Resolved .

[Reality4711]

Don't you lot have something more useful/fun/creative to do?

I tried to follow this thread in the hope of insight brain got fried with boredom oil.

 

[namethisfile]

Go in to an Apple store, and ask the genius if the CPU is a "user serviceable part", same as the RAM is a "user serviceable part".

Guess what, it's not. The fact that you can get to it and disassemble it and DIY, doesn't mean it's what Apple classifies as "user serviceable". A technician can pull and replace the 2013's GPUs as well, going to claim they're "user serviceable"?

Just like the cMP's bluetooth and wifi aren't "user serviceable", while the drive sleds, ram and PCI cards, are. User Serviceable, are the parts Apple designs specifically to be replaced by the end user, not an Apple Certified Technician.

Yeah, it's not official. But, the user can service them if they wanted to. Hence, "user-serviceable." And, one would DIY, a RAM upgrade for example, because it's cheaper and not that hard to do.

If, you sent a 2013 Mac Pro in to Apple for a RAM upgrade, you will have to pay for the Apple RAM upgrade price since you can't bring in third-party RAM at the Apple store to use for the upgrade. So, you will pay for Apple RAM price; the gas for your car to bring it there; the time you needed for that; then, your Mac might sit in the store for a couple days, if they don't have the Apple RAM and have to order it in; then, they will tack on labor cost...

So, for RAM upgrades, most people on here would agree that the best route is DIY. And, it's safe and not hard to do. So, even though it's not "official." It's also totally official. Know what I mean? Like, everybody does it.

You wouldn't care, Apple would care. Splitting the userbase only makes sense when the separate products produce a market enlargement sufficient to cover the increased expenses of two different machines. A sealed appliance Mac Pro is not going to appeal to any significant market outside of markets already addressed by the iMac Pro, while stealing sales, and therefore component volume discounts from the iMac Pro.

The question, "I," was also referring to, "Anyone." Like, why would anyone care? And, by anyone, thereby, implying you, as well. You get my drift there, sir? So, let me ask that question, again:

Why do you care about splitting the consumer base?

The more screens Apple sells, the cheaper the screens are, and the better either their margins, or their ability to lower prices - that means they can make the iMac Pro cheaper than it would otherwise have to be, or the iMac, depending on where they want to push around their pricing.

Yeah, we know about bulk order economics.

But, selling a lot of iMacs, like, even if they sold out their inventory will not necessarily correlate to iMacs becoming cheaper for the consumer.

The gains Apple makes in this scenario, will just go into the Apple bank. They have, after all, a future to move forward in. And, I heard they wanna make cars!

So long as Apple buys enough of them that LG wouldn't be better off using the production capacity to build something else that sells in greater volume, and is therefore more profitable.

I think it's clear Apple buys enough of them!

Lots of reports are the ultrafine is pretty much no longer stocked - it's an order-ony product. Given how badly it was received, quality issues etc, and being TB3 (so basically Mac)-only, I'll go out on a limb and say the entire product was probably a rush job, created to soak up overproduction on panels, not as a serious attempt to set up an independent LG 5k display line.

It's stocked. You can order it online, right now.

I am not sure about it being a rush job without knowing and being there when they made it.

I am probably one of the minority who thinks that the LG 27" UltraFine 5K monitor is fine and a handsome piece of monitor. And, sleek in its own way.

It's not enough for 5K panels to be profitable for LG to make them, it has to be more profitable to make them, than it would be to make something else, and 5k is basically an orphan resolution that only Apple are invested in, and even then, the only reason they've invested in it, literally the only reason, is because they couldn't get a proper resolution independent vector-based UI to work out. 5K / retina is a kludge, pure and simple.

Aha. I see you use words here like "orphan" and "it's not enough to just be profitable," it has to be "even more profitable."

Aha.

I heard about another industry where the word "orphan" was used and "profits..." did you wanna allude to that? If, so, I don't wanna comment.

8k is a mainstream video format, and 8k displays are going to be mainstream displays in a few years, similar to the transition from 1080p to 4K. 5K is its own little wilderness, existing only by the chance of fate that it's double the 1440p resolution of a non-retina 27" display.

5K is Apple being Apple, probably. And, chose that resolution because of the scaling that you mentioned. So, in that sense, it wasn't an act of fate. But, more an act of feat, or as Apple likes to call it, courage.

IIRC the iMac Pro has more in common with the 2013 Mac Pro, than the normal iMac in terms of its hardware system design. I guarantee you, the way it was announced was not the way it was intended to be. If there had not been the fever pitch in the mac community about Apple abandoning the pro space, if you hadn't had Gruber, Ritchie etc openly saying "something has gone wrong in Apple" about the 2013, there would have been no mea culpa meeting, the iMac Pro would have just launched as "The new (i)Mac Pro", and that would have been Apple's sole pro desktop.

Yeah, the iMac Pro has workstation-grade parts like the 2013 Mac Pro but updated to 2017 tech.

