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

[deconstruct60]

I was thinking the same thing, the Mac Pro is gonna be the star of the WWDC keynote.

Highly liklely not. WWDC 2018 consistent of no Mac announcements.

https://www.macrumors.com/2018/06/04/everything-apple-debuted-at-wwdc-2018/

Apple has the same number ( perhaps more) of OS variants to cover in approximate the same time allocation. Apple is suppose to be trying to start the path to merge App distribution but merging app stores.

It must have so much different about it, they will need to give it an hour to explain all the changes, accessories, pro displays, demos from content creators who have had a chance to work with early prototypes.

macOS itself may not get an hour let alone one Mac product. If would take over an hour to explain why the Mac Pro was relevant then it would be far more useful to do it with its own ( or coupled with iMac Pro ) own event. If it isn't largely self evidence then WWDC is largely the WRONG place to do it. WWDC has a relatively long list of stuff to cover. People's attention spans are only so long and can only absorb at modest rates. Apple isn't going to be able to "blipvert' watchOS , tvOS , iOS into a much smaller time slot.

The majority of developers at WWDC are not Mac developer ( flavors of iOS dominate). The bulk of the time is going that way. WWDC isn't going to turn into a primarily Mac only show. That is just gross mis-expectation setting.

 

[AidenShaw]

They're not going to back away from the Mac Pro being some sort of sacred sealed object.

Sadly, I suspect that you are right.

 

[Pressure]

The majority of developers at WWDC are not Mac developer ( flavors of iOS dominate). The bulk of the time is going that way. WWDC isn't going to turn into a primarily Mac only show. That is just gross mis-expectation setting.

As long as Xcode still require MacOS everyone needs a proper Mac for it.

 

[AidenShaw]

Intel does like to change up their sockets pretty regular, gotta push those motherboard sales...

This is unfair.

There are real, useful reasons for socket changes.

If a new CPU has six memory channels - are you going want to use a motherboard/PCH which has three or four channels?

If a CPU has 48 PCIe 3.0 lanes, do you want to use a motherboard/PCH which supports 40 PCIe 2.0 lanes?

If the CPU supports 3.0 TiB of DDR4 LRDIMM memory, do you want to use a motherboard/PCH that supports 384 GiB of DDR3 memory?

And, considering that Intel is not a huge force in the motherboard market - it's not only unfair, but also silly.
[doublepost=1558470419][/doublepost]

Good read!
https://www.zdnet.com/article/six-ways-the-new-mac-pro-could-go-terribly-terribly-wrong/

And they missed #7 - " 'no Nvidia' means 'dead on arrival' ".

 

[Mr. Dee]

Highly liklely not. WWDC 2018 consistent of no Mac announcements.

https://www.macrumors.com/2018/06/04/everything-apple-debuted-at-wwdc-2018/

Apple has the same number ( perhaps more) of OS variants to cover in approximate the same time allocation. Apple is suppose to be trying to start the path to merge App distribution but merging app stores.




macOS itself may not get an hour let alone one Mac product. If would take over an hour to explain why the Mac Pro was relevant then it would be far more useful to do it with its own ( or coupled with iMac Pro ) own event. If it isn't largely self evidence then WWDC is largely the WRONG place to do it. WWDC has a relatively long list of stuff to cover. People's attention spans are only so long and can only absorb at modest rates. Apple isn't going to be able to "blipvert' watchOS , tvOS , iOS into a much smaller time slot.

The majority of developers at WWDC are not Mac developer ( flavors of iOS dominate). The bulk of the time is going that way. WWDC isn't going to turn into a primarily Mac only show. That is just gross mis-expectation setting.

Huh?

I am sure the the 2007 Intel Mac Pro and XServe were both announced WWDC, as well 2013 Mac Pro was previewed at WWDC 2013, 2017 MacBook Pro's and iMacs and the iMac Pro were previewed at the 2017 WWDC. Why not the same for the next Mac Pro?

 

[AidenShaw]

Huh?

I am sure the the 2007 Intel Mac Pro and XServe were both announced WWDC, as well 2013 Mac Pro was previewed at WWDC 2013, 2017 MacBook Pro's and iMacs and the iMac Pro were previewed at the 2017 WWDC. Why not the same for the next Mac Pro?

