Desktop =/= workstation. A 14900K is a desktop CPU. A Max is a desktop CPU and an Ultra is low end workstation.
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.
Why does Apple not bring out desktop CPU?
- Thread starter Bubble99
- Start date
- Sort by reaction score
I'm not really in that world anymore but I doubt it would have any impact that wouldn't be easily overcome by simply keeping up with relatively modern hardware. There are plenty of custom chiplets onboard to help with tasks like these, and you can always just edit at a lower resolution for 95% of the time and switch over to full-res for a final pass and touchups. And if you're doing big 8K productions, you likely have some sort of production infrastructure for doing the heavy final render.
It’s only got two benchmarks for the Mac, neither of which are used anymore and one of which is run under Rosetta imposing a 15-40% performance penalty. Even the PC benchmarks are old. Also if you’re concerned with video rendering, scientific workloads, or graphics they are completely outmoded or irrelevant.
I actually agree with this OP. The post makes sense. If Apple came out with a real desktop-class CPU/GPU, it would be amazing. Think a true Mac Pro that has Nvidia 4090-level graphics and real AI performance.
The issue seems to be as Steve Jobs once compared laptops as cars and desktops as trucks. The iPhone is now the motorcycle, the iPad is the car, and the MacBook is the pickup truck. What’s a real desktop look like? A 18-wheeler semi-truck?
Who wants that? There is the issue. Very few people need what is all ready available. The iPad is running an SoC capable enough for 95% of users. But the OS sucks. See Tim isn’t smart like Steve. Tim cares more about his shares and stock grants annually than truly making great products. Tim knows that okay, if we let people run Mac-capable apps on their iPads we won’t sell as many MacBooks.
And now comes the money. When the money is calculated it just doesn’t make sense to make desktop-class SoCs when the current M-series SoCs do everything over 99% of users need. We might be talking less than .01% of the Mac population that needs a faster SoC. Researchers? AI? Servers? Video production studios? We are pretty limited here. Apple could absolutely compete in this category if it wished to. But financially, it just doesn’t make sense.
Timmy’s AAPL is afraid of cannibalism of its products. That’s because Tim doesn’t understand technology. He’s not a product person. He gives users what they ask for. Look at the M-series 14” & 16” MBPs. They’re thicker than they need to be. They have ports most don’t use to stop the dongles. He can’t stand the criticism. AAPL shareholders are far better with Tim in charge but Apple users would be much better if someone like Steve was in charge. In reality there needs to be give and take. It shouldn’t be as shutdown as it is with Tim, but his strategies make sense since it’s a public company.
I would think that if Tim allowed the cannibalism of Macs by iPads, they would more than make up for it with new users coming from Chrome and Windows/Intel platforms. But that’s a risk Tim doesn’t want to take. Steve wouldn’t care what the shareholders think, and that’s not necessarily good either.
But at the end of the day the answer to your question is there’s no money in it for AAPL.
The issue seems to be as Steve Jobs once compared laptops as cars and desktops as trucks. The iPhone is now the motorcycle, the iPad is the car, and the MacBook is the pickup truck. What’s a real desktop look like? A 18-wheeler semi-truck?
Who wants that? There is the issue. Very few people need what is all ready available. The iPad is running an SoC capable enough for 95% of users. But the OS sucks. See Tim isn’t smart like Steve. Tim cares more about his shares and stock grants annually than truly making great products. Tim knows that okay, if we let people run Mac-capable apps on their iPads we won’t sell as many MacBooks.
And now comes the money. When the money is calculated it just doesn’t make sense to make desktop-class SoCs when the current M-series SoCs do everything over 99% of users need. We might be talking less than .01% of the Mac population that needs a faster SoC. Researchers? AI? Servers? Video production studios? We are pretty limited here. Apple could absolutely compete in this category if it wished to. But financially, it just doesn’t make sense.
Timmy’s AAPL is afraid of cannibalism of its products. That’s because Tim doesn’t understand technology. He’s not a product person. He gives users what they ask for. Look at the M-series 14” & 16” MBPs. They’re thicker than they need to be. They have ports most don’t use to stop the dongles. He can’t stand the criticism. AAPL shareholders are far better with Tim in charge but Apple users would be much better if someone like Steve was in charge. In reality there needs to be give and take. It shouldn’t be as shutdown as it is with Tim, but his strategies make sense since it’s a public company.