As for the mea culpa (I didn't know what this was at first; but, I presume it was when Apple made that public apology where they blame themselves for not updating the Mac Pro because of its design); then, I have already posted multiple entries in MR about this, wherein, I thought that it was probably Apple blaming themselves to assuage the fact that they won't be seeing an updated Mac Pro. But, an updated iMac, instead. And, I continued that why would Apple do this? Why make it public? And, I thought, "Maybe by making it public, they are calling to attention a certain company (Something M D) into the event without saying their name..." Thereby, IMO, it has something to do with A Something D... and, in the end, the iMac got the A M Something part in it and not the Mac Pro because the Mac Pro is getting something else better. And, it is waiting since it had already waited long enough for another company that is not Something M D to improve their yields, improve their drivers, perhaps, lower the price of their latest lineup before Apple jumps ship....

 

[mattspace]

Yeah, it's not official. But, the user can service them if they wanted to. Hence, "user-serviceable." And, one would DIY, a RAM upgrade for example, because it's cheaper and not that hard to do.

No, RAM updates are "User Serviceable" because Apple publishes the instructions on their site for users to do it themselves. Its got nothing to do with convenience - RAM in the 2013 is designed to be installed by end users, the same way cables are designed to be plugged into the back of it by end users, or pci cards and storage were designed to be installed by end users in the cMP era.

So, for RAM upgrades, most people on here would agree that the best route is DIY. And, it's safe and not hard to do. So, even though it's not "official." It's also totally official. Know what I mean? Like, everybody does it.

Absolute nonsense.

This https://support.apple.com/en-au/HT205044 is why the 2013's RAM is "User Serviceable". Beginning, middle and end of story. There is no such customer guide for the CPU.

Not going to bother continuing this.

 

[ActionableMango]

If the nMP couldn't edit videos, we would have seen dozens and dozens of similar stories. But we haven't. We just keep seeing the several years-old deadpool example repeated over and over again.

Well maybe you haven't seen them, but I certainly have.

There are numerous people here who've been through more than one nMP, and at least one user who's been through four of them costing him tens of thousands of dollars in equipment and lost jobs. The 2013 Mac Pro failure thread has nearly a quarter of a million views and is in the top 15 replied-to threads of all time in the Mac Pro forum. If you read through that thread you will see many people who had their defective GPUs replaced with GPUs that then went right ahead and failed again. There was also a class action lawsuit filed over the GPUs. And as you've pointed out, there were enough failures for Apple to have an extended repair campaign for them. You don't have an extended repair campaign if the problem is insignificant.

No they can't satisfy every pro. But most are or will be served just fine.

I think most people weren't well served by the nMP. You think most were. This was debatable up until Apple's press event. Apple didn't serve the pros that, by their definition, they were trying to reach with the 6,1 Mac Pro.

Craig Federighi speaking about the 6,1:

“It’s good for some; it’s an amazingly quiet machine, it’s a beautiful machine… But it does not address the full range of customers we wanna reach with Mac Pro.”​

Apple bet heavily on Dual GPUs gaining traction, and they lost big time on that one.

The 2013 Mac Pro has 2 GPU's for a number of reasons. But, I don't think those reasons include gaining traction into whatever thing your thinking about.

@Silencio is correct.

Craig Federighi again:

"We designed a system that we thought with the kind of GPUs that at the time we thought we needed, and that we thought we could well serve with a two GPU architecture … But workloads didn’t materialize to fit that as broadly as we hoped."​

 

[Joestanxx]

Don't you lot have something more useful/fun/creative to do?

I tried to follow this thread in the hope of insight brain got fried with boredom oil.

wow this is rabbit hole

 

[shaunp]

I quite liked mine, but in hindsight once I'd moved back to a PC it was a very messy solution as everything was external. The marketing photos look very good, but once you get a couple of monitors plugged in, a keyboard (there wasn't a wireless extended keyboard at the time), a couple of Thunderbolt storage arrays and the usual USB gubbins (card readers, etc) then there's just a big pile of crap all over your desk. That and it can't really be expanded. Otherwise it was a nice quiet machine.

 

[deconstruct60]

....

I think most people weren't well served by the nMP. You think most were. This was debatable up until Apple's press event. Apple didn't serve the pros that, by their definition, they were trying to reach with the 6,1 Mac Pro.

Craig Federighi speaking about the 6,1:

“It’s good for some; it’s an amazingly quiet machine, it’s a beautiful machine… But it does not address the full range of customers we wanna reach with Mac Pro.”​

There is nothing in Federighi comment that breaks down firmly the percentages. 'Some' means a significant quantity. It isn't 'all' or 'every' so it isn't the whole/full targeted market. It doesn't necessarily mean a relatively small one. Some of the Mac Pro 2013 target market Apple addressed with the iMac Pro. That subset also includes some folks coming off of previous Mac Pro in addition to that MP 2013.