Apple has made many hardware announcements at the spring San Francisco event (née MacWorldSF, now called WWDC).

• 2003 - PowerMac G5
• 2004 - 23" and 30" Cinema displays
• 2005 - Intel transition announced, and the Intel development platform
• 2006 - Mac Pro announced, Intel XServe
• 2008 - Iphone 3G international
• 2009 - 13" MacBook Pro, updates to 15" and 17" MBPs, Iphone 3GS
• 2010 - Iphone 4 (and Antennagate)
• 2012 - update MacBook Air, and MacBook Pro models - including first "retina"
• 2013 - MP6,1, AirPort Time Capsule, AirPort Extreme, MacBook Air
• 2017 - Imac, MacBook and MacBook Pro, Imac Pro, 10.5" Ipad Pro, HomePod

https://en.wikipedia.org/wiki/Apple_Worldwide_Developers_Conference#2000s

 

[Boil]

Apple has made many hardware announcements at the spring San Francisco event (née MacWorldSF, now called WWDC).

Why do you keep cross-referencing WWDC & MacWorld...?

The former is a developers conference, the latter a (now defunct) trade show...

They were never one & the same...

And excepting the 2002 - 2015 span when it was in San Francisco, WWDC has been in San Jose since its inception in 1987...

 

[AidenShaw]

The former is a developers conference, the latter a (now defunct) trade show...

Q: Why is Apple making hardware announcements at a "developer's conference"?

A: Because Apple misses the spring west coast trade show, and is corrupting the "developer's conference" with hardware announcements.

 

[Peace]

Why do you keep cross-referencing WWDC & MacWorld...?

The former is a developers conference, the latter a (now defunct) trade show...

They were never one & the same...

And excepting the 2002 - 2015 span when it was in San Francisco, WWDC has been in San Jose since its inception in 1987...

1987 Santa Clara.

wink wink

 

[AidenShaw]

And excepting the 2002 - 2015 span when it was in San Francisco, WWDC has been in San Jose since its inception in 1987...

San Jose and San Francisco are "next door" in the Bay Area - about 45 minutes by one map's estimate to get from McEnery to Moscone. Many WWDC attendees will fly to SFO for the conference, not to SJC.

If I'm in Kuta and someone asks me where I'm from, I'll say "San Francisco". Question answered.

If, instead, I replied with the name of the Peninsula town where I live - I'd get a "where's that".

In the 5 km view, San Jose and San Francisco (and Santa Clara) are the same city.

​

 

[Boil]

Q: Why is Apple making hardware announcements at a "developer's conference"?

A: Because Apple misses the spring west coast trade show, and is corrupting the "developer's conference" with hardware announcements.

Or because developers need to know about the hardware they are developing for...?!?

1987 Santa Clara.

Yeah, I let that slip by, even after reading it in Aiden's wiki link about WWDC...

San Jose and San Francisco are "next door" in the Bay Area - about 45 minutes by one map's estimate to get from McEnery to Moscone. Many WWDC attendees will fly to SFO for the conference, not to SJC.

If I'm in Kuta and someone asks me where I'm from, I'll say "San Francisco". Question answered.

If, instead, I replied with the name of the Peninsula town where I live - I'd get a "where's that".

In the 5 km view, San Jose and San Francisco (and Santa Clara) are the same city.

None of this makes your constant statements of WWDC San Jose = MacWorld SF true...

Two separate conferences (one long gone), with two separate focuses...

Two separate cities, just ask the USPS...

Too much time spent with this side bar...

 

[AidenShaw]

Two separate conferences (one long gone), with two separate focuses...

But Apple has been blurring the focus, by making hardware announcements at the "developer's conference" year after year.

Two separate cities, just ask the USPS...

If I ask the USPS, a house four houses away from mine is in a different city - even though we're in the same school district and share many other services.

For someone in Boston, is there any real difference about flying to SFO and taking the shuttle south to San Jose or the other shuttle north to San Francisco?

Too much time spent with this side bar...

Agreed. Ciao.

 

[fuchsdh]

The bottom end of the Xmac crowd from 5-8 years ago is far more covered by the Mini now then they were. The top end of the Xmac crowd would pretty easily be covered by "year before last's" Mac Pro model is Apple got back on track of doing regular updates and somewhat grew the refurbished/used market. And also keep the iMac upgrades coming. There will still be a gap but high priced "lego" blocks probably won't close that.

