Will there be a new Mac Pro in 2015?

[deconstruct60]

Apple likes surprises - the GM204 chips have good OpenCL performance and sip electricity. The nnMP could have OpenCL+CUDA!

Apple likes to surprise customers. Apple does not like suprises from their suppliers. It isn't symmetrical. Apple also doesn't like suppliers who don't get on board with their strategic objectives.

While AMD, Intel, Imagination Tech (PowerVR) , and just about everyone except Nvidia is rolling out OpenCL 2.0 GPUs for the late 2014-2015 era, Nvidia is kneecapping OpenCL.

Apple was OK when Nvidia was putting about as much effort into OpenCL as it was CUDA. Now that Nvidia is on the track of "OpenCL has to loose for CUDA to win" tactics, they are not given themselves a competitive edge with Apple. Even if Apple likes the lower power prospects, they aren't going to shoot OpenCL in the head to get to it.

There is an established set of OpenCL 1.1 benchmarks that Nvidia sales marketing will increasing play misdirection games with: "Oh look we're now faster on OpenCL from 5 years ago ... we are 100% on board".
It is just another phase to the lure folks into the proprietary tarpit that want to draw more customers into.

You're responding to someone who's been bad-mouthing the Haswell-EP for months.

The Mac Pro isn't just the Xeon processor ( or the GPU or the Thunderbolt controller(s)). it is a collection. Whether a new system is better than an older system is multidimensional.

The long term trend is that the Xeon E5 (and up ) models are going to be more design optimized for more cores at good speeds far more than fewer cores at the fastest speeds. Intel has kneedcapped the E5 1620 v2 and 1630 v3 a bit. 4 cores (with 4 switched off) and a TDP budget of 140W shouldn't be backsliding on top end Turbo speeds. That appears to be more lack of AMD competition and market segmentation from Intel's mainstream line up than necessary by the micro architecture. It isn't a Haswell design flaw... just marketing.

It does make Apple's justification of a new Mac Pro harder than simply graphing geekbench drag racing scores. But if that is the limit of their marketing abilities they are more problems than just limited Haswell-EP single threaded scores.

We have some integer-heavy apps that are seeing 20% improvements on our E5-v3 systems vs E5-v2. AVX2 brings "loop unrolling" into the hardware. It's huge.

It isn't exactly the same but these two are similar in relationship to each other AVX (new) vs SSE (old) x86-64 vs. x86-32 . The "loop unrolling is going to make it easier for compilers to write good-to-very-good numerical kernels for AVX2 (whereas AVX is far more limited and needs custom hand coding). Old code targeted at compilers from 4-5 years ago... sure no speed increases. Folks using modern tools will see something.

When things are made easier for software developers ( compiler and/or OS library ) does most of the work the uptake over time is usually pretty good.
[ there is a long software development track record to support that . ]

----------

Yes.

All I will say right now is that, what we see from Mantle right now is only being Low-Level API for gaming.

Future is long...

Sometimes the past is repeated in the future. It isn't like there wasn't a round of vendor specific low level APIs before. There some cyclic stuff going on here. Low-level APIs come in and kick the high level stuff in the shins on performance. There is an adjustment and the pendulum swings again. windows 10 and the DirectX updates seems to be answering relatively rapidly on the Windows side. OS X .... never was a sprinter when it comes to graphics. iOS .... actually would like more vendor lock-in so probably not in a hurry either.

AMD has much bigger problems that Mantle really doesn't resolve all that well. As an interim way point, it will probably be helpful. As a long term strategic solution..... it probably is not. The other low level folks have way more money to throw at the issue long term. Each of the console vendors are going to want their own. Windows is adjusting. AMD will have to track all of those closely if it wants to stay alive.

 

[deconstruct60]

That's true. I'm assuming TBv3 will be changed to use DP 1.3, because if it isn't, we really all are truly screwed.

TBv3 should allow the same "hack" that is internal to the Retina iMac to be employed. (bond two DP v1.2 connections at the display controller). It isn't as 'clean' as a DP v1.3 solution but it will work. In fact, the vast majority of set ups in 2015 and much into 2016 will probably work just that way.

