MP 7,1 Mac Pro Apple Silicon is going to eat our lunch

[mattspace]

I also think Apple is transitioning from the common "computers". Apple has been cohesively integrate all their devices. Now that Macs are also ARM based, it can now finally do the same.

No more bicycles for the mind. From now on, it's mental travelators and trams. You'll go where Apple decides, at the pace they decide.

Think Alike.

Don't Think.

As compared to Microsoft's new pitch for windows "the platform for platform creators", and emphasis on user agency being the foundation of computing.

 

[Sophisticatednut]

That's Intel Mac Pro though.

I think Apple will have Intel Mac Pro available for a couple more years (if you follow my train of thoughts in previous posts)

If apple’s roadmap is to be believed intel Mac will be dead in the end of 2022. So I’m skeptical of them keeping it at all. And it being an intel mac doesn’t negate the need of modularity and stupid amounts of memory for 3D animation and 8k rendering etc.

Anyway, it's debatable if Apple will have DIMMs in the coming Smaller Mac Pro. People's guess will be as good as anyone else perhaps

It’s debatable, but unlikely unless they want to downgrade the Mac Pro again to a expensive piece of aluminum.

Sorry..I believe you don't see the problem based on what you've posted in this thread

EDIT:

I also don't agree with what goMac said about UMA in the past few messages. I chose to disengage.

Indeed I don’t see the problem as no obvious technical one exist. I just said my idea. Not your idea

 

[strangerthanlight]

I hope the next (and final) Intel Mac Pro will be equipped with Sapphire Rapids and skip Ice Lake. If Intel enters Sapphire Rapids volume production in 2022 Q2 with no further delays, that would mean a late 2022 Intel Mac Pro which may come out together with a cheaper M1 Supreme/M1 Max Quad mini Mac Pro.

Up to 56 core with 64GB onboard HBM2 memory, support up to 4TB DDR5 ram (can’t imagine what Apple would charge for these but the nice thing about modular Mac Pro is you can find it elsewhere for much cheaper prices), future-proof PCIe 5 support and potentially up to 30x in AI performance. Granite Rapids will likely be using the same LGA4677 socket which means the 2022 Mac Pro can easily upgrade CPU along the roadmap. Apple just need to make available PCIe 4 based MPX 7800X Duo and Afterburner II with new ProRes encoder and decoder, throw in higher rated PSU to incorporate the new Intel chips and this will be hard to beat. No need to change the form factor even.

 

[Sophisticatednut]

I'm not sure you're reading all of what I'm posting.

You're right, that at basics, it's a cache coherency problem. I even used the words cache coherency in one of my recent posts.

But the problem is if you have multiple "expansion cards" with multiple banks of RAM, you no longer have unified memory. You have different pools of memory.

This is literally a non existent issue.
I could jus call it
Unified memory and expandable cash. The point of the cash is just to speed up data transfer from main storage to CPU. this cash being made up of 2-20 DDR sticks have zero relevance.

And again, yes, that's a cache coherency problem. Except now it's much bigger. You have tens or hundreds of gigs of data that all need to be managed and kept coherent.

Nothing needs to be kept coherent. The CPU or GPU will never read from the DDR sticks. They are just intermittent storage for the unified memory when it’s full.
The M1 max have two options now
1: purge the ram for new information and rewrite it from slow storage
2: use the SSD as RAM(this is when it slows down to a crawl)

With expandable DDR dimms available it would not kill the efficiency.

If you're going to have to solve that problem, then the current Apple Silicon architecture no longer makes sense. If you can synchronize multiple discrete banks of memory across multiple cards quickly, just make the GPU separate again.

If you can solve that problem, you've "solved" the same problem Apple Silicon is trying to fix in the first place.

Apple already solved it, and it’s nothing about syncing. The reason they don’t have a GPU memory is because PCIe-5 16x is still 2-4 times slower than DDR5 and with horrible latency. x16 link of pcie5 ~128GB/s vs. pcie4~64GB/s.

