MP 7,1 Waiting for Mac Pro 7,1 A1991 (no more)

[goMac]

In 2007, I started and ran my LLC with a MacBook. With the power of the new 16" MacBook Pro - I suspect that's an easy thing to do in 2019 as well.

Initial reviews are indicating the throttling issues are much improved. If the Macbook Pro does have much less throttling, that would definitely be a substitute for a lot of users.

It seems like Apple is making a pretty conscious choice to move the Macbook Pro back out of trendy laptop territory into pro territory. They even *gasp* made the new Macbook Pro a little thicker.

 

[deconstruct60]

The 9600 crowd has always wanted a massive slot successor, and they finally got one. It makes a lot of sense with high end GPUs getting larger and more power intensive....

the 9600 was a 1996-1997 product. The crowd waiting around since the last century makes a lot of sense. ...... errrrr No. It really doesn't. The processor in the 9600 was fabbed at 350nm. We are at about 14nm now even on suck in the mud Intel process here. The world changed in the last 20 years.
The level of integration now is about two orders of magnitude better. Limited by the same design constraints is something that is contrived; makes a lot of sense.

 

[deconstruct60]

Apple does appear to be tailoring their Pro-labled products to the feedback they are getting from the Pro Workflow Teams and those folks are probably closer to the upper-end of the user spectrum than the lower.

The Pro Workflow group only does small scale (highly likely statistically insignificant ) sampling. The feedback comes from Apple hiring folks to do on site projects. That is only going to be a relatively low number of contractors. And probably zero controls for demographics sampled ( it is also likely folks who Apple filtered out to be highly Mac friendly and trustworthy enough not to talk too much).

Apple stopped reporting units sold and started focusing reporting on margins. That's highly more likely one of the drivers here. More "fat, juicy" products over a maturing user base. Price sensitive customers dropped. Chase customers with deeper pockets and/or less price sensitivity.

So by default, that should mean that their needs represent a majority of general user's needs so what Apple makes "for them" should also work "for the rest of us".

It isn't going to pragmatically work if it is out of budget. Having an 1.4KW power supply when only really need a 700-800 W isn't going to work "well" for someone throwing extra money at that extra 700W for no material impact.

That being said, as deconstruct60 has noted, that means the final product will be more expensive and that price may very well put it out of the league of many of "the rest of us". The Mac Pro appears to be the extreme case for this where the base model price and configuration is of effectively no value,

The base has effective value for some. It just isn't aimed at "average' folks. Wrapped in a rack enclosure and allocated to running a hypervisor with 3-4 VMs is probably a good value match. Don't need a large boot hypervisor drive. Can add either higher end networking or storage for the VMs (that can be provisioned by a couple of slots ) and if mainly headless grunt work for the VMs then GPU makes no material difference.

Is that configuration going to help John Doe with a MP 2010 with 4 HDDs and 2.5" SSD boot who just wants a tinker box and grab the latest GPU off the shelf from the Fry's discount bin? Nope.

though the upper-end is as close to "no limits" as a Mac has ever been.

If the Pro workflow group has done anything, I suspect it is more likely is skewing the Mac Pro farther into being a LogicX and/or FCPX machine. The projects that Apple hires folks do are extremely skewed to those to apps or at least the solution space those apps play in. That is acutally not going to be "no lmits" but it is a limited solution space being explored.

Closer to what they have is chasing a customer base that has has "no limits" on what they spend ( or pragmatically are largely spending other people's money). This Mac Pro has limits. There are a bit higher than what they offered in 2010 but not really a "biggest of big" . If try to slot up several x16 cards and use them concurrently for high bandwidth will also probably see some limits on the other slots/ports.

 

[goMac]

the 9600 was a 1996-1997 product. The crowd waiting around since the last century makes a lot of sense. ...... errrrr No. It really doesn't. The processor in the 9600 was fabbed at 350nm. We are at about 14nm now even on suck in the mud Intel process here. The world changed in the last 20 years.
The level of integration now is about two orders of magnitude better. Limited by the same design constraints is something that is contrived; makes a lot of sense.

Errrrr that wasn't really what I meant. The 9600 wasn't it's own market that just disappeared when Apple stepped out. PC vendors kept shipping high end, high slot systems. This is just Apple's first step back into that market since they left.

