Apple Silicon Mac Pro could combine two M1 Ultra chips for speed
The Apple Silicon Mac Pro may follow the lead of the Mac Studio's M1 Ultra, a rumor claims, by effectively combining two M1 Ultra chips into a single 40-core SoC.
-xl-xl.jpg)
Apple's "Peek Performance" event included a brief tease for a Mac Pro using Apple Silicon, opening the door to speculation about the inbound update. In one rumor, it is said that Apple will go one stage further than it did with the Mac Studio in reusing its existing chips in new ways.
For the Mac Studio, Apple introduced the M1 Ultra, a chip that connected together two M1 Max chips with a die-to-die interconnect called UltraFusion. The concept effectively makes two chips work as one singular powerful version, complete with 20 CPU cores, a 64-core GPU, and 32 Neural Engine cores.
An image leak by "Majin Bu" on Twitter claims to show a schematic for an interconnect that will connect "2 M1 Ultra together," extending the concept by another level. The leaker says the bridge will be "found in the new 2022 Mac Pro," with a processor name of "Redfern," and is scheduled for a September release.
If the image is correct, the supposed four-chip assembly will practically introduce a new long bridge that sits the two M1 Ultra assemblies side by side. Three interconnects would be used in total to connect the four M1 Max chips, including the two used to form a pair of M1 Ultra chips.
However, the interconnect as shown would have some limitations. Without a newer bridging technology, RAM would be limited to the same 128GB as the Mac Studio supports.
The idea of a Mac Pro with such a high number of cores has been presented in earlier rumors. In May 2021, there were claims the Mac Pro could use chips with 20 or 40 computing cores, and graphics options with either 64 or 128 cores.
If the leak is genuine, the proposed Apple Silicon chip will offer considerably more cores than the existing Mac Pro offers. Under the existing Intel version, the maximum amount of cores is offered by a 28-core Intel Xeon W processor. But, maximum RAM would be less on the Apple Silicon Mac Pro versus the Intel Mac Pro.
Read on AppleInsider
Apple's "Peek Performance" event included a brief tease for a Mac Pro using Apple Silicon, opening the door to speculation about the inbound update. In one rumor, it is said that Apple will go one stage further than it did with the Mac Studio in reusing its existing chips in new ways.
For the Mac Studio, Apple introduced the M1 Ultra, a chip that connected together two M1 Max chips with a die-to-die interconnect called UltraFusion. The concept effectively makes two chips work as one singular powerful version, complete with 20 CPU cores, a 64-core GPU, and 32 Neural Engine cores.
An image leak by "Majin Bu" on Twitter claims to show a schematic for an interconnect that will connect "2 M1 Ultra together," extending the concept by another level. The leaker says the bridge will be "found in the new 2022 Mac Pro," with a processor name of "Redfern," and is scheduled for a September release.
If the image is correct, the supposed four-chip assembly will practically introduce a new long bridge that sits the two M1 Ultra assemblies side by side. Three interconnects would be used in total to connect the four M1 Max chips, including the two used to form a pair of M1 Ultra chips.
However, the interconnect as shown would have some limitations. Without a newer bridging technology, RAM would be limited to the same 128GB as the Mac Studio supports.
The idea of a Mac Pro with such a high number of cores has been presented in earlier rumors. In May 2021, there were claims the Mac Pro could use chips with 20 or 40 computing cores, and graphics options with either 64 or 128 cores.
If the leak is genuine, the proposed Apple Silicon chip will offer considerably more cores than the existing Mac Pro offers. Under the existing Intel version, the maximum amount of cores is offered by a 28-core Intel Xeon W processor. But, maximum RAM would be less on the Apple Silicon Mac Pro versus the Intel Mac Pro.
Read on AppleInsider
Comments
John Turnes of Apple quite literally said that the M1 Ultra completes the M1 line up just days ago
It is more likely that the MacPro will be based on M2 silicon or even a completely different design
Stuart
And, fingers crossed, some expandability.
We shall see…
The Mac Pro can be offered with M1 Ultra at the entry level and an M1 Ultra Duo. There is no higher name than Ultra. Pro = better, Max = maximum, Ultra = beyond maximum. They can only call it Ultra something like infinity + 1. I expect they will be able to offer 256GB RAM on the Ultra Duo. The amount of RAM they need to offer is just what people have been installing, they don't need to support 1TB+ just because other computers do.
I don't want to violate copyrighted images so check out max tech, Apple's M2 Ultra DUO Mac Pro, 7:34 mark.
