Agreed Apple is becoming the king of E-waste.. Expecting ppl to throw away every computer in the line-up every 3 yrs?
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.
MP All Models 40 core ASI with PCI-E slot is around the corner!!
- Thread starter Amethyst
- Start date
- Sort by reaction score
I'm also probably crazy, but I think the Mac Pro.. in this horrible configuration might actually see regular updates.. since its is the same SoC as the Studio... unless Apple actually spend the time to make the MP "special" with the next iteration.. M3 Extreme? it will fall behind if it is not updated every 15? months like the Studio.
@Amethyst you were right again! Please keep the info coming, I will definitely be watching for your next murmurs from your friend.
Hopefully now the dreamers will finally understand that "ComputeModule" is a device to run iOS!
I'm disappointed there's no new monster Apple GPU for the Mac Pro like I expected, nor an expandable RAM option. But personally speaking, I honestly don't need either of those features. As a working editor of feature films and broadcast / streaming series, my current M1 Ultra is awesome with "just" 64GB RAM and the 48-core GPU. Film / video editors specifically should be perfectly satisfied with current options. It sounds like music producers and other audio professionals are more or less covered by this also, although I'm sure more RAM could always be better if available.
Heavy 3D rendering like VFX needs more GPU power than M2 Ultra, but everyone says VFX industry already has standardized on NVIDIA and Linux anyway. Looks like Apple just decided to cut their losses there. Resolve colorists could always use more GPU, but I'm guessing they're probably covered for most of today's 4K-8K camera media with the top-spec Ultra GPU.
And despite not yet having a "4090 killer" high-end 3D GPU, really cool to see that Apple is pushing harder into gaming with the DirectX 12 Wine translation layer in the "Game Porting Toolkit," macOS Sonoma Game Mode, the partnership with Kojima etc. Nobody can say now they don't have any interest in games, as games are a huge revenue driver in their App Stores.
While this first Apple Silicon shredder is only just the "minimum viable Mac Pro" it's nonetheless cool to see so many PCIe 4.0 slots, dual 10Gbe, 8 Thunderbolt ports on Apple Silicon for the first time! I hope we see much more frequent annual-ish updates on Mac Pro now, and in future versions I'm still hopeful we could eventually see that "Extreme" SoC, even bigger GPUs, expandable RAM, and other Pro wishlist features we didn't get this year.
The Mac Studio is known to have power issues with its TB ports, for bus-powered peripherals it can choke as soon as the 2nd one is plugged in, namely NVMe SSD enclosures. It seems the machine has a shared "pool" of wattage available for all buses, and once you reach a certain point, macOS won't let you add more; I frequently have to unplug and re-shuffle what are plugged into my Studio when I try to have two NVMe enclosures mounted at the same time.
Then with I/O cards and network cards there are issues with latency when done externally. Really sensitive audio equipment also can pick up electrical noises with crappy PSU (on 3rd party PCIe enclosures). And the usual clocking issues with audio.
The problem with the new Mac Pro is that it only solves the above problems but nothing else.
Chances are that egpu box doesnt support bifurcation, so you’ll need that board with nvme carriers to have either a hardware raid on the card or a pcie switch if you want to stuff a pile of fast SSDs in that way - which is probably more expensive than just buying the MP
Hardware raid cards are about $400-ish and they will saturate the thunderbolt bus (which is fast enough for my use case as I said). I already have the box and a platinum power supply. There are some terrible implementations that bifurcate each SSD to its own lane, a lot of OWC offerings are like that and effectively force each drive down to 1x but plenty of others don't and will dynamically adjust so you're always getting the full thunderbolt speed. My use case involves almost entirely doing a one-time (per project) read from SSD into RAM, never writing to those volumes, so even a 20-30 second load is fine to get everything moved around since it's a one-time thing per session. I use Optane drive(s) without RAID for working project storage (and OS drives for my non-T2 Mac computers) due to their incredible write endurance and random read performance, the latency difference is noticeably better.
