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.
2018 Mac Pro is the Phoenix Mac
- Thread starter cohen777
- Start date
- Sort by reaction score
I'll wait until it flies to call it a Phoenix. Nobody, probably not even anyone at Apple, knows what the mMP will be. From that article: "What’s coming in the future, we don’t know". How can you crow that it's a phoenix when it might be a dodo?
If Apple doesn't come up with a new Mac Pro that has the power and flexibility of the cMP tower - it will be dead on arrival. If it doesn't accept standard off-the-shelf PCIe graphics cards, it will be dead on arrival. (The MP6,1 D300/D500/D700 fiasco will mean that Apple will have a very hard time convincing pros to buy a system with single-source proprietary graphics.)
If it doesn't support CUDA, it will be dead on arrival - especially since OpenCL is now dead.
(Come on Apple - get rid of your nonstandard bastard EFI boot and use the UEFI standard that everyone else uses. EFI was deprecated in 2005.)
I guess I should give up on Alti-Vec at this point, as well.
people from Phoenix are Phoenicians
[doublepost=1492562956][/doublepost]
i'm pretty sure the capabilities i was hoping to get from openCL (or CUDA) are just not going to come from there.
(the capability being real-time raytracing and/or photoreal rendering of 3D models *)..
i'm now thinking these capabilities are going to be born out of VR/AR technologies.
i'm not saying GPGPU itself is dead.. (and for all i know, this will be a very important factor with VR?)..
but i'm saying they're no longer something i'm interested in as a problem solver for my particular needs/wants.. better things on the horizon i'm thinking.
---
*edit- another way of expressing this ability to a wider audience would be something like:
an animator could simply move a character on screen in real time (this is already possible).. while they are doing that, they could basically screen-record their display and that's the final product.. no more stories of "i heard disney-movie part IV took 2 years to render omg"
[doublepost=1492562956][/doublepost]
heh, as far as i'm concerned, they're both dead.. (cuda & openCL)
i'm pretty sure the capabilities i was hoping to get from openCL (or CUDA) are just not going to come from there.
(the capability being real-time raytracing and/or photoreal rendering of 3D models *)..
i'm now thinking these capabilities are going to be born out of VR/AR technologies.
i'm not saying GPGPU itself is dead.. (and for all i know, this will be a very important factor with VR?)..
but i'm saying they're no longer something i'm interested in as a problem solver for my particular needs/wants.. better things on the horizon i'm thinking.
---
*edit- another way of expressing this ability to a wider audience would be something like:
an animator could simply move a character on screen in real time (this is already possible).. while they are doing that, they could basically screen-record their display and that's the final product.. no more stories of "i heard disney-movie part IV took 2 years to render omg"
Last edited:
Do you want to meet up for a glass of wine at the GPUtech conference in early May?
There may be some VRWorks announcements that will interest you.
Last edited:
i like wine.. i like California.. i like you too..
but i don't like GPUs that much so i'm going to have to pass
Too bad - by 9 May the ATI fans will have to shift from touting the non-existent Vega against the GP100/GP102 to touting the non-existent Vega against the non-existent Volta GVxxx.i like wine.. i like California.. i like you too..
but i don't like GPUs that much so i'm going to have to pass
Unless we're surprised by Volta shipping in May.
haha. sounds fun.touting the non-existent Vega against the non-existent Volta GVxxx.
Sure hope so! If it can essentially be a high-end workstation-class PC, but with macOS, then that would be SO incredible! (And obviously, being more beautiful below my desk than a PC). Seriously, video professionals such as myself just need a tower with PCIe slots, expandable internal storage, and user-upgradeable components. It's as simple as that!
Good? It sets some expectations that really [sic don't] have a firm grounding.
1. Apple said nothing about 2018. They said not 2017, but that isn't necessarily a tag of 2018. ( Yes there would be a tons of even more pissed off folks if they slide through 2018. )
2. This isn't the first "die and reborn' experience. From 2010 to 2013 the Mac Pro went into a long comatose state. It was withdrawn from EU markets in early 2013 due to having a rather old, obsolete design constraint. I doubt Apple planned in 2010-2011 to be in that context if Mac Pro's were a major contributor to the projected Mac market.