I would think that if Tim allowed the cannibalism of Macs by iPads, they would more than make up for it with new users coming from Chrome and Windows/Intel platforms. But that’s a risk Tim doesn’t want to take. Steve wouldn’t care what the shareholders think, and that’s not necessarily good either.
But at the end of the day the answer to your question is there’s no money in it for AAPL.
Your comparison is garbage - you are comparing the lowest end 3 year old 1st gen M1 chipset with its lowest core count - to the top of the line workstation CPU.
Try the comparison to a fully loaded M3 Ultra.
Forget Apple, if Nvidia doesn’t increase RAM in their 5090 GPU, Apple will eat its sales. I am a heavy 4090 user, and the 24 GB RAM runs out of memory. I am forced to use a slower M1 Max MBP for anything that needs more than 24 GB GPU memory, its not much for todays workloads. I haven’t bought an Apple workstation in past 12 years, but will move to Studio if Nvidia doesn’t get its stuff together with low memory options. I will take a 256 GB unified memory which can run heavy GPU workflows.
I know.
If performance is key then you want native benchmarks on latest top end processors.
There are no fully loaded M3 Ultra there aren’t any Ultra beyond the M2 Ultra today
Still king of workloads, Hopefully Apple has an M4 ultra ready to go soon.
Here, the point has been phrased out now !
thats the question.
we´ll see how they´ll handle the things with the M4.
let´s hope they´ll give the desktops some more speed.
This is one of the primary reasons why I think there’s actually a market for an Extreme or larger Ultra, especially if Apple increases matmul performance on its GPUs (for some even just just the ray tracing cores + huge VRAM is big). Apple has a big opportunity here.
As a fan of tech, this would be awesome. Apple has all the money in the world, it would be cool for them to make the most powerful computer just because they could. But, as you mentioned, there's just no money in it. It would be prohibitively expensive and serve a wildly small niche of their already relatively small desktop pro consumer base. It would have be something they did for the prestige of just being able to say they did.
It depends … right now there’s a massive market probably a bubble for AI. Further Apple is rumored to be using its own chips for servers. So an Apple Extreme or even just the Ultra with ray tracing and increased GPU matmul with large vram could be a huge seller for graphics professionals and AI trainers despite that it would be high cost. But even if Apple is their own best customer they could sell the leftovers.
Unified memory is a game changer. Apple MLX team has done very good job bridging the gap with consistent additions and improvements. PyTorch had a big for almost 18 months where it was doing core matmul on cpu, though it was supposed to run on GPU. MLX still has some functional differences but supports some core functions. I do hope that M4 Ultra or extreme or high end version to be more than 2X M4 Max.
As someone with 1 M1 Ultra and 2 M2 Ultra Mac Studios, and M1 Max and M3 Max Macbook Pro systems, the Ultra variants are fine......We could all appreciate higher performance but they are fine.
Compared to my 13900k/4090 System, even my M1 Ultra does some things better than it.
However, what Apple needs to do is address when they refresh their desktop line. I find it increasingly frustrating that in a lot of ways my M3 Mac Macbook Pro is better than my M2 Ultra.
$3,499.00 Laptop can outperform my $6,599.00 Mac Studio. And if the rumors are true that we won't see an updated desktop line until 2025, that is just ridiculous. The laptops make me partially regret getting Apple's desktops. In some cases the Mac Studio does perform better than it, but in others the laptop does.
I do think Apple should have a true desktop chip for the Mac Pro to make the $3,000 price increase over the Studio worth it.
Compared to my 13900k/4090 System, even my M1 Ultra does some things better than it.
However, what Apple needs to do is address when they refresh their desktop line. I find it increasingly frustrating that in a lot of ways my M3 Mac Macbook Pro is better than my M2 Ultra.
$3,499.00 Laptop can outperform my $6,599.00 Mac Studio. And if the rumors are true that we won't see an updated desktop line until 2025, that is just ridiculous. The laptops make me partially regret getting Apple's desktops. In some cases the Mac Studio does perform better than it, but in others the laptop does.
I do think Apple should have a true desktop chip for the Mac Pro to make the $3,000 price increase over the Studio worth it.
