Not yet? This is the first time that Nvidia couldn't support new MacOS and it's been more than 3 months already.
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 7,1 Waiting for Mac Pro 7,1 A1991 (no more)
- Thread starter Mago
- Start date
-
- Tags
- macpro 2016
- Sort by reaction score
Not yet? This is the first time that Nvidia couldn't support new MacOS and it's been more than 3 months already.
'Could not' is probably too broad. Apple is moving to a primarily Metal and several GPU features focused stack. In the older set up Apple did the top "half" of OpenGL and the video hardware folks did the bottom "half". Metal somewhat does two things. First those "halves" are potentially much smaller. Down to the point where have two chefs working the same stove.
In short, if Nvidia isn't writing/creating drivers the "right way" Apple will probably not sign them. Apple kicks folks out of AppStore all the time for busting the rules. If Nvidia did drivers that didn't get along with other GPUs (Intel , Nvidia) and kneecapped Metal in some way ( Nvidia engaging in some "embrace, extend, extinguish" campaign against Metal ) Apple probably isn't going to go along with that going forward.
Similarly Nvidia probably sees Apple getting into the GPU market as greater exposure to an Intellectual Property (IP) leak. (all the more so if Apple is trying to squeeze them out of low level driver work with more demands to low level IP information. ) [ If Nvidia is back to saber rattling on suing on GPU implementation IP royalties front, then all the more so. ]
Pretty good chance this is a "want to" issue on at least one and probably both sides. More interest in being the "800 lb" Gorilla to the other side than in getting a joint project done. Nivida using their CUDA 'club' as to how they can set their own requirements and Apple using their own macOS 'club' to do similar. The "want to" aspect would perhaps being that they form a joint/sharing IP facility where both sides contribute to sharing space between the two. The way they were doing things before doesn't work at this point. ( and Apple may need to pay more money to get things done the way they want. While Nvidia takes on a bit more risk )
Yet no responds from Nvidia and Apple for several months. Apple made OpenCL to against Nvidia's CUDA btw.
There is a difference between a whole new variant die and simply just turning on/off features that are on the die in the first place and binning results. macOS requires different drivers anyway. Some certifications and software/firmware optimizations isn't really all that much different than getting spun up on a new OS. Dx00 has some Windows driver baseline with the pro drivers also for that OS configuration side, but that's a bit secondary.
Binning for clock speed wouldn't be a modification either.
The no response for several months is one reason I suspect this is primarily a "want to" issue and not not some kind of either technical glitch or "we didn't get/understand the docs/API " so we can't comply issue. Nvidia's initial 100% point the finger at Apple probably didn't help at all if this a negotiation. Apple probably isn't going to get drawn into public finger pointing exercise.
Apple kickstarted OpenCL solution so that most systems (Apple or not) wouldn't be hooked to one and only one hardware vendor. Very similar to Apple's push of HTML5/javasrcipt over Flash (minus the crufty implementation problems that Flash interpreter had) .
Nvidia's overriding objective is to lock-in more of their own hardware. (CUDA didn't have major glitched implemented issues. But there lots of work that is to basically to dig a deeper moat. And fair amount of "embrace and extinguish" non work on OpenCL. ) If Nvidia had submitted CUDA into the Khronos and opened it up as a standard there zero indication that Apple wouldn't have gone along with it. The problem at the time was that Nvidia, AMD , Microsoft, and I think on of the common ARM GPU designs ( and/or Google? ) were all off proposing silo solutions in their own camps. Few with enough clot to get a standard kickstarting were trying. Most of the other players saw it as a inflection point where they could control it if used the leverage they had.
However, the last couple of years where the drivers have a "stop and halt" feature to trap on every OS update is indicative that the development process is broken on some relationship/responsibility level. A kernel extension/driver that properly asks the kernel resources and sticks to minding its own business with its own allocated stuff shouldn't stumble over the rest of the system at every dot release. If not getting early dot releases then really should be worried about what did to get into the "dog house". That actually is a bit like Flash (apparently brittle code that has problems more than several times a year. ) I'm a bit surprised that folks didn't see that problem would result in a major problem eventually.
Apple's GPU stack driver release process probably isn't perfect either. Again if it is one of those "they have to loose for us to win" attitudes it is a problem. Apple can reasonably demand that Metal have first class status but moves to cripple alternatives so can 'win' by default would be a problem.
I'm gonna guess WWDC19 because I ( personally) believe Apple will be putting forward looking tech into the nMP.
Im pretty sure that you need to wait till WWDC to see the leak and then the real announcement will be in October like Mac Pro 2013.
Marginally better than you buying a lottery ticket and winning.
Probably, February also.
Picks up a bit in March. By end of April should be enough new info that probabilities will need recalculating. (some narrow metric of "sneak peak" at least or some resetting of expectations (e.g., not sliding into 2020). ). Apple already has two cycles of "manage some expectations on the Mac Pro" done in April ( 2017 and 2018 ). At the very least, they will do that again.