I don't think there are many 9600 users lurking around, but there are a lot of people who have been wishing for Apple to jump back into that high end workstation market. The cMP, G5 and G4 were all fine, but they weren't really that sort of workstation.

I think the only place those sort of users would fault the Mac Pro is no dual socket option.

 

[deconstruct60]

The counterpoint to Apple abandoning the upper-middle end is that they've pushed further into the high end, as some customers were demanding.
.....

Four high end GPUs running at full link speed is impossible on a mid end tower, and was not possible on the classic Mac Pro.

In Apple's "dog ate my homework" meeting in April 2017 they stated that more customers were looking for one bigger GPU not two ( as in MP 2013 ) and that some stuff didn't scale out to two GPUs. Here this is about not just two but four. The Mac Pro is sweeping out outliers more so than the core of the market if going past two up to four. Metal with the new augments makes it look like four not one uber GPU.

Unless there is a huge connector hidden by Apple's and AMD's pictures there are four full links to each other GPU, but not really to the CPU and main RAM. Probably not a limiting issue in most cases where the data is loaded up and then crunched on for a longer amount of time than it took to load. but the Duo card's GPUs are sharing the same x16 PCI-e v3 host link. That is bit scaled back from what they could do individually in another system.

The Infinity Fabric 'back channel' though is full speed. But like the MP 2013 "bet the farm" on OpenCL move this is to a new software that Apple's is just rolling out. FCPX is on board. they probably got Blackmagic DaVinci on board but if the API doesn't get picked up by most apps then they could run into a similar stumbling block.

This also falls into the "not possible on classic Mac Pro" largely because never did do GPU interlink fabric support before. That isn't really a hardware thing. That is a software thing. ( which probably does add to the MP 2019 cost since pragmatically not any other Mac are going to leverage this new API edition for at least a while. And Apple may keep it locked on only on the Mac Pro. )

 

[goMac]

In Apple's "dog ate my homework" meeting in April 2017 they stated that more customers were looking for one bigger GPU not two ( as in MP 2013 ) and that some stuff didn't scale out to two GPUs. Here this is about not just two but four. The Mac Pro is sweeping out outliers more so than the core of the market if going past two up to four. Metal with the new augments makes it look like four not one uber GPU.

I'm aware of where Metal is with multi GPU. But as edge case as it sounds... We have four GPU systems and the Mac Pro was never an option until now. I don't think anyone is looking at replacing anything right now, but the Mac Pro is now an option in the future.

I don't think we're alone and we don't even have that big of a budget.

This also falls into the "not possible on classic Mac Pro" largely because never did do GPU interlink fabric support before. That isn't really a hardware thing. That is a software thing. ( which probably does add to the MP 2019 cost since pragmatically not any other Mac are going to leverage this new API edition for at least a while. And Apple may keep it locked on only on the Mac Pro. )

Infinity Fabric isn't that simple. It's not just upgraded Crossfire. Crossfire was never really suitable for compute for a lot of reasons.

Yes, Apple could have added Crossfire support to their GPUs. But it wouldn't have really enabled the same use cases the way Infinity Fabric does.

Edit: Should mention that Apple has had API to link memory of two GPUs for a long while. But as far as I know, even on a 2013, GPU synchronization was done over the PCIe bus, and not through a Crossfire link.

 

[mattspace]

iMac Pro.

As if it needs to be said, and yet it seems to have to be said, over and over.

Fixed display and no user-upgradable graphics (thunderbolt shortbus is not a substitute) = not an equivalent to a Mac Pro, no matter what Apple Marcom would like the world to believe.

Hey, everyone remember when the argument was that user-upgradable GPUs were going away, that the 2013 was the future, and we just had to get on with things and accept the inevitability that the everything in the entire industry was "going that way". Haha, fun times.

 

[teagls]

Yes, Apple could have added Crossfire support to their GPUs. But it wouldn't have really enabled the same use cases the way Infinity Fabric does.

Edit: Should mention that Apple has had API to link memory of two GPUs for a long while. But as far as I know, even on a 2013, GPU synchronization was done over the PCIe bus, and not through a Crossfire link.

I used to develop software for the 2013 MP and if I remember correctly, if you booted into Windows you could use Crossfire, but on MacOS you couldn't. Apple's GPU sharing implementation was also done at the software level. Making it painfully slow and basically impractical for real-time rendering.

