Yes, it will get faster - but this is not due to any major accomplishments by Apple.
This is largely due to the work being done by very bright people at AMD/Intel/nVidia - who are enabling Apple to benefit. It would be interesting to benchmark the price/Performance of the A-series chips from Apple in a desktop/server application, against AMD/Intel/nVidia and work in that direction.
But, currently - to give credit to Apple, is akin to giving credit to the cock that crows every morning at sunrise.
Bullcrap.
Apple designed and composed a complete, turnkey, ready to rock computer SYSTEM from components that, in total and as an integrated system, is very well positioned to tackle computational workloads that push the limits of what current workstation class computers using current state of the art compilers and (OS and application) execution strategies can handle. Because they made the Mac Pro system highly modular and extensible, individual components can be swapped out in the future with newer and faster ones, extending the productive life of a current Mac Pro far beyond that of current computing systems that were not architected with similar provisions for modular extensibility.
No single component in the Mac Pro architecture is the "lone hero on a white horse riding in to save the day" (sorry, Kim Jong-un) in the system performance and quality equation. The system as a whole, and the underlying architectural design that enables the quality attributes, including performance, that make the Mac Pro system so spectacular is the closest thing to being a major accomplishment, but in aggregate and not due to any single contributing hardware or software component. Because it's a system, there's a large system team making it happen, not just a few bright people working on the CPU (which itself is also a system with a sizable team).
I also must add that software is a huge piece of the total system, and in many ways, is usually the laggard and weakest link that's holding the rest of the system back from its full potential. Since the very first multiprocessing capable computing hardware found its way into general purpose computing systems there's always been a huge quantity of unrealized computing potential stagnating in these computing systems. This is largely caused by a sizable impedance mismatch between the type of problems that a computing system is capable of solving based on its processing architecture and the fairly primitive means that programmers armed with programming languages, algorithms, and compilers are able to present their problems to these computing systems running on general purpose operating systems. If we had to justify our computer purchases based on our ability to actually consume the full processing capabilities of the computer we wanted to buy, most of us would be hard pressed to justify a Raspberry Pi single board computer.
For the past couple of decades most of us have been carrying around massive sledge hammers and using them to drive in tiny little nails. One reason for having so many tiny little nails is the software and humanware "problem." Yes, compilers, special purpose libraries, and APIs like Metal are getting better at covering up and working around some of our programming, algorithmic, and problem solving shortcomings, e.g., by taking advantage of parallelizing opportunities. However, we're now starting to see with greater frequency a lot more of the "big nail" problems emerging, e.g., AI, AR, ML, big data analytics, etc., that are stressing the limitations imposed by general purpose computing systems running general purpose operating systems with suboptimal (for the problem space) compilers building apps for an underlying system architecture that hasn't changed radically in over 40 years. At some level the Mac Pro and macOS are caught in this as much as any other computing system, but the Mac Pro's modular extensibility should allow it to adapt to future changes in both hardware and software components that are designed to overcome anticipated limitations and position it to take on more big nail problems.
The Apple Mac Pro system team has done their homework and they have built a very nice platform, but they don't know what they don't know. Only time will tell how much productive life they can squeeze out of their system. The fact is that they are playing a game where the goalposts are always moving. Yeah, it sucks to be them. There's nothing they can say or do to change that fact, but they had to take a shot and ship a product that they felt very comfortable with, and the Mac Pro is what they came up with.
If I had a nickel for every time I have heard "and it will get faster" when programmers do something…over the last 30 years. Heck, I'd be happy if touch ID would just work everywhere the System asks for a password.
Yes, it will get faster - but this is not due to any major accomplishments by Apple.
This is largely due to the work being done by very bright people at AMD/Intel/nVidia - who are enabling Apple to benefit. It would be interesting to benchmark the price/Performance of the A-series chips from Apple in a desktop/server application, against AMD/Intel/nVidia and work in that direction.
But, currently - to give credit to Apple, is akin to giving credit to the cock that crows every morning at sunrise.
The reporting and info so far has been mixed but it is hard to say any of the above with respect to the Afterburner card.
Yes, it will get faster - but this is not due to any major accomplishments by Apple.
This is largely due to the work being done by very bright people at AMD/Intel/nVidia - who are enabling Apple to benefit. It would be interesting to benchmark the price/Performance of the A-series chips from Apple in a desktop/server application, against AMD/Intel/nVidia and work in that direction.
