M2 Pro and M2 Max

[exoticspice1]

They will fare better obviously but still at a huge power consumption cost. The leaked TDPs of mobile series 40 are hardly inspiring

What cost? The 40 series are more efficient than previous gen and uses 100 watts less power while being more powerful in raster, RT and also having lower temps. If Nvidia uses the same 5nm arch and the same Ada arch I expect efficiency improvements as well for the laptops cards.

Not everyone is Intel pumping up more more watts every gen to make speed bumps.

 

[Jimmyjames]

Interesting article on the P cores and E cores in the M2 Max.

M2 Max: Under the Hood With the Latest Apple Silicon

Many years ago, I used to give a presentation to business school students specifically about the semiconductor industry. Part of that presentation focused on the wonder of semiconductors where we can marvel at the bigness and smallness of these magical pieces of science. Appreciating the bigness...

creativestrategies.com

 

[Jimmyjames]

More proof of how close the M2 Max is to the M1 Ultra. This time Pugetbench.

 

[Jimmyjames]

3D Wildlife Extreme test. No idea how it compares with Nvidia and AMD.

 

[exoticspice1]

3D Wildlife Extreme test. No idea how it compares with Nvidia and AMD.

View attachment 21472

The 3DMark tests are different on PC and also incorporate RT for some tests and and are also more demanding but at least with wildlife extreme we can compare with past Apple GPUs.

 

[exoticspice1]

More proof of how close the M2 Max is to the M1 Ultra. This time Pugetbench.
View attachment 21471

This shows the M1 GPU was a testing ground. I expect Apple to catch up to AMD soon with M2 Ultra if it scales wells.

 

[theorist9]

3D Wildlife Extreme test. No idea how it compares with Nvidia and AMD.

What's the link for that? I can't tell from the graphic if that's the 30-core or 38-core M2 Max. [Always good to post links anyways, to give the person who created the graphic credit.]
If it's the 30-core, the scaling from the 10-core M2 is 98%; if it's the 38-core, then it's 77%.

Edit: I emailed the author, and he said it's the 38-core. So 77% scaling relative to the M2. He also said that they didn't use high-power mode when testing it.

 

[Jimmyjames]

What's the link for that? Can't tell from the graphic if that's the 30-core or 38-core M2 Max. [Always good to post links anyways, to give the person who created the graphic credit.]

Sure. It's the Arstechnica review

2023 MacBook Pro review: A refined second generation

The M2 Pro and Max and improved connectivity headline Apple’s laptop refresh.

arstechnica.com

 

[leman]

Yeah that would VERY hard. Unlike AMD and Intel, Nvidia is competent and uses the lastest nodes from TSMC now. Plus cuda and Optix are great too. Their RT is also industry leading.

My take is on Apple's Rt will be on par with the 30 series at I hope it will be.

Yeah, Nvidia has been working on it for a while. Apple is a newcomer. But judging by published patents their approach is different enough to be promising. And they have some of the advanced features of the 40x series like ray compacting etc.

What cost? The 40 series are more efficient than previous gen and uses 100 watts less power while being more powerful in raster, RT and also having lower temps. If Nvidia uses the same 5nm arch and the same Ada arch I expect efficiency improvements as well for the laptops cards.

What you write here is different from the leaked product specs. But I’d like to wait and see.

The 3DMark tests are different on PC and also incorporate RT for some tests and and are also more demanding but at least with wildlife extreme we can compare with past Apple GPUs.

Wildlife is specifically designed to be a cross-platform test, it only uses features available on all platforms (at least according to benchmark description).

BTW, that result is comparable with a 3070 (laptop). Not too bad.

 

[Andropov]

My take is on Apple's Rt will be on par with the 30 series at I hope it will be.

Don't know about that, Apple is good at being late to the game and still beat everyone else playing.

 

[exoticspice1]

What you write here is different from the leaked product specs. But I’d like to wait and see.

I was referring to their RTX 40 desktop lineup. We can already see the benefits of Ada and TSMC 5nm on desktop. No reason that would not apply to the laptop lineup as well it's based on Ada and 5nm too.

 

[theorist9]

.

 

[exoticspice1]

Don't know about that, Apple is good at being late to the game and still beat everyone else playing.

Hopefully, I have no doubt they will best AMD and tbh I don't think they will compare with Nvidia unless they beat them.

 

[exoticspice1]

Wildlife is specifically designed to be a cross-platform test, it only uses features available on all platforms (at least according to benchmark description).

Ahh my bad. I was thinking of the steam 3DMark which is completely made for PCs

New 3DMark benchmark for Windows, macOS, iOS and Android

Find out more at benchmarks.ul.com

benchmarks.ul.com

The Wildlife one that I linked shows that your right. Someone has a RTX 4080 to compare with?

 

[exoticspice1]

Interesting the Steam version also has Wildlife extreme along with a bunch other ones. I must say 3DMark sounds infinitely more fun than Geekbench when it comes to benchmarking GPUs.

Save 75% on 3DMark on Steam

3DMark is for gamers, overclockers and system builders who want to get more out of their hardware. With its wide range of benchmark tests, 3DMark has everything you need to test your PC's performance.

store.steampowered.com

 

[leman]

The Wildlife one that I linked shows that your right. Someone has a RTX 4080 to compare with?

Probably north of 70k points, judging by the Ultra score. An entirely different beast to be sure. But also almost an order of magnitude higher power consumption.

 

[exoticspice1]

Probably north of 70k points, judging by the Ultra score. An entirely different beast to be sure. But also almost an order of magnitude higher power consumption.

It's seems the 4080 is in the 60K to 65K range.

3DMark.com search

3DMark.com search

www.3dmark.com

 

[leman]

It's seems the 4080 is in the 60K to 65K range.

3DMark.com search

3DMark.com search

www.3dmark.com

Seems a bit low given that the 4090 is in 80k+, but who knows.

 

[exoticspice1]

Seems a bit low given that the 4090 is in 80k+, but who knows.

I think that's alright cause the 4090 has 16000 cuda cores while 4080 has around 9500.

EDIT: looks like the 4070 Ti is around 45K. If the M2 Ultra scales well in Wild life it should around 40K. Kinda sad that Apple can't even match the 4070 Ti unless Apple increase clocks.

If the M2 Ultra is high as it goes in the Mac Pro then the GPU bound tasks excluding the ones that benefit from the large unified memory will be slower than a RTX 4070 Ti.

That's not even mentioning the price. The 4070 Ti is $800 - $850. Apple charges $1000 to go from 48 cores to 64 cores on the M1 Ultra. Very greedy.

 

[Jimmyjames]

I think that's alright cause the 4090 has 16000 cuda cores while 4080 has around 9500.

EDIT: looks like the 4070 Ti is around 45K. If the M2 Ultra scales well in Wild life it should around 40K. Kinda sad that Apple can't even match the 4070 Ti unless Apple increase clocks.

If the M2 Ultra is high as it goes in the Mac Pro then the GPU bound tasks excluding the ones that benefit from the large unified memory will be slower than a RTX 4070 Ti.

That's not even mentioning the price. The 4070 Ti is $800 - $850. Apple charges $1000 to go from 48 cores to 64 cores on the M1 Ultra. Very greedy.

The M1 Ultra already matches the 4070 in gfxbench 4K Aztec. The M2 Ultra should beat it easily.