So M1 Max with 32 Core GPU matches RTX 3080 Mobile.

[Surne]

Do you have a list of the power plans? I’m having trouble finding a detailed description of them and their effects on power.

Quiet mode
Balance mode
Performance mode.

I'm out at the moment but when I get home I can check Vantage on my laptop for some info for you.

Let me know if there's anything I can help you with.

 

[Homy]

 

[crazy dave]

Quiet mode
Balance mode
Performance mode.

I'm out at the moment but when I get home I can check Vantage on my laptop for some info for you.

Let me know if there's anything I can help you with.

Just saw your edit: Yeah I was curious if you knew what wattages the CPU and GPU were put in for each state?

 

[altaic]

That's a pretty fair review. I have to say, though, estimates on what gains will be had from optimizing a game for the Jade SoCs are anyone's guess. I suspect such optimization would entail a pretty extensive rewrite of a rendering engine (i.e. a lot more work than just plugging in MoltenVK and D3DVK).

 

[jeanlain]

I suspect such optimization would entail a pretty extensive rewrite of a rendering engine (i.e. a lot more work than just plugging in MoltenVK and D3DVK).

Not necessarily. The game already uses Metal. I think some key optimisations to take advantage of TBDR could make a sizeable difference.

 

[vir]

Mac shadows were off.

 

[jeanlain]

Mac shadows were off.

Of course not.

 

[Surne]

Just saw your edit: Yeah I was curious if you knew what wattages the CPU and GPU were put in for each state?

I'm not sure on the exact voltage/wattage.

 

[diamond.g]

Of course not.

Yeah they seemed fine to me. I kind of wish he would run the test at 1440p (assuming the PC panel supports the resolution).

 

[crazy dave]

Not necessarily. The game already uses Metal. I think some key optimisations to take advantage of TBDR could make a sizeable difference.

My limited understanding is that’s not necessarily a trivial rewrite either depending on the renderer. But not my field so I’ll let someone else comment on how much would have to change.

 

[diamond.g]

My limited understanding is that’s not necessarily a trivial rewrite either depending on the renderer. But not my field so I’ll let someone else comment on how much would have to change.

On the same token it there could be some performance left on the table with the DX12 renderer.

 

[crazy dave]

On the same token it there could be some performance left on the table with the DX12 renderer.

In that shadow’s engine doesn’t have one or that it might not be very good?

 

[diamond.g]

In that shadow’s engine doesn’t have one or that it might not be very good?

They use DXR (well for shadows) but no mention of VRS, Mesh Shaders, or Sampler Feedback. Which would (should?) improve performance when used (properly).

 

[vir]

Compare the settings closely.

 

[anticipate]

View attachment 1901826

Yes it’s off. I’ve tested with them on, similar scores.

 

[vir]

That's pretty cool.

 

[altaic]

Not necessarily. The game already uses Metal. I think some key optimisations to take advantage of TBDR could make a sizeable difference.

That's the point I was making. When people say a ported game "uses Metal," they are most likely referring to a game with a DirectX or Vulkan engine using libraries (e.g. D3DVK and MoltenVK) to get to Metal.

That's a far cry from an engine designed specifically for Metal and the M1 Pro/Max; TBDR, unified memory with gobs of bandwidth, on-die accelerator units, etc. So, "optimizations" belies the amount of work involved, as well as the potentially massive benefit.

 

[crazy dave]

That's the point I was making. When people say a ported game "uses Metal," they are most likely referring to a game with a DirectX or Vulkan engine using libraries (e.g. D3DVK and MoltenVK) to get to Metal.

That's a far cry from an engine designed specifically for Metal and the M1 Pro/Max; TBDR, unified memory with gobs of bandwidth, on-die accelerator units, etc. So, "optimizations" belies the amount of work involved.

In this case in between: the game is native Metal but is not optimized for the M1 TBDR.

 

[mr_roboto]

My limited understanding is that’s not necessarily a trivial rewrite either depending on the renderer. But not my field so I’ll let someone else comment on how much would have to change.

It's not really my field either, but every time I've watched one of Apple's WWDC videos about optimizing ported graphics code, it's clear there's usually low hanging fruit with easy fixes. There's lots of ways for naive Metal ports to accidentally declare dependencies which aren't real, forcing unnecessary serialization points on TBDR GPUs. Another common one is accidentally flushing buffers to DRAM instead of keeping them on-chip for the next render step.

Here's a video on the topic from WWDC '21, if you've got time to watch:

Optimize high-end games for Apple GPUs - WWDC21 - Videos - Apple Developer

Optimize your high-end games for Apple GPUs: We'll show you how you can use our rendering and debugging tools to eliminate performance...

developer.apple.com

At about 18:45 he starts talking about a simple shader compiler flag setting which decreased frame rendering times in Metro: Exodus by 21%. The only reason it was ever set to the slower option was that the unspecified Metal translation layer 4A Games used in their port defaults to it.

 

[mi7chy]

That's great and all but who's going to spend $3500+ along with cost of Parallels and Crossover subscriptions and a lot of trial and error time just to play a limited number of older games so they can say they get 1.5 hours or less unplugged and unthrottled? How realistic is that vs buying a Lenovo Legion with 3060 or even 3070 for $1400 or less when on sale that can play all games including recent titles, faster with better quality plugged in and seamless without loss hair?

 

[jeanlain]

My limited understanding is that’s not necessarily a trivial rewrite either depending on the renderer. But not my field so I’ll let someone else comment on how much would have to change.

This is what Brad Oliver, developer at Feral, said about results from Anandtech showing the M1 Max being beaten at SoTR.

Shadow of the Tomb Raider was tuned for discrete GPUs so it doesn't have any optimizations for M1s, which are a pretty different beast from "normal" GPUs.

It's a reasonable takeaway to say the benchmark represents what you might see if you don't actively take advantage of any of the unique features in Apple GPUs. In that light, it still vastly outperforms Intel GPUs with no effort. I'm pretty sure it'd get a lot closer to a discrete GPU with a bit of tuning and rewriting here and there.

 

[jeanlain]

That's great and all but who's going to spend $3500+ along with cost of Parallels and Crossover subscriptions and a lot of trial and error time just to play a limited number of older games so they can say they get 1.5 hours or less unplugged and unthrottled? How realistic is that vs buying a Lenovo Legion with 3060 or even 3070 for $1400 or less when on sale that can play all games including recent titles, faster with better quality plugged in and seamless without loss hair?

Who said they'd buy a MacBook Pro for gaming?

 

[Homy]

View attachment 1901826

Good catch! According to Nvidia ""Screen Space Contact Shadows" on High would reduce the fps by 6-7 frames compared to Off at Max settings and 1440p. That would make MB Pro M1 Max on par with RTX 3080 on battery.

 

[Homy]

That's great and all but who's going to spend $3500+ along with cost of Parallels and Crossover subscriptions and a lot of trial and error time just to play a limited number of older games so they can say they get 1.5 hours or less unplugged and unthrottled? How realistic is that vs buying a Lenovo Legion with 3060 or even 3070 for $1400 or less when on sale that can play all games including recent titles, faster with better quality plugged in and seamless without loss hair?

Then you haven't really understood the target group for M1 Max. People buying that don't buy it for gaming. They buy it for other tasks but it's nice to be able to take a break and also have some fun with your Mac without having to buy an extra pc or console. They buy it because they want a Mac of course, not a pc, no matter how cheaper it would be. Is gaming your primary goal then you should buy a pc or console.

 

[vir]

My goal is "everything".