, even M1 pro GPU should pack many more transistors than mobile RTX 3070 ( 33B with majority dedicated to GPU vs 17.4B with 5/6 of shaders enabled)

By looking at the M1 Pro die shot, it looks like 1/3 of the area is occupied by the GPU.

https://images.anandtech.com/doci/17019/M1PRO.jpg

 

[Kpjoslee]

By looking at the M1 Pro die shot, it looks like 1/3 of the area is occupied by the GPU.

https://images.anandtech.com/doci/17019/M1PRO.jpg

Well, pure compute area is about 2/3 of the GPU on GA104. I am just throwing out rough estimation of performance compared to other GPUs by looking at the total area and how fast the frequency is. While not very scientific way of doing it, I find it pretty decent way of estimating where certain GPU should fall in terms of performance against other GPUs with similar transistor count.

 

[eltoslightfoot]

I blow glass for about 4-5 hours and then waste the rest of my day doing random stuff.
I also invested in $TSLA which allows me to be a real lazy f*ck when I want to
gives me time to post on here and make silly youtube testing videos ?
I also have a kid so he keeps me pretty busy

Hahaha! This was a fantastic response. Bravo!

 

[hefeglass]

The sarcasm was supposed to imply 1: in the end, all anyone does is waste their time, and 2: the most successful people are the ones who can afford to sit around and do nothing but entertain themselves (and if you find joy in learning a language, or you find joy in helping other people, or you find joy in creating things, you are on equal ground with the person finding joy in movies and video games, because we're all on the clock and nobody and nothing matters aside from happiness). Sorry if that didn't come across.

Getting a little too deep here for a thread about M1 gaming, apologies. Carry on.

I want this quote on the wall of my studio

well said

 

[theotherphil]

Interesting … do you have a sense of wether the performance loss on each low power mode (battery/AC) is the same? Are the low power modes equivalent?

Aleksander Ziskind tested this but this was only on CPU bound tasks (compiling code) and found some interesting differences.

 

[theotherphil]

The sarcasm was supposed to imply 1: in the end, all anyone does is waste their time, and 2: the most successful people are the ones who can afford to sit around and do nothing but entertain themselves (and if you find joy in learning a language, or you find joy in helping other people, or you find joy in creating things, you are on equal ground with the person finding joy in movies and video games, because we're all on the clock and nobody and nothing matters aside from happiness). Sorry if that didn't come across.

Getting a little too deep here for a thread about M1 gaming, apologies. Carry on.

I understood what you meant...wrong audience though ?

 

[ikir]

Amazing. Just wish more games would support Metal. Sigh. I guess gaming on macOS isn't that popular enough for game developers to justify. I was using Boot Camp to game before my M1 Max and I'm missing it. Tried Parallels but it is still not as responsive as Boot Camp. Bought a Dell XPS 8940 just for gaming.


Haha. ? That's the first thing I noticed actually. I was gonna PM OP about it but never mind.

Games will come if us (users) will support Mac gaming. It takes time. Crossover is very promising.

 

[nathansz]

i would edit title to read “mobile rtx
3070” to be fair

 

[hefeglass]

i would edit title to read “mobile rtx
3070” to be fair

fixed

 

[jdb8167]

Aleksander Ziskind tested this but this was only on CPU bound tasks (compiling code) and found some interesting differences.

Interesting tests. One thing he doesn’t specify is how large is the SSD on each machine. This will have significant impact on the Xcode benchmark. I’m guessing at least 2TB on the M1 Max.

Just for reference, on my M1 8-core/16 GB/1 TB MacBook Air I get about 135 secs on this test in normal mode. I haven’t tested in low power mode.

 

[theotherphil]

Interesting tests. One thing he doesn’t specify is how large is the SSD on each machine. This will have significant impact on the Xcode benchmark. I’m guessing at least 2TB on the M1 Max.

Just for reference, on my M1 8-core/16 GB/1 TB MacBook Air I get about 135 secs on this test in normal mode. I haven’t tested in low power mode.

The absolute performance difference between each individual machine isn't being tested here - that's a prior video where full specs are detailed.

This video is purely testing the drop in performance on the same model, not between models in response to the power setting. If you're looking at the performance delta between the same model with low power vs high power, the SSD won't have any bearing....because it will be the same SSD each time.

eg, If I buy a 14" Max, I can expect to see a ~10% performance drop in low power mode.

What wasn't taken into account was GPU workloads but it makes an interesting starting point for further testing.

 

[hefeglass]

The absolute performance difference between each individual machine isn't being tested here - that's a prior video where full specs are detailed.

This video is purely testing the drop in performance on the same model, not between models in response to the power setting. If you're looking at the performance delta between the same model with low power vs high power, the SSD won't have any bearing....because it will be the same SSD each time.

eg, If I buy a 14" Max, I can expect to see a ~10% performance drop in low power mode.

What wasn't taken into account was GPU workloads but it makes an interesting starting point for further testing.

I did a comparison a couple weeks ago between the 14 m1 pro and 16 m1 pro in normal mode and lower-power mode to see if gpu performance is affected. I ran GFXbench Metal on each machine and the 14 has lower scores for just about every test ..the 16 had some lower scores, and some exactly the same. Apparently low-power mode is more aggressive on the 14 which makes sense. Here is the video and some screenshots of the graphs if you dont want to watch the video.

 

[theotherphil]

I did a comparison a couple weeks ago between the 14 m1 pro and 16 m1 pro in normal mode and lower-power mode to see if gpu performance is affected. I ran GFXbench Metal on each machine and the 14 has lower scores for just about every test ..the 16 had some lower scores, and some exactly the same. Apparently low-power mode is more aggressive on the 14 which makes sense. Here is the video and some screenshots of the graphs if you dont want to watch the video.

Thank you for this! Your results jive with Aleksander's in that the 14" is more aggressive in it's low power mode. As you say, it makes sense given the smaller battery.

Given the 100WH battery vs the 70WH....1% on the 16" is actually 30% larger in absolute capacity than 1% on the 14".

So if they both enable low power mode at 20%, that means the 14" is at 14WH whilst the 16" still has 20WH left.

 

[jdb8167]

The absolute performance difference between each individual machine isn't being tested here - that's a prior video where full specs are detailed.

This video is purely testing the drop in performance on the same model, not between models in response to the power setting. If you're looking at the performance delta between the same model with low power vs high power, the SSD won't have any bearing....because it will be the same SSD each time.

eg, If I buy a 14" Max, I can expect to see a ~10% performance drop in low power mode.

What wasn't taken into account was GPU workloads but it makes an interesting starting point for further testing.

Yes, that is true but the absolute performance is interesting as well.

I couldn't find any video where he shows the full specs. He ignores the SSD size in all of his Xcode benchmarks as far as I can tell. If you go to the Xcode Benchmark on GitHub that he is using, you can see that SSD size makes a fairly large difference in performance (you might have to look at the pull requests since the Readme hasn't been updated in a while.)

 

[mi7chy]

Still fishy how a 3070 in a thick laptop is lower than 3060 in thin laptop. Plus, these comparisons are silly since they're comparing API overhead. Metal should be compared with other bare metal APIs like Vulkan and DirectX 12 instead of 2015 game using older non-bare metal DirectX 11.

Dirt Rally 1440p ultra 3060 100W

Dirt Rally 1440p ultra 3060 70W

 

[Adarna]

Check again by year 2023-2025 if AAA games are announced/out for the Mac. By then 2nm iPhones/iPads/Macs will be out.