Still runs smoother than my MBP M1Max even when unplugged. Apple Silicon's power savings is mostly in the idle and mundane tasks like browsing or if you're using the media accelerators. Once you're pushing it while gaming or running something general purpose like handbrake CPU only, the power savings advantage is significantly reduced.
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.
M3 Max GPU vs NVIDIA RTX 4090 Laptop
- Thread starter AnonFr13n45
- Start date
- Sort by reaction score
Sure, an RTX 3080 could perform very well even on battery, since power increases (and decreases) exponentially with clock speed, hence performance.
It doesn't mean that Apple couldn't put an M3 Ultra in a laptop if they didn't care about size, noise and weight.
Not true. This spring I did some benchmarking with a Lenovo laptop (12th Generation Intel® Core™ i7-12800H vPro® Processor and 8GB Nvidia RTX 3070 Ti GPU) we use for 3D rendering and CAD vs. my M2 Max running applications like Adobe Lightroom and Photoshop, both of which can hammer all of the CPU cores and the GPU on many tasks. On battery power the Lenovo got crushed unless I turned off all of the power management. When I did that, the battery in the Lenovo was killed in less than 20 minutes. Meanwhile the M2 used less than 10% of its battery power to complete the tasks.
I’m talking about mostly gaming here. In productivity the Mac’s better efficiency wins but in gaming it’s not much better. My Lenovo has the AMD cpu also which is much more efficient than the Intel version. Handbrake cpu encoding is very slow and inefficient as it doesn’t use the media accelerators.
Not sure how accurate the comparison is, but playing Dwarf Fortress unplugged I still get very good battery life. Same with using Illustrator and Photoshop.
I fail to see a difference. The testing I did with Lightroom exporting several thousand raw files to jpeg fully pegs out the cores and the GPU--arguably even more so than many games--and did so for almost 15 minutes straight. That nearly drained the battery on the Lenovo, but only used up about 10% of the capacity on the Mac. I honestly can't see how a game would push the system any harder.
I don't see how you can assume that gaming is a lighter load than doing productivity. A quick google search or actually trying some AAA gaming on your macbook will quickly give you an answer.
I can easily pull over 200+ watts on my 13900K/4090 system running Lightroom AI denoise or Davinci Resolve, but the rig easily pulls over 500 watts when gaming with titles such as starfield, bg3, CP2077.
Now you can do tricks like locking fps to 30 or running at very low resolutions and eye candy to get better battery life, but I can do the same to the windows laptop also to extend the battery life.
Dwarf fortress is much less graphically intense than BG3. I'm using BG3 as a comparison as it's a newer AAA title that runs on AS natively.
To compete with RTX 4090, M2 GPU has to be at least 300 cores, not 76 cores. But then, at 300 GPU core, it will consume as high as RTX 4090 in terms of power consumption.
Wtf are you talking about? In what world is the 3060ti faster than the M2 Ultra?
The M2 is faster in Wildlife Extreme, Gfxbench, geekbench etc. what is the 3060ti faster in?
I'm not assuming it's a *lighter* load. I'm saying I can 100% peg the cores and GPU doing certain tasks in apps like Lightroom. On large batches I can sustain that load for 20+ minutes. Pegging out the CPU and GPU are the same, game or no game. That was my point. And contrary to what the person I responded had said, when I benchmarked performance and battery life on the Lenovo PC and the M2 Max it was night and day differences in battery life, noise, heat, everything. I literally killed the battery in the Lenovo before it could finish the task.
Cinebench 2024 Scores [CPU & GPU] (Updated Results)
The new Version of Cinebench, 2024, benchmarks CPUs and GPUs in the popular Redshift render engine. We benchmarked all modern CPUs and GPUs and list them in our score overview.
Those benchmarks are NOT reliable and not represent actual results which has been proven.
Dishonest garbage. So all other benchmarks which measure on an equal footing are not reliable except your cherry picked one which happens to use optix on the 3060. What a joke.Cinebench 2024 Scores [CPU & GPU] (Updated Results)
The new Version of Cinebench, 2024, benchmarks CPUs and GPUs in the popular Redshift render engine. We benchmarked all modern CPUs and GPUs and list them in our score overview.www.cgdirector.com
Those benchmarks are NOT reliable and not represent actual results which has been proven.
ArtIsRight did the same with importing 1000 Nikon D850 RAW pictures, which you can see the result of here.
The specs are wild on that Windows laptop, yet...