I feel like that’s still going to be quite niche (although I admittedly don’t follow it that closely and am not very interested in it, so maybe I’m wrong).It depends … right now there’s a massive market probably a bubble for AI. Further Apple is rumored to be using its own chips for servers. So an Apple Extreme or even just the Ultra with ray tracing and increased GPU matmul with large vram could be a huge seller for graphics professionals and AI trainers despite that it would be high cost. But even if Apple is their own best customer they could sell the leftovers.
I expect AI needs for most users will be handled by dedicated AI chiplets, while more intensive stuff would still be done in the cloud, again with specialised hardware.
If Apple does end up making server-class chips, maybe those could be binned down to become a higher tier of desktop chips, but I still think the chances of that are fairly slim.
Yeah I get it, the “AI” market appears over heated and oversold with more caveats than it’s most zealous proponents will admit. Having said I do think there’s some potential there, and some of this growth is real and Apple’s peculiar approach could actually be incredibly useful.
Right now, and who knows how long it’ll continue, every AI startup under the sun is buying every Nvidia H/B GPU/Superchip they can get their hands on but there aren’t enough to go around so even desktop AMD GPUs strung together are selling. Right now the market is huge. Again probably a bit of a bubble and I’m not suggesting that Apple try to compete with big iron. That’s not their business and never really has been but if Apple wants to train their own models and allow users to do so, even the ultra with GPU tensor cores would actually be theoretically very performant and an Extreme would be even more so. The combination of UMA and TOPs would be hard to beat in a market currently ravenous for it.
Again I don’t know how long that will last. But orders for Nvidia’s systems are reportedly backlogged months if not longer. Even now Apple SOCs are well reputed for their high memory bandwidth, it’s just batched inference and training performance that they lack. So there’s definitely a much larger than niche market.
But Apple is rumored to not be shipping M4 Ultras until next year, with possible rumors of the Extreme, and unless they add capabilities to the larger M4 chips than the base, which seems unlikely, we probably wouldn’t see Apple chips with better GPU matmul until the M5 and it’s Ultra is unlikely to come out until 2026 if the current rumors are accurate (I generally don’t trust Gurman beyond a few weeks but his predictions here are … reasonable even if I would prefer Apple move faster). And who knows what the market will be by then.
Last edited:
My guess is that Apple may have little interest in competing in this space directly.
Apple silicon is optimised for power efficiency, making it great for laptops, but arguably less for for desktops where users may prioritise pure performance and less so on energy consumption. That said, Apple does sell a lot more laptops than desktops, and they may simply feel it's not worth it to come out with a desktop version of Apple Silicon solely for say, their Mac Pro.
Another take is that apple silicon allows Apple to ship desktops in form factors you normally don't see with the competition, such as the iMac and Mac Studio (since intel / AMD processors tend to generate more heat and require more room for cooling). I suspect that if Apple had to make a choice between optimising their processor for the Mac Pro vs dropping the Mac Pro altogether, they would sooner just scrap the latter as a product category.
Nvidia's moat is not their GPUs. It's their software, CUDA.
Apple machines can already have boatloads of VRAM due to unified memory. Do you see the AI/ML/gaming industry flocking to Apple? No.
I say this as a CUDA guy: CUDA isn't the moat it used to be. And Apple only just added ray tracing (with no Ultra option) and has no tensor cores. But yes they need to continue to improve their software. No question. MLX is new so I would expect further improvements and for it to be plugged in to more backends.
Last I checked we are in 2024, that was few years ago, you were tied to CUDA. MLX, PyTorch and other libraries are plenty for those who don’t want to get their hands dirty with Metal or CUDA. There is plenty of support for AMD GPUs too. Nvidia biggest strength and astronomical growth is in data center market. Nvidia rightfully is focusing more on data centers than consumer GPUs. They can add clusters of powerful GPUS, which Apple nor others cant match in that market. That leaves a big chunk for Apple and AMD for on device computing.
Last edited:
Right now it's the cheap consumer GPUs like the 4090 that are eating their sales. Nvidia would be making more money by focusing on the data center market. It's a crazy market out there, and the demand for GPUs far exceeds the supply. But because it's uncertain if the AI boom is sustainable, Nvidia is hedging their bets by maintaining presence in other markets. Gamers need powerful GPUs, but they can't afford paying anything resembling market prices. Which means that consumer GPUs must be intentionally crippled to avoid cannibalizing the real market.
I'm not sure if anyone really cares about workstation GPUs anymore.