Apple hit some with the Mac Pro 2013 (6,10) and the missed some with the 6,1. You've conveniently trimmed out this statement also.
"... The original iMac, you never would’ve thought as remotely touching pro uses. And now you look at today’s 5k iMac, top configs, it’s incredibly powerful, and a huge fraction of what would’ve traditionally required the Mac Pros of old and are being well addressed by iMac — whether its audio editing, video editing, graphics, arts and so forth. But there’s still even further we can take iMac as a high performance, pro system, and we think that form factor can address even more of the pro market. ..."
https://techcrunch.com/2017/04/06/t...-john-ternus-on-the-state-of-apples-pro-macs/

Apple's comments is that they have seen substantive customers coming out of the Mac Pro onto other Macs also.


There are also two other things.

First, there are several things in the that and other talks Apple has done that indicate that the next Mac Pro isn't targeting the same market they had intended to target with the Mac Pro 2013 ( and that the Mac Pro 2013 was targeting 100% of the previous Mac Pro market. It wasn't for example the two processor package set ups. Apple probably still isn't going to targets two processor package set ups but the dynamics 6-9 years later, depending upon your perspective is different now. )

Are they some folks that skipped both the 6,1 and iMac Pro who are circling the airport on 5,1 (and previous)? Yes. Is that the only (100%) folks Apple is going after with the next Mac Pro? Probably not. A large chunk? Yes. But Apple in no way has indicated that they are focused on building a HP z8 (or Dell 7920 ) clone either.


Second, all that Apple has really indirectly indicated is that there is "enough" folks that that scope outside the iMac Pro (and rest of Mac line ) to be interesting enough to take another stab at a Mac Pro. That doesn't mean that they most of the folks aren't covered already, just that there is enough 'left over' to do something with. From a 11-15 trend line most folks are dropped out of desktops. (the evidence is widely available). The real issue is how many are left.... not that there is some massive, growth market for the Mac Pro that is untapped.

@Silencio is correct.

Craig Federighi again:

"We designed a system that we thought with the kind of GPUs that at the time we thought we needed, and that we thought we could well serve with a two GPU architecture … But workloads didn’t materialize to fit that as broadly as we hoped."​

But Apple didn't just 'loose' here just because of hardware. More accurately Apple 'beat the farm' on both dual GPUs and OpenCL. The hardware GPUs ran into problems but OpenCL ran into a buzzsaw of shifting technical strategic plans also. Apple didn't put their full weight behind OpenCL, Nvidia (and Microsoft ) pragmatically did a variation of "embrace, extend , extinguish" with other solutions variants. Apple shifted all their weight behind Metal (and metal on iOS pervasive shared memory architectures first priority).

Again some workloads surfaced to match the moves they made and some stuck to the status quo. ( the quote is that they didn't broadly appear. Not that they didn't appear at all. ). In some of the workloads where the GPUs failures were more common, lots of indicators is that it was too much dual usage was a contributing factor to the problem; not too little.

Apple didn't quite get the two GPUs right either at the higher ends of those workloads. The bigger issue there was more so going into "Rip van Winkle" mode. Not the hardware or the software specifics at any one time. That's where there still seems to be a problem ( if measuring on tasks finished and accomplished. )

 

[namethisfile]

wow this is rabbit hole

In my defense, I tried to be a ferret.

 

[ActionableMango]

There is nothing in Federighi comment that breaks down firmly the percentages. 'Some' means a significant quantity. It isn't 'all' or 'every' so it isn't the whole/full targeted market. It doesn't necessarily mean a relatively small one.

Yes we don't know firm percentages, but that's a nearly impossible standard to ask for.

What we do know is that Federighi said they didn't reach all of the pro customers that they wanted to, it was a number significant enough to mention, and it was important enough that the redesign that will address "expandability" and "upgradability" in some way--the exact two things that those of us left behind have been endlessly complaining about.

 

[namethisfile]

There is nothing in Federighi comment that breaks down firmly the percentages. 'Some' means a significant quantity. It isn't 'all' or 'every' so it isn't the whole/full targeted market. It doesn't necessarily mean a relatively small one. Some of the Mac Pro 2013 target market Apple addressed with the iMac Pro. That subset also includes some folks coming off of previous Mac Pro in addition to that MP 2013.

Apple hit some with the Mac Pro 2013 (6,10) and the missed some with the 6,1. You've conveniently trimmed out this statement also.
"... The original iMac, you never would’ve thought as remotely touching pro uses. And now you look at today’s 5k iMac, top configs, it’s incredibly powerful, and a huge fraction of what would’ve traditionally required the Mac Pros of old and are being well addressed by iMac — whether its audio editing, video editing, graphics, arts and so forth. But there’s still even further we can take iMac as a high performance, pro system, and we think that form factor can address even more of the pro market. ..."
https://techcrunch.com/2017/04/06/t...-john-ternus-on-the-state-of-apples-pro-macs/

Apple's comments is that they have seen substantive customers coming out of the Mac Pro onto other Macs also.

There are also two other things.

