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.
Apple is falling behind, M4 Max nowhere close to RTX 5090.
- Thread starter Zest28
- Start date
- Sort by reaction score
I tasked myself with anchoring some hanging file-holders in my classroom, today.
Unfortunately, I forgot to bring my hammer...
I used my cordless drill&tape dispenser to hammer-in the anchors after I drilled the holes.
It was a success!
But, it would have been a heck of a lot easier if I had remembered my dedicated
Aye; I was seriously contemplating purchasing the 6000 Blackwell Max-Q yesterday, and then decided that the better choice was to wait for my DGX Spark (and maybe also the M4/5 Ultra Studio!) to arrive.
Sometimes Functionality supersedes Performance.
I’m very curious what you’re doing. I have a Linux workstation with a 16 core AMD and 5090. I have barely touched my Mac since the purchase. With very big LLMs I can’t keep them all in VRAM, but 32GB is still plenty big for some reasonable models.
What use is speed if Nvidia GPU doesn’t have enough memory, it will run out of memory and errors out. 24-32 GB GPU isn’t much, the data sets these days are much larger. Sure I can spend $$$ on cloud or just wait 20 more minute for my M4 max to finish. Either you have enough GPU memory or you don’t.
Why stop at that, you could run 6 4090x. Good look maintaining and powering that thing. I need to pretty much revamp my whole workstation to be compatible with RTX 6000 Pro. I may get 192 GB but I can run 100 GB swap and can easily scale to around 210 GB on M4 max in case I need larger memory for a spike or shorter time. I will probably consider spark when I need to replace my workstation. I may also consider and ultra with 512 GB.
Last edited:
I don’t run lot of LLM. In fact if I wouldn’t spend thousands of $$$ to buy laptops/workstations to run LLM. Cloud costs or subscriptions are much cheaper to run LLM. I have large data sets I work with, and M4 Max gives me ability to lower cloud costs.
I can wait on my workstation but most likely my choice would be Spark vs Ultra studio.
I must say am a bit confused about all the 5090 vs. M4 Max comparisons.
The 5090 RTX mobile (~ 10K shader cores, 160 watts) scores ~52K in 3DMark Wild Life Extreme
The M4 Max (~ 5K shader cores, around 70-80 watts) scores ~34K in 3DMark Wild Life Extreme
I'd it's about as expected? The other question is whether it makes sense for Apple to use >100W GPUs in their laptops.
I do get the laptop vs. desktop discussion, at the same time it's important to keep in mind that these are ultimately mobile chips. That's a business decision, not a technical limitation.
The 5090 RTX mobile (~ 10K shader cores, 160 watts) scores ~52K in 3DMark Wild Life Extreme
The M4 Max (~ 5K shader cores, around 70-80 watts) scores ~34K in 3DMark Wild Life Extreme
I'd it's about as expected? The other question is whether it makes sense for Apple to use >100W GPUs in their laptops.
I do get the laptop vs. desktop discussion, at the same time it's important to keep in mind that these are ultimately mobile chips. That's a business decision, not a technical limitation.
And then there's the AMD Ryzen AI Max+ 395 which goes up to 128GB of unified memory. How does its iGPU stack up against Apple's?
Should be a bit better than the M4 Pro. Depends on what you want of course. For gaming or some types of ML work, the Ryzen would work better.
While I agree that Apple Silicon was designed primarily for the mobile market. However, Apple markets Apple Silicon as a viable desktop solution. We can't say out of one side of our mouth that its a mobile chip, so don't compare it to nvidia desktop gpus, and at the same time promote how awesome their desktop computers are (compared to PC desktops)
Apple sells the Mac Pro as a high end workstation class solution, that can easily exceed 10,000 dollars, its only natural to see how Apple Silicon matches up to Intel/AMD computationally and Nvidia for the GPU. Likewise for the Studio, its another high end desktop solution that many people are using, instead of buying a PC desktop. Apple is marketing the Mini, Studio, Mac Pro (and even the iMac) as desktop computer solutions that are in direct competition to PC desktops. Comparing desktop CPU and GPUs is warranted because that's what apple is doing
Apple themselves stopped comparing their GPU's to nvidia after they (mistakenly IMO) compared the M1 Ultra to the 3090 (thus causing this perpetual comparison chain).
