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

[Mago]

Apple is not going to hurt it's relationship with Intel so any dream of AMD CPU is just that. A dream.

Apple don't cares to hurt Intel, just killed Intel's cellular modem business and kicked Intel from mobile market.

Intrl only strength is the latest core i5/i7 IPC way ahead AMD and hard to reach by any arm CPU, simple not even arm cortex A77 (available for consumer in 2020< is close to be competitive with AMD in single Thread, much less with Intel now upto 30% faster with it's 10th gen core CPU.

Apple would switch to AMD on better TDP, PCIe lines and PCIe 4 but for raw single Thread performance (the one most people actually sees) Intel even with delays hot shortage, is still the king of the hill .
[doublepost=1559211385][/doublepost]Apple could build a Mac pro classic based in the trashcan with full AMD internals (ryzen 3, x570, dual Vegas), and sell 3u Rack/desktop Mac pro based on some server MB with apple EFI and wireless/TB3 via aic, sell it as the Mac pro Lab for those willing to pay 20k+ $ for dual socket 4 nVidia GPU etc.

 

[Boil]

...but for raw single Thread performance (the one most people actually sees) Intel even with delays hot shortage, is still the king of the hill.

Hard to really say that when there are brand new Ryzen 3000-series CPUs that have yet to be benchmarked, and yet to be compared to the latest Intel paper launch...

Apple could build a Mac pro classic based in the trashcan with full AMD internals (ryzen 3, x570, dual Vegas), and sell 3u Rack/desktop Mac pro based on some server MB with apple EFI and wireless/TB3 via aic, sell it as the Mac pro Lab for those willing to pay 20k+ $ for dual socket 4 nVidia GPU etc.

That would satisfy a vast majority of the prosumer class of "Pros", while also satisfying the demands of the ML / AI / CUDA crowd...

 

[apolloa]

I did hear a rumour, on some guys You Tube video, that the Mac Pro will cost from 15 grand UK Sterling... I did lol as that would price the majority of its customers out of the market! That’s around 18,900 dollars US, if you were charged 20% tax.

Some crazy speculation out there..

 

[Simon R.]

I did hear a rumour, on some guys You Tube video, that the Mac Pro will cost from 15 grand UK Sterling... I did lol as that would price the majority of its customers out of the market! That’s around 18,900 dollars US, if you were charged 20% tax.

Some crazy speculation out there..

I heard a rumour that it will only work during full moon and when the wind is blowing from the north east.

Point being that all of those "rumours" that we have read about is nothing but speculation that is made up by John Does. And then because it's the internet some of that speculation spreads and goes from "something someone who knows nothing concrete mad up" to "rumour" for no good reason. Nobody knows ANYTHING besides some people at Apple. I mean, HOPEFULLY Apple knows enough to publish concrete information come WWDC.

 

[Manzanito]

Apple haven't been ahead of the curve regarding computers since 2006. Totally unlikely.

Yes. It hurts to remember the passion Steve and Phil showed while introducing the first mac pro. Or was it the power mac? The one with the comparisson between the mac and a dell workstation running photoshop.

 

[Simon R.]

Yes. It hurts to remember the passion Steve and Phil showed while introducing the first mac pro. Or was it the power mac? The one with the comparisson between the mac and a dell workstation running photoshop.

The 2006 Mac Pro was ahead of the curve, top of the line. They dont have to/need to do that again to keep the pro/prosumers with Apple though, just stay current, in line with other desktop PC manufacturers.

Apple should eat their pride and stop focusing so much on form with the Mac Pro, and realize they would gain far more in long term sales and goodwill from the pros/prosumers by providing an easy-to-upgrade, standard component machine instead of trying to lock down the hardware upgrades and force people to buy new machines all the time. That just doesn't work for this customer group. They should realize they have far more to gain from a more open platform (like the cMP), despite not being able to cash in on that customer group every 2-3 years. They can afford to break even or even lose a little money on the MP to keep pro customers in the loop.

Apple brought this on themselves. Had they kept the MP up to date and upgradeable, the user base would be much bigger today and they would not have even think about this being a money losing platform. You snooze, you lose.

 

[jscipione]

That is why Apple will use Cascade Lake Xeon W and not SP.