First, there are several things in the that and other talks Apple has done that indicate that the next Mac Pro isn't targeting the same market they had intended to target with the Mac Pro 2013 ( and that the Mac Pro 2013 was targeting 100% of the previous Mac Pro market. It wasn't for example the two processor package set ups. Apple probably still isn't going to targets two processor package set ups but the dynamics 6-9 years later, depending upon your perspective is different now. )

Are they some folks that skipped both the 6,1 and iMac Pro who are circling the airport on 5,1 (and previous)? Yes. Is that the only (100%) folks Apple is going after with the next Mac Pro? Probably not. A large chunk? Yes. But Apple in no way has indicated that they are focused on building a HP z8 (or Dell 7920 ) clone either.

Second, all that Apple has really indirectly indicated is that there is "enough" folks that that scope outside the iMac Pro (and rest of Mac line ) to be interesting enough to take another stab at a Mac Pro. That doesn't mean that they most of the folks aren't covered already, just that there is enough 'left over' to do something with. From a 11-15 trend line most folks are dropped out of desktops. (the evidence is widely available). The real issue is how many are left.... not that there is some massive, growth market for the Mac Pro that is untapped.

But Apple didn't just 'loose' here just because of hardware. More accurately Apple 'beat the farm' on both dual GPUs and OpenCL. The hardware GPUs ran into problems but OpenCL ran into a buzzsaw of shifting technical strategic plans also. Apple didn't put their full weight behind OpenCL, Nvidia (and Microsoft ) pragmatically did a variation of "embrace, extend , extinguish" with other solutions variants. Apple shifted all their weight behind Metal (and metal on iOS pervasive shared memory architectures first priority).

Again some workloads surfaced to match the moves they made and some stuck to the status quo. ( the quote is that they didn't broadly appear. Not that they didn't appear at all. ). In some of the workloads where the GPUs failures were more common, lots of indicators is that it was too much dual usage was a contributing factor to the problem; not too little.

Apple didn't quite get the two GPUs right either at the higher ends of those workloads. The bigger issue there was more so going into "Rip van Winkle" mode. Not the hardware or the software specifics at any one time. That's where there still seems to be a problem ( if measuring on tasks finished and accomplished. )

I am more or less on your side here and I agree with the things you mentioned, especially about target audience, etc.

What I am still gray on is towards the end of your post where you specifically talk about the 2013 Mac Pro, OpenCL, Metal, and GPU failures.

I have read of the GPU failures. Since I don't have a trashcan MP, I am only relying on ppl posting on here about them. Has anyone ever figured out the cause of them?

I am also gray on the whole OpenCL, Metal thing since I am not a developer and when I hear Metal and OpenCL, my experience with them is from a UI interface perspective and in FCPX where those API's might be leveraged. Basically, they're suppose to make things faster, make UI smoother since MacOS is so UI extensive in terms of visual/graphical elements.

And, so, Apple basically switched to Metal because it merges the OpenCL and OpenGL thing, right?

So, it doesn't seem like Apple abandoned OpenGL or OpenCL, they sort of, merged the two? Or am I wrong on this?

So, I don't see how Dual GPU usage would lead to GPU failure. Can you elaborate more on this? It sounds interesting....

 

[ActionableMango]

@deconstruct60 is one of the very few posters around here who really knows what he's talking about, and all his posts are completely ignored... none of the typical crowd wants to engage discussion with him because they're all in make-believe land while he's talking reality)

For sure @deconstruct60 is one of the smartest and most analytical people around here. I say this even though I'm often in strong disagreement with him.

But your interpretation of why people don't engage with him is a bit one-sided and ignores a very real technical issue...he often deletes the username when quoting, which prevents notification for that user. The lack of notification is a big barrier against conversation, not just if you are disagreeing with him, but even if you are in agreement or just want discussion.

To ensure you are staying in the conversation with him, you'd have to manually take the time to periodically check those threads where you'd engaged in conversation and examine his numerous and lengthy posts to see if anything you've written was quoted.

Who's going to do that? That's an extraordinary high burden for internet forum discussion, and in a lot of cases I'm sure that it gets him the last word. To say that others don't respond to him only because they are in make-believe land is insulting, and it ignores the extra burden that he places on people.

 

[ixxx69]

For sure @deconstruct60 is one of the smartest and most analytical people around here. I say this even though I'm often in strong disagreement with him.

That is what's so weird... what are you in "strong disagreement" with him about? 90% of what he says are just the facts and/or common understanding by people in this industry who know what the heck they're talking about, unlike most of the crowd that's still left to frequent this forum.

Look, I get that a lot of people just want to hang out here to shoot the ****, talk about the golden years of the cMP (and a lot of tinkering and upgrades), and hang with other people who see things the same way they do, whether it's based in reality or not. It's just a little annoying when those same people who really don't know what they're talking about attempt to lecture those of us who do.

But your interpretation of why people don't engage with him is a bit one-sided and ignores a very real technical issue...he often deletes the username when quoting, which prevents notification for that user...