And LPDDR5 is normally just 50Gbps/Chanel. Dual Chanel (100Gbps)memory is normal. And sometimes quad memory forr servers(200Gbps). But apple went the extra mile and use 8 memory Chanel’s directly on the die with 400Gbps

There whole purpose of RAM is to eliminate SSD usage as it sucks.

That's why it doesn't make any sense. If Apple figures that out they can go back to a standard architecture with separate GPUs and CPUs. Why bother with any of this? If you get rid of the need for unified memory then _you don't need unified memory any more._ That's why I'm saying it doesn't make sense to combine a unified memory architecture with a discrete one. Either the discrete components will drag unified memory down, or you've designed discrete components that operate just as fast as unified memory, and then why bother with unified memory.

You already have this problem.
SSD storage right now drags unified memory speed as it’s 50 times slower.
Or you could store things in DDR5 moving the bottle neck up depending how much extra memory you have.
Or apple implement 8 Chanel memory with higher latency.

(My hunch again is that physics will probably be a problem here, and that one reason everything exists on the same package is to minimize the distance signals need to travel.)
Memory speed has nothing to do with the definition of UMA. At that speed, that would be faster than vanilla M1 anyway.

You are right that is why

There are UMA Ryzens with fast memory. There are UMA Ryzens with slow memory. Just depends what you're willing to spend. Ryzen systems like the Xbox Series X use faster UMA memory, faster than even M1 Max AFAIK.

Yes because it’s a custom solution. Apple will never use it tho as it’s x86. And they have their own solution.

You can even build PC with a Ryzen 4700S. That has DDR6, which is faster than what Apple uses. (AMD only sells it as a kit because it's literally PS5 CPUs/boards that didn't pass binning. Technically it's a UMA chip, but unfortunately the GPU is what didn't pass binning on it. Lol.)

DDR6 doesn’t exist. Your talking about GDDR6x and it could be used by apple. It’s just more expensive, more power hungry and much more latency than LPDDR5.

And no, you can build now computers with the same speed as apple M1 max, they just happen to be server class CPUs and not ever used in laptops.

And why would you ever want the Max pro to have a smaller form factor? Just remake the Mac mini instead unless you want the trash can crisis again.

 

[mattspace]

And no, you can build now computers with the same speed as apple M1 max, they just happen to be server class CPUs and not ever used in laptops.

Pointing out again, that Alder Lake, Intel's next laptop chip is already posting benchmarks exceeding the M1 Max, which would seem to bode well for Desktop versions.

 

[Melbourne Park]

....

With expandable DDR dimms available it would not kill the efficiency.

Apple already solved it, and it’s nothing about syncing. The reason they don’t have a GPU memory is because PCIe-5 16x is still 2-4 times slower than DDR5 and with horrible latency. x16 link of pcie5 ~128GB/s vs. pcie4~64GB/s.

And LPDDR5 is normally just 50Gbps/Chanel. Dual Chanel (100Gbps)memory is normal. And sometimes quad memory forr servers(200Gbps). But apple went the extra mile and use 8 memory Chanel’s directly on the die with 400Gbps

There whole purpose of RAM is to eliminate SSD usage as it sucks.

....

?I enjoyed your post.

On the RAM channels though, my understanding is the following:

2020 iMac - 2-channel memory
2017 iMac Pro 4 Channel Memory
2019 Mac Pro 6 Channel Memory
m1 - 8 Channel Memory
m1 Pro - 16 Channel Memory (2 x 8)
m1 Pro Max - 32 Channel Memory (4x8)

Pic of comparative size and layout of the 3 M processors shows the memory channels, in gold, I think right hand side on the M, bottom right and left on the Pro and the same but double the channels on the Pro.

Pic of M Pro (from Apple): One can see the memory channel chips on the lower left and right of the Pro chip, in between the black RAM on either side:

Pic of M Pro Max, and note that there are two memory channel chips on either side of the processor.

I can still imagine putting another Max Pro on the left and right of that memory and having three processors separated by memory, I guess sitting on common silicon? I thought main benefit of the faster LPDDR5 used in the Pros (compared to that used in the base M SoC) is that while faster LPDDR5 still uses little energy and runs cool. I thought that the other options were faster but hotter and less efficient. Which would be a lower priority in desktop.

