Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
What microarchitecture choices have made M2 less efficient than M1?
- Thread starter Xiao_Xi
- Start date
- Sort by reaction score
Nah, you should stop. You're the dude posting word salad. "the cycle goes up to and across peak more smoothly"? "gating was being designed to respond properly at lower differentials"? I'm a professional in the field and none of that makes sense.
If you are designing a 3300 MHz circuit for TSMC 5nm and you're having trouble closing timing, one of your options to make your design work without further alteration is to bump supply voltage up. If all other variables are held constant, digital logic does in fact respond to increased voltage by running faster. The thing is, process node and design details are hugely important variables, so comparing dramatically different nodes and designs to each other this way (as you have by bringing in ancient 1980s 5V stuff) and using that to claim that clocks go up because voltage goes down makes no sense.
Technically it was possible to do this kind of thing in the 1980s too, but back then variable-voltage power supply rails and variable clock generator circuits were a bit too exotic for things like mass market personal computers. So you got one frequency. If you wanted to attempt an overclock, you'd have to find a replacement crystal and swap it in, and if you ran into problems it wasn't very practical to try to increase from 5V to 6V.
Thanks for taking the time to reply.
I misread your original post and assumed (my bad) that the same formula was being used to compare different architectures and say the m2 was less efficient. Again you didn’t say this, I misread.
I think this paragraph really helped provide clarity of understanding for your viewpoint. Thank you!
If launching a game and setting AA to MSAA 8x is too difficult then keep waiting. Seems to be a MacOS issue/limitation since Google also ran into the issue.
You'd be surprised.
Just tried it with my old GTX 1080 MSI Gaming X in Cyberpunk.
Stock: 1,049 Volt, 1949 Mhz, 5000 Mhz Memory.
64 fps, 180 Watt
UV: 0,800 Volt, 1823 Mhz, 5500 Mhz Memory OC.
62 fps, 120 Watt.
Not to mention, much less heat, less fan noise etc.
Let's to back to M1.
M1 has 561 Points per Wat in CB r23 Multicore. I personally got 516.
But put the M1 in Low Power Mode (whole SoC consumes not more than 4 Watt), points drop to 4500, but points per Watt goes up to 1125.
M1 in 4w low Power Mode has 1125 points per Watt there.
it's insane how efficient Chips can be, if you don't go for maximum Performance.
I’ll wait. It’s up to you to show that MSAA 8x is « fake » on macOS. Until then, I’ll consider this to be just another BS claim from you.
You know the rules here, make a claim, back it up.
You need to show at least a good faith attempt at conversation/debate.
Regarding the topic at hand, I don’t see any path forward in the next 2-3 years to match the low power lead that Apple has. Intel and AMD’s business models simply don’t allow for this radical a change in capabilities.
Apple has a 2-3 year runway to run away in overall performance AND insane battery life, and consumers will start to notice in the not too distant future. I have no idea if Apple actually has an interest in capturing massive market share (playing exclusively in the premium end of the market has paid off for decades now), but the capabilities of the M1 5 years out at sub-$750 price point would result in a staggering coup.
That said, Apple’s focus is higher quality materials so idk if the Mac will ever have an “iPhone 5c” moment.
Not trying to take the burden, I wonder if setting a game to 8x MSAA doesn't actually apply 8x MSAA, like it caps out at 4x MSAA.
Or at least games that are using OpenGL (IIRC the example given is an OGL game right?).
More like setting MSAA 8x is like no AA so benchmark comparisons are misleading compared to other systems with working MSAA 8x.
I can't force people to open their eyes. Just look at the screen shot comparisons.
Last edited:
You know that MSAA on Apple hardware is essentially free, right? One of the benefits of TBDR architecture.
It’s not really free, but much cheaper than on other architectures. MSAA is bandwidth-heavy, but you don’t have to pay the bandwidth cost on Apple hardware because everything stays within tile memory.
Free like 8GB = 16GB? MSAA 8x is only free now since it's essentially no AA. Wait until it's working before making that claim.
Make up your mind. Is it free or not free?
On AS MacOS it appears to be a wider issue and not per game. Discovered it in early July but apparently there's a similar MSAA 8x Google bug report opened since 3/12/2022. No Intel Mac heaters allowed here.
I know AA is broken on the Uningine stuff (OpenGL). I haven't redownloaded BG3 to check if AA there works or not.
opengl is broken on macOS. Apple has not updated it in a long time.
all future games will be made with Metal 3 like No Man's sky and RE: Village
all future games will be made with Metal 3 like No Man's sky and RE: Village
The story is all over the place. Now it's Apple Silicon macs can't do msaa, a few weeks ago this person was accusing people of misleading benchmarks because of a 10 year old Intel mac game!
Which issue? Which bug report? Can you be more concrete? It's like pulling teeth with you.
FWIW, Apple Silicon only supports 4x MSAA. Maybe that's what this is all about.
So using Batman as the example, what happens if you choose 8x MSAA? I know in Uningine Heaven you have to disable AA to get anything to show up on the screen, does Batman do the same thing?
How would I know? All I can tell you that Apple Silicon only supports up to 4x MSAA (which is clearly exposed in both Metal and OpenGL), if the game gives you other options than it must be doing something arbitrary in the background.