2016 nMP

[Hank Carter]

Since that post was referring to both consumer and Pro versions, My answer was aimed at both.

I think the using updated/overpriced/underpowered is the constant comparison with its PC counterparts. The nMP is not quite two years old yet since being shipped. Being on a longer refresh cycle on its workstation than on its consumer versions.

The technology in the Mac Pro is 3-4 years old. And if you are not going to compare it to a PC/Linux workstation than what else is there? Apple doesn't exists in a separate universe where it is the sole maker of workstations.

If having up to date equipment is a priority, its way to late by now as Apple is usually the last with new products forth coming. It should be no surprise at all by the Pro's who have been using them for many years already.

There is some truth to that, but the situation has never been this bad. Apple has a long history of lagging behind in workstation performance, but at least the 5,1 could be upgraded with faster graphics cards and the dual processors gave it a lot more punch etc. The trashcan is a dead end that is about as upgradeable as an iMac and isn't exactly a CPU powerhouse.

As I always mentioned, I compare Apples newer products to what Macs I already have and if its fast/new technology enough for what I need it to do. I don't compare it to Windows/Linux PC's because I don't plan to change platforms any time soon.

Well, if you had to compete directly with other people or companies sitting on Linux/PC boxes that allow them to turn a job around multiple times faster than you on a 12c/D700 then you may think differently.

 

[tuxon86]

I don't compare it to Windows/Linux PC's because I don't plan to change platforms any time soon.

And yet your alias is LINUXcooldude...

 

[linuxcooldude]

The technology in the Mac Pro is 3-4 years old. And if you are not going to compare it to a PC/Linux workstation than what else is there? Apple doesn't exists in a separate universe where it is the sole maker of workstations.

When the nMP was in development they used the current technology at the time. A few years later when it was shipping, its no longer the most current. As we know, Apple does not always use the most current hardware but makes the best use of it compared to some others.

Well, if you had to compete directly with other people or companies sitting on Linux/PC boxes that allow them to turn a job around multiple times faster than you on a 12c/D700 then you may think differently.

If your currently on a Mac platform and planning on staying, there is usually a more updated/faster one already available than the one you are currently using. If you think windows/PC better for updates/faster, than by all means move to that platform if you wish. But if your still comparing platforms, they won't match.

And yet your alias is LINUXcooldude...

Yes, I've been using it for around 16 years...but.....since going into broadcasting/videography, Linux was far from the ideal platform during that time for me. Until fairly recently a limited selection of broadcasting, video editing and post production software available. Screencasting software was in a a very proprietary format, forcing me to convert to a standardized one.

A username rarely tells the whole story. Some still want to use it as a derogatory term.

 

[tralfaz]

It's not "nerdy" to expect your system supplier to give you the latest tools to do your job.

You can buy a 44-core system with a TebiByte of RAM and one to four GPUs of your choice from other vendors. Ten internal drive bays.

Or a 4-core system with 4 GiB, single low end GPU, and a disk or two. (Spinner or SSD)

Apple offers a 4 to 12 (outdated) core system with max support for 64 GiB, and a pair of outdated midrange consumer GPUs (whether you need a second one or not). One internal SSD drive with an unusual connector and an unusually low max system capacity. At a premium price.

You do realize while everything you say is true, and I understand all of it, Apple would consider it all nerd techno-babble.

I too have held out hope that Apple would throw our types a single bone to alleviate our anguish, but I have just managed to extinguish that hope. It has been a long time coming, but I have told myself, "It's dead Jim!" in McCoy's most emotive voice.

 

[tuxon86]

When the nMP was in development they used the current technology at the time. A few years later when it was shipping, its no longer the most current. As we know, Apple does not always use the most current hardware but makes the best use of it compared to some others.



If your currently on a Mac platform and planning on staying, there is usually a more updated/faster one already available than the one you are currently using. If you think windows/PC better for updates/faster, than by all means move to that platform if you wish. But if your still comparing platforms, they won't match.