3. With respect to modularity with an external display the current Mac mini is just as modular as the current Mac Pro. ( that display modularity was the context used in Apple's discussion). The implied premise that they will simply just bring back a relatively slightly modified version of the old enclosure is highly questionable. The Phoenix implies the return on the same bird; a regeneration. There is nothing but quicksand to support the "old case, reborn" premise in what Apple outlined in that round table.
4. "There was no way to add different GPUs or upgrade CPUs. " That isn't true. No way to trivially do a CPU upgrade with you normal trusty screwdriver? Yes. However, "no way" (i.e., impossible ) isn't true. There are/were turnkey services. There was no blockage in the design that barred Xeon E5 v3 or v4 (other than needed a new logic boards for new socket and PCH chipset). Nothing in Apple's commentary about the problems they faced said anything about that in the slightest. Nor are they particularly 100% blocked on the GPU class they started out with. They said couldn't fit a different class of GPUs, but blocked on same class ? Nothing in the commentary there either.
It is more a "form over function" driver from the user base than Apple. Standard PCI-e card form is a form ( not function) argument. It is not the function of the PCI-e data transmission connection nor the function of the GPU itself. It is picking a rigid form.
5. The article paints the Cube being withdrawn driven clearly by a sales failure. In Apple's commentary on the market for the current Mac Pro they said they didn't meet as many as they wanted to but that were some users that it worked for. The fact they didn't pulling after two years means that it was far closer to "mixed signals" than it was a very clear no by the whole targeted market. The equivalency with the "Cube" is not a tight fit at all.
Yes, there are a sizable number of people who have been stomping their feet in the "no" camp all along. The unsupported assumption is whether they are enough to be a viable market by themselves. Getting 100% of what they are asking for ( return of old case) probably isn't going to happen. There is a decent chance for a single, secondary GPU slot that can optionally hold a standard form card. But Apple selling what pragmatically would be a 'bare bones" like box where replace all of the major parts, that is unlikely.
Another evolutionarily iteration on design is far more likely than a regeneration of an old design. Technically, that isn't a Phoenix.
Last edited:
That was certainly true of the announcement last March, where the only thing you could read into "modular" was "no built-in display" and the references to "updatable" were about Apple being able to keep the design up to date. However, since then, a December press release for the iMac pro has at least added the word "upgradeable" to the mix:
In addition to the new iMac Pro, Apple is working on a completely redesigned, next-generation Mac Pro architected for pro customers who need the highest performance, high-throughput system in a modular, upgradeable design, as well as a new high-end pro display.
...which is a small concession to user upgradeability... still not exactly promising PCIe slots though.
Not really - a standard PCIe card slot vastly reduces the bar to the availability and user-upgradeability of alternative GPUs and other internal expansions - PCIe cards made for PCs may still be dependent on Mac software/firmware support (although that pretty much exists for, e.g. NVIDA GPUs) but that is a far easier prospect than having to produce physically different hardware to fit a nonstandard, possibly proprietary slot.
I seem to recall that there was also the small problem of the case cracking - pretty fatal to a product that demands a premium mainly on the grounds of cosmetics. That's a rather different sort of "design failure" to misreading the market.
The heat produced by G4 450-500Mhz was terrible and to make the matters worse it meant Cube and Power Mac was stuck with this CPU for more than one year until IBM/Motorola could move to a new process.
In that same time frame Intel and AMD were in a megahertz war and their CPUs were become faster, cooler and more power efficient every month. Their progress rate was crazy so Apple had to launch dual processor models to keep up.
Even then Apple was being crafty at WWDC. When they did live demonstrations of the Power Mac next to a PC they would disable to cache on the Intel CPU to slow it down massively. It created a lot of arguments online because people would replicate the same benchmarks themselves with a very different result. Today all the stories are known about the onstage demos always involving magic tricks.
Now Vulkan support's macOS Metal2, so dont care on CUDA (AMD also promise CUDA support in coming months)...
2018 Macs have the opportunity of using the Titan Ridge Thunderbolt 3 controller and Intel are working hard to ensure TB3 spreads further by making it royalty free for motherboard manufacturers.