I might be mistaken, but I remember reading Apple stating "the mMP is not a product for this *fiscal* year". Companies close their FY generally around March/April, therefore I would not expect them to tell us anything before.
On the other side, Intel is shipping Xeons to early adopters, AMD has in hand the hypothetical Vega2 and surely some Navi cards and Apple had time enough to develop a T3 chip and some other fancy stuff to put in a less thermal throttled system mounted in a super-stylish metal case.
SSDs cost is going down and DDR4 RAM is widely available.
Shall we all go to Cupertino and show them how to assemble them on a motherboard?
Kidding aside, it could even be possible that we will be seeing the mMP ready to ship by WWDC. They need to gain a bit of credibility and the mMP given to developers at WWDC like Apple once used to do, could help giving the impression they want to go back on track.
Besides, if they don't want to f%&$ it up, I am pretty sure this time they thought of a system that has at least a feasible upgrade path. Whether it's going to be new machines or upgrade parts, what I mean is that they surely thought of a system that can host future chips and future GPUs so that the design costs can be less of a failure than those of the tMP. We all know the mMP is not a mass product, therefore - as much as they seemed to understand that it's a product that has to be produced to give completion to a lineup - I could expect them to try to use the budget better than they did before.
Hence, their engineers might have developed a prototype with a Xeon W from an iMP, but updating the board and the chipset to the new CPU/GPU available at launch shouldn't take much longer than what a typical PC OEM would need to come up with a new product. They had (at least) three years to develop it. I get they want to do it Apple-style, therefore needing time to develop a concept and its USPs, but if they need more than a few months to update the mMP to a newgen CPU/GPU, then the project is already a failure, IMHO.
On the other side, Intel is shipping Xeons to early adopters, AMD has in hand the hypothetical Vega2 and surely some Navi cards and Apple had time enough to develop a T3 chip and some other fancy stuff to put in a less thermal throttled system mounted in a super-stylish metal case.
SSDs cost is going down and DDR4 RAM is widely available.
Shall we all go to Cupertino and show them how to assemble them on a motherboard?
Kidding aside, it could even be possible that we will be seeing the mMP ready to ship by WWDC. They need to gain a bit of credibility and the mMP given to developers at WWDC like Apple once used to do, could help giving the impression they want to go back on track.
Besides, if they don't want to f%&$ it up, I am pretty sure this time they thought of a system that has at least a feasible upgrade path. Whether it's going to be new machines or upgrade parts, what I mean is that they surely thought of a system that can host future chips and future GPUs so that the design costs can be less of a failure than those of the tMP. We all know the mMP is not a mass product, therefore - as much as they seemed to understand that it's a product that has to be produced to give completion to a lineup - I could expect them to try to use the budget better than they did before.
Hence, their engineers might have developed a prototype with a Xeon W from an iMP, but updating the board and the chipset to the new CPU/GPU available at launch shouldn't take much longer than what a typical PC OEM would need to come up with a new product. They had (at least) three years to develop it. I get they want to do it Apple-style, therefore needing time to develop a concept and its USPs, but if they need more than a few months to update the mMP to a newgen CPU/GPU, then the project is already a failure, IMHO.
I don’t think a new round table would be a good idea if they don’t want to piss customers off even more. They already had two of those. Next time they speak about the mac pro has to be to disclose details of the product and announce availability.
Your presumption is that the round table is not a far more full fledged "sneak peak" of an actual pre volume production, engineering model sitting in the middle of the round table and much of the discussion focused around talking about a real physical object. ( It doesn't have to run software or a demo, but simply a physical artifact to ground the discussion. )
That would highly useful. If they have one in April, but are waiting until some relatively (to the product schedule) arbitrary point in time of WWDC to do that in a big "dog and pony show" context , then that would be highly dubious. That would have far more of a "piss off " customer impact. If Apple is completely radio silent in April that will also piss off a large number of customers at this point also. They have actually backed themselves into a corner at this point. Both options are bad (more "mad" folks among other issues) options.
Even just nailing down that they are going back to "desk side" and have removed the "literal desktop" constraint that pegs the Mac Pro to a Mac Mini desktop footprint would be useful.
March-April Apple probably has some new iPads event that they could weave into. March-April is a decent chance have something new for iMac that they could weave into (it too is overdue and is possibly waiting on 2018 parts that slid into 2019. ). That on top of the previous April pronouncements is why the more generally available information is likely to change during April . In short, there are a couple of things to point to Apple doing active "reach out" to the tech press in the April time frame. For Apple to be 100% radio silent in that span would be odd.