But, currently - to give credit to Apple, is akin to giving credit to the cock that crows every morning at sunrise.
Oh, so did they design the Afterburner card and implement it in Apple software? Did they write the Metal framework? Also, considering thermal throttling, thermal management is the name of the game -- and that is designed entirely by the OEM, and Apple should get the credit for how their design handles heat.
Tho by the same token then Apple is not to blame for lack of updates. On Gruber's The Talk Show he said the 2017 iMac Pro has no suitable Xeon chips to update to, thus no updates...but I see on the forums Apple gets the blame.
Apple has upgrade options for the IMac Pro. That might mean going AMD but that could put Apple in the odd position of having an iMac that is faster than the Mac Pro. The processor world is a bit strange at the moment with AMD having an advantage In performance and thermals. This hasn’t been the case in decades.
Great job by Apple hands down. For those who don’t quite grasp what this machine is for, its not for you. For those who need it, this is a grand slam.
It is certainly a bold statement for AI to make. Frankly the only way that could happen is if more software becomes multithreaded, more work is done on the GPU or the afterburner card is programmable. These are not impossibilities but we all know that the easy stuff has already been done.
I'm old enough to remember buying computers in the late 80s, when a nice machine would run about $3K. That's about $6500 in today's dollars, so the Mac Pro does not feel expensive in any sense to me.
Now Apple has produced a powerful compute workstation, their past decision to deprecate OpenCL ( not OpenGL) will isolate Apple even more from the community that builds code leveraging GPUs for fast computational work (typically either using OpenCL or CUDA from nVidia).
There might be some hope if the Intel OneAPI initiative gains traction - the idea is developers can write C++ code to target all GPUs using one code base. But I see no indication yet Intel is including Metal as a target.
Great job by Apple hands down. For those who don’t quite grasp what this machine is for, its not for you. For those who need it, this is a grand slam.
It is certainly a bold statement for AI to make. Frankly the only way that could happen is if more software becomes multithreaded, more work is done on the GPU or the afterburner card is programmable. These are not impossibilities but we all know that the easy stuff has already been done.
Developers implementing Metal 2 better (or at all), Afterburner implementation, and better multithreading (as evidenced by Adobe Photoshop) will fit the bill, as you say.
And I agree - the easy stuff has mostly been done.
Yes, it will get faster - but this is not due to any major accomplishments by Apple.
This is largely due to the work being done by very bright people at AMD/Intel/nVidia - who are enabling Apple to benefit. It would be interesting to benchmark the price/Performance of the A-series chips from Apple in a desktop/server application, against AMD/Intel/nVidia and work in that direction.
But, currently - to give credit to Apple, is akin to giving credit to the cock that crows every morning at sunrise.
NVidia you say? The company that’s dropping existing Mac support from their next round of drivers? Okay.
Have I missed an announcement?
I misspoke — they're dropping macOS support in their CUDA Toolkit in the next version, which means no more future support for renderers that supported CUDA on Mac. That's more or less not unexpected given the lack of Nvidia drivers in Catalina etc., and in many cases the renderers in question like octane and Redshift are coming to Metal & AMD. Seems to also answer the question as to whether or not Nvidia support was going to come back to macOS with the Mac Pro at some point, unless they'd somehow get to the point where they're releasing a Metal-based driver/toolkit. ¯\(°_o)/¯
After decades of gray and white, the black mouse and cables seem like a significant design departure, and a break from the reign of Ive. Of course lots of people don’t care about these details, and I can see an argument for their position, but I hope the Ive design aesthetic is generally maintained, at least until Apple comes up with something better. I know it’s illogical, but even after all this time, I think the gray and white can still feel futuristic in a way that black cables and peripherals never will.
I am tired of white and I am glad to see more black returning.
I bought a Black Macbook back in 2007 or thereabouts and I liked that machine but I never liked that it came with the same white power adapter and cable that all other white Macs used. It just didn't feel right, because the two didn't go together well.
I recently bought a Space Grey Magic Trackpad and I was surprised and also very happy to see a black Apple lightning cable in the box. I couldn't even tell you how many white Apple lightning cables I have amassed throughout the years. That was my first ever black one from Apple.