Not true on the quotes. Anyone can check out his post history for themselves - they all include usernames in the first quote. What he often does is emit the same username in further quotes within the same post.

However, I wasn't questioning why he isn't the most popular poster - there's any number of reasons, some of which you point out. I said that his posts were "ignored". This is the same crowd that will fall all over themselves to argue about literal gibberish with some AMD fanboy, but don't seem to ever want to take on posts they "strongly disagree" with because?...

 

[CC88]

I own it and I love it.

 

[Killerbob]

The nMP is NOT a failure - no matter what anyone is saying.

I have used Mac Pros since 2008, and got my nMP when it came out in spring of 2014 - equipped with the 6C CPU, 64GB RAM, 1TB SSD, and a bnunch of addons (external Sonnet Thunderbolt enclousure w. Sonnet Tempo Pro and Accelsior SSDs). I use mine every day and it still plows through anything I throw at it. I am a photographer, and speed (CPU and/or GPU) is only an issue when I render or work with 4K videos. For everything else it is plenty fast for me.

If anything, MacOS is getting a bit long in the tooth. I also run Win10 (in BootCamp and via Parallels), and Win10 is actually faster to boot up, and no, it is not MacOS, but for working in anything but Adobe CS, it is actually pretty nice.

I know the nMP cannot be upgraded, and yes that sucks, but for me it is a great machine.

 

[jinnyman]

For personal level, nMP may be a great machine for some. But business wise, and in terms of valueing professional user wise, it is a total failure.

 

[Polymorphic]

The "trash can" Mac Pro failed because Apple had a chip on their soldier, tried to make a statement, and chose the wrong product as the means to do that.

When it was unveiled in 2013, Phil Schiller made the comment "can't innovate anymore, my ass," in response to criticism at the time that Apple as out of ideas.

So Apple released a computer for the high-end professional market that was totally different from what preceded it.

The problem was that different isn't always good, and the professional market is especially sensitive to the removal of useful features (see also: Final Cut Pro).

Not only did the "trash can" Mac Pro remove useful features, but the new features it brought to the table (small size, inadequate single-fan thermal solution, proprietary GPUs) were not things that most professionals cared about.

In effect, the Mac Pro was inherently compromised so that Apple could make a point. It failed because Apple didn't seriously consider the needs, desires, and workflows of professional users.

Professional users like this: https://twitter.com/JBowdacious/status/698662679540428802

 

[namethisfile]

DELETED***

Found my high horse****

 

[th0masp]

Not only did the "trash can" Mac Pro remove useful features, but the new features it brought to the table (small size, inadequate single-fan thermal solution, proprietary GPUs) were not things that most professionals cared about.

I've worked in a few offices over the years (in art departments). Small form factor and near silent operation is something that would have been appreciated in all of them. Most of the big towers we had were never opened or upgraded but simply phased out after a few years.
I suspect that's the crowd Apple had in mind, not the ones with a forest of cables sticking out of their tower case, side panels off and an extra fan pointed at the innards.

 

[fuchsdh]

I've worked in a few offices over the years (in art departments). Small form factor and near silent operation is something that would have been appreciated in all of them. Most of the big towers we had were never opened or upgraded but simply phased out after a few years.
I suspect that's the crowd Apple had in mind, not the ones with a forest of cables sticking out of their tower case, side panels off and an extra fan pointed at the innards.

Yep. I'd say it's the usual issue of Apple going too far in a direction (and then never following through in either walking it back gradually or iterating enough that it doesn't matter.)

I would absolutely love a Mac Pro that maintains the same basic performance capabilities as the cheesegrater but is quieter and smaller. And that's eminently possible. But that also doesn't equal "and we made it 1/6th the volume and removed upgrade potential for the people who wanted it".

Put it this way: Apple makes a Mac Pro that's 7" x 17" x 17" and you've made a pro machine that's theoretically rack-friendly, smaller than a Z6, and the smallest Mac pro desktop they've ever made in more than two decades, leaving aside the 6,1.

 

[ixxx69]

I am more or less on your side here and I agree with the things you mentioned, especially about target audience, etc.

What I am still gray on is towards the end of your post where you specifically talk about the 2013 Mac Pro, OpenCL, Metal, and GPU failures.

I have read of the GPU failures. Since I don't have a trashcan MP, I am only relying on ppl posting on here about them. Has anyone ever figured out the cause of them?

I am also gray on the whole OpenCL, Metal thing since I am not a developer and when I hear Metal and OpenCL, my experience with them is from a UI interface perspective and in FCPX where those API's might be leveraged. Basically, they're suppose to make things faster, make UI smoother since MacOS is so UI extensive in terms of visual/graphical elements.

And, so, Apple basically switched to Metal because it merges the OpenCL and OpenGL thing, right?

So, it doesn't seem like Apple abandoned OpenGL or OpenCL, they sort of, merged the two? Or am I wrong on this?