 

[Sophisticatednut]

?I enjoyed your post.

On the RAM channels though, my understanding is the following:

2020 iMac - 2-channel memory
2017 iMac Pro 4 Channel Memory
2019 Mac Pro 6 Channel Memory
m1 - 8 Channel Memory
m1 Pro - 16 Channel Memory (2 x 8)
m1 Pro Max - 32 Channel Memory (4x8)

i cant say how many channels the iMac/pro or MAc pro uses, it depends on the bandwidth used.
the M1 have a quad memory channel 68GBps
M1 128bit 4 channel 68GBps LPDDR4
And actually M1 MAx have only 8 chanels
M1 pro 256bit 4 channel 204GBps LPDDR5
M1 Max 512bit 8 channel 406Gbps LPDDR5
more reading
The yellow ones on teh edge each heve two channels to each memmory die.
M1 have two LPDDR4 dies
M1 Pro two LPDDR5 dies
And M1 Max four LPDDR4 dies
That you can see on the pictures you provided
link to 4K resolution of your picture

Pic of comparative size and layout of the 3 M processors shows the memory channels, in gold, I think right hand side on the M, bottom right and left on the Pro and the same but double the channels on the Pro.

View attachment 1902753

Pic of M Pro (from Apple): One can see the memory channel chips on the lower left and right of the Pro chip, in between the black RAM on either side:

View attachment 1902754

Pic of M Pro Max, and note that there are two memory channel chips on either side of the processor.

View attachment 1902755

I can still imagine putting another Max Pro on the left and right of that memory and having three processors separated by memory, I guess sitting on common silicon?

And technically with the current presentation apple could use at max 256Gb DDR5 ram.
DDR5 can have 64Gb maximum die density. or ad two more memory modules on the top and bottom bringing it up to 512Gb maximum ram and theoretically Mind melting 800GBps bandwidth on a slightly bigger M1 Max.

imagine Dual M1 Max with 512-1Tb Ram

 

[Lammers]

Very true.

I also think Apple is transitioning from the common "computers". Apple has been cohesively integrate all their devices. Now that Macs are also ARM based, it can now finally do the same.

Apple computers will be less of a "computer" or "laptop" that we think of and more as IoT devices.

Like a Mars rover to the mission. Apple is creating IoT devices to provide a certain experience or "mission", rather than a "computer" for different things, with no real "mission".

I think if that were true then Apple wouldn’t have made the Mac Pro.

 

[Melbourne Park]

i cant say how many channels the iMac/pro or MAc pro uses, it depends on the bandwidth used.
the M1 have a quad memory channel 68GBps
M1 128bit 4 channel 68GBps LPDDR4
And actually M1 MAx have only 8 chanels
M1 pro 256bit 4 channel 204GBps LPDDR5
M1 Max 512bit 8 channel 406Gbps LPDDR5
more reading
The yellow ones on teh edge each heve two channels to each memmory die.
M1 have two LPDDR4 dies
M1 Pro two LPDDR5 dies
And M1 Max four LPDDR4 dies
That you can see on the pictures you provided
link to 4K resolution of your picture


And technically with the current presentation apple could use at max 256Gb DDR5 ram.
DDR5 can have 64Gb maximum die density. or ad two more memory modules on the top and bottom bringing it up to 512Gb maximum ram and theoretically Mind melting 800GBps bandwidth on a slightly bigger M1 Max.

imagine Dual M1 Max with 512-1Tb Ram

I read the memory channel differences, at another site. I copied the figures into text edit. Perhaps they're wrong, but if one looks at the chips, you can see the 8 channel chips and also the differences between the Pro and Pro Max visually - one having half the bandwidth of the other, and half those channels. The problem though is that all the pictures are sourced from Apple, and evidently they are not quite accurate!!! Some things are missing on the bottom of their illustrations!

However, I'm posting because of Anandtech's article on the new processors. They say that for single core, the usage of the RAM is hits 102 GB/sec. I presume, like the M1.