I'm not saying it's perfect, but it would at least be portable to a new machine. I have a $40 usb-c carrier for a single M.2 drive that gives me 1200mb/sec on an old intel Mac now and have used that for SDX Libraries without any problem. The nice thing with Audio sample libraries is I can put them all over the place so I don't need one large storage volume. I DO need a ton of RAM if I'm working with a lot of Orchestral samples, just one instance can use 7-11GB of memory so my $ would go to RAM vs a huge permanent SSD that I can't take to my next computer. I would absolutely choose the Mac Pro if it was "only" $1k-2k more, but $3k is really pushing it too far especially when M3 is expected to be such a large jump in 18-24 months.
There's also spec confirmation now direct from Apple that the M2 Studio is 10-11 decibels quieter than the M1 Studio, so they did something to fix the fan curve which makes it even more appealing.
I hope someone gets an M2 Mac Pro and M2 Studio and compares them at max clock etc. because I'm even more intrigued now. Maybe there really is a significant clock difference at full load given the massive thermal and power headroom difference. It would at least be something "extra" to justify the price.
Last edited:
Apple expects its users to replace every 4 years. Intel every 5-6 years.
And you can recover the residual cost by selling it to someone who is looking for used.
They likely run at exactly the same speed. This has been the case across all the other machines; the M-series is no faster in the Studio vs. the MBP, despite the massive heatsink and mains power. We've also seen that CPU cores are essentially identical from M1 to M1 Ultra - you just get more of them in the latter. Even if you could increase the frequencies, the difference would be negligible and would hardly constitute a selling point. Why go to the bother of binning the chips?
Who is going to want a 4 year old un-upgradeable computer in an incredibly large and expensive case?
Imagine being a value conscious user in 2027 and you can get a used M2 Ultra Studio for ~$1500 or buy from some incredibly hopeful dude trying to sell his M2 Ultra Mac Pro for $5000.
The Mac Pro has always been an incredibly hard computer to sell used, and the 8,1 is going to be nearly impossible.
This is the unfortunate truth.
Apple SoC are not designed to do well at higher clock speeds.
Due to our half a century with Intel/AMD we expect laptop chips to be slower than desktop chips.
It does not make sense to us that laptops would be equally as fast a desktop if they have the same Apple SoC.
Given the design priorities of Apple going forward I think it was a good call for them logistics-wise to keep Mac Pro "M2 Ultra"-only.
Granted, it's a pain not being able to increase RAM or storage. If Apple's prices for either were reasonable, you could simply over-specify for your needs at purchase and be done with it. But when you're being gouged, the option to upgrade later would be very welcome. Though Apple was obviously happy to close that loop-hole.
Selling a computer is a hassle, since it needs to be wiped, advertised etc. The other issue is that the range has some pretty big steps. Say you bought a Studio Max, then e.g. got into Unreal Engine and decided you needed more GPU. Where do you go from there? The only option would be to step up to an Ultra at twice the cost. And if that wasn't sufficient (it's probably roughly equivalent to an RTX4070 for 3D, minus the RT), you'd be done with the Mac platform and would need to move to a PC.
That argument could be made on any and all used goods.
People buying into a Mac Pro are largely expected to make money on it. So when charging for goods and services using a Mac Pro as the tool then price it in.
Yes, depreciation is a thing that varies from product to product that you can take into account when making a business decision.
I believe that the 8,1 will be incredibly hard to sell and will depreciate heavily, even compared to previous Mac Pros which have generally maintained value via novelty or utility as a "classic" Mac.
Also I've always had an issue with the "I'm making money with my computer, therefore I can set my money on fire because I'm a professional businessperson" line of reasoning. A professional businessperson will likely factor in total cost of ownership, in which case depreciation and the ability to sell your asset at the end of life will be fairly important variables to account for.