TBv3 shouldn't be changed; at least short term. Going from 20Gb/s to 40Gb/s and getting all of the other USB Type C "Alternatve Mode" stuff to work is good enough. That is a significant amount of work without throwing more complexity on top. The other down side is that certifications will get more complicated for two port TB peripherals the more complicated the "mode" matrix is. The peripheral vendors don't want and largely don't need the additional complexity and cost.


There was a "Redwood Ridge" TB controllers before ( TB v1 with pass-thru DP 1.2 support ) the "Falcon Ridge" TB v2 controllers came. Apple by-passed those but Intel did them.

Probably same progression with TB v3 if Intel stretches out the transition from TB v3 to TB v4. [ If TB keeps churning every 2 years it will croak over time. At some point have to get to a point where have a "large" established base. ]


first iteration legacy "alternative modes".
second iteration more 'pass through' compatibility with newer protocol.
3rd iteration weave more newer protocols into newer, faster TB base layer.

It has been a two year cycle.

Like I said before 2015-2016 time frame Apple more likely can get a simpler more focused "railroad" switch that just moves between USB and DisplayPort to put behind a Type C connector.

I don't know exactly what Apple is planning with USB type C. I can't believe they'd abandon Thunderbolt.

USB Type C has Apple's fingerprints all over it. Symmetric plugs on both sides and reversible connector. One cable to power and output from a system (docking cable with just one head ... not some 2-3 headed hydra ). The chance to spin some "one port to rule them all " spin. (if not simplified supply chain of just one connector type) ..... all very Appleesque.


I don't think they are going to abandon Thunderbolt but they can certainly do stop gap for a year with a socket that just does two instead of three things. The rumored MBA 12" is probably a temporary backslide. [ similar to the aluminum Macbook that got stripped of Firewire and then next iteration it came back. ]

[qoute] Maybe another option is that Thunderbolt gets some sort of bypass that when you plug in a DP display directly, it bypasses the Thunderbolt controller. [/quote]

Effectively, that is how it works now. There is some physical coupling of the TB involved but the TB specific guts are essentially by passed.

I'm pretty sure what will happen is TBv3 will get support for DP 1.3 though.

Over time will there be a TBv3 controller that works with it. Sure. The first one on the market? Probably not.

 

[koyoot]

Sometimes the past is repeated in the future. It isn't like there wasn't a round of vendor specific low level APIs before. There some cyclic stuff going on here. Low-level APIs come in and kick the high level stuff in the shins on performance. There is an adjustment and the pendulum swings again. windows 10 and the DirectX updates seems to be answering relatively rapidly on the Windows side. OS X .... never was a sprinter when it comes to graphics. iOS .... actually would like more vendor lock-in so probably not in a hurry either.

AMD has much bigger problems that Mantle really doesn't resolve all that well. As an interim way point, it will probably be helpful. As a long term strategic solution..... it probably is not. The other low level folks have way more money to throw at the issue long term. Each of the console vendors are going to want their own. Windows is adjusting. AMD will have to track all of those closely if it wants to stay alive.

Let me put this to you in this way.

Nvidia does not support Mantle because in future it will be rival for CUDA.

Everybody supports Mantle: Intel, Apple, Imagination, ARM. Everybody started working on Mantle. Everybody, apart from Nvidia.
Not because AMD doesn't let them.

 

[goMac]

TBv3 should allow the same "hack" that is internal to the Retina iMac to be employed. (bond two DP v1.2 connections at the display controller). It isn't as 'clean' as a DP v1.3 solution but it will work. In fact, the vast majority of set ups in 2015 and much into 2016 will probably work just that way.

OOoooo. That's a good one. I hadn't thought about that. Even if DP1.3 doesn't get adopted by TB3, that should at least let a 5k Cinema Display get out the door.

 

[MacProCard]

The updates will be driven by economics. Plain and simple. Or at least that's what their shareholders expect.

 

[Umbongo]

The updates will be driven by economics. Plain and simple. Or at least that's what their shareholders expect.

Of course, but that should mean that a new Mac Pro is imminent.

The idea that Apple designed a new form factor and built a brand new plant and process to produce it, but can't switch to the next generation of hardware because it isn't economically viable is ludicrous to me.

For clarity I am not inferring you are suggesting that is the situation, but others have.

 

[MacProCard]

Of course, but that should mean that a new Mac Pro is imminent.