The bandwidth being achieved depends on the demands of the various SoC's requirements, but in fact, the actual draw they measured was not much more than half the capacity of the memory bandwidth. I think the CPU could only call on around 224 max bandwidth.

I presumed when reading that article when it came out a few weeks ago, that Apple had put in a great deal of bandwidth perhaps, for multiple SoC usage ... that was my optimistic interpretation! So my view was that adding an extra CPU would mostly just utilise bandwidth that is mostly only being half utilised at the moment.

I'l quote Anandtech https://www.anandtech.com/show/17024/apple-m1-max-performance-review/2 :
Quote:
From a single core perspective, meaning from a single software thread, things are quite impressive for the chip, as it’s able to stress the memory fabric to up to 102GB/s. This is extremely impressive and outperforms any other design in the industry by multiple factors, we had already noted that the M1 chip was able to fully saturate its memory bandwidth with a single core and that the bottleneck had been on the DRAM itself. On the M1 Max, it seems that we’re hitting the limit of what a core can do – or more precisely, a limit to what the CPU cluster can do.

The little hump between 12MB and 64MB should be the SLC of 48MB in size, the reduction in BW at the 12MB figure signals that the core is somehow limited in bandwidth when evicting cache lines back to the upper memory system. Our test here consists of reading, modifying, and writing back cache lines, with a 1:1 R/W ratio.

Going from 1 core/threads to 2, what the system is actually doing is spreading the workload across the two performance clusters of the SoC, so both threads are on their own cluster and have full access to the 12MB of L2. The “hump” after 12MB reduces in size, ending earlier now at +24MB, which makes sense as the 48MB SLC is now shared amongst two cores. Bandwidth here increases to 186GB/s.

Adding a third thread there’s a bit of an imbalance across the clusters, DRAM bandwidth goes to 204GB/s, but a fourth thread lands us at 224GB/s and this appears to be the limit on the SoC fabric that the CPUs are able to achieve, as adding additional cores and threads beyond this point does not increase the bandwidth to DRAM at all. It’s only when the E-cores, which are in their own cluster, are added in, when the bandwidth is able to jump up again, to a maximum of 243GB/s.

While 243GB/s is massive, and overshadows any other design in the industry, it’s still quite far from the 409GB/s the chip is capable of. More importantly for the M1 Max, it’s only slightly higher than the 204GB/s limit of the M1 Pro, so from a CPU-only workload perspective, it doesn’t appear to make sense to get the Max if one is focused just on CPU bandwidth."

 

[Sophisticatednut]

?I enjoyed your post.

On the RAM channels though, my understanding is the following:

2020 iMac - 2-channel memory
2017 iMac Pro 4 Channel Memory
2019 Mac Pro 6 Channel Memory
m1 - 8 Channel Memory
m1 Pro - 16 Channel Memory (2 x 8)
m1 Pro Max - 32 Channel Memory (4x8)

Pic of comparative size and layout of the 3 M processors shows the memory channels, in gold, I think right hand side on the M, bottom right and left on the Pro and the same but double the channels on the Pro.

View attachment 1902753

Pic of M Pro (from Apple): One can see the memory channel chips on the lower left and right of the Pro chip, in between the black RAM on either side:

View attachment 1902754

Pic of M Pro Max, and note that there are two memory channel chips on either side of the processor.

View attachment 1902755

I can still imagine putting another Max Pro on the left and right of that memory and having three processors separated by memory, I guess sitting on common silicon? I thought main benefit of the faster LPDDR5 used in the Pros (compared to that used in the base M SoC) is that while faster LPDDR5 still uses little energy and runs cool. I thought that the other options were faster but hotter and less efficient. Which would be a lower priority in desktop.

like this

 

[Sophisticatednut]

I read the memory channel differences, at another site. I copied the figures into text edit. Perhaps they're wrong, but if one looks at the chips, you can see the 8 channel chips and also the differences between the Pro and Pro Max visually - one having half the bandwidth of the other, and half those channels. The problem though is that all the pictures are sourced from Apple, and evidently they are not quite accurate!!! Some things are missing on the bottom of their illustrations!