So, I don't see how Dual GPU usage would lead to GPU failure. Can you elaborate more on this? It sounds interesting....

I'm not speaking for D60, but regarding the GPU failures...

1. When the 2013 Mac Pro's were released, there were still a number of bugs with firmware/software, both Apple's and some of the popular video editing products. A lot of software issues were blamed on the hardware.

2. 2013 Mac Pros manufactured for several months in early 2015 with defective D500/D700 GPU's. Apple launched a repair program for them in early 2016. I'm not sure anyone outside of Apple knows the specifics of the issue, and I don't know why it took Apple so long to figure it out - maybe they weren't selling enough of them to see a pattern on the defective units being brought in for service. Who knows how many of the GPU issues (and other system issues) we've heard about are directly related to those defective GPU's. Once a story gets out, it can circulates the internet for years.

3. There's been a lot of reports of GPU issues, the vast majority related to very heavy processing of video. Might be defective units (just like any product), might be software issues, might be inadequate cooling capacity for specific usages, might be user error... it's certainly some amount of each of those, but we don't know to what degree (and MacRumors forum is not indicative of the real world). My guess is that there were some real issues with certain usages... if not, Apple's certainly didn't try to put the rumors to rest. But the issues are widely overblown, and affect mostly a niche usage.

Regarding OpenCL/Metal... yes, Metal kind of merges OpenCL and OpenGL.

Whether or not the GPU cooling issues were exacerbated by dual GPU's (more strain on the cooling system), the 2013 MP's dual GPU's was a different kind of failure and probably more indicative of why the 2013 Mac Pro failed in the marketplace... Apple bet the farm that there was a bigger market for workflows dependent on multiple GPUs (using OpenCL, now the focus on Metal). That never materialized in the way Apple was hoping.

In other words, most of the market that would want to purchase the 2013 MP weren't gaining anything from dual GPU's (I would have much rather had a single D700 or even a D500 than dual D300's). They were buying MBP's and iMacs instead - computers that actually offered better performance-value for their workflows. And at least a good portion of the market that would benefit from dual GPU's didn't want the 2013 MP.

The 2013 MP is a great little computer for its time, and if it had parts and price more akin to the iMac in that form factor, it would have sold like hotcakes. But Apple has so far refused to make that computer - they're stubbornly just not interested (the super long-odds bet is the 2019 MP is basically that, the xMac, and they finally give up on workstation-class hardware... it's just such a tiny niche that needs it). At its price point, it just wasn't appealing to a large enough market to warrant further updates by Apple.

 

[deconstruct60]

I
I have read of the GPU failures. Since I don't have a trashcan MP, I am only relying on ppl posting on here about them. Has anyone ever figured out the cause of them?

I don't think there is one single cause for all of the failures. A sizable class though seems to be sustained workloads which light up one or both of the GPUs for relatively long periods of time along with moderate to high utilization of the CPU.

Too much coupling of heat sources combined with not quite enough airflow and no overall 'balancing' global fallback mechanism probably was/is a major contributing factors to those. It seems doubtful that Apple extensively torture tested the MP 2013 model primarily as a high performance computational node as opposed to a single user interactive "viewer" workstation.


There were software/firmware glitches too. Lots of stuff piled up, but the 'differentness' of the Mac Pro was an easy thing to point a finger at in the blame game.

I am also gray on the whole OpenCL, Metal thing since I am not a developer and when I hear Metal and OpenCL, my experience with them is from a UI interface perspective and in FCPX where those API's might be leveraged. Basically, they're suppose to make things faster, make UI smoother since MacOS is so UI extensive in terms of visual/graphical elements.

And, so, Apple basically switched to Metal because it merges the OpenCL and OpenGL thing, right?

No. Pragmatically Apple switched to metal because there wasn't a "open standard" that was evolving quickly enough (and with the priorities Apple wanted emphasized). OpenGL had some previous development stages where it turned into more of an exercise of herding cats and than a diligent, focused committee solving problems. So did their own and didn't ask for consensus input on direction. Nvidia was primarily pushing CUDA ( and slow rolling OpenCL to allow CUDA to develop more traction/inertia ). Microsoft as usual was not particularly interested in either OpenGL or CL ( they have alternatives to both to push; which muddles the water further ). Google Android was off hemming and hawing. Imagination Tech, Intel, and AMD played along with Apple requests, but they apparently were also open to Metal (to keep the Apple checks coming).

Similar to Thunderbolt where Apple pitched the notion to Intel and they sprinted off getting something done without trying to gather a large group into some sort of standard body. Standards are funny things. It isn't just technical issues. The timing has to be right for them to work. Too many cooks in the kitchen in the extremely earrly stages can mess can defeat consensus building. Too late in the process and others have cooked up their own alternatives to participate well.