The iMac Pro just recently got "spec touch" updated. There is probably nothing new coming soon or they wouldn't have 'kick the can' in March.

The Xeons aren't actually out yet. Since the iMac Pro is an established product it'll probably get the updates when they ship, which most likely isn't until after WWDC timeframe.

Given the recent iMac, iMac Pro, and MBP updates, it's possible Apple is shifting back into a more regular pattern of spec bumps, either because they've realized that neglecting the Mac certainly hasn't helped their financials, and/or want to prop up Mac sales as Intel continues to miss and under-deliver. In which case there's far less pressure on these events for minor updates.

On the Mac mini, that's a valid point. An eGPU + Mac Mini isn't an integrated minitower but it's still closer in capabilities to an xMac than Apple's put out since the G4 days.

But Apple has been blurring the focus, by making hardware announcements at the "developer's conference" year after year.


If I ask the USPS, a house four houses away from mine is in a different city - even though we're in the same school district and share many other services.

For someone in Boston, is there any real difference about flying to SFO and taking the shuttle south to San Jose or the other shuttle north to San Francisco?


Agreed. Ciao.

Apple has released or announced hardware at WWDC only half the time. And they've *never* had repeated back-to-back years of big product announcements.

 

[jscipione]

Apple has just announced the 2019 mbp with 9th gen Intel processors which means there's nothing standing in their way from releasing the final Intel machine at WWDC. Hype!

 

[Boil]

Apple has just announced the 2019 mbp with 9th gen Intel processors which means there's nothing standing in their way from releasing the final Intel machine at WWDC. Hype!

Yes, this could be the final x86-64 Mac Pro...!

(...note I said x86-64, which leaves room for AMD...?)

 

[AidenShaw]

Yes, this could be the final x86-64 Mac Pro...!

Did you mean x64 ? That's the common term for 64-bit Intel - unless you're doing assembly or low level C in Linux.

​

 

[Blair Paulsen]

The MBP spec bump announcement ahead of WWDC, instead of at it, might indicate they want to preserve some time/mindshare for mMP reveal... Or not.

Assuming they can get at least a couple of the new Navi cards by late December they could still technically keep their 2019 ship date, even though waiting for Navis in quantity could push deliveries well into 2020. As others have noted, there are potential buyers who will hold out for the mMP as long as they have some idea what it will do, what it will cost and when they can get it - which makes saying nothing problematic.

Does anyone think that Apple and AMD might have been working on a custom form factor version of the Navi architecture for quite a while now? Could that project have been placed ahead of getting "regular" Navi cards out the door? If Apple was willing to pay a premium and promise to buy x number of units over the next x months would AMD say no? What if Apple was also willing to dump Intel and go with AMD for CPUs...

If Apple does switch horses, an mMP announcement at WWDC makes even more sense as they would want to get people coding for the new ecosystem ASAP.

At the end of the day, even a well designed and reasonably priced mMP on Intel might seem like an overpriced version of similar systems on Windows. OTOH, by going full AMD and promoting the apps that optimize for more available cores, marketing can cherrypick metrics that make the OSX/AMD option look pretty good...

 

[AidenShaw]

OTOH, by going full AMD and promoting the apps that optimize for more available cores, marketing can cherrypick metrics that make the OSX/AMD option look pretty good...

I'm glad that you mentioned "cherry pick" - because that's the only way that you'll find any apps that use more than 8 to 12 cores on a workstation.

 

[Blair Paulsen]

Since Apple can't count on developers making the effort to really exploit the potential advantages of a high core count CPU, they will likely stay with Intel's better cores - especially since many tasks will only utilize one core anyway.

Sure, Apple could potentially recode their own apps - Final Cut and Logic in particular - to really leverage 32 or more cores. They could also throw engineering resources at optimizing Metal 2 performance on AMD GPUs. For a particular sub-set of buyers such a machine would be most welcome. That said, it seems unlikely that Apple would be willing to do the heavy lifting involved for a niche of that size...

 

[Boil]

Did you mean x64 ? That's the common term for 64-bit Intel - unless you're doing assembly or low level C in Linux.

Yes, I did mean x64...