If they don't have a physical sample product at the April minimally what will happen is more information on timing ( however vague. ) and general scope/size. They don't have a "better" good choice option to go with. If they aren't even looking at a "big box container" class of system at all, letting the folks looking for that know would cut down on the amount of venom they are going to get. They technically don't have to break the corporate rules about talking about future products to nail it down to what general portion of the workstation space they are shooting for.
Just killing off the expectations about a "z8" sized killer box or a "sealed and just about as small as the MP 2013" would be useful and productive at this point. The range expectations on the Mac Pro are spinning out of control at this point.
Apple does not have a "Mad Customer" problem at this point. This thread has 10+K posts in it and it is about the 3rd ( or 4th ? ) generation 10+K post thread about a "Next Mac Pro 7,1" product. Apple had "mad customers"issue back in mid-late 2014. Years ago. It is bigger now, but they had a substantive one at the first April 2017 session. What these April sessions are far more so in the managing the problem and the expectations solution space.... they never were about avoiding the problem. As long as the problem is active it still needs management.
What Apple has at this point is an increasing number of people who are buying something else problem. It isn't "mad" if have hard business requirements that need solutions; that is just solution finding business. It is straight up product defections. "less information" isn't going to plug the hole on that front in any significant way at this point. ( some folks are planning in March-April what they will do in June-October. It is a long pipeline to getting a new system approved/allocated and in place. ). Once those people have bought something new and are on a 3-6 year renewal cycle, those sales are pragmatically gone for a relatively long time.
Also, people who were on 5-7 year workstation product upgrade cycles are already almost 2 years into "circling the airport". At this point, a very significant chunk of them are going to be either looking for an alternative place to land or a very definite/concrete point in the queue to "land". Lack of information is only going to push more to the first option as a clear, pragmatic path. That isn't a "mad", that is just doing professional operations.
Apple playing "bury head into sand" ostrich in April would kill a subset of potential Mac Pro sales with a larger caliber weapon than necessary. It is simply just digging an even deeper hole. If they have no new product, then what they can do is stop digging an even deeper hole at the same rate they have been. That would be progress.
Last edited:
I'd generally agree (and also agree with the sentiment we have to hear something a year after the last time we heard something) which makes me think we're at least going to get a tease by April.
Given that they did screw the pooch the last time they brought it up at WWDC for an unveiling ultimately, I don't think repeating a "we'll barely ship it in time to fulfill our promise" strategy is a sound idea. Especially since by that point there will presumably already be chips for the iMac Pro and Apple really has to keep its entire pro line more regularly updated.
For the WWDC 2013 that was a much about the new Mac Pro as announcing the 'death watch' of the old Mac Pro. People were served notice that there was only a fixed number of months left to buy one of the old ones if they wanted it. ( along the lines of the notice served at the end of the XServes. ). Probably not a huge line of folks trying to buy but the 2013 is probably in a very similar boat ( more so that Apple wants to officially discontinue manufacture as to start the support count down clock on it). And Apple had "unofficially" let slip in WWDC 2012 that new Mac Pro coming in 2013 ( which they hit a 12 month timer on at WWDC 2013. )
Part of 2013 was parts too. Thunderbolt 2 wasn't suppose to ship in volume until 2014. I don't think that was a 'want to do" as much as the portions out of Apple's control pushed them into that early "sneak peak". They were more so in a "had to" state.
For the WWDC 2017 tease gap on the iMac Pro was parts. Apple doesn't have 100% control on that. While not an official replacement for the MP 2013 there was implied overlap which would help with the situation (image problem) the MP 2013 was burning with (the price cut in April 2017 wasn't going to cut it).
If they have hitched the Mac Pro to 2019 (back in 2017) then this "barely ship it in time to fulfill our promise" is more so a good way of making lemonade out of a lemon product management choices.
if they feel there is no good GPU for the iMac Pro by Spring-Summer for iMac Pro it won't be surprising if Apple drags their feet on the iMac Pro. The Apple "dog ate my homework" machine could slide into very late 2019 not doing anything with the iMac Pro.
I don't think Apple is going to dramatically pick up their tempo. Something closer to 24 month iteration cycle would be "up tempo" given their track record over the last 10 years. Regular and yearly cycles are two different things.
Never really thought about it that way, but you are exactly right. That's exactly when I bought my cMP 5,1. I was just so disappointed in the nMP
At the time I had a cMP 3,1 that was only slightly modified. In the 5+ years I've had my 5,1, I've done much to keep it relevant. Now, with my MVC flashed GTX 1080, if I could only run Mojave
Lou
well they can easily do cpu changes on the imac pro. Not so much video.
Now the new mac pro needs to have video on some kind of card so that it can be upgraded. MXM cards where a part of that idea but the cooling part mad that not really take off.
Hmm, RTX 2060 is quite nice product, from Nvidia. A little overpriced, a little inefficient, but that is understandable from 445 mm2 die. I still am completely not buying RT cores(meme tech, To be honest), but rasterization performance, is quite decent.