The idea that Apple designed a new form factor and built a brand new plant and process to produce it, but can't switch to the next generation of hardware because it isn't economically viable is ludicrous to me.

For clarity I am not inferring you are suggesting that is the situation, but others have.

Only in time.

They can but why? You and a few others are unhappy. Get real. It's about numbers my friend. Nothing more than that. If it's not about the current numbers, it's about the future numbers.

 

[AidenShaw]

Only in time.

They can but why? You and a few others are unhappy. Get real. It's about numbers my friend. Nothing more than that. If it's not about the current numbers, it's about the future numbers.

Exactly.

And if Apple lets the MP6,1 wither on the vine even though new generations of CPUs and GPUs are available - the future numbers will suffer. People will see the MP6,1 as a half-assed end-of-line "pretty" system.

Do you realize how much credibility Apple would regain in the professional space if they offered the MP7,1 systems with Haswell-EP CPUs and GM204-based GTX960/970/980 graphics? It would be huge. CUDA plus often better OpenCL performance than ATI can sell.

"Wait for whatever Mantle eventually becomes" won't be a consideration.

 

[deconstruct60]

Let me put this to you in this way.

Nvidia does not support Mantle because in future it will be rival for CUDA.

Mantle competes with DirectX/Direct3D , OpenGL , and console box's GPU APIs . For example, Anandtech's evaluation of Mantle one of the benchmarks was the below:

http://www.anandtech.com/show/7868/evaluating-amds-trueaudio-and-mantle-thief
The two items compared are Direct3D and Mantle. Direct3D isn't a rival of CUDA either.


Even AMD's slides align it up as a graphics rendering pipeline, not a general purpose computational one.

http://www.anandtech.com/show/7728/battlefield-4-mantle-preview

Nvidia has no low level graphics pipeline leveraged from optimization work on next generation consoles because Nvidia is in no next generation consoles ( at least at the volume levels of three largest players. )

AMD on the other hand
"... As the supplier of the APUs in both the Xbox One and PS4, AMD is in a very interesting place. Both of these upcoming consoles are based on their GCN technology, and as such AMD owns a great deal of responsibility in developing both of these consoles. ...
...
This brings us to the crux of the matter: what’s not being said. Simply put, what would happen if you ported both the high level and low level APIs from a console – say the Xbox One – back over to the PC?
... "
http://www.anandtech.com/show/7371/understanding-amds-mantle-a-lowlevel-graphics-api-for-gcn

Everybody supports Mantle: Intel, Apple, Imagination, ARM. Everybody started working on Mantle.

Looking at Mantle or working on Mantle? Apple has "Metal". Intel will probably look at it as Mantle has/had major clues as to what Microsoft was going to do long term with DirectX/Direct3D. [ MS can move Xbox One APIs back to Windows too .... since they control both it won't be a Manhattan Project scale hard problem to address. ]

Low-level APIs are a dual edge sword. Mantle is hooked to AMD's GCN specifics. So the overhead is low, but the portability is also low.

Mantle may expand a bit over time on computational shaders, but that isn't the space that CUDA really plays in. Just because have "raw" access to GPU doesn't mean getting a clean GPGPU API to work with.

If Mantle is expanded to fit the different styles and architectures of 3-4 different GPU vendors then it probably would stop being a low-level API. Unclear weather clean up and adjustments by DirectX and OpenGL largely get to roughly the same point.

Everybody, apart from Nvidia.
Not because AMD doesn't let them.

Nvidia doesn't mess with Mantle largely because they are the dominate Windows GPU gaming vendor. If MS does all the hard work of upgrading DirectX/Direct3D then Nvidia will largely remain the dominate Windows GPU gaming vendor. Similar on Android, Nvidia will likely ride OpenGL and the the OS GPU API. Nvidia instead is plowing more money into a wider set of GPU implementations per generation over a broader set of platforms.


In contrast, AMD needs to grow share. They are hoping that the insights they have from their console wins into an incrementally larger gaming share on Windows PC. That will create a nice feedback loop where Mantle work will feed back into better chance to win follow on console design competitions (i.e., better honed experience writing low level drivers focused on games between console design competitions ). Rise and repeat (as long as MS Xbox and Windows are slightly different graphics stacks).

AMD's compete against CUDA is their Heterogenous System Architecture http://www.hsafoundation.com/ (Nvidia is missing there; no surprise. Intel is missing; somewhat no surprise. Apple is missing but given Intel is missing; no surprise) . A component of that is OpenCL.