The MBP spec bump announcement ahead of WWDC, instead of at it, might indicate they want to preserve some time/mindshare for mMP reveal... Or not.

Assuming they can get at least a couple of the new Navi cards by late December they could still technically keep their 2019 ship date, even though waiting for Navis in quantity could push deliveries well into 2020. As others have noted, there are potential buyers who will hold out for the mMP as long as they have some idea what it will do, what it will cost and when they can get it - which makes saying nothing problematic.

Does anyone think that Apple and AMD might have been working on a custom form factor version of the Navi architecture for quite a while now? Could that project have been placed ahead of getting "regular" Navi cards out the door? If Apple was willing to pay a premium and promise to buy x number of units over the next x months would AMD say no? What if Apple was also willing to dump Intel and go with AMD for CPUs...

If Apple does switch horses, an mMP announcement at WWDC makes even more sense as they would want to get people coding for the new ecosystem ASAP.

At the end of the day, even a well designed and reasonably priced mMP on Intel might seem like an overpriced version of similar systems on Windows. OTOH, by going full AMD and promoting the apps that optimize for more available cores, marketing can cherrypick metrics that make the OSX/AMD option look pretty good...

I really think Navi is going to be shipping in Q3 of this year...

AMD very well may have been working on a custom Navi for Apple, much as they have been doing for Sony & the PS5; but AMD is in the custom silicon business...

I'm glad that you mentioned "cherry pick" - because that's the only way that you'll find any apps that use more than 8 to 12 cores on a workstation.

Those extra cores do not have to go to one app, they can also assist in speeding workflow when using multiple apps at once...

Since Apple can't count on developers making the effort to really exploit the potential advantages of a high core count CPU, they will likely stay with Intel's better cores - especially since many tasks will only utilize one core anyway.

Sure, Apple could potentially recode their own apps - Final Cut and Logic in particular - to really leverage 32 or more cores. They could also throw engineering resources at optimizing Metal 2 performance on AMD GPUs. For a particular sub-set of buyers such a machine would be most welcome. That said, it seems unlikely that Apple would be willing to do the heavy lifting involved for a niche of that size...

I dunno, having a personal desktop workstation that blows anything else away in FCPX & LPX can only be good for both end users & shareholders alike...?

Less than two weeks & hopefully we will know more about a possible modular Mac Pro, Ryzen 3000-series CPUs & Navi GPUs...!

 

[Biped]

Since Apple can't count on developers making the effort to really exploit the potential advantages of a high core count CPU, they will likely stay with Intel's better cores - especially since many tasks will only utilize one core anyway.

Sure, Apple could potentially recode their own apps - Final Cut and Logic in particular - to really leverage 32 or more cores. They could also throw engineering resources at optimizing Metal 2 performance on AMD GPUs. For a particular sub-set of buyers such a machine would be most welcome. That said, it seems unlikely that Apple would be willing to do the heavy lifting involved for a niche of that size...

I read something interesting a while back, about hw/sw packages that would 'smart schedule' processes over multiple cores... even if the software wasn't compiled to take advantage of multiple cores. I think the mutterings were about weaving it into the hardware abstraction layers of next operating systems. I vaguely remember it still being beholden to amdahl's laws and other constraints; the performance scaling wasn't linear.

Modest core counts at commodity level is still a relatively new thing. Maybe the opportunity looks more appealing for folks to develop around it now; especially since clock speeds have more or less have hit a plateau, and cute tricks like predictive branching have shown to be significant security concerns, on a nearly weekly basis.

 

[mattspace]

I do wonder a bit if a stackable design, or something like it, is Apple's way of dealing with that.

They could have "consumer" brain and "pro" brain modules. i7/i9/Xeon options and eventually ARM. That would take care of the XMac crowd and cover a very wide range of price points.

All just speculation. But while we're here trying to figure out how Apple is going to push into the super high end, maybe they're going the opposite direction and trying to build a pro machine which can also dip into the mid end.

See when I hear "stackable", I've never thought "every single component in a separate module" - ie every GPU in a proprietary individual box, I'm thinking the existing cMP PCI backplane, IO ports, power supply, and processor boards in separate boxes, using the same sort of megaconnector that attaches the CPUs to the backplane.