 

[tevion5]

Yeah, the whole observation that the base 7,1 will have roughly the same computing horsepower as a next gen console went right over everyone's head.

Isn't that very speculative? We're talking about comparing a yet unreleased and un-benchmarked Mac Pro with an unreleased and un-benchmark PS5 or so. Just because something has the same number of cores doesn't mean it's equal in computing power...

 

[Mago]

The ncgmMP market is nicher than ever:

1. Video studios
2. 3-D animation

While it used to include:

1. Programmers
2. Researchers
3. Audio producers

Programmers tied to Apple likely will hold their iMac iMac Pro, while audio producers should be happy with higher end Mac mini or iMac, while researchers migrated to windows/Linux.
Machine learning (and all related HPC applications) should have been another market but without CUDA it's not possible to attract any serious researcher, even those loving macOS will work from an iMac on a local or remote Linux based cluster (what I do).

BTW Apple scheduled another NY friends event on December 2, likely not to introduce new Apple watch bands

 

[tobiastimpe]

But what about the people that are music producers, video editors and iOS developers at the same time? Or just people that need multiple monitors for music production? An iMac Pro with the current design and an LG 5K display just look awful side-by-side.

 

[Stephen.R]

The ncgmMP market is nicher than ever:

1. Video studios
2. 3-D animation

I'm neither of those. My work is a combination of writing software (libraries, dev tooling, ops/infra setup tools, etc, and some web application dev) and managing infra using that software.

As things stand I intend to get one (or a follow up model if it's out) once I move (I don't intend to stay in this country for another 7 years and I don't want to try to ship a ~20kg Mac Pro internationally).

Theoretically an iMac Pro could work for me (more CPU cores + more memory is what I need) but it's got a high (by my needs) base GPU, and I prefer using two ~mid-twenty inch displays compared to one huge one. It'd also be nice to have more choice for very fast storage. TB3 is great, but it'd be nice to just be able to stick a M2 or U2 PCIe card into a slot and get a heap of crazy fast space without dealing with another wall adapter, another case, etc.

None of it is earth shattering stuff - I get by on a Mini with 64GB now, and maybe by the time I move, they'll have ~10-12 core/128GB Mini's, and that'll suit what I need it for. But for now, it can show its limits at times.

 

[deconstruct60]

The ncgmMP market is nicher than ever:

1. Video studios
2. 3-D animation

Apple putting substantive efforts into creating a rack mounted version means that folks in the cloud services ( e.g. macOS VM hosting ) are a significant market also.

While it used to include:

1. Programmers
2. Researchers
3. Audio producers

"used to include" is an overgeneralization. There are still going to be representatives from those groups buying a MP 2019. Not the same percentage numbers because of two factors. One, there probably were not equally represented across the full price range in the old Mac Pro systems spaces. Any demographic that was more couched in the first 40% percentile of the old price range is likely not buying.

Programmers tied to Apple likely will hold their iMac iMac Pro,

There are gobs more programmers on MBP than either of those two. That is one reason there was a Holy War over the physical escape key on the MBP 15" that Apple reversed course on.

Most programmers being on "big box with slots" is a 2005-2009 era notion that hasn't been true in creeping up in a decade.

while audio producers should be happy with higher end Mac mini or iMac,

The ones that are wedding to using PCI-e add in card augments for work, probably aren't. Folks don't "have to" get a Mac Pro though to be serious. Don't really have to get a Mini or iMac either for a reason number of tracks ( MBP 16" ) if interfaces are just Thunderbolt or USB accessed.

while researchers migrated to windows/Linux.
Machine learning (and all related HPC applications) should have been another market but without CUDA it's not possible to attract any serious researcher, even those loving macOS will work from an iMac on a local or remote Linux based cluster (what I do).

If have a Mac Pro with a decent chunk of memory in it they can run Windows/Linux on it concurrently with macOS.

BTW Apple scheduled another NY friends event on December 2, likely not to introduce new Apple watch bands

which they basically said they wanted to talk about their favorite apps and games for 2019. ( AppStore ... Apple Arcade part of a service biz that is about as big as the Mac division. Apple probably could use another Arcade bump .... especially a couple of weeks after Stadia stumbles out of the gate. ).

 

[CWallace]

Mac Pro prosumers were wanting to upgrade their mac pro since 2013. They got to be frustrated.