Can Apple double the shader cores and still only use 70-80 watts? How much of a die shrink would be required for this to happen? Could they boost the clock to make up the 20K deficit and still stay around the same power usage? How much power would be required to boost the GPU clock to make up the difference?
This makes sense to me. What I want to point out is the nature of this discussion is marketing and business strategy rather than technology. Apple is pursuing very specific hardware design goals, and their product lineup is a direct consequence of this.
I fully agree with you that the Mac Pro is really underwhelming for its form factor and does not offer a good value proposition in the current iteration. Maybe for a very small group of audio professionals that use Macs and need internal PCI expansion.
Studio etc. seem quite successful in their segment, so much actually that Nvidia and AMD start copying the form factor. Of course, this is a very fast paced market, so who knows what the status quo will be tomorrow.
I'd say that their architecture still has room to grow. Not sure about doubling the performance, but I think that 50% general-purposehigher performance with comparable power and die size (on the current process) should be achievable. They still have some tricks up their sleeves, for example there was a recent patent describing how to fetch data from GPU registers with lower power consumption.
For ML workloads, 4x improvement or higher should be doable at the current power consumption.
It would really surprise me if they go that route. Who knows, maybe their upcoming tech will be more flexible...
My Studio is the best Mac and desktop I've owned, it harkens back to the days when Apple made great desktop computers at an affordable price.
My Studio occupies. a tiny percentage of waht my mini-tower took and is significantly faster. Another plus, is that it doesn't heat up my room
and I get all of the benefits of using macOS.From the looks of it the 5090M and the M4 Max run at about the same clock rate; the official boost for 5090M is lower by like 300Mhz, but that is assuming you are capping TGP to 95W, which no one does. Without that 65W headroom, I imagine the GPU's are a lot closer.
Its fairly competitive. As it should be, on similar process tech, and AMD having 30 years of GPU development including their acquisition of ATi.
I think i understand WHY apple is doing what they're doing.
Their core market is laptops. Those running Apple desktops are a rounding error really, especially PRO users (a rounding error on the rounding error). They exist, and it sucks for them that Apple haven't really pushed the envelope there, but its a lot of R&D for a limited ROI. Most of their desktop users just don't care and are fine with even M series base.
Apple building more powerful general end user desktop GPUs would also segment the market. Part of the reason consoles work so well is that there's a few fixed configurations to optimize for. Apple isn't quite THERE but they're much closer than the PC world. And software support wise that's good.
So yeah, Apple Silicon is a mobile/laptop chip with the single exception being the Ultra, which really appears to be a casual side project for Apple unfortunately.
Is it a high end gaming machine? No. Is it a competent portable gaming machine that can be used on battery and not carry a 3kg power brick around? Yes.
I play games on my m4 max. BG3 runs fine. Cyberpunk runs great. Not too far away from how it runs on my desktop 6900XT with RT on and 14" screen appropriate upscaling.
Does it run every AAA game at 120 FPS full native res? No. Do i care? Also no.
I bought the thing to get work done, any gaming it can do is a bonus.
Yes, that's specifically why i mentioned cars.
They had a side project and thankfully sanity prevailed and they killed it.
AMD did a great job on the HX 3* line and the number of laptop and mini-desktop design wins has been surprising as those used to be ruled by Intel.
The M4 mini with 16 GB of RAM at $450 (Microcenter) is a game-changer. Those interested in a Mac for basic home use have a really great option here in terms of efficiency and performance. The Mac Studio at $2K is a harder sell but you can get M1 and M2 Studios for a lot less and these are nice for efficiency and performance. I'm on the MacStudio subreddit and I see people asking about them, even when they are way more than they need. The price-points on the M1 models get people interested in them.