The raytracing is nonsense, in that considering their performance now it makes no sense you'd buy these cards when next year or the year after that they'll probably perform so much better for equivalent or lower prices, and the 60/70 cards can't push decent frame sizes at 60fps with the tech on.
As a future technology? It's seriously impressive. I remember when Adobe tried to make CUDA raytracing a thing in After Effects but it was dog-slow.
What's very annoying is that this decission has already been made by Apple by now - unless there is no 2019 MP release, or no future MP at all - and yet prospective customers are being kept in the dark .
They treat the Mac line like their iProducts, which has customers with a completely different purchasing mentality .
I guess, everybody who talks about RT Cores, haven't seen or used Radeon Prorender.
Im not sold at all on RT Cores. They are meme technology, regardless of GPU we are talking about.
350$ for GTX 1070 Ti performance is actually good value, but its painful that you have to pay for SKU, that last generation did cost 249$.
Also. Nvidia does make by far the best reference design in the industry. If I would be buying RTX 2060(very likely) I would 100% pick Founder Edition.
In the context of 'dust up' between Apple and Nvidia. No recent break on the drivers on that front points to that not being pragmatic at this point as a near-intermediate term option.
There is a "drop in" , CPU socket compatible solution coming (Cascade Lake derivative) , but there isn't much of a change in the context of Apple's iMac Pro focus.
i. Meltdown and Spectre fixes ( Apple will probably be receptive to the incremental speed bump and fewer band-aids need to add macOS. But still several macOS workarounds required so not revolutionary better. ).
ii.. Some "AI" machine learning VNNI instructions (which Apple libraries probably won't use in 2019-2020 since focused work on Apple's engine and GPUs. )
iii. Optane DIMMs . ( Again requires an OS and libraries changes to macOS . Is Apple going to do those)
iv. Some modest clock bumps ( Apple is already running older Xeon W at reduced clocks in iMac Pro ... chips that clock even higher than the clocks Apple didn't want to use are useful how? )
i & iv would be useful for a new "greenfield" Mac Pro. Less legacy Meltdown/Spectre kludges have to deal with over the machine's service lifetime the better. The Mac Pro also probably won't be "down clocking" the W offerings (i.e., designed to substantively larger thermal envelope). It would makes sense to start there. That there are a decent number of cloud hosting folks with current Mac Pro who will probably throw new ones in also in a VM centric context means that 'i.' issue is incrementally even higher utility.
Cooper Lake derivative which uses the next socket is due toward end of 2019. If move the iMac Pro to that then could goose two updates out of that with no substantive logic board changes. Cooper Lake will have
i. some clocking changes. May be able to do more incrementally improved dynamic clocking. Or more problematical a signifcant bump up in TDP range. ( which means Apple will need some new cooling sytsem refactoring to do).
ii. more AI stuff. (bfloat16) which Google's Tensor can leverage.
iii. more DIMMs support options. (which Apple will likely ignore on the iMac Pro )
iv. PCH chipset updates to go along with new Socket.
Late 2019 there will probably be some new options for the iMac Pro in GPU space.
If Apple is trying to use the same team to do both Mac Pro and iMac Pro updates then I easily see them assigned in most of 2018 for Mac Pro ( Cascade lake to catch the security and timeline of late 2018) and then shifted to a iMac Pro revision/upgrade for Cooper Lake and new socket change on a ramp from 2H 2018 up to full time 2019. Then flip again in 2020 back to Mac Pro to either pick up Cooper Lake or Ice Lake depending if Intel has hiccuped on 10nm yet again. ) Any new socket/PCH quirks tackled in the imac Pro board folded into the Mac Pro board.
Even if Apple shifts the iMacs onto 6 cores in 1H 2019 they can probably float the update of the iMac Pro into late 2019 without major problems ( grumbling sure but not major problems ).
MXM isn't really designed for heavy duty coolers. If in laptop range of TDP then fine. But if want to go up into the mid-upper range of desktop TDP then not so optimal. Especially if trying to keep noise at below average levels.
Some of the characteristics ( putting video back into the logic board) and no external edge connectors are on the right track, but Mac Pro probably is a bit too big for that if Apple shifts to deskside focus. If they are still laser focused on literal desktop then sure. Painting themselves into just a broader corner and oddball overlap with the iMac Pro.
P.S. Apple should be able to upgrade the iMac Pro and the Mac Pro at approximately the same time. However, their track record on being able to walk and chew gum at the same time over the last 5-6 years is horrible. For whatever reason they have decided that isn't a priority so these two will probably be single tracked with only modest overlapping pipeline work done. Even with Apple adding a single, leapfrogging track that is active year round , ever year , that is still way better than Rip van Winkle mode where they go completely to sleep for years at a time. That's progress.
Last edited: