Apple unveils Mac Studio, a new desktop powered by the gargantuan M1 Ultra SoC

nanoguy

Posts: 1,217   +21
Staff member
In brief: The Mac mini is here to stay (for now), and so is the Mac Pro. However, Apple is adding a new member to the Mac family called Mac Studio, which is a new category that fits right between those two devices. It can be configured with either an M1 Max or M1 Ultra chipset, and if you pair it with the new Studio Display, it's almost like a spiritual successor to the 27-inch iMac.

Apple has announced the Mac Studio, a new desktop system that can only be described as a Mac mini on steroids. The newest member of the Mac family is aimed at professionals who are looking for something more compact than a Mac Pro, but also more powerful than the Mac mini.

The base model Mac Studio is powered by Apple’s M1 Max chip that debuted in the MacBook Pro 14 and MacBook Pro 16 last year, but it can also be configured with a new chipset called the M1 Ultra. The Cupertino company claims this will be up to 50 percent faster than a specced-out Mac Pro, and far beyond even the fastest 27-inch iMac. The latter comparison is even more relevant, as the 27-inch iMac has seemingly disappeared from Apple’s online store.

You can think of the M1 Ultra as essentially a pair of M1 Max chips fused together into a humongous chip. Apple has crammed no less than 114 billion transistors into it, which is more than three times the amount of transistors found in an Nvidia GeForce RTX 3090 GPU and an Intel Core i9-12900K, combined.

This new chipset packs a 20-core CPU with 16 high-performance cores and four efficiency cores, a 64-core GPU, and a 32-core Neural Engine in a package that’s roughly eight time the size of the M1. This is made possible by Apple’s UltraFusion architecture with an interposer spanning some 10,000 interconnects, which begs the question of whether the company is interested in joining others in building the future of semiconductors through the Universal Chiplet Interconnect Express (or UCIe) standard.

Apple believes the M1 Ultra will “once again shock the PC industry,” which is certainly possible in terms of performance-per-watt, as previously shown with the rest of the Apple Silicon family. It can also be configured with a whopping 128 gigabytes of high-bandwidth unified memory for a total of 2.5 terabytes per second of low-latency bandwidth shared across the CPU, GPU, and Neural Engine.

More importantly, developers won’t have to rewrite code to tap into the full potential of the M1 Ultra.

Speaking of performance, Apple said the CPU in the new chipset is up to 90 percent faster than Intel’s Core i9-12900K paired with DDR5 memory, at least within a 60-watt power envelope. Or put differently, the M1 Ultra is able to match the performance of the Intel Alder Lake desktop flagship processor using 100 fewer watts.

Another area where the M1 Ultra is expected to shine is GPU compute performance for things like professional graphics rendering and video editing applications. Apple said the M1 Ultra’s 64-core GPU is capable of more than matching Nvidia’s RTX 3090 GPU while using 100 watts — three times less power than the latter part.

Beyond the CPU and GPU, as well as the oversized Neural Engine, the M1 Ultra has a media engine capable of playing back up to 18 simultaneous streams of 8K ProRes 422 video, which makes this the most powerful Mac to date.

Circling back to the Mac Studio, the chassis isn’t exactly large at 3.7 inch tall and 7.7 inch wide, so it should easily fit under most displays. The added space allowed Apple to include two blower fan cooling systems, but the company claims the Mac Studio will remain quiet even under heavy workloads.

You can configure the system with up to 8 terabytes of SSD storage that can deliver speeds of up to 7.4 gigabytes per second, which is on par with some of the fastest PCIe 4.0 SSDs in the PC space.

Connectivity options are also plentiful for a device this size, with four Thunderbolt 4 ports, a 10 Gb Ethernet port, an HDMI port, an audio jack, and two USB-A ports (working at 10 Gb per second)— a rare sighting in recent Apple product releases. There’s also support for Wi-Fi 6 and Bluetooth 5.0, and you have two additional USB-C ports on the front, as well as an SD card slot.

The Mac Studio is available for pre-order at $1,999, or at $1,799 if you’re eligible for an education discount. The Mac mini will continue to be sold, and shipping for the Mac Studio starts on March 18. Pair the new Mac with Apple’s newly-announced 27-inch 5K Studio Display, and you’ll suddenly wonder why the 27-inch iMac still exists. Or does it?

Permalink to story.

 

kiwigraeme

Posts: 1,052   +780
Seems to be a creators dream - if used 24/7 will really test onboard storages life span .
Wonder how this compares to Google's encoding engines - Not sure google is worried about a few watts - but they are re-coding lots of stuff to the new codecs - which are incredibly processor intensive .
I still think the OS, software is also a huge challenge - however parallel computing - was once considered very hard .
Not my favourite company - but Apple are good at this stuff - whether Apple hits mainstream is another matter .
Everyone is upping their game - so exciting times

Would be interesting if Apple can bypass the need for a GPU and match Nvidia/AMD - again software/drivers skill needed - but Apple has money
 

BigRedPDX

Posts: 271   +191
Would be interesting if Apple can bypass the need for a GPU and match Nvidia/AMD - again software/drivers skill needed - but Apple has money
At least you can use an eGPU on this machine. Match that with the mega processor they built and it should do pretty decently. I don't think Apple will come up with anything that'll match NVIDIA or AMD unfortunately.
 

Dimitriid

Posts: 2,212   +4,262
I was somehow wrong but not entirely: It is focusing a lot on just GPU horse power to make most of it's claims but it is just 2 chips together when I thought they wouldn't go this way.