"A recent Intel client roadmap indicates that the new Cascade Lake W parts should land in the third quarter of the year." ding ding ding.

 

[Macintosh IIcx]

"A recent Intel client roadmap indicates that the new Cascade Lake W parts should land in the third quarter of the year." ding ding ding.

Yep

 

[now i see it]

My prediction:

Xeons
Latest I.O.
Plenty of RAM potential (easily added)
Larger enclosure (single enclosure)
Room to add extra drive/s
Room for big graphics card
Headroom for future updates by APPLE
T2 chip
Looks nice
Not too noisy

That's it! Bet I'm not wrong.

 

[deconstruct60]

Hard to really say that when there are brand new Ryzen 3000-series CPUs that have yet to be benchmarked, and yet to be compared to the latest Intel paper launch...

So caught up in AMD fanboy fever that have forgotten that AMD has already benchmarked the 3000 series? Every single tech press article write up I've seen that wasn't pure superficial spin has included the results from the AMD benchmarks.

https://www.anandtech.com/show/1440...-cores-for-499-up-to-46-ghz-pcie-40-coming-77

What is missing is independent and wider benchmark set of the 3000 series; not all benchmarks. AMD probably spent months putting together theirs. It is highly unlikely that the 3rd party ones are going to turn in better AMD results than what AMD got itself. ( perhaps some compiler optimization changes come out for AMD later that move the results slightly but ball-park these are pretty close what will see on similar workloads. ). Most likely AMD is going to be a little lower on several broader benchmarks and in the 3rd party testing.


Single threaded AMD has closed the gap. Ice Lake ( Sunny Cove microarch ) puts a bigger gap between them ( Intel touts double digit gains but have limited clock bump available and MDS/Meltdown/etc overhead that is still probably still around and/or part of that speed bump. ). However, for Apple that gap ( if relatively narrow +/- 2-3% ) probably isn't as important as steady, predictable progress. Apple doing "round robin" Mac system upgrades means they need future parts to be ready when roadmaps say they'll be ready. That way Apple can finish up product X and move on to product Y. Otherwise screw up more than one products schedule.

Won't be able to do "head to head" marks for the next 6-18 months anyway because Intel is going "small mobile" first and AMD "desktop GPU-less" first. Their "new" microarch implementations aren't going to overlap in being deployed for more than a few months.

So if Intel got back on the steady execution track again, any short term benchmark gap that AMD had they'd probably pass on. Apple isn't selecting parts 100% solely on benchmarks (e.g., Nvidia) .

That would satisfy a vast majority of the prosumer class of "Pros", while also satisfying the demands of the ML / AI / CUDA crowd...

It is probably going to be hard enough to do one additional Mac model, let alone two more. If Apple building everything for everybody, they would make every " make exactly what I want or we'll storm and burn the castle down" group happy. Apple isn't going to do that. The explicitly have said they aren't going to do it. A limited number of Macs to cover has many folks as they can ( but not everybody).

Unless Ryzen 3 is going into the iMac isn't probably isn't coming to the Mac line up. Period.
[doublepost=1559234803][/doublepost]

"A recent Intel client roadmap indicates that the new Cascade Lake W parts should land in the third quarter of the year." ding ding ding.

The second "half" of that article that points at the SIPP ( stable image platforms program) is substantively flawed. I'm not sure why they keep trotting that out as some prominent doom and gloom roadmap. it is for "stable" ( long term platforms for embedded or extended lifecycle business stuff).

It is the same first one from this article.
https://www.tomshardware.com/news/intel-roadmap-10nm-14nm-gpu-cpu,39163.html

the more relevant graph is the second one there. the "Stable Image' one trails because it doesn't stat until "stable". It isn't Intel suddenly flip flopping their roadmaps.

They are on track though that it is missing Intel's 10nm move. Probably about the same time they make the 2nd Cascade Lake-W move with Apache Pass added. The 10nm Xeon W will probably be a diffrent socket. Those who who want a boost on Memory capacity will go for the Apache Pass. Those looking for IPC bump (and maybe clock bump) will probably go with the Ice Lake move. ( the Ice lake move may not move the core count much at all. Unless go double dies in W line up. )

 

[filmak]

My prediction:

Xeons
Latest I.O.
Plenty of RAM potential (easily added)
Larger enclosure (single enclosure)
Room to add extra drive/s
Room for big graphics card
Headroom for future updates by APPLE
T2 chip
Looks nice
Not too noisy

That's it! Bet I'm not wrong.