I'm glad that Apple is starting to explore other color options besides white.
Of course, Apple charges a premium for their darker colors and I have paid extra for it each time (Black Macbook, Space Grey Magic TrackPad). I'm not complaining, as long as they keep offering it as an alternative to the white.
Now Apple has produced a powerful compute workstation, their past decision to deprecate OpenCL ( not OpenGL) will isolate Apple even more from the community that builds code leveraging GPUs for fast computational work (typically either using OpenCL or CUDA from nVidia).
There might be some hope if the Intel OneAPI initiative gains traction - the idea is developers can write C++ code to target all GPUs using one code base. But I see no indication yet Intel is including Metal as a target.
Is Metal 2 not an appropriate replacement for those things as it gets more widely supported? I don't actually know, my knowledge is more based in 3D rendering and not compute. I know they're positioning it as such but don't know much beyond that.
With all the Apple TV+ original shows, series, and movies (especially the SciFi stuff) in the works Apple has quite the impetus to develop their own high performance 3D graphics toolkit.
I foresee Final Cut Pro X like 3D tool chain being developed for internal projects with release for external use over time.
Renders, Modelers and a. line of Apple designed GPU hardware to support it all as very likely.
Curious what the top end users these machines are thinking is the expected life before some modules start getting replaced with newer hardware?
I assume the design means Apple can and will want to update the kit of parts as they can get their hands on them. Also assume they would like to create an aftermarket for refurbished parts down the track. How long before the demand for a prosumer MacPro might be satisfied by second-hand market. Or even a smaller case that can only handle say one refurb MPX module as a Mac sub-pro.
Yes, it will get faster - but this is not due to any major accomplishments by Apple.
This is largely due to the work being done by very bright people at AMD/Intel/nVidia - who are enabling Apple to benefit. It would be interesting to benchmark the price/Performance of the A-series chips from Apple in a desktop/server application, against AMD/Intel/nVidia and work in that direction.
But, currently - to give credit to Apple, is akin to giving credit to the cock that crows every morning at sunrise.
NVidia you say? The company that’s dropping existing Mac support from their next round of drivers? Okay.
The dropping of support has got nothing to do with Nvidia. Nvidia writes the Mac drivers for their cards, Apple writes the AMD drivers. Apple is refusing to sign the Nvidia drivers. My 2012 iMac with a Geforce 680MX still gets (or got - until Apple put their middle finger up at Nvidia) regular graphics driver updates. This is not the case for AMD cards from that era. Also, the AMD drivers Apple writes are utter crap. Back in 2006 I had a Mac Pro with an AMD card, and it would drop frames when any UI element showed over a window with 3D graphics in. Switched to a Nvidia card, problem vanished. Now on my top of the line 2019 i9 iMac with a Radeon Pro Vega 48, I get the same frame drops I got 15 years ago. Apple fell out with Nvidia (again) and refuses not only to use their cards but refuses to sign Nvidia drivers too. Nvidia cards are better at most things than AMD cards. Until Apple grows up and stops this ridiculous spat, we won't see any Nvidia drivers or cards on Macs, which only has the result of hurting Apple's customers.
I'm old enough to remember buying computers in the late 80s, when a nice machine would run about $3K. That's about $6500 in today's dollars, so the Mac Pro does not feel expensive in any sense to me.
So am I. My first “Mac” was a Radius clone. This PowerMac 8100, with a 110 MHz processor, retailed for over $6000.00 in yesterday’s dollars! We are so spoiled by modern computers and still complain about them.
This is a stunning machine.. I will admire it from afar. There is no way I could ever justify it.
I myself am waiting to see what Apple does with the 2020 iMac. I'm hoping for a major design change and going for serious CPU cores. It's time. My understanding is that the current 5K iMac exceeds the performance of the (now-aging) iMac Pro.
Crossing fingers. My 2015 iMac is beginning to slow me down.
Comments
Apple designed and composed a complete, turnkey, ready to rock computer SYSTEM from components that, in total and as an integrated system, is very well positioned to tackle computational workloads that push the limits of what current workstation class computers using current state of the art compilers and (OS and application) execution strategies can handle. Because they made the Mac Pro system highly modular and extensible, individual components can be swapped out in the future with newer and faster ones, extending the productive life of a current Mac Pro far beyond that of current computing systems that were