They're also making some funny claims on the presentation saying how this interconnect is unrivaled in the entire industry yet their own terabytes per second claim is not as fast as AMD's infinity fabric. Which granted doesn't needs to be since this is dependent entirely on UMA where AMD still has traditional memory bus instead so it's probably a bit faster in practice but in theory, if I was AMD I would focus on that and become a reply guy to them with an "Actually, our interconnect tech is even faster" but again since there's no AMD UMA devices they probably won't bother to counter the claims.

This is still not enough horse power to counter a pc HEDT system that can add things like more than 1 GPU (Which is about what's needed to exceed what they pack) or more than 8tb of ssd storage to avoid the almost required external storage needs for these macs or more than 10gb ethernet because again, if there's only 10gb then you will also have that bottleneck to get to actually large storage needs and such.

The only thing I like it's how small they made it for the gpu cores they're packing inside that does look good in my book but as usual, not all that impressive besides that.
 

GettCouped

Posts: 39   +61
TechSpot Elite
I was somehow wrong but not entirely: It is focusing a lot on just GPU horse power to make most of it's claims but it is just 2 chips together when I thought they wouldn't go this way.

They're also making some funny claims on the presentation saying how this interconnect is unrivaled in the entire industry yet their own terabytes per second claim is not as fast as AMD's infinity fabric. Which granted doesn't needs to be since this is dependent entirely on UMA where AMD still has traditional memory bus instead so it's probably a bit faster in practice but in theory, if I was AMD I would focus on that and become a reply guy to them with an "Actually, our interconnect tech is even faster" but again since there's no AMD UMA devices they probably won't bother to counter the claims.

This is still not enough horse power to counter a pc HEDT system that can add things like more than 1 GPU (Which is about what's needed to exceed what they pack) or more than 8tb of ssd storage to avoid the almost required external storage needs for these macs or more than 10gb ethernet because again, if there's only 10gb then you will also have that bottleneck to get to actually large storage needs and such.

The only thing I like it's how small they made it for the gpu cores they're packing inside that does look good in my book but as usual, not all that impressive besides that.

I saw your post in the Threadripper article and you were very correct in Apple's interests. However, saying that Apple 'just' put two together is a bit of an injustice to how impressive this accomplishment is and making it transparent to the OS. This is the holy grail of GPU chiplets and why Apple can factually claim this is unrivaled. Let's see how it ends up but it is VERY impressive!
 

Dimitriid

Posts: 2,212   +4,262
I saw your post in the Threadripper article and you were very correct in Apple's interests. However, saying that Apple 'just' put two together is a bit of an injustice to how impressive this accomplishment is and making it transparent to the OS. This is the holy grail of GPU chiplets and why Apple can factually claim this is unrivaled. Let's see how it ends up but it is VERY impressive!
Not saying it isn't impressive but AMD has been doing it for a while with Ryzen chiplets and I remember there was very early on issues with their interconnet not being fast enough, depending on very fast ram speeds and latency and on the threadripper launch Windows not being able to prioritize the cores correctly.

I'm assuming Apple has been taking notes and is probably pushing all of that a step further in their software implementation and also on a hardware level since depending more on UMA and not just on large cpu cache means they're able to get a better version.
 

GettCouped

Posts: 39   +61
TechSpot Elite
Not saying it isn't impressive but AMD has been doing it for a while with Ryzen chiplets and I remember there was very early on issues with their interconnet not being fast enough, depending on very fast ram speeds and latency and on the threadripper launch Windows not being able to prioritize the cores correctly.

I'm assuming Apple has been taking notes and is probably pushing all of that a step further in their software implementation and also on a hardware level since depending more on UMA and not just on large cpu cache means they're able to get a better version.

Well what AMD does with Serial CPU processors and cores is MUCH MUCH easier than a single logic block presented to the OS in the form of a dual GPU.

BTW what AMD did was very hard just Apple's accomplishment is degrees more difficult because it's a GPU.
 

Dimitriid

Posts: 2,212   +4,262
Well what AMD does with Serial CPU processors and cores is MUCH MUCH easier than a single logic block presented to the OS in the form of a dual GPU.

BTW what AMD did was very hard just Apple's accomplishment is degrees more difficult because it's a GPU.
Oh ok I think I didn't catch the part about the GPU side I was just thinking CPU. I know AMD is working on that for GPUs too but it's not out yet so it makes sense that it's a very different kind of deal, thanks for clarifying btw.
 

ypsylon

Posts: 510   +522
In a way I'm a bit disappointed with Apple. They continue to glue M1 into bigger and bigger packages, but there is no real progress with something new and better. We already know what MacPro will be like: 4 or 8 M1 Ultras, but... it still IS the M1, chip which is 2 years old already.

Make no mistake. I love the Studio. Perhaps Jony Ive era is finally truly dead at Apple and about bloody time. Probably first Mac I really, really want, but in current market will never buy it. With war going on, inflation and currency exchanges in flux I keep my money no matter what. Even most basic version is already ~2500USD equivalent and I'm pretty certain it'll go up weekly just looking at the cluster...war in Ukraine; and speccing it up to what I would consider sensible that's 3400, but in two weeks it could be equally well 5000. :(

For studios you have to be deep inside Mac-ecosystem. Like that mentioning of rendering power, with lots of marketing BS. Well you basically are forced into C4D which has exclusive features on Mac, but then if you want to compare rendering in Redshift on Mac M1 silicone and CUDA, then better find some popcorn because it'll be murder. Even Blender on M1 vs CUDA is. That's the biggest problem of new Apple silicone. Inability to utilize and leverage CUDA ecosystem which is basically a monopoly in 3D space. And rendering on the cloud where you use offsite Redshift or Octane running on nVidia cards doesn't count as "rendering power".

As for built-in storage I wouldn't worry about longevity. Any reasonable studio won't operate on system drive to do all the heavy lifting + Mac backup capability is in its own class. Usually at least DAS or more commonly Cloud is used for all the work.