Now you see it!

 

[deconstruct60]

"We have a team working hard on [the modular Mac Pro] right now, and we want to architect it so that we can keep it fresh with regular improvements, and we’re committed to making it our highest-end, high throughput desktop system, designed for our demanding pro customers."

That makes it sound like the Mac Pro will not be low-end - it will either utilize Threadripper or Xeon SP, not Ryzen. Then again Phil "my ass" Schiller said it, so who knows.

Your are coaching "throughput" to mean some parallel CPU computations per second. From a overall system perspective "throughput" can ( and should mean) moving more data around the system through all the layers involved. Not just from L3 CPU cache zone, but from persistent storage , through computation and back to results being stored.

That is something that isn't purely x86 core count based. I know x86 count count wars are a favorite topic around here but that isn't Throughput. If running up against stuff like Amdahls law there are more things that exist than just x86 cores.

There is zero need to go to Xeon SP parts when the iMac Pro leaves x20 lanes of bandwidth on the floor. Nor is some 2x increase count in AMD PCI-e lanes going to lead to a 2x increase in bandwidth much if they are provisioned from a internal switch. ( once the infinite fabric is saturated you can add PCI-e lanes on the outside until the cows come home... the overall system bisection bandwidth isn't going to go up. )

What Apple is missing if look at the rest of the Mac line up is more so bandwidth; not core count. If can't keep the cores feed than not going to get maximum throughput.

Ryzen 9 3000 with two Memory channels trying to keep 12-16 cores feed versus Threadripper 2 (or 3) with four memory channels trying to keep 12-16 cores feed, the latter has higher max throughput. It isn't the core count, but the core count relative to the rest of the system.

Does that means Apple "needs" Xeon SP and AMD Epyc for the Mac Pro? No. They just need something that is faster than the rest of the stuff they are using in the other parts of the line up. That's it ... completely in compilance with Schiller's statement. Could they build an even bigger monster? Sure. They "could" lower their prices by 5-10% too.
Given the latter, it doesn't make much sense to push the Mac Pro entry sky high. If the sales drop far enough below 1% of the Macs sold they will probably be more inclined to kill it than keep it.

 

[beaker7]

As long as it has AltiVec

 

[ssgbryan]

You're probably right but there's definitely a market for a cheaper MacPro using the X570 motherboard,

Most content creators don't need over 64Gb of RAM or more than 8 or 12 Cores, PCIe 4 gives us more graphics power

So, what content do you create? Please see sig.

I create 3d Art (as an amateur) and in that field, you need every byte of ram and every core you can throw at at your software - that includes all of the hobbyist level software, btw. My render engines are CPU based, and with my 12 core 4,1, I saw a major improvement in render times over my 8 core 4,1.

I like having a 32Gb ram disk to throw everything into before I start to work - makes things faster. Not to mention that I normally have 7 or 8 programs all open at the same time and switching between the apps as needed.

I understand that the Apple folks were the last to the party afa preemptive multitasking, but some of us have had workflows that take advantage of that since the 386 chip rolled out the door (an AMD 386DX40 in my case).

It isn't just about today, however - If you haven't noticed, the tcmp is about to hit 2,000 days. If one stays with OSX, you have to plan on the upcoming mp will be your computer for the next 5+ years.

You don't know what you will need 24 months from now, much less 5 years from now.

 

[kiiso]

For GPUs maybe, but for storage, how does 15GB/s read/writes sound? That would be from a RAID card that holds four PCIe 4.0 x4 NVMe M.2 SSDs & takes full advantage of a PCIe 4.0 x16 slot

Most M.2 NVMe SSD’s use PCIe 3.0 4x
so there is plenty of room to improve without PCIe 4.0.

Apple will use for sure a proprietary connector and they could utilise 8x or even 16x.

 

[partsofspeech]

What will they be in WWDC 2019?