Yes, I've been using it for around 16 years...but.....since going into broadcasting/videography, Linux was far from the ideal platform during that time for me. Until fairly recently a limited selection of broadcasting, video editing and post production software available. Screencasting software was in a a very proprietary format, forcing me to convert to a standardized one.

A username rarely tells the whole story. Some still want to use it as a derogatory term.

You see it as derogatory, I see it as funny... Especially since Apple is a locked down closed platform, more so than even Microsoft. I've been on and off Linux since Slackware 2.1 in 1994 myself. I presently have Antergos on my laptop.

 

[--AG--]

I think that it may be ok not using the latest tech and having longer upgrade cycles. But if so, the price should be adjusted when the tech becomes outdated. However, when I bought my cMP 3.1. the price/performance was actually very competitive compared with other workstations and the performance was in the front line and frequently updated – a totally different approach to the product compared to what we see now. At the time many colleagues were choosing MP in their labs, not so anymore. I believe that there are added values from having a large base of developers using its system. For me this is the number one argument for why apple should have a high end workstation that is priced aggressively. E.g. apple spent a lot of time of its latest event talking about health oriented apps and its bright futures, but who will develop such apps and perform the bioinformatics research needed for such software if there are no apple workstation that can do the work? Moreover, maybe a fallacy of apple's marketing department is believing that buyers of MP are tickled by the same features that buyers of laptops and iOS devices are, e.g. slim design, when (I claim) other attributes typically have higher priority (performance, flexibility, upgradeability, etc... ). Today the nMP is in nowhere land, not powerful enough for serious computing and to expensive for enthusiasts and perhaps not even meeting the criteria that either serious developers nor enthusiasts have.
[doublepost=1460705498][/doublepost]

You do realize while everything you say is true, and I understand all of it, Apple would consider it all nerd techno-babble.

I hope that Apple realizes that MP's target group, its buyers, are full of "nerd techno-babble"

 

[Zarniwoop]

http://9to5mac.com/2016/04/14/macos-name-change-apple/

Rumors that OS X is going to be macOS are getting stronger!

So this and the aging Mac line can really mean the big change in Mac lineup.

Update:
I think macOS will start, most likely, from version 1.0 and it works only on a quite new computers. Maybe same specs as Metal has (Mac's 2012 or later) and El Captain will get updates for one year more (not just security updates, but 10.11.xx versions.

This would be perfect time to start to bring ARM based Macs too. Not at this WWDC, but maybe after macOS beta testing ends and when there's more software supporting macOS directly.

Maybe next Apple Pro apps are going to work only with macOS.. that is why the whole Mac line has been on the wait state along with the software

It has been speculated, that Siri will come with next OS X (macOS) revision, and that might need a special hardware in order it to work properly. HSA 2.0* (Apple branded of course) and Metal v2 compatible.

And this is relevant for nMP too, if it is the reason we haven't seen an update lately..

* The HSA Foundation is defining key interfaces for parallel computation utilizing CPUs, GPUs, DSPs, and other programmable and fixed-function devices, thus supporting a diverse set of high-level programming languages and creating the next generation in general-purpose computing. ARM, Imaginations and AMD are key supporters. - www.hsafoundation.com

 

[Zarniwoop]

There's just one problem to fully implement HSA 2.0, and its name is Intel. Since Haswell, there has been a basic implementation of unified memory for CPU and Iris iGPU, but I don't think it is open for others.

So what is left? ARM or AMD APU.

Next autumn.. when macOS would be ready for shipping.. is also a time for Apple A10 and AMD Zen.

 

[koyoot]

Well, if Apple will abuse Open Source initiative to sell hardware to people, I will mock them exactly how I mock Nvidia.

P.S. Small hints of HSA are currently in both versions of Metal... Interesting thought you brought Zarniwoop.

 

[Zarniwoop]

This slide tells the future:



The Multi-core era is at its sunset.. HSA is coming! Enabled by power efficient GPUs'.
[doublepost=1460729415][/doublepost]

P.S. Small hints of HSA are currently in both versions of Metal... Interesting thought you brought Zarniwoop.