The source is from this video:
I'd be curious to see M2 Ultra framerate in Baldurs Gate 3. I've not found anyone that has done so yet. I've seem impressions (folks say it is butter smooth) but no numbers to go along with it.
I guess you never used RTX 4090 and M2 Ultra on GPU intensive software huh? It's TOO slow or not even usable. If you wanna prove your point, bring actual tests results with 3D software or game. Guess what? most of them dont even support Mac cause of poor GPU performance.
Seriously, since when Apple Silicon is as great as RTX 40 series? I'm freaking tire of whining from Mac users claiming that M2 Ultra = RTX 4090 without any proofs and even so, that would be only a few great example just for Mac such as GFXbench which Apple claimed that M1 Ultra = RTX 3090 which is not even close. Video software are exceptions cause they have dedicated chips.
Yes, that’s his goal.
Yeah that thought entered my head. I don’t really play graphically intensive games, so I really don’t have a point of reference for that.
Yea you can see on notebook check the M2max can draw pretty significant power if maxed out. So sure on a game like BG3 i suppose you can drain the battery the same as a windows machine. Where the two are worlds apart are in strategy games. I can easily game for 5 hours or more in hi resolution on my paradox games with low noise on mac where a windows machine would be lucky to get 2.5 hours at best even on something like say the 7940hs using igu only.
NB. that this is just for fun!
No doubt there will be some M3 Max Metal benchmark numbers on Geekbench along soon. Meanwhile ....
Taking the ratio of the early GB6 OpenCL numbers for the M3 Max (40 core) to the M2 Max's OpenCL score and applying it to the M2 Max's Metal score suggest an M3 Max Metal figure of somewhere between 150 and 160,000. As previously mentioned, the 4090 Laptop's OpenCL number (sadly, there are no GB6 CUDA numbers) is just under 182,000.
A similar projection for the M3 Ultra (I do realise it doesn't exist in the wild yet ...) gives somewhere in the region of 240,000 for Metal, vs. the 4090's OpenCL score of c.322,000.
I make absolutely no claim about the validity or usefulness of these numbers .... 😬. (Even if I haven't made mistakes with the arithmetic, which is entirely possible.) Although GB OpenCL, Metal and Vulkan scores are supposed to be comparable, I have my doubts about that. Some comparisons using real workloads will be much more interesting than GB scores.
Incidentally, based on the early Metal numbers, it's looking like the M2 to M3 GPU performance increase is c.5%. Adjusting for the number of cores, I wouldn't be too surprised if the M2 Max to M3 Max increase (i.e. for a single GPU core) is closer to 3%.
No doubt there will be some M3 Max Metal benchmark numbers on Geekbench along soon. Meanwhile ....
Taking the ratio of the early GB6 OpenCL numbers for the M3 Max (40 core) to the M2 Max's OpenCL score and applying it to the M2 Max's Metal score suggest an M3 Max Metal figure of somewhere between 150 and 160,000. As previously mentioned, the 4090 Laptop's OpenCL number (sadly, there are no GB6 CUDA numbers) is just under 182,000.
A similar projection for the M3 Ultra (I do realise it doesn't exist in the wild yet ...) gives somewhere in the region of 240,000 for Metal, vs. the 4090's OpenCL score of c.322,000.
I make absolutely no claim about the validity or usefulness of these numbers .... 😬. (Even if I haven't made mistakes with the arithmetic, which is entirely possible.) Although GB OpenCL, Metal and Vulkan scores are supposed to be comparable, I have my doubts about that. Some comparisons using real workloads will be much more interesting than GB scores.
Incidentally, based on the early Metal numbers, it's looking like the M2 to M3 GPU performance increase is c.5%. Adjusting for the number of cores, I wouldn't be too surprised if the M2 Max to M3 Max increase (i.e. for a single GPU core) is closer to 3%.
Exactly, and he didn't do the testing on battery only.
I'm just calling BS on the guy that said the game will kill a Mac and PC in roughly the same time. At max power draw even an M2 Max will draw way less wattage than an Intel or AMD processor and a mobile high-performance GPU. It doesn't matter what you are doing: There's a max wattage a system can and will draw and there's the battery capacity. Their claim that the Apple silicon power savings is mostly from the efficiency cores and is also bunk; the performance cores also use massively less power. On top of that, when I benchmarked my M1 Max and M2 Max MBP's doing real-world intensive editing on battery they performed almost identically to plugged in.