 

[deconstruct60]

I could see a Ryzen in a low end Mac Pro, most audio producers and video editors don't need more than 12 core machines (or 8 core -16 core), the only problem would be the cannibalisation of the Mac mini and iMac Pro. Epyc and Threadripper for the high End

It isn't purely only the lowest end entry model nor the higher end BTO option that is going to be primary driver of the Mac Pro. Since Apple only does a lmited number of Mac Products they need something with a range.

They need something bigger than 16 and but not the biggest possible number imaginable ( by throwing more and more CPU packages at it). The lower bound can't be too high as to drag the entry price too high.

If they went AMD they could go TR2 ( or TR3 if wait until end of the year) they could start at 12. TR2 12 core is ~$580. It wouldn't a big difference in Bill of Material (BOM) costs to the Ryzen 9 3000 series. Better overall system throughput. And since that is the "bottom" of the core count spectrum more head room for core count increase at the top of the BTO.

Similar with regular Xeon W. They can start at 8 and scale up to 18. Intel needs to revisit their pricing on those though, because they have real problems. AMD has two lines ups that beat them in the 12-16 range. That's is going to start to bleed if they treat Xeon W pricing as dogma.

It isn't just core count. A substantive number of video folks are probably going to want more than just one internal drive. Read source from disk system 1 record out to disk system 2. The princess and the pea latency audio folks ... same thing.

It also really doesn't help to pretend that the iMac (and iMac Pro) don't exist. The Ryzen 5/9 3000 line up best fit is with the iMac. Yes, they'll get xMac pressure, but that isn't new. But it also isn't the relative "Mac Pro" space either.

 

[Macintosh IIcx]

My prediction:

Xeons
Latest I.O.
Plenty of RAM potential (easily added)
Larger enclosure (single enclosure)
Room to add extra drive/s
Room for big graphics card
Headroom for future updates by APPLE
T2 chip
Looks nice
Not too noisy

That's it! Bet I'm not wrong.

My prediction:

A fairly powerful CPU
RAM
Motherboard with I/O
A fairly powerful GPU
Thunderbolt
A computer case of some sort
Looks nice
Not too noisy

Bet I'm not wrong!

 

[deconstruct60]

t for storage, how does 15GB/s read/writes sound? That would be from a RAID card that holds four PCIe 4.0 x4 NVMe M.2 SSDs & takes full advantage of a PCIe 4.0 x16 slot

Most M.2 NVMe SSD’s use PCIe 3.0 4x
so there is plenty of room to improve without PCIe 4.0.

M.2 is a bit of the limiting factor (number of NAND packages is limited and 3D stacking has a shared bus along 3d stack), but there isn't much of a gap left on PCIe v3 x4.

PCIe v3 x4 is 32Gb/s or 3.98 GB/s or ~3980MB/s

If trying to get to 4-5GB/s then x4 isn't going to cut it on v3 whereas on v4 it would ( since about a 2x increase).

But if go to double sided M.2 blades and double the number of dies the SSD controller can talk to then busting past 4.0Gb/s is quite doable.

Most M.2 SSD controllers to date haven't wanted to go to double sided blades ( many boards can't really deal with those well) or even be confident really going to get x4 worth of bandwidth. ( x4 v3 hanging of a switch and/or PCH then not really getting x4 in the first place if another SSD or something substantive on the same feed. )

Apple will use for sure a proprietary connector and they could utilise 8x or even 16x.

For an SSD? Highly likely not. Apple has commingled the SSD controller with the Power management in the T-series. That means they have a pretty descent need to be attached to the primary PCH (AMD's or Intel's).

Going "wider" isn't really an option. First that isn't going to scale across all of the Macs. Second, the PCH doesn't scale that way.

Apple use some 3rd party SSD controller attached to the CPU? Again not that likely at all. Not going to scale scale across Macs and not going to build some "Mac Pro only" SSD controller.

Provisioning some M.2 slots ( that users fill in with 3rd Party SSD and controller) or a x8-x16 slot basically could just solve the issue with limited work from Apple.

Apple going to some x4 PCI-e v4 could be on the roadmap for the Macs but if looking at more NAND chips then diverging away from the iOS devices. The larger the divergence means a larger impediment to getting it into the T-series. I wouldn't bet on that.

 

[H2SO4]

I heard a rumour that it will only work during full moon and when the wind is blowing from the north east.