So what could differentiate the Mac Pro? In a word: expandability.
1) PCIe slots. The M1 Ultra seems to have plenty of I/O potential, and a fast PCIe bridge chip would easily enable a lot of expansion potential.
2) Drive bays. The Mac Pro would have the same built-in super fast SSD, but in a large case a whole lot of additional storage can be accommodated.
3) RAM. This is where it gets tricky. The Apple Silicon approach is to use in-package memory, and there are real constraints on how much can be put into a single package. Some Pros just need more than can be fit into a single package, or more than is worth building in the TSMC production run. So conventional DIMMs are needed to supplement the super fast in-package memory. The question is, how does OSX use it? Apple seems to want to keep the programming model simple (i.e. CPU/GPU shared memory with a flat/uniform 64-bit virtual address space), so having some fast vs slow areas of memory doesn't seem like the direction they want to go in (although they could and just rely on the M1 Ultra's ENORMOUS caches). They are already doing virtual memory paging to flash, however... so why not do virtual memory paging to the DIMMs instead? Big DMA data transfers between in-package and on-DIMM memory across the very fast PCIe 5.0 lanes would ensure that the available bandwidth is used as efficiently as possible, and the latency is masked by the big (page-sized) transfers. A 128GB working memory (the in-package RAM) is huge, so doing VMM to get to the expanded pool is not as bad as you might think. Such a memory scheme may even just sit on PCIe cards so buyers only need to pay for the DIMM slots if they really need it. Such "RAM disk" cards have been around for ages, but are usually hampered by lack of direct OS support... and issue Apple could fix easily in their kernel.
I think Apple will comeup with a version of Ultra fusion that allows using multiple SOCs without performance issues. They might even come up with GPU only chips that users can add as an extra GPU, or maybe give users the ability to swap entire SOCs in future.
I think it's safe to say this isn't happening.
I agree.
This was speculated on shortly after the M1 Pro and Max reveal:
How about putting all the M1 Max interconnects facing towards each other with Apple obviously making another Interconnect that would tie all 4 together with another chip? I think it's doable, I just don't know if the number of Mac Pros being sold would warrant the cost of development and production.
I have no idea what Apple will do, but I'm not going to say it's impossible simply because it hasn't been done yet.
1. It could in fact still be called an M1 Ultra, just adds two new upgrade options…
20/48, 20/64, 40/96, 40/128
2. That’s definitely possible. Maybe they break apart the SoC into a few discrete chips; X1 CPU (32/48/64), G15 GPU (96/128/160), and a new “Afterburner” card with all the other bits; Neural Engine, Video Codecs, ISP, etc. that way they could still offer Intel XEON and AMD GPUs for professionals that needed it.
The faster the better of course but for machines like the Mac Pro where the workloads will be highly parallel, they can treat them independently and it will still scale very well as it does for the AMD GPUs:
https://barefeats.com/pro-w6800x-quadruple.html
Apps like After Effects and Da Vinci can send the same data to each Ultra chip and render different frames and it will give a linear speedup. Even single frames can be split and processed separately, including real-time. Multi-GPU systems do this all the time.
From what Apple described, it doesn't sound like they'll make a 4-die chip with a 4-way UltraFusion unless they figured out a way to stack it. They can just connect two separate M1 Ultras with a lower bandwidth connection. They can call it Ultra Duo or Ultra II. They could have more than two if they have a bigger box but I don't think they need to.
With the introduction of the Mac Studio, I wouldn't be surprised to see the Mac Pro go primarily rackmount. Very few people need more computing power than the Mac Studio offers at their desks. Almost everyone who does need more computing power is in an environment where they can rack the computer in a closet. For example, recording studios, film studios, scientific labs, and so on all have 19" rack space for other equipment, so putting specialist workstations in there isn't a stretch. That said, rackmount would mostly be relevant for a box with several full-height, full-length PCIe slots (e.g., to add hardwired audio and video inputs), and I'm not yet convinced Apple is interested in that at all. They might say the future is a rackmount interface which connects to the system via Thunderbolt. I'd be curious to know what they have seen the current rackmount Mac Pro doing.
Called the idea of using one high-end die in the laptops, then multiple high-end dies in the high-performance desktops, but I expected it to be the M1 Max Duo or something rather than a whole new name. I did get the dGPU very wrong, though. I didn't quite understand at that time just how powerful Apple's GPU cores are.
Not quite sure that’s physically possible as the memory interfaces are on the sides of the chip…