Yes, it seems a bit bizarre. I guess if the projects you're working on have huge budgets, and you're spending tens of thousands on cameras / lenses etc., then maybe spending a bit more on a workstation is unremarkable in the grand scheme of things. It starts to add up though if you need a bunch of them. You also have to question how essential it is that you (and any employees) use macOS, if you're spending all day in Maya, Nuke, Premiere etc. At the end of the day, it's a tool, and money saved on tools is money in your pocket.
Wondering the same thing. The MacPro had been updated anywhere from 420 to 2182 days with an average of 938. The Studio has now been updated at 454 days, let's call it a year. Maybe we are on a new schedule that updates for both happen regularly, slightly over a yearly schedule, when the next M version is released, M3, M4, ... ?
Also, are the clients that do business with you expect you to have the latest versions of the software all the time? Meaning do they specify that you need to use version X or have deliverables of the latest Y?
If that is the case then price appropriately.
But if say you're a wedding photographer/videographer with couples who haven't a clue on the difference between dSLR vs MILC or 1080p vs 4K vs 8K or SDR vs HDR then why keep upgrading the hardware or software at a 4-6 year clip?
This reminds me of a printing press that still uses PPC Macs and Quark. Their clients have unchanging requirements so why keep upgrading when the current system fits?
Back in 2015 Canon came out with the cheapest way to get 50MP RAW images. They did this as there was a client demand for 50MP RAW images that photographers need to fulfill. Prior to Canon the only way to achieve this was through MF cameras that cost 5 digits.
2013-2023 Mac Pro suffered longer than 1.3 year refresh due to tepid demand relative to even the 2021 iMac 24" M1.
If it sold as well as it was from 2006-2013 it would have had annual refresh like other Macs.
2023 Mac Pro being limited to only two M2 Ultra chip SKUs is another indicator of how in-demand PCIe slots are among Mac users.
The $1k price hike from 2019 to 2023 model is another indicator of how important the Mac Studio is to the pro desktop space.
It is likely when a refreshed Mac Studio comes out a Mac Pro will do so to. Supply chain-wise it is cheaper that way.
Sure, so $400+the cost of an eGPU chassis to have the extra box on your desk, avoid buying a mac pro, and save… what? maybe $300 all told?
If mobility is impt and assuming super low speeds are fine for you you’re better off just using a plain jbod/raid usb chassis, a thunderbolt chassis and raid card is absolute ridiculous overkill. You’re talking about hardware that in an 8x PCIe slot unbound by TB you could hit more than 10x the speed you say you need.
Your argument seems suspiciously like “I don’t need a mac pro so no one does”
Last edited:
Certainly feasible and could be true - but so far there is zero evidence that there are any extra chiplets in the M2 Ultra. "Twice the i/o bandwidth of the M1 Ultra" would be an obvious bragging point. I also think that if the M2 Ultra in the Mac Pro were "different" to the one in the Mac Studio (or even if the extra bit was disabled in the studio) Apple would have taken the opportunity to call it the "Ultra+" or something to get rid of the "Its just a Studio in a big box" image.
What we know is that the M1 Ultra SoC had at least 6 and probably 8 TB4 ports, each driven internally by 4 PCIe lanes, which physically consist of 4 serial lanes and so, theoretically, could be configured as PCIe pass-through with no visible extra hardware or used to drive TB to PCIe adapters. That way, Apple wouldn't need to make any custom silicon (or add redundant features to the Studio) for the MP. We'll see.
...but (if you just mean UHD at 24-30fps) so would 6 cards each with its own 40Gbps thunderbolt connection. Unless you mean 24 streams each?
The point is, without more information we can only speculate what the bandwidth will be and how much advantage this will have over TB4.
If you are trying to propose the "hackery TB " source for PCI-e slot backhaul that is already false. The PCI-e v3 lanes of the TB controllers can't do PCI-e v4. So can't hand wave at the TB controllers and say they are backstopping the slots. Furthermore, Apple hooks up ALL of the TB controllers to the eight ports on the box. Those eight TB controllers are 'gone' just on port provisioning.
The PCI-e controller is pragmatically logically inside the TB controller.