Point being that all of those "rumours" that we have read about is nothing but speculation that is made up by John Does. And then because it's the internet some of that speculation spreads and goes from "something someone who knows nothing concrete mad up" to "rumour" for no good reason. Nobody knows ANYTHING besides some people at Apple. I mean, HOPEFULLY Apple knows enough to publish concrete information come WWDC.

What you mean is nobody can CONFIRM anything except Apple.
There are likely people in the know that have spilled their guts.
People talk, this is not the military.

 

[Manzanito]

My prediction:

A fairly powerful CPU
RAM
Motherboard with I/O
A fairly powerful GPU
Thunderbolt
A computer case of some sort
Looks nice
Not too noisy

Bet I'm not wrong!

You’re stretching your luck on on the gpu side .

 

[CWallace]

The nasty issue with Intel's Xeon “Cascade Lake SP” CPUs is that their Turbo boost kinda sucks for single/few Cores load work. Not great when the iMac next to you can do work 20-25% faster on single Core load which a lot of CPU tasks still are.

I would hazard a guess the Mac Pro (like the iMac Pro) would be aimed at workloads that are naturally highly-paralleled to benefit from multiple cores. The tasks that are not would be "light" enough that the slower single-core speed would be sufficient.

As such, if your workload was more optimized for single core, you would be using an iMac or MacBook Pro with the top-end i5/i7/i9.

 

[OS6-OSX]

How does the saying go ? If you make a mistake, do it again ... then call it jazz.

Since you did not make up this "saying", this reply is to the premise of that nonsensical saying. Not at you or you quoting it!

Dear sir or madam, this is an invitation for you to go on stage, make a couple of mistakes and call it Jazz. This of course would be done with the instrument of your choice.
1. If it's the Trumpet, hit the high notes of Cat Anderson and Maynard Ferguson
2. If it's the Saxophone, play through the changes like Charlie Parker and John Coltrane. Or make the Baritone sax the melodic tool like Gerry Mulligan was able to do.
3. If it's the Clarinet, play it flawlessly like Benny Goodman
4. If it's the Piano, "tickle those ivory 88's" like Bud Powell, Oscar Peterson, Keith Jarret, Dave Brubeck etc
5. If it's the Bass, walk like Ron Carter, Ray Brown, Charles Mingus etc
6. If it's the Drums, Tony Williams, Gene Krupa, Billy Cobham, Philly Jo Jones etc will be the ones you are compared to

We will place you on stage with players of this caliber and if you perform on their level
Then call it Jazz!

To make it fit into this thread, of course there will be PCIE slots on the mMp to place a "real" audio card. You can then attach your Neve pre amp and Telefunken for recording.

 

[Biped]

Since you did not make up this "saying", this reply is to the premise of that nonsensical saying. Not at you or you quoting it!

Dear sir or madam, this is an invitation for you to go on stage, make a couple of mistakes and call it Jazz. This of course would be done with the instrument of your choice.
1. If it's the Trumpet, hit the high notes of Cat Anderson and Maynard Ferguson
2. If it's the Saxophone, play through the changes like Charlie Parker and John Coltrane. Or make the Baritone sax the melodic tool like Gerry Mulligan was able to do.
3. If it's the Clarinet, play it flawlessly like Benny Goodman
4. If it's the Piano, "tickle those ivory 88's" like Bud Powell, Oscar Peterson, Keith Jarret, Dave Brubeck etc
5. If it's the Bass, walk like Ron Carter, Ray Brown, Charles Mingus etc
6. If it's the Drums, Tony Williams, Gene Krupa, Billy Cobham, Philly Jo Jones etc will be the ones you are compared to

We will place you on stage with players of this caliber and if you perform on their level
Then call it Jazz!

To make it fit into this thread, of course there will be PCIE slots on the mMp to place a "real" audio card. You can then attach your Neve pre amp and Telefunken for recording.

Just quickly, I think this is called an association fallacy. Potentially an appeal to purity ...

But, Krupa <3 ! Although I'm more in the MMW camp than traditional, but much respect.

Its more a statement about spin, and to some degree why dissecting the statements and quotes is largely a wasteful exercise. Whatever product, whenever it will be released, and subject to whichever update/maintenance schedule will communicate enough to let a consumer make a fair decision.

 

[AidenShaw]

As long as it has AltiVec

You mean Velocity Engine® ?