Surely the wheels are the joke, and Apple still make them.
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.
Apple Silicon Mac Pro Configurations: Everything We Know
- Thread starter MacRumors
- Start date
- Sort by reaction score
I actually own a "512K Mac". I was something like a 10-inch monochrome screen and no hard drive, just a single floppy drive. It still works, but the machine is useless except as an antique.Every single system I have disagrees with this so much.
This si what they mean by Apple stuff lating only a few years, it will still work, but you don't care because they have dropped software suport for it.
I (and others) have posted about this before, but they really will need to offer at least one case that has as many PCIe slots as the 2019 MP, so that users can do this:
Instead of this:
Source:
I imagine there are a lot who aren't in that category, and for whom the the processing power/memory of the Studio Ultra would be adequate, but decided to stick with their Intel Mac Pro's for now because they're still waiting for all their software to fully transition to Apple Silicon (including getting most of the kinks worked out)—and/or they wanted to keep those PCIe slots.
Seems that a good item that is missing is user installed storage and dedicated cards (audio / video, etc.).
Perhaps creating a new method of high speed paths would be in order and simply insert the SSD or cards in.
This is not unlike traditional other than much much faster and far more integrated.
Perhaps creating a new method of high speed paths would be in order and simply insert the SSD or cards in.
This is not unlike traditional other than much much faster and far more integrated.
Music production hardware like MPCs, EMU rack samplers, etc tend to be old computers and use these old computer parts.
And they have only become more expensive and popular unlike ancient general computers cause their workflow and sound will never really go obsolete.
Since I used to assemble PCs for fun back in the 90s (after picking up hardware at the computer show), it’s actually really fun to dig into these old systems and refurbish them. Actually can be profitable too.
That is probably not true. The number of folks who have > 1TB + quad GPU configurations is highly likely too small for Apple to build a product just for them.
The Mac Pro starting off with a revised Ultra class SoC and going up to an "Quad Extreme" likely won't cover the > 1TB zone. At best probably looking at revisions to at least get perhaps x32-40 PCI-e v4 lanes , but the memory is likely limited to LPDDR5 density improvements.
The MP 2019 was pretty likely built so that the folks who needed > 1TB RAM and quad GPUs could buy one in the 2021- early 2023 time frame and then squat on it for 3-5 years. Apple really won't be trying extremely hard to sell them anything new in the 2023-24 time frame, because knows they very likely are not moving ( in the system cost recovery phase).
For folks who have 200-400GB RAM workloads footprints, a MPX Vega II or Vega Duo or W6800 GPU , 16-28 cores the "extreme" will likely work for a heft fraction of them. Those folks will greatly outnumber the > 1TB + QuadGPU folks. Apple will make a profit and Mac Pro will iterate once or twice further before folks riding on extremely high MP 2019 sunk coasts start looking around.
If there is a 6 month lag between "sneak public preview" and actually taking order then that is a roadmap. It is probably a better roadmap than Intel has been dribbling out from 2018-2020. Roadmap , snoadmap ... what you are actually going to do is what really matters.
The easiest way is that they aren't built-for-laptops Max chips.
Something that is Max sized ( or less if using TSMC N3 ) that is structured a bit different.
What Apple needs is a "Max for laptops" that doesn't have any connector.
A "max for desktops" that probably has two (if not three) connectors that is bigger than than the laptop version in similar fashion to how the Pro is largely just a "chopped down" version of the Max. If structure it so that what you remove from the design doesn't greatly perturb the floor plan of the subsystems, you can get a very high amount of design reuse ( and an increment change to the die mask for making the different chips).
The other issue is that UltraFusion having 10,000 connection pads seems like a bit of overkill for just two dies. Apple is going "wider and slower" to control the power drain os shipping 1TB/s of data from die to die but even 5,000 pads would be alot. ( for perspective the next gen Epyc Zen 4 and Xeon SP gen 4 have packages with 5000-6000 pins/pads. 10K is double that. )
Apple sandbagged folks on the intial pictures of the Max die where they photoshopped out the connector. Apple could be slightly sandbagging on the UltraFusion connector also. There probably isn't connections to three dies there but 1.5 or 1.25 would not be a stretch.
If put a slightly different interposer in there could possibly "point" in more than just direction without loosing much die-to-die bandwidth in the pair. So something that points off to another die with 'chiplet' implementation of a PCI-e controller or some other I/O.
Unless Apple massively shrinks the Max class die , they can't use the same 2.5D/3D packaging technology on the quad SoC solution as they did on the "Ultra Duo". Seems problematical starting with a laptop optimized die if going to build a mega package. Apple really needs something that is mostly the same (to save costs) but has at least a few desktop , "mainly or solely used in chiplet/tile SoC" properties to it. The Studio + Mac Pro + (maybe iMac Pro ) could aggregate into some decent numbers without relying on MPB 16" sales to solely pay for this design.