The Titan Ridge controller will allow Displayport 1.4 signals and up to 8k displays to be connected with a single TB3 cable assuming a discrete GPU. I believe a single cable is exactly what Apple have been looking for to reintroduce their own retina Cinema Display range even though 5k at 60Hz would take up a large proportion of the bandwidth of a TB3 connection.
This is not to say that a modular Mac Pro couldn't come with an internal PCIe 3.0 x16 slot for the inclusion of a graphics card - external GPU cases are very pricey and would make graphics cards very pricey indeed - but if they intend to ship a version without the ability to add a PCIe card it has to start cheap - perhaps with an SKU that takes in high end Mac Mini SKUs.
The Titan Ridge controller will allow Displayport 1.4 signals and up to 8k displays to be connected with a single TB3 cable assuming a discrete GPU. I believe a single cable is exactly what Apple have been looking for to reintroduce their own retina Cinema Display range even though 5k at 60Hz would take up a large proportion of the bandwidth of a TB3 connection.
This is not to say that a modular Mac Pro couldn't come with an internal PCIe 3.0 x16 slot for the inclusion of a graphics card - external GPU cases are very pricey and would make graphics cards very pricey indeed - but if they intend to ship a version without the ability to add a PCIe card it has to start cheap - perhaps with an SKU that takes in high end Mac Mini SKUs.
All those considering the mMP to include bare PCIe slots are naive as 3rd grader Apple will never again allow DIY upgades as long they can, and high ticket components as the GPUs are perfect target for propietary tricks.
1st technically TB3 needs DP1.4 signal back line, it dont comes with PCIe3, it will require either a dirty solution as a an DP->MB feedback cable (As on some Gygabyte MB with TB), or an new GPU slot including both PCIe (data) and DP1.4(video) signals.
the trash can, included that PCIe+DP1.2 interface, just the GPU where doomed by the Thermalcore flexibility (or by Apple need to delay tcMP updates).
What is reasonable to expect:
Either Apple opens and updates the tcMP GPU interface (PCIe3+DP) so the Industry can adopt it even for non-apple products or Apple follows they historic proceedings and keep the GPU as closed as possible, but an new design that allow they to commission own or licensed GPU upgrades that includes Apple's tax and technical blessing you may install by means authorized partners (as the iMac Pro ram upgrades) if dont want to void the warranty.
Would you really think professionals will allow themselves to be locked into Apple GPU card options again though?
Ironically, at the moment it might be that Apple's volume deals with AMD will allow them to sell the MMP with a GPU (probably built in) at very reasonable prices compared to what folks with are having to contend with at the moment. For example, a Vega 56 or 64 built into some SKUs of the MMP could make the first generation very reasonably priced compared with the folks who would be looking to add their own GPU afterwards to an SKU possibly without a GPU.
I don't imagine for a moment that Apple would release a big box like the classic Mac Pro again but they must have had discussions with professionals they feel would be the target for the modular Mac Pro and have settled on eGPU as the primary upgrade option for users due to the way that macOS development seems to be going. Any new Mac Mini, for example, could be a variant of MMP that comes with integrated graphics only (or at least just a low power dGPU to satisfy people who need DP1.4 for a 5k monitor).
It's probably too soon for Thunderbolt 4 to be a factor in the new modular Mac Pro - I believe the bandwidth possible there could be a much better solution for eGPU, but don't forget the Titan Ridge TB3 controller I mentioned in my earlier post (follow the link) which I believe solves the issues you raised.
You don't care if your CAD/CAM/AI/ML/rendering software supports Apple's proprietary API of the month?
Apple's proprietary API of the month?
I'm pretty sure the last proprietary API that Apple had for graphics was in 1998.
Do you mean Metal 2? That's just branding. It's just Metal 1 + more features. It's not a new API.
Yes, there's a term for that: Steve Jobs famous Reality Distortion Field.
Which Windows, Android and Linux systems support Metal?
Is the Metal 2 API identical to the Metal 1 API? If it's not binary compatible forwards and backwards - it's a new API.
[doublepost=1519695240][/doublepost]
Yes, there's a term for that: Steve Jobs famous Reality Distortion Field.
Outright cheating, however, is a Trumpian Fact Distortion Field.