Unified memory is one of the key features of Metal for iOS. I think it is emulated on OS X, because Intel CPU's don't really support it. That could be a reason why the speed advantage is not always so great on Metal for OS X.

 

[Hank Carter]

If your currently on a Mac platform and planning on staying, there is usually a more updated/faster one already available than the one you are currently using. If you think windows/PC better for updates/faster, than by all means move to that platform if you wish. But if your still comparing platforms, they won't match.



Yes, I've been using it for around 16 years...but.....since going into broadcasting/videography, Linux was far from the ideal platform during that time for me. Until fairly recently a limited selection of broadcasting, video editing and post production software available. Screencasting software was in a a very proprietary format, forcing me to convert to a standardized one.

A username rarely tells the whole story. Some still want to use it as a derogatory term.

Again, Apple does not exists in a bubble. If you are not going to compare the Mac Pro to a Windows or linux workstation what are you going to compare it to? Last years model? Nothing? Should Apple just get a pass, because they are special or something? I've been on Mac's for over 20 years and use mine professionally in post production for film and TV and the current situation is the worst it's ever been. Moving to Windows is not necessarily an option for those of us who need things like FCPX and Quicktime and have invested a lot of money in a Mac based production pipeline.

 

[Zarniwoop]

HSA should make OS more CPU independent.. with its own libraries and JIT compilers.
HSAIL as a virtual ISA. And promotes GPU as a real co-processor
A single source code base for the CPU and GPU!
[doublepost=1460730076][/doublepost]

So Metal is like a combination of HSA and Mantle. Maybe in its next revision more so.
[doublepost=1460731285][/doublepost]Btw, Nvidia has chosen openPower as its HSA architecture.

So there are two trains going different tracks; AMD (and its allies) with HSA, IBM Power & Nvidia (+ allies) on a second track called openPower.

And Apple is already very deep in HSA with ARM and Imagination. And AMD's business partner.. so.. HSA sounds more likely, and this could be the reason Nvidia has been out atm. Intel's future doesn't look so great either, if there's not going to be a full compatibility.

 

[tuxon86]

HSA should make OS more CPU independent.. with its own libraries and JIT compilers.
HSAIL as a virtual ISA. And promotes GPU as a real co-processor
A single source code base for the CPU and GPU!
[doublepost=1460730076][/doublepost] View attachment 626918

So Metal is like a combination of HSA and Mantle. Maybe in its next revision more so.
[doublepost=1460731285][/doublepost]Btw, Nvidia has chosen openPower as its HSA architecture.

So there are two trains going different tracks; AMD (and its allies) with HSA, IBM Power & Nvidia (+ allies) on a second track called openPower.

And Apple is already very deep in HSA with ARM and Imagination. And AMD's business partner.. so.. HSA sounds more likely, and this could be the reason Nvidia has been out atm. Intel's future doesn't look so great either, if there's not going to be a full compatibility.

Competition is good for everybody. And frankly it isn't really different then what we had a few years back with NVidia flavoured mobo and AMD flavoured mobo. It forced Intel to invest more in chipset design and research that gave us the great chipset that we have today.

Mono-culture is a BAD thing. Diversity brings strength.

 

[ManuelGomes]

Funny, that Apple page seems to have been rectified from MacOS back to OS X...
I wonder if someone got canned for this slipup?!
[doublepost=1460732082][/doublepost]http://www.apple.com/environment/answers/

 

[koyoot]

HSA should make OS a CPU independent.. with its own libraries and JIT compilers.
A single source code base for the CPU and GPU!

Have you ever imagined about connecting a touch sensitive device wirelessly to GPU rack/cluster and operate on the data that are connected to it?

Imagine yourself being on Oahu on vacation and at the same time if emergency comes, you could connect through iPad to your company's compute cluster, and do your work, that needed intervention

One of the possibilities of HSA foundation.

 

[tuxon86]

Have you ever imagined about connecting a touch sensitive device wirelessly to GPU rack/cluster and operate on the data that are connected to it?