However, I'm posting because of Anandtech's article on the new processors. They say that for single core, the usage of the RAM is hits 102 GB/sec. I presume, like the M1.

The bandwidth being achieved depends on the demands of the various SoC's requirements, but in fact, the actual draw they measured was not much more than half the capacity of the memory bandwidth. I think the CPU could only call on around 224 max bandwidth.

I presumed when reading that article when it came out a few weeks ago, that Apple had put in a great deal of bandwidth perhaps, for multiple SoC usage ... that was my optimistic interpretation! So my view was that adding an extra CPU would mostly just utilise bandwidth that is mostly only being half utilised at the moment.

I'l quote Anandtech https://www.anandtech.com/show/17024/apple-m1-max-performance-review/2 :
Quote:
From a single core perspective, meaning from a single software thread, things are quite impressive for the chip, as it’s able to stress the memory fabric to up to 102GB/s. This is extremely impressive and outperforms any other design in the industry by multiple factors, we had already noted that the M1 chip was able to fully saturate its memory bandwidth with a single core and that the bottleneck had been on the DRAM itself. On the M1 Max, it seems that we’re hitting the limit of what a core can do – or more precisely, a limit to what the CPU cluster can do.

The little hump between 12MB and 64MB should be the SLC of 48MB in size, the reduction in BW at the 12MB figure signals that the core is somehow limited in bandwidth when evicting cache lines back to the upper memory system. Our test here consists of reading, modifying, and writing back cache lines, with a 1:1 R/W ratio.

Going from 1 core/threads to 2, what the system is actually doing is spreading the workload across the two performance clusters of the SoC, so both threads are on their own cluster and have full access to the 12MB of L2. The “hump” after 12MB reduces in size, ending earlier now at +24MB, which makes sense as the 48MB SLC is now shared amongst two cores. Bandwidth here increases to 186GB/s.

Adding a third thread there’s a bit of an imbalance across the clusters, DRAM bandwidth goes to 204GB/s, but a fourth thread lands us at 224GB/s and this appears to be the limit on the SoC fabric that the CPUs are able to achieve, as adding additional cores and threads beyond this point does not increase the bandwidth to DRAM at all. It’s only when the E-cores, which are in their own cluster, are added in, when the bandwidth is able to jump up again, to a maximum of 243GB/s.

While 243GB/s is massive, and overshadows any other design in the industry, it’s still quite far from the 409GB/s the chip is capable of. More importantly for the M1 Max, it’s only slightly higher than the 204GB/s limit of the M1 Pro, so from a CPU-only workload perspective, it doesn’t appear to make sense to get the Max if one is focused just on CPU bandwidth."

it comes from the specifications of how DDR5 and DDR4 is implemented. apple have stated the Bus size of 512bit and 400Gbps bandwith this makes it obligatory to be 8 Chanel memory

And the bandwith limitations is more likely a cap inorder that the CPU ant use 400Gbps in total, but allow the GPU to use extra bandwith.
Link
as you can see in the right hand corner 4 chanels= 256Bit as the M1 Pro have.

2 chanels on the left= 128Bit and 2 chanels on the right= 128Bit. giving it a 256Bit 50Gbit(DDR5 single chanel bandwith) X 4 chanels= 200Gbps

And the M1 Max have double this giving us 512Bit memory controlers devided on 8 chanels. 8 channels X 50Gbps(DDR5 single chanel bandwith)= 400Gbps bandwith.

 

[Melbourne Park]

Apple and experts all talk about how many memory channels.

I thought that Apple was using 4 x 8 memory channels on Pro Max, and 2 x 8 on the Pro. Each bank being 128. And one 8 channel on the M1. Think about it! I did think though that the M1's was not 128, I thought 64, I forget why. Maybe I am wrong about that ...

And Apple do state they have 6 channels on the Mac Pro 2019 https://support.apple.com/en-us/HT210405.

I thought too that the reason why the iMac Pro was faster at several tests despite having slower CPU etc., was due to it's having 4 memory channels compared to a max spec iMac's having two.