Apple probably felt they didn't have 'time' for all of that courtship stuff. iOS needed to move to a "next generation" graphics stack sooner rather than later, so Apple just largely did their own based on some factors that were being discussed as to what to do "next' for OpenGL and some common baseline approaches for the video gaming systems ( Playstation and Xbox/DirectX ) , and AMD's Molten. (What was 'wrong' with OpenGL from the 3D library developer perspective was relatively well researched and disccused at that point). In 2012-2014 timeframe none of that was lining up for a solid well formed consensus, so Apple opted to just do it themselves ( i.e.., similar to Microsoft model where leverage your market inertia to get other folks to row to their cadence/design dictations. ) Vulkan and OpenCL 2+ eventually arrived in 2015-2016, but that was probably at least two years too late for Apple.


Metal incrementally picked up some compute/computational aspect but it isn't trying to be a "general usage' as OpenCL. ( it is a shader/shading language that is taken on few more general compute aspects ). That aspects Apple is probably a bit behind the alternatives in.

So, it doesn't seem like Apple abandoned OpenGL or OpenCL, they sort of, merged the two? Or am I wrong on this?

No. Metal ( and Vulkan and the latest 'next gen' ) graphics stacks are much more 'low level' than OpenGL is. There is really now more a presumption that most applications are going to use another 'portable' stack on top of the lowest level one ( Apple's Foundation libraries , QT , gaming 'engine' , company proprietary porting library , etc.).
Metal as a substitute for OpenGL requires a very substantive rewrite (or port to 3rd party) work.

Similarly, OpenGL (and DirectX ) had shader languages attached to them previously. Metal having a shader language doesn't really replace/merge OpenCL. Again folks can mutate their OpenCL solution to fit the Metal abilities but it is a substantive change. ( Khronos which manages the development of OpenGL/Vulkan/OpenCL do coordinate Vulkan and OpenCL so that compute aspects tend to get funneled through OpenCL, but it is a somewhat different approach. )

GPU compute and classic GPU work all have to get along and 'share' the physical GPU so the management has to be merged at some point (both can't do whatever they want at any time they want. )

OpenGL/OpenCL versus Metal is pretty similar to putting a round peg in a square hole. For some apps the round peg going to be smaller than the square hole so it will fit with some small shim/scaffolding adjustments. For some apps the round peg is bigger than the square hole and just hammering on the peg isn't going to make it fit. ( When Apple just waves their hands and say "it is a simple port" ... it is mostly that ... hand waving. )

So, I don't see how Dual GPU usage would lead to GPU failure. Can you elaborate more on this? It sounds interesting....

"... The triangle you mentioned, the thermal core, is designed to have three fairly similar loads – similarly balanced in power. And so the overall size of the product and the fan, that defines the overall thermal capacity for the enclosure. ...
... . The way the product was constructed was with these three balanced loads in mind. So, again, it served its purpose well and created a really quiet and powerful machine using it for a certain class of things, it just didn’t necessarily have the flexibility that we need to have. ..."
https://techcrunch.com/2017/04/06/t...-john-ternus-on-the-state-of-apples-pro-macs/

If the heat sources are not all balanced I suspect there is more horizontal tranfer than vertical transfer in the thermal core. Instead of the heat mostly going vertical , you get some going to the other side to heat up the other component(s) that aren't quite as hot. That spreads the head over a smaller surface area in the core and thermal transfer isn't as good to the air flow.

Also even if get heat spend out evenly over the internal surface area of the thermal core if don't flow enough air through the center then have a problem . A fan that devotes a sizable fraction of its potential exhaust area to delivering Wi-Fi signals isn't particularly optimized to moving the most air possible.

If the thermal core radiates enough heat back to the other components on the video board ( capacitors , power management , etc.) over an extended period of time then can run into problems over the long term. If can get some heat flows that are not primarily "up and out" and you'll get issues.

It isn't just dual per se (purely the number two). It is dual with those design constraints.

What Apple wanted to do could have worked better with some reasonable adjustments. That would have made what they had work better. The problem that AMD ran into with process shrink availability and design trend blows outs on perf/W and overall trend to crank all high performance desktop GPU boards ( GPU + RAM power load) to higher TDPs was a disconnect from the baseline design.

 

[deconstruct60]

...Apple bet the farm that there was a bigger market for workflows dependent on multiple GPUs (using OpenCL, now the focus on Metal). That never materialized in the way Apple was hoping.

There is a difference between market for general market for multiple GPUs and what Apple was "hoping for".
When the Mac Pro 2013 launched OpenCL was behind the curve (CUDA had more momentum and AMD was still conflicted with proprietary vs open ). Part of what Apple was hoping for is that alot of other players would put down their own discretionary money to move OpenCL forward. That Mac Pro would be a leading OpenCL platform without investing much Apple money into making general, overall OpenCL much more healthy and that enabling that Mac Pro had close to best-in-class OpenCL toolchain and environment. (Meanwhile Apple threw at least as much, if not more, money are Metal as a diversion from that goal. )

Apple had Swift development going on at that point too. It really wasn't a "bet the whole farm" in terms of actual capital investments. It was more like let's choose these products constraints and it will all magically fall into place with non-holistic investment ( Oompa-Loompas will show up while we're in Rip van Winkle mode and flush out the stuff we're not going to work on all that hard. )

I think the Mac Pro got caught partially in the strategic shift from OpenCL to Metal. It was was too far along (and Metal way too nascent ). It wasn't just the Mac Pro, the uptake of OpenCL across the general Mac stack was going to be an issue. For much of that too Apple wasn't completely fully baked into a clear vision.