Imagine yourself being on Oahu on vacation and at the same time if emergency comes, you could connect through iPad to your company's compute cluster, and do your work, that needed intervention

One of the possibilities of HSA foundation.

Already a reality on NVidia side with their Grid system. If you can stream hidef games in realtime you can also stream less demanding pro-applications.

 

[ManuelGomes]

NVidia Pascal launch?
http://wccftech.com/nvidia-launch-pascal-editors-event-may/

AMD Polaris:
http://wccftech.com/amd-radeon-r9-480-470-polaris-10-polaris-11/

Usual bag of salt recommended, but seems in line with what's been rumored already.
Polaris 10 TDP looks to fit nicely in nMP, if accurate.
Maybe we're holding on for the GDDR5X version of Polaris 10, until it gets really mainstream.
What baffles me is that I'd figure that Apple this time around wouldn't make the same mistake and deliver non-ECC mem GPUs (gaming cards if you will) in the nMP. OK, come on, attack me as you will saying ECC is not required, but this is a workstation and that should be something to consider, even if only for some workloads. I see AS coming up with another one of her comments... come on, say something smart.
[doublepost=1460733814][/doublepost]I could imagine myself in Oahu for sure
Even better if no emergency calls are coming in though

 

[koyoot]

What if they will use Fiji based GPUs?


Fiji device ID is linked to the Can frame buffer in last versions of OS X. So who knows? Someone is testing something: it can be iMac with Fury, Mac Pro, new version of Mac Mini? Only Apple knows what they will do next with GPUs.

HSA should make OS more CPU independent.. with its own libraries and JIT compilers.
HSAIL as a virtual ISA. And promotes GPU as a real co-processor
A single source code base for the CPU and GPU!
[doublepost=1460730076][/doublepost] View attachment 626918

So Metal is like a combination of HSA and Mantle. Maybe in its next revision more so.
[doublepost=1460731285][/doublepost]Btw, Nvidia has chosen openPower as its HSA architecture.

So there are two trains going different tracks; AMD (and its allies) with HSA, IBM Power & Nvidia (+ allies) on a second track called openPower.

And Apple is already very deep in HSA with ARM and Imagination. And AMD's business partner.. so.. HSA sounds more likely, and this could be the reason Nvidia has been out atm. Intel's future doesn't look so great either, if there's not going to be a full compatibility.

Now read this post

I will tell you this, but supposedly nobody will believe me. I have written something similar 1.5 years ago on this forum.

AMD is currently working on their own CUDA-like API for professional, compute applications. It will be based on Mantle v2. What is more important it will be very similar to Metal. It will most probably be Open Source. That is very reason why AMD launched Boltzmann Initiative, and CUDA compilers for OpenCL.

You see connection?

 

[Stacc]

What baffles me is that I'd figure that Apple this time around wouldn't make the same mistake and deliver non-ECC mem GPUs (gaming cards if you will) in the nMP. OK, come on, attack me as you will saying ECC is not required, but this is a workstation and that should be something to consider, even if only for some workloads. I see AS coming up with another one of her comments... come on, say something smart.

It depends on how AMD positions Polaris 10. Tahiti was unique in that it had a 1:4 DP ratio on its consumer cards. This was a really great consumer DP card and why so many people bought it for bit coin mining. When it came to Hawaii, consumer cards had a 1:8 ratio for SP-DP and 1:2 on workstation cards. The markup on consumer to workstation cards (that include ECC) is something like 10x.

Apple chose to go with the consumer variant of the chip that didn't include ECC but was very good at DP compute. They could have gone with the workstation variant with ECC memory but that would have seen a huge increase in cost for those dual video cards for a relatively narrow niche market. Just like Apple chose to not appeal to the niche of mac users who want dual processors they also chose not to appeal to the small niche that needs ECC VRAM. Given that they aren't really positioning this as a high end GPU compute box this is probably the right move. I think they imagine that these GPUs are being used to render intermediate codecs in FCPX and doing test renders for 3d modeling. Things like that. Once you need 100% accuracy and quality you export it using the CPU.