 

[MacProCard]

Exactly.

And if Apple lets the MP6,1 wither on the vine even though new generations of CPUs and GPUs are available - the future numbers will suffer. People will see the MP6,1 as a half-assed end-of-line "pretty" system.

Do you realize how much credibility Apple would regain in the professional space if they offered the MP7,1 systems with Haswell-EP CPUs and GM204-based GTX960/970/980 graphics? It would be huge. CUDA plus often better OpenCL performance than ATI can sell.

"Wait for whatever Mantle eventually becomes" won't be a consideration.

Apple is a mass producer. That means they have to find the largest market possible for the product. Or they lose money. Or the product gets too expensive.

While you might wish the nMP had everything and then some...it's not really possible for a company the size of Apple. And if you feel they once targeted say the top 5% power users, I propose they've let that notion go and are actually targeting the top 30% now. You as a consumer should want that. You as a techie maybe not so much.

 

[koyoot]

cut...

I know what Mantle is.

Everything what I can say right now about it to add something is that AMD is in the same place, as was Nvidia when they were bringing CUDA to market.

Only the difference is that AMD is doing it better. Way better.

Lets end this off top, shall we?

 

[handsome pete]

Apple is a mass producer. That means they have to find the largest market possible for the product. Or they lose money. Or the product gets too expensive.

While you might wish the nMP had everything and then some...it's not really possible for a company the size of Apple.

So wait, they're not able to compete in the same market as Dell, HP, etc.?

And if you feel they once targeted say the top 5% power users, I propose they've let that notion go and are actually targeting the top 30% now. You as a consumer should want that. You as a techie maybe not so much.

That doesn't make sense. Opening up a product line to more mass appeal can still alienate that top 5%. So they're not necessarily included in that top 30% figure. And in that circumstance, you don't want that as a consumer, because they no longer offer a product you want.

 

[VirtualRain]

Apple is a mass producer. That means they have to find the largest market possible for the product. Or they lose money. Or the product gets too expensive.

While you might wish the nMP had everything and then some...it's not really possible for a company the size of Apple. And if you feel they once targeted say the top 5% power users, I propose they've let that notion go and are actually targeting the top 30% now. You as a consumer should want that. You as a techie maybe not so much.

What are you talking about?

 

[goMac]

Apple is a mass producer. That means they have to find the largest market possible for the product. Or they lose money. Or the product gets too expensive.

While you might wish the nMP had everything and then some...it's not really possible for a company the size of Apple. And if you feel they once targeted say the top 5% power users, I propose they've let that notion go and are actually targeting the top 30% now. You as a consumer should want that. You as a techie maybe not so much.

Mass producers make things for niche markets all the time. I don't disagree that times have changed at Apple. But I don't think "Apple is a mass producer" is alone enough to convince anyone that the Mac Pro is being abandon.

 

[MacProCard]

Mass producers make things for niche markets all the time. I don't disagree that times have changed at Apple. But I don't think "Apple is a mass producer" is alone enough to convince anyone that the Mac Pro is being abandon.

All I meant by that was Apple is bound to the same constraints as all other corporations their size. The entire leadership team can be ousted by the shareholders. Who only want to make as much money as possible now.

I never said abandon it. They just invested money in it. I've said expand the market. Might be even redefining it to some extent as well. That's why they needed a good introduction/splash. Couldn't be the same old box. Had to be a silent cylinder that is portable. Take it over to your buddies house and process your video.

So as long as the item is selling well, there's no need to really update it to keep up with a Dell. That is until Dell starts making something similar.

----------

What are you talking about?

Here we go again with the Apple is a rebel corporation and plays by their own rules?

 

[AidenShaw]

I remember when Apple did lots of things for the "halo effect" - product X might not make a high profit, but it helps drive sales of other items. (Mercedez will run a Super Bowl ad for a $131,000 car. Think that they'll sell very many?)

It's hard to quantify, and the bean counters hate the halo effect.

 

[MacProCard]

I remember when Apple did lots of things for the "halo effect" - product X might not make a high profit, but it helps drive sales of other items. (Mercedez will run a Super Bowl ad for a $131,000 car. Think that they'll sell very many?)