I suspect some folks are focused on a path where this Quad/Extreme SoC isn't very expensive. If the MBP 14/16" have essentially paid for the Max chip design then simply 'gluing' four paid for chips will be cheap. Pretty good chance that "glue together" process is going to turn out to be far more expensive than they hope.
The primary reason the Ultra consists of two exact twin dies is to save money. There is wasted silicon in doing that. ( no need for two secure enclaves. two SSD controllers , 8 TB controllers, etc. ) . It is cheaper because use the MBP 14"/16" unit volume to drive down economies of scale. Scaling past two is pretty likely to generate lots of 'wasted' die space of internal function units never use.
The Mac Pro 2019 > 20 CPU and ProVega/W6800 options are expensive. The quad package perhaps won't be the combination cost of those to put together but likely larger than either one of those BTO options.
I can't wait to see what Apple Silicon can do for the Mac Pro.
Are you serious?
You think the M Ultra has the best GPU in the world?
And what about ram? The limit is quite low, so far.
They should reach 1TB of ram to make it a beast.
An Ultra/Quad M2 SoC likely take care of the need for two MP 2013 in the picture above. If there are 32 P cores that will swap anything those Xeon E5 v2 cores were doing. Same way the MP 2019 made those two systems disappear as a CPU resource constraint.
Those are all single width slots with relatively low bandwidth requirements ( some old audio cards are really on PCI (before PCI-e ) foundations. And more than a few solely PCI-e bus powered ( < 75W ) also. That all could be done with a rackmount case that had an internal slot extender. For example could have 4 slots in a shorter tower. If turn that board horizontal there is extra space in rack chassis. Use the same 'mainboard" and 'snap on' a slot extender onto the port close to edge and run that through PCI-e switch.
I don't think that is going to be make the "I want a 3x width GPU firebreathing monster card" folks happy.
There are a number of sub groups that may or may not leave behind.
There are likely a lot of folks for which if you gave them just 2-4 of those cards it would work out also. (even more so if Apple provisioned 1-2 M.2 SSD slots so not soaking up a slot with storage. )
The Mac Pro 2008-2012 model had just four slots and somehow the world didn't implodes. That 6-8 is some kind of "minimal' slot number is a pretty big stretch. Even over in the Windows workstation world systems like the HP Z4 and the Dell mid range sell lots of units and don't have 6-8 slots.
An "Utlra class" Mac Pro with 2-4 slots would have substantive market segmentation over an "Ultra class" Mac Studio that had zero slots. Especially if the Mac Pro starts $1-2K higher price point.
P.S. a substantive issue though for some of these much older design PCI-e/PCI cards is whether their drivers have kept up with macOS on M-series and with the move away from kernel extensions ( it is deprecated).
Some folks have a large collection of PCI cards filling up slots because they never toss/retire any card. As the driver support drops for those cards they won't fill long term future M-series powered Macs as the Intel support dries up and goes away.
The MP 2019 being the end of the line on x86-64 , old school drivers is more likely going to be the "end of the road" for those sorts of legacy card collections and collectors.
There is a healthy number of card that do with on macOS on M-series ( via Thunderbolt PCI-e extension enclosures for now). But there is also a number of cards that have made zero movement in about two years. There is a decent overlap there of "age of card" with "lack of movement". Some non Audio cards that are associated with early boot environments ( boot off RAID or boot GPU screen) probably are not going to make the cut either. Pointing at large collections of those that "need" to be inserted into a M-series Mac Pro is just pointing at zombies. They are dead anyway.
There are enough of the cards that work already though that it would be more than a bozo move if Apple doesn't put slots into the next Mac Pro. It is an ecosystem that is just already there and going to be there anyway because Apple has a decent range of support for PCI-e cards already in macOS. ( not GPUs or boot or abandoned/no-budget drivers. )
Low budget? No.
Any budget where "power" is largely computational horsepower ? Yes.
quad a100, 64 CPU cores , 320GB of VRAM ( and another 512GB of system RAM).
If primarily want to play upcoming Metal 3 optimized video games as a measure of "power" then not as much.
or
T-Class
Versatile platforms for demanding 3D content creation and motion media workflows, the APEXX T-class features blazing fast AMD Ryzen™ Threadripper™ Pro processors.
www.boxx.com
I…didn’t say that. I explicitly asked if there is anything that doubles the performance.
It was a question.