Rumors have Polaris being another card with a good SP-DP ratio. Something like 1:3. If this is indeed the case I see Apple again going with the consumer chips and not paying for the markup required to get "workstation" variants that include ECC. It will be interesting to see if they can get a polaris 10 chip with 16 GB of GDDR5(X).

What if they will use Fiji based GPUs?


Fiji device ID is linked to the Can frame buffer in last versions of OS X. So who knows? Someone is testing something: it can be iMac with Fury, Mac Pro, new version of Mac Mini? Only Apple knows what they will do next with GPUs.

I don't really see this happening. Fiji is a high end gaming chip that has little use as a compute card. The poor SP-DP (1:16) ratio and limited VRAM make it not very compelling. If Polaris 10 ends up in a similar performance ballpark at a lower TDP then there isn't much benefit to using Fiji. If Apple was going to use Fiji they could have released a Haswell-EP and Fiji Mac Pro almost a year ago.

 

[ManuelGomes]

Hope not, leave Fiji alone now. Polaris is the way to go.


Still going strong:
http://anandtech.com/show/10253/amd-announces-32gb-firepro-w9100

 

[Mago]

This would be perfect time to start to bring ARM based Macs too. Not at this WWDC, but maybe after macOS beta testing ends and when there's more software supporting macOS directly.

Beyond servers and itoys ARM CPU isnt viable to replace x86, the reason its simple, Singlethread execution still half way the slowest single thread on x86, this its an big taxation on most apps that relies on sigle threads also most user interaction on pc will experiment an evident lag.

ARM macs, not yet, later maybe.

 

[ManuelGomes]

Stacc, right on. I feel that way too but the nMP price is already steep and should allow for having real WS cards. Of course, Apple considers that every penny in their pocket is better than in ours, not that I find that criminal, I can even understand, but would given them extra credibility in the WS market. Anyway, it's a business and we need to accept that they exist to make money for the shareholders 8i'm not one of them but some of us might be).
Or we could see cards with extra mem to do a "soft" ECC.

 

[Zarniwoop]

ARM macs, not yet, later maybe.

Maybe so, this could be on their roadmap for laptops.. but not yet. Rumors say, that Macbook will get Skylake update very soon, so maybe next year (for ARM).

 

[Mago]

Maybe so, this could be on their roadmap for laptops.. but not yet. Rumors say, that Macbook will get Skylake update very soon, so maybe next year.

before Apple jumping to ARM on Macs we will see AMD cpus on Macs, much more likely, while at servers its another history server software use to be multithread optimized, so an OS/X server on ARM could arrive sooner, maybe this year since Apple recognized re-starting its server hardware development at leas for internal services (iCloud, SIRI etc), maybe Apple ends selling some rack-mounter blade based on ARM running OSX server, or maybe they just build that for internal consumption.

right now the mac mini as NAS its fair good as long you attach external thunderbolt drives, a Mac mini based server appliance would need HDD trays at least two, or upto 5 to be competitive against popular NASes

 

[tuxon86]

NVidia Pascal launch?
http://wccftech.com/nvidia-launch-pascal-editors-event-may/

AMD Polaris:
http://wccftech.com/amd-radeon-r9-480-470-polaris-10-polaris-11/

Usual bag of salt recommended, but seems in line with what's been rumored already.
Polaris 10 TDP looks to fit nicely in nMP, if accurate.
Maybe we're holding on for the GDDR5X version of Polaris 10, until it gets really mainstream.
What baffles me is that I'd figure that Apple this time around wouldn't make the same mistake and deliver non-ECC mem GPUs (gaming cards if you will) in the nMP. OK, come on, attack me as you will saying ECC is not required, but this is a workstation and that should be something to consider, even if only for some workloads. I see AS coming up with another one of her comments... come on, say something smart.
[doublepost=1460733814][/doublepost]I could imagine myself in Oahu for sure
Even better if no emergency calls are coming in though

What was "stupid" was releasing a workstation with only half of the data being protected by ECC, especially one that is to be used for GPGPU processing... Once the data leaves the system ram to be process on the GPUs all bets are off about its accuracy.