It's hard to quantify, and the bean counters hate the halo effect.

True to some extent. Though the nMP isn't a luxury purchase.

Then why change anything if the goal was to be the same ole box? Target the same ole market. Worry about the next processor. Clearly half the "pros" are pissed.

Someone in Apple had to pitch a concept that would increase sales for Apple in some capacity. Or they wouldn't invest in it. Market expansion is the only way to increase growth/sales for the nMP.

I'm not a bean counter, but being a developer you have to work with plenty of them. It does pay to learn their language.

 

[goMac]

True to some extent. Though the nMP isn't a luxury purchase.

Then why change anything if the goal was to be the same ole box? Target the same ole market. Worry about the next processor. Clearly half the "pros" are pissed.

Someone in Apple had to pitch a concept that would increase sales for Apple in some capacity. Or they wouldn't invest in it. Market expansion is the only way to increase growth/sales for the nMP.

I'm not a bean counter, but being a developer you have to work with plenty of them. It does pay to learn their language.

Oh jeez, the last time we had this conversation...

The Mac Pro went up in price, not down. That doesn't signify them wanting to capture a larger market. They have the Retina iMac now, which is clearly their prosumer machine.

 

[scottsjack]

Oh jeez, the last time we had this conversation...

The Mac Pro went up in price, not down. That doesn't signify them wanting to capture a larger market. They have the Retina iMac now, which is clearly their prosumer machine.

It looks to me like the Mac Pro may be going in the same direction as the X-Serve. For a while the Mac Pro was competitive with its competition to some extent. Now while the new Mac Pro may be a great FCP machine it cannot be compared in any way to the current offerings from HP and Dell.

Like the X-Serve the Mac Pro may soon be forgotten about by Apple.

 

[MMcCraryNJ]

It looks to me like the Mac Pro may be going in the same direction as the X-Serve. For a while the Mac Pro was competitive with its competition to some extent. Now while the new Mac Pro may be a great FCP machine it cannot be compared in any way to the current offerings from HP and Dell.

Like the X-Serve the Mac Pro may soon be forgotten about by Apple.

It's also a great Pro Tools machine. I can run complex sessions and peg every processor thread at nearly full load for many hours at a time and not worry about the machine throttling down on me and my audio stuttering about, as I would have to worry about when working on an iMac (especially the Retina one) or a Macbook Pro. It also lets me use whatever screen I want.

It all depends on your needs. Yes, you can run Pro Tools in Windows on a custom-built PC...there are many reasons why people usually don't, though.

 

[Larry-K]

While you might wish the nMP had everything and then some...it's not really possible for a company the size of Apple.

How big a company are you looking for?

 

[MacProCard]

Oh jeez, the last time we had this conversation...

The Mac Pro went up in price, not down. That doesn't signify them wanting to capture a larger market. They have the Retina iMac now, which is clearly their prosumer machine.

Yep. You said an update was right around the corner. Been saying an update is right around the corner for almost a year now. Intel getting ready to ship this and that. Someday you'll be right.

----------

How big a company are you looking for?

For Apple to do anything it cost them a lot of money... Why am I trying to teach business to a bunch of "pros". Enjoy waiting and speculating.

 

[goMac]

It looks to me like the Mac Pro may be going in the same direction as the X-Serve. For a while the Mac Pro was competitive with its competition to some extent. Now while the new Mac Pro may be a great FCP machine it cannot be compared in any way to the current offerings from HP and Dell.

Like the X-Serve the Mac Pro may soon be forgotten about by Apple.

Depends. The Xserve sales numbers were a fraction of the Mac Pro sales, as sad as that already is. If the Mac Pro dips under 50,000 units a year, it might be in trouble. I think Apple is perfectly happy to have the Mac Pro just be a great FCP machine as long as it continues to sell.

 

[Larry-K]

For Apple to do anything it cost them a lot of money... Why am I trying to teach business to a bunch of "pros". Enjoy waiting and speculating.

So you're saying Apple doesn't have any money?

What kind of business are you trying to teach us? Plumbing?

 

[Nosferax]

True to some extent. Though the nMP isn't a luxury purchase.

At $4k a pop, yeah, it's a luxury item. This is especially true since plenty of people who bought them use them for things that wouldn't even stress a Mac Mini or an MacBook Air.