And Anandtech stated that the M1 had 8 channels ( each 16 bit ).

https://images.anandtech.com/doci/16252/M1_575px.png

My main interest with all this, is why would Apple have over 400 width for its memory, when Anantech's tests show a max usage capability of the CPU to use just half that, and a max usage all up of around just 250?? It seemed to me and still does, that a duel CPU desktop could use the same 400 width and that it would be able to use all that bandwidth / capability. Fine tune / balance the performance with faster memory that is available. I can certainly see an easy to produce twin CPU Apple desktop. If a Pro Max costs $400, then I can see a duel CPU desktop costing CPU wise, $600 extra for the duel Max desktop compared to a single Max desktop. Wouldn't it be nice if a four processor one cost an extra $1200 one top of that. And it would be profitable for Apple to do so. Pricing is important too because IMO competitive pricing would greatly increase volumes, which would mean more profits. it would be nice to see Mac Pros as being more affordable, and with less expansion capabilities then the prices could ... I am getting too off topic.

Frankly though, the main issue for me is, I have not seen that the Max's GPU capabilities are being utilised that much. But we'll be finding out I guess.

 

[4wdwrx]

I think if that were true then Apple wouldn’t have made the Mac Pro.

Apple Silicon was unveiled after the Mac Pro. With the Apple Silicon, they can integration that they never could have done.

Imagine your iPhone will be also your Mac. You bring your iPhone close to your Mac, it becomes one. When you walk off, the iPhone has everything. Then go to your office Mac, everything is there.

 

[Sophisticatednut]

Apple Silicon was unveiled after the Mac Pro. With the Apple Silicon, they can integration that they never could have done.

Imagine your iPhone will be also your Mac. You bring your iPhone close to your Mac, it becomes one. When you walk off, the iPhone has everything. Then go to your office Mac, everything is there.

thats an absolute nightmare for profesional users. let the Mac pro be a Pro machine. and the Mac mini, iMac and macbook air etc be consumer devices.

the Mac pro must be modular. and thankfuly Apple have stated that iOS will stay seperate from Mac os. plus the yhate sideloading and would never allow that to happen

 

[4wdwrx]

No more bicycles for the mind. From now on, it's mental travelators and trams. You'll go where Apple decides, at the pace they decide.

Think Alike.

Don't Think.

As compared to Microsoft's new pitch for windows "the platform for platform creators", and emphasis on user agency being the foundation of computing.

We all follow the flow. Marketing (society) makes people believe they are unique, in reality we all follow the same pace.

You weren't a creator if Microsoft didn't say you will become one if you use windows "the platform for platform creators".

 

[Boil]

Just seeing that render at the end, I could go for that with quad SoCs; but PLEASE Apple, maybe offer Space Gray or Space Black; and since there is no intake on the bottom, and the smaller Cube shape would be intended for the desktop (not deskside, aka on the floor), I feel the "feet" from the new MacBook Pro laptops would fit the design...!?!

We might get a preview of an Apple silicon powered Mac Pro at WWDC 2022, but if actual availability is not until December of 2022, I would think the first Apple silicon Mac Pro would be derived from the M2 / M2 Pro / M2 Max series of SoCs...?

I can see Apple showcasing the incredible power of dual & quad SoC Mac Pro workstations; and hinting that single SoC M2 Pro / Max laptops, desktops, & AIOs will be available early 2023...?

Mn Single SoC Macs

• 12" MacBook
• 14" MacBook
• Mac mini (shorter chassis)
• 24" iMac

Mn Pro/Max Single SoC Macs

• 14" MacBook Pro
• 16" MacBook Pro
• Mac mini Pro (taller chassis)
• 27" iMac Pro

Mn Pro/Max Dual/Quad SoC (SiP) Macs

• Mac Pro Cube (for those who don't need PCIe slots)
• Mac Pro (for those who need numerous PCIe slots)
• 32" iMac Pro

 

[Sophisticatednut]

Just seeing that render at the end, I could go for that with quad SoCs; but PLEASE Apple, maybe offer Space Gray or Space Black, INCLUDING the feet...!?!