This is intel's basic logical overview of their implementation
I covered here why Apple's isn't going to be too much different.
By briefly, the PHYS path out for those PCI-e controllers is only through the TB path/socket. That delivers way lower power overhead, thereby deliverying much better Perf/Watt. Apple is all about preaching sermons on benefits of better Pref/Watt. There is not other path out but through TB. And to pass TBv4 standards you can't 'steal' the PCI-e aspect away .
Apple could have thrown out some TB controllers from the die and put bigger PCI-e v4 ones in, but they'd have less than 8 TB ports.
two x16 PCI-e v4 controllers is going to be more die area. Not the area they got now for TB subsystems.
There are no 'spare' Thunderbolt connections!
You are trying to throw away information you already have. TB controllers are consumed, so they are not it.
Until M1 came along, each pair of TB3 ports on an Intel Mac shared a single controller... PCIe-4 TB5 isn't out yet, but M2 could be half-way to implementing it. Intel's TB3 implementation may not have PCIe passthrough - that doesn't make it impossible that Apple does.
I'm not saying "I'm right, you're wrong" here - you're describing a "high end" implementation that would add excellent internal PCIe4 without compromising the existing TB4 capabilities - and you might be right, but would require in major visible changes to the SoC that would be noteworthy enough, so its surprising that Apple didn't highlight them. I'm spitballing possible kludges that would save Apple a dime. We'll find out when someone tears one of those down.
It will also be interesting to see if the ASi Mac Pro is using a higher clocked M2 Ultra.. vs. the Studio.. or if they are totally identical machines with and without PCIe.
Doubtful. On most of Apple's Pref/Watt sermons Apple posts a graph where the Super overclockable, single thread drag racing tuned , Intel CPU has a very long tail of diminishing returns. More power buys less and less along the tail. For example,
Introducing M1 Pro and M1 Max: the most powerful chips Apple has ever built
Apple today announced M1 Pro and M1 Max, the next breakthrough chips for the Mac.
www.apple.com
Whereas Apple's system has a really short 'tail'. The ramp up to where they top out and the curve starts to bend over and Apple 'quits'.
Petty likely that is because the design is highly tuned to not to enable overclocking into the diminishing returns zone. There are several 'wins' they can get with that.
AMD hasn't spelled out all the details on Zen 4c but if this breakdown is mostly on track :
"... However, the truly stunning thing here is the die size. 16 Zen 4c cores are barely larger than 8 Zen 4 cores. At ISSCC 2023, AMD disclosed Zen 4’s CCD to be 66.3mm². This is the design area without die seal and scribe lines at the edges. Zen 4c’s CCD design area is just 72.7mm², not even 10% bigger! Keep in mind that there are double the cores, double the L2 cache, and the same amount of L3 cache on each die. The cores must have shrunk greatly to fit even more cache per die with only a small area increase. ..."
Zen 4c: AMD’s Response to Hyperscale ARM & Intel Atom
Bergamo Volumes, ASP, Performance, Hyperscale Order Shift, Die Shot, Floorplan, Physical Design, and Future Use of Dense Core Variants
www.semianalysis.com
AMD got a dramatic reduction in 'core' size by throwing high overclocking out the window (and 3D stacked cache). So they get a die that has more CPU cores and behaves more consistency , but can't win the 'top fuel' , single threaded drag racing crown.
Apple has to have a CPU core that can fit inside of phones and tables and small thin laptops. Throwing away die space to consume substantially more battery energy even faster really isn't the primary objective. New fab processes and design updates lets them incrementally expand the envelope, but they are interested in expanding faster than what the fab processes 'sweet spot' is really provisioning.
Apple isn't trying to go slowest, but also not going to throw Perf/Watt under the bus to get there either. With Apples design objectives, the CPU cores have to share die space buget with the GPU cores ( and NPU cores and ... ) . And the vast bulks of the die area budget goes to the GPU cores ( as die area increases so at the "max class" die stage basically have CPU cores bolted to a largish GPU complex. ). The P-cores have to be good die space neighbors and not suck up as much die area as possible.