But they also should know by now that Apple is never going back to that form factor ("inexpensive" upgradeable mid-level tower).

If a prosumer will only accept that form factor, then they're stuffed unless they have the technical expertise or support base for a Hackintosh. But if they're willing to "compromise" on the form factor, the iMac Pro and MacBook Pro are more powerful than the Mac Pro 6,1, much less 5,1.

 

[ManuelGomes]

Just got an email from AMD announcing the Radeon Pro W5700, RDNA based. Navi coming to WS. This should be the entry level Pro option for the mMP.
[automerge]1574190109[/automerge]
Those prices on iBuildMacs should be placeholders but if true they're not so bad.

 

[deconstruct60]

Just got an email from AMD announcing the Radeon Pro W5700, RDNA based. Navi coming to WS. This should be the entry level Pro option for the mMP.

Zero macOS support mentioned on their site. ( only Windows , not even Linux). I wouldn't get hopes up for "real soon now". Apple does tend to trail behind the "Pro card" release in the priority queue. So couple of months it would be surprising if this card got official support ( but not a MPX module).
But late next Spring, Apple should do something if they are serious about the new Mac Pro commitment to upgrades.

$799 price point though (initially) . That many not replace the 580X soon. I highly doubt Apple and AMD are going to agree to the same point as a replacement in the near future.

 

[ManuelGomes]

I know, just pointing out that it would be a much better fit but still too early.

 

[ManuelGomes]

The only way I see Apple throwing us a bone right now (since the base price really seems set in stone) would be up the SSD to 512GB (dual blades) and fill all 6 mem channels to 48GB (hopefully @2933).
And throw in the XDR stand for free, as a buy now and get it free bonus (for a short initial period at least, eventually for good).
But all this is unlikely of course, just wishful thinking.
It sure would help some of us to pull the trigger I guess, would help me for sure.
Wouldn't help them (Apple) though. Less revenue in the "almost mandatory" upgrades.

 

[Mago]

If have a Mac Pro with a decent chunk of memory in it they can run Windows/Linux on it concurrently with macOS.

I know we have how to virtualize or even boot windows and Linux in a mac, but if you have the knowledge (or are used to suffer those os) what's the point to spend more in a much more complicated setup (no direct GPU in macOS vm), other options are cloud and local clusters thru remote development, this case a Mac mini is up for the task.

 

[RyanXM]

I'm of the thought that the delay MIGHT be to change the GPU options. Could we possibly see the 5700 as the base and the 5700 XT offered as upgrade before getting to the Radeon VII?

 

[deconstruct60]

I'm of the thought that the delay MIGHT be to change the GPU options. Could we possibly see the 5700 as the base and the 5700 XT offered as upgrade before getting to the Radeon VII?

The 5700's are skewed more toward video output graphics than toward compute. I highly doubt Apple is going to put a higher priority on faster frame rates over bigger grunt compute. The latter is mainly what Apple is aiming at with the Vega II cards. So perhaps the late arrival of some 3rd party add-in card options isn't likely to disrupt that in any significant way at all because they aren't really replacements for those cards.
Apple may certify some Navi10 based card(s) within a couple of months timeframe, but it is highly doubtful they are waiting for the either as a base configuration option or as a build/configure to order one.

This system is way, way , way past late. There is zero good reason to jam up the release with "oh wait throw one more new thing on the pile. " stuff. Apple is already loosing sales with the delays they have already piled up. Digging an even deeper hole isn't going to help. Waiting for a Navi10 variant would likely do as much damage and any handwaving "help" it was suppose to add later.

There are empty x16 slots on this machine. The card that is coming "real soon now" just need drivers and it can get added by end users after the launch. If can't do that have deeper problems with the product line.

 

[ManuelGomes]

Unfortunately it's quite unlikely that Apple will make any changes at this time concerning the GPUs.
Maybe next year they'll start popping up.

 

[thisisnotmyname]

I'm of the thought that the delay MIGHT be to change the GPU options. Could we possibly see the 5700 as the base and the 5700 XT offered as upgrade before getting to the Radeon VII?

no, they already have built systems sitting on shelves in TX.

 

[RyanXM]

no, they already have built systems sitting on shelves in TX.

Do tell....

 

[Stephen.R]

Do tell....

A shelf is a flat surface that you stack things on, to allow storing more items in a given amount of floor space.