We might get a preview of an Apple silicon powered Mac Pro at WWDC 2022, but if actual availability is not until December of 2022, I would think the first Apple silicon Mac Pro would be derived from the M2 / M2 Pro / M2 Max series of SoCs...?

I can see Apple showcasing the incredible power of dual & quad SoC Mac Pro workstations; and hinting that single SoC M2 Pro / Max laptops, desktops, & AIOs will be available early 2023...?

Mn Single SoC Macs

• 12" MacBook
• 14" MacBook
• Mac mini (shorter chassis)
• 24" iMac

Mn Pro/Max Single SoC Macs

• 14" MacBook Pro
• 16" MacBook Pro
• Mac mini Pro (taller chassis)
• 27" iMac Pro

Mn Pro/Max Dual/Quad SoC (SiP) Macs

• Mac Pro Cube (for those who don't need PCIe slots)
• Mac Pro (for those who need numerous PCIe slots)
• 32" iMac Pro

completly agree with some small deviations.
Mn Single SoC Macs

• 12" MacBook
• 14" MacBook
• Mac mini (shorter chassis)
• 
  
  
• 24" iMac

Mn Pro/Max Single SoC Macs

• 14" MacBook Pro
• 16" MacBook Pro
• Mac mini Pro (taller chassis)
• 
  
  
• 27" iMac Pro

Mn Pro/Max Dual/Quad SoC (SiP) Macs

• Mac Pro Cube (for those who don't need PCIe slots) server version and why make two different motherboards? if you need this much power PCIe slots are needed
• 
• Mac Pro (for those who need numerous PCIe slots)
• 
• 32" iMac Pro

 

[kvic]

Apple could use something like Infinity Fabric to link multiple SoCs together, like AMD is doing. But that's not true unified memory space, it just makes the CPUs and GPU's act like they're in a single unified space. Still interesting, but it's not the same performance outcome.

It's also not really necessary. The rumored "quad" M1 Max chip will be plenty hot and plenty fast. And probably extremely expensive.

"Unified memory" is over hyped. But indeed, Infinity Fabric (or Apple's proprietary creation) can be used on connect multiple SoCs, and also connect multiple MPX-like SoCs modules with a jumper cable. I also doubt it Apple will do or have to do so in their first incarnation of Mac Pro.

Recent rumours saying Apple is doing a server chip. If so, it'll completely change people's imagination about subsequent Mac Pro models.

 

[mattspace]

We all follow the flow. Marketing (society) makes people believe they are unique, in reality we all follow the same pace.

You need to spend time with a more diverse group of people.

You weren't a creator if Microsoft didn't say you will become one if you use windows "the platform for platform creators".

I've tried reading this a dozen times, and I can't pull the meaning from it. Are you arguing that Microsoft limited the creative choices of Windows users in the past, or limited what types of business you could build on Windows?

 

[Boil]

Mn Single SoC Macs

• View attachment 1903058 Silver Mac mini

Mn Pro/Max Single SoC Macs

• View attachment 1903057 Space Gray Mac mini

I would be alright with current Mac mini chassis, Silver (or Starlight...?) for Mn-series models & Space Gray (or Space Black...?) for the Mn Pro/Max-series models...

Mn Pro/Max Dual/Quad SoC (SiP) Macs

• Mac Pro Cube (for those who don't need PCIe slots) server version and why make two different motherboards? if you need this much power PCIe slots are needed

Just a single example, but a 3D artist (modeling/shading/lighting/etc.) could use "this much power" and not need a single PCIe slot (assuming dual 10Gb Ethernet)...?

Gotta have that Quad ASi Cube...! ;^p

 

[mattspace]

Just a single example, but a 3D artist (modeling/shading/lighting/etc.) could use "this much power" and not need a single PCIe slot (assuming dual 10Gb Ethernet)...?

And in 12 months time, when that power has caused them to double their poly count, and want to run their viewport fully ray traced on a larger display, and consumer-level PCI GPUs are doing all that for less than the cost of bumping his RAM spec up one level, and they're only 25% of the way through the repayments on the machine, and the secondhand market has tanked because the hardware isn't upgradable by secondhand purchasers, and Apple has released a new model with upgraded GPU...