Further, I could easily see a version that has all IO on PCI cards (including something like the oldschool "personality card", which had several IO bundled together), and you just configure it with a "lane budget" if you don't want a preconfigured option. It's a small volume product, Apple can afford to run it as a more complex BTO-Only product with several options for PCI expansion numbers, or make parts like that a third party opportunity.

They tried the most integrated, most-appliance-like workstation, and it was a failure. Making the most componentised, most dis-integrated workstation, that strikes me as "completely rethinking the mac pro" as an actual change in direction, and I don't think we've really seen any Macs yet that weren't a (preplanned) part of the 2013 philosophy.

 

[Boil]

I really, Really, REALLY hope Apple does not actually go the 'Lego-esque' modular approach...

Just think of the extra cost for those proprietary connectors & separate module enclosures & power distribution...

I would really rather buy a 'done right' Apple Mac Pro than assemble a Hackintosh, knowing I would (probably) pay less for the Hackintosh, but would not have to worry about software updates breaking a build...?

 

[danwells]

Nobody actually wants more cores (at least outside of some highly granular tasks in the server market).

Intel would rather build a 7 GHz 8-core CPU with a 20% improvement in instructions per clock.
Apple would rather build a Mac Pro with such a CPU - HP would love to throw it in a Z8 at whatever cost.
Adobe and others would rather code for that CPU.

End users would beat a path to the door of whomever introduced that CPU in a computer - even if the CPU alone cost $3000, and it was in a $7000-$10,000 system (there are plenty of applications where paying a 6x premium on the CPU that translates to 2x the price on the entire system is VERY worth it for 80% more performance per thread).

The problem is that Moore's Law is failing and that CPU doesn't exist. Somewhere not very far in front of us, physicist Werner Heisenberg is standing in front of all the engineers with a stop sign in his hand. Professor Heisenberg is saying "aha, zee Uncertainty Principle izz now proven - you cannot go faster, or you will not read the result".

Building a CPU that is much faster per thread seems to require some tricky work on the border of physics and engineering - and I'm not sure it isn't pretty much pure physics, since the engineers don't seem to be getting anywhere fast.

All they can give us is more of the same kind of cores - the fastest standard CPU per thread in the world is already in a Mac (I'm not sure whether some supercomputer has a faster chip with exotic cooling) - it's the i9-9900K in the top non-pro iMac. Some of the HEDT chips and Xeons offer more memory and I/O bandwidth without losing much clock speed, and are faster at I/O intensive tasks - but pure grunt per thread goes to the i9. It goes faster in some PCs than it does in the iMac, because of better cooling (somewhat better cooling means it doesn't throttle, while much better cooling means it can be overclocked). Any AMD chip is significantly slower per thread - they can have a lot of cores, though.

Can anybody point to an ARM chip with anything like 5 GHz and the kind of instructions per clock the i9-9900K offers?

We may not have hit the limit, but we're very close to it, and we should expect that progress close to it will be painfully slow. We're not going to get something that's a bunch faster and reasonable to code for - we're stuck with highly parallel CPUs and (worse yet) specialized processing in GPUs and the like.

I suspect that not only do we never see a CPU with a +40% clock speed and +20% efficiency per clock, we may never even see a CPU with +20% clock speed (~6 GHz) and +10% instructions per clock.

 

[Boil]

...the fastest standard CPU per thread in the world is already in a Mac (I'm not sure whether some supercomputer has a faster chip with exotic cooling) - it's the i9-9900K in the top non-pro iMac. Some of the HEDT chips and Xeons offer more memory and I/O bandwidth without losing much clock speed, and are faster at I/O intensive tasks - but pure grunt per thread goes to the i9. It goes faster in some PCs than it does in the iMac, because of better cooling (somewhat better cooling means it doesn't throttle, while much better cooling means it can be overclocked). Any AMD chip is significantly slower per thread - they can have a lot of cores, though.

AMD may very well beat the i9-9900K with their forthcoming Zen 2 Ryzen 3000-series CPUs...

We may not have hit the limit, but we're very close to it, and we should expect that progress close to it will be painfully slow. We're not going to get something that's a bunch faster and reasonable to code for - we're stuck with highly parallel CPUs and (worse yet) specialized processing in GPUs and the like.

If we are hitting the core clock limit, then multi-core / multi-threading is the way forward; which means maybe moving to AMD would make sense...?