The new Mac Pro is several years newer then the older models, so in monetary terms it’s not actually 3000 dollars more expensive, if you work out what 3000 dollars from the time of the older machines is worth now. Animation studios also did use the trash can Mac Pro? and I’m also not sure who considers 3000 dollars as an impulse buy. Think you’ve got that a bit wrong.
I bet the new Apple Silicon machines start from a higher price though.
Last edited:
I’d be really surprised if Apple released another Intel Mac. However I’d like it because it’d mean that my current Mac will have a longer support period.
Maybe not "just" for them - but part of the high price of the Mac Pro comes from the extreme RAM expandability and large number of full-bandwidth PCIe slots. E.g. in the 24 and 28 core versions Apple are only offering the M-suffix CPUs that cost twice as much as the regular versions (which already cost thousands), just to get support for >1TB RAM. I'm sure that they could have sold a lot more MPs if they'd made more of a mid-range Xeon tower with maybe 768GB max RAM and space for a single high-end GPU + a handful of lower-bandwidth slots...
Well, yeah, but to put it bluntly:
Intel/Windows/x86 3-year roadmap: we'll definitely maybe start using 3mm grade stone for the road surface by 2023. Plus we've extended the extended support period for cobbled donkey tracks.
Apple Mac Pro 3-year roadmap: Roads? Where we're going we don't need roads! (Whaddya mean 'yes we do'?)
I mean, I'm gonna go out on a limb and say that, apocalypses permitting, in 3 years' time you'll still be able to buy an Intel or AMD x86 tower with PCIe & DIMM slots with incrementally better processors and GPUs than are available today, whereas Apple have - since ~2011 - failed to produce two successive Mac Pro models that don't assume a major workflow re-think. Heck, here we are already speculating on whether the next Mac Pro will have PCIe and/or support discrete GPUs...
The scaling and performance of the M1 Ultra and above failed to achieve the design goals as the way cross-die data is received causes an unexpected interaction; the raw interface between them is not the limiting factor. Apple seemed to have hoped that this bottleneck (albeit more akin to an instability at higher workloads) could be ironed out in firmware but this did not come to pass. As a result the M1 Ultra is unable to flex its muscle, the maximum power draw is constrained well-below the design's TDP and, as a side-effect, the thermal load never made it to the level that required the expensive and rather exotic cooler in the Studio.
This unexpected issue is undoubtedly a major driver to the delay of the ASi MP and effectively forced the move to M2-based processors for the multi-SoC Mac Pros. As I understand it, the M2 multi-SoC systems do scale proportionally and should provide exceptional performance. Those I speak too are very happy with their work.
I know, but don't throw something out as a fact when you're obviously pulling it out of your ass.
Saying we know something when we don't is an outright lie. It's a pretty simple concept.
The Mac Pro entry price doubled overnight from $3000 to $6000 in 2019. However you compare the specs, there was no longer a headless Mac in the ~$3000 price range.
General inflation between 2013 and 2019 was about 10% and from 1980 until the last year or two deflation has been the norm for electronic goods - don't forget that the $3000 entry level trashcan you could buy in 2019 was not the same quad core model that cost $3000 at launch in 2013 but the 6 core, better GPU model that originally cost $4000.
Yes, "Animation studios" etc. did buy the trashcan and the original cheese grater - the prices went all the way up! However they were still just about accessible to mortals plus - at least when they were launched - the entry model started off more powerful than the best iMac and actually made sense on its own - whereas the $6k 2019 Mac Pro makes no sense unless you're going to fill it with upgrades.
As for $3000 impulse buys... I think you'd be surprised, and where I'm probably wrong is in saying that $6k wouldn't be an impulse buy for some people... but we're talking about rapidly diminishing numbers. Maybe "impulse buy" was the wrong term though and "on the spot purchase" would be better. It's the difference between "I got the milk and they had a Mac Pro on offer so I picked one up" and "I drove to the Apple Store to look at computers and came back with a Mac Pro". I can't imagine doing the latter with a $6k Mac Pro - apart from the fact that I don't currently need one (a) it's over my credit card limit, (b) it makes no sense stand-alone and for any hypothetical need I'd have to research and track down another $6k worth of kits and (c) have you seen the size of the crate...?
I wouldn't rule out the $6k price point if there's going to be a M2 Ultra version. Hard to justify more than that for what's going to have an only incrementally better CPU and GPU than the $4k Studio Ultra and probably less expandability than the current $6k MP.
From my understanding of the current situation is that much of the software still isn’t coded to actually be able to get data to the GPU fast enough, so the Ultra never gets to actually show its uplift.
There are workflows that do showcase phenomenal performance, but that seems to only be from vendors who have re-architected how their software works.