 

[singhs.apps]

And in 12 months time, when that power has caused them to double their poly count, and want to run their viewport fully ray traced on a larger display, and consumer-level PCI GPUs are doing all that for less than the cost of bumping his RAM spec up one level, and they're only 25% of the way through the repayments on the machine, and the secondhand market has tanked because the hardware isn't upgradable by secondhand purchasers, and Apple has released a new model with upgraded GPU...

Some users pretty much want an AIO (minus display) cube Mac Pro or Mac mini pro and call it a day.

But the whole point of the 2019 Mac Pro was going back to what really works for all in this segment (minus portability)

I fail to see how expansion in a Mac Pro or extra DDR 5 modules off chip is a bad thing. It didn’t make sense in the Cylinder days and it doesn’t make sense now.

What happens after Apple releases the last Intel Mac Pro ? How long will they support that system ? 5 years ?
After that will all Mac pro users be left with a closed cube/cylinder/whatever and you spec it out based on your need at the time of order and remain stuck with that system until you feel the need to buy a whole new system again ?

What about multi GPU ? Or extra ram ? Or add in cards ?

The quad m1 max GPU could be a good as say a 4090 (minus RTX) but what about 2x 4090 ? Or other newer ones down the road ?

Why not keep the expansion feature as is ? What’s to stop Apple releasing its own discreet GPUs to plug into the Mac Pro over PCIe 5 ( 2x 128 core GPU with 256GB ram on the same module )? Heck it could develop its own GPU interconnect like nvlink (say 200 GBPs) and you can plonk in 2 such modules.

Apple could well insert its own ray tracing cores on these GPUs to plug the performance gap.

They could well repurpose these modules as eGPUs for non Mac Pro Apple users.

Or Dram over PCIe (like Samsung is trying )
PCIe Gen 6 will have 128 GBPs uni directional speed. That’s more than 50% above the max memory bandwidth of my intel CPU

Apple’s strategy seems to be building CPUs around powerful GPUs… at least that’s what I am guessing, as a way around the 64 thread limit of macOS (nothing to stop them going higher with a new version with higher thread support in the future )
But with tasks that can scale upto so many parallel processing…why not switch to GPUs ?
The only bottle neck I can see is the traditional discreet GPU requiring fast on board ram which, in my limited understanding, I think apple can solve very easily (the UMA thingie) or that MPX module with gobs of onboard RAM over PCIe

 

[Boil]

And in 12 months time, when that power has caused them to double their poly count, and want to run their viewport fully ray traced on a larger display, and consumer-level PCI GPUs are doing all that for less than the cost of bumping his RAM spec up one level, and they're only 25% of the way through the repayments on the machine, and the secondhand market has tanked because the hardware isn't upgradable by secondhand purchasers, and Apple has released a new model with upgraded GPU...

And they still won't need a PCIe slot, because Apple is not using discrete GPUs; but they can soldier on another year, get a new machine & the old machine becomes a render node...

 

[mattspace]

And they still won't need a PCIe slot, because Apple is not using discrete GPUs; but they can soldier on another year, get a new machine & the old machine becomes a render node...

Every new machine purchase is an opportunity to lose an existing customer. Every upgrade purchase is a chance to ensnare them more deeply.

 

[Boil]

Some users pretty much want an AIO (minus display) cube Mac Pro or Mac mini pro and call it a day.

Exactly, I (personally) do not need expansion slots; a Mn Max-powered Mac mini would be good, but a dual or quad Mn Max-powered Mac Pro Cube would be better...!

But I plan to get a M1 Max (32-core GPU/64GB RAM/1TB SSD/10Gb Ethernet) Mac mini (Pro) when they become available (fingers crossed)...!

I fail to see how expansion in a Mac Pro or extra DDR 5 modules off chip is a bad thing.

Never said it was, but not everyone needs all the expansion the current (massive) Mac Pro provides...!

Why not keep the expansion feature as is ?

See above, a smaller Mac Pro without SO many expansion slots would serve many...!