That 'top fuel dragster' design trade offs aren't entirely all that good for Intel Xeon SP server line up either. There is significant space being thrown at making just one lonely single thread go faster. And one of the foundational reason why got E-cores on the mainstream line up ( coupled to being stuck on Intel 7 far longer than they wanted to be. But basically still a 'die-area' cap issue. ).
Apple's 'non-monomanically focused on top fuel dragster' status , P cores means that when have a SoC with 10 , 20 ,30 ,etc cores the single threaded Turbo speed doesn't go down without binning down to some crazy 'princess and the pea' die that just 'happens to still work' in mixed mode. Apple can ship MILLIONs of their dies with no big problems ( because not binning down to rare dies. If an Apple silicon die can't ship millions then it is probably in deep trouble.
Mac Pro is likely going to sag less under extremely heavy , sustained load over a wider variety of workloads.
No. I'm not decribing a 'high end' implementation.
Intel shipped this before Apple did with Gen11 . ALL the mobile Gen 11 units have this all the way down to the lowly i3.
The 11 Gen i3 mobile line up.
11th Generation Intel® Core™ i3 Processors Product Specifications
11th Generation Intel® Core™ i3 Processors product listing with links to detailed product features and specifications.
The lowest end i3 on the list :
i3 1110g4
" ... Launch Date Q3'20 ...
... Intel® Thunderbolt™ 4 Yes "
Intel® Core™ i3-1110G4 Processor (6M Cache, up to 3.90 GHz, with IPU) Product Specifications
Intel® Core™ i3-1110G4 Processor (6M Cache, up to 3.90 GHz, with IPU) quick reference guide including specifications, features, pricing, compatibility, design documentation, ordering codes, spec codes and more.
Desktop can't find it at all ( no TB 3 or USB either) , but the iGPU is different there also.
It is a $309 chip. But so is the Gen 10
Intel® Core™ i3-10110U Processor (4M Cache, up to 4.10 GHz) Product Specifications
Intel® Core™ i3-10110U Processor (4M Cache, up to 4.10 GHz) quick reference guide including specifications, features, pricing, compatibility, design documentation, ordering codes, spec codes and more.
Which means Intel added this in for no increase in costs. So that that 'high end' at all. ( a lot of baseline costs is the bigger iGPU allocation and being on heavy multipatterning 10nm family that was newer at that point. )
[ if go back to Gen 8 era mobile i3's were cheaper.
Intel® Core™ i3-8100B Processor (4M Cache, 3.60 GHz) Product Specifications
Intel® Core™ i3-8100B Processor (4M Cache, 3.60 GHz) quick reference guide including specifications, features, pricing, compatibility, design documentation, ordering codes, spec codes and more.
But also lost big when it came to iGPUs also. Apple didn't touch them with 10 foot poles. ]
The majority of the SoC that Intel has shipped over last 3 years are Thunderbolt 4 capable. It doesn't look that way for many low end Windows PC laptops because the system vendors have gimped on the PHYS chips needed to correctly pass TBv4 standards testing. Heck, on the desktop side at the low end retail 'box with slots" can't even find even a nominal Type-C USB 3.1 port hardly anywhere due to the penny pinching on PHYS parts... doesn't mean the PCH controllers don't support it.
If you want to tag this as a 'high end' feature we've got different definitions of 'high end'. The Core i3 ain't it for me.
Last edited:
https://www.macrumors.com/2023/06/11/apple-exec-discusses-mac-pro-lack-of-egpu-support/
I think the answers to our questions are on this interview. And the future is not so promising and bright for us who have relied on the Apple solutions for professional needs. Actually coming events cast their shadows before. I think we are quite blind to see what's going on at the Apple side.
I think the answers to our questions are on this interview. And the future is not so promising and bright for us who have relied on the Apple solutions for professional needs. Actually coming events cast their shadows before. I think we are quite blind to see what's going on at the Apple side.