There is also a decent chance that the "dual GPU" had a Scrooge McDuck motivation behind them also. In order to hit a higher volume of custom GPUs made, the limited number of folks who are going to buy get two. Part of Apple didn't want to be in the video card making business so cranked up the numbers to settle that debate. Need software to leverage it too (which is also an investment. Scrooge McDucks will moan on that too. )

Apple could have had good reasons to drop Nvidia from the solution mix for the system. However, for the Mac Pro that was really going to mean they were going to need to work "twice as hard" to make that 'bet the farm' choice work. There is nothing to indicate that they did that. In fact, almost everything about the Mac Pro has been either 'slow motion' or "Rip van Winkle" like in response. The other parts of the Mac business were doing good and the iOS doing extremely well so they simply focused on other stuff as opposed to working twice as hard to make the Mac Pro 'work' .
( highly likely there is probably a significant faction inside Apple that periodically gets the upper hand on "don't do Mac Pro anymore" debates. )

The 2013 MP is a great little computer for its time, and if it had parts and price more akin to the iMac in that form factor, it would have sold like hotcakes. But Apple has so far refused to make that computer

But they did do the iMac and iMac Pro. The iMac Pro is probably a bit over inflated but that is as much air cover for the iMacs they are selling in the upper BTO range.

iMac Pro isn't selling like hotcakes, but it appears to be selling "good enough". The iMac Pro is to a large extent the follow on to the MP 2013. ( not most of the dual GPU users that were extremely unhappy or happy, but a large overlap. ). the iMacs at some point will pick up six cores and get even more traction from the old Mac Pro market user space from 5-10 years ago.

- they're stubbornly just not interested (the super long-odds bet is the 2019 MP is basically that, the xMac, and they finally give up on workstation-class hardware... it's just such a tiny niche that needs it). At its price point, it just wasn't appealing to a large enough market to warrant further updates by Apple.

The iMac Pro was were Apple put their effort (time and money) into.

Doubtful that it is an xMac. The mini is far more so a 'bone' thrown at many of the xMac folks. Not that it will keep most happy but it is extremely doubtful Apple is going t spend all this time just to crank up the fratricide in the core iMac price zone.

Apple has been shifting to higher average selling prices so they aren't going to replace the Mac Pro with a range of system selling a far lower average selling price. All of Apples moves of late point in the opposite direction. The next Mac Pro might not see a big raise from the MP 2013 prices, but Apple will trade one less GPU bill-of-material (BOM) cost for more of something else. ( bigger capacity, default SSD at Apple mark up pricing or just more base level RAM or both). The iMac Pro selling "OK" on volume but some more volume to spend out the BOM cost over a bigger base is probably what they are going to do for some of the major components in both.

The problem with the Mac Pro (and iMac Pro probably also) is the 'dancing on a pin' exercise of being big enough to be interesting to Apple (and their target minimal volumes ) and aligned enough with rest of the product line. The "further updates" is really back in the space that is more so covered by what they previously shifted away from ( the 2008-2012 Mac Pros that folks are circling the airport with). It is a smaller market than it was about a decade ago when they kind of decided it might be too small to be in. Hence the really slow motion effort to jump back in.

 

[AidenShaw]

The triangle you mentioned, the thermal core, is designed to have three fairly similar loads – similarly balanced in power. And so the overall size of the product and the fan, that defines the overall thermal capacity for the enclosure.

I realize that this is a quote and not your words - but the MP6,1 thermal core is anything but balanced.

​

There are literally twice as many fins cooling the CPU as there are fins per GPU. Hmm, and the GPUs often burn out.

 

[rrl]

The 2013 MP is a great little computer for its time, and if it had parts and price more akin to the iMac in that form factor, it would have sold like hotcakes. But Apple has so far refused to make that computer - they're stubbornly just not interested (the super long-odds bet is the 2019 MP is basically that, the xMac, and they finally give up on workstation-class hardware... it's just such a tiny niche that needs it). At its price point, it just wasn't appealing to a large enough market to warrant further updates by Apple.

Wow, this paragraph even makes Apple apologists cringe.

 

[ActionableMango]

That is what's so weird... what are you in "strong disagreement" with him about? 90% of what he says are just the facts and/or common understanding by people in this industry who know what the heck they're talking about, unlike most of the crowd that's still left to frequent this forum.

Well I respectfully disagree. You can look at recent post #97 and see speculation and opinion, just like everyone else does.

Not true on the quotes. Anyone can check out his post history for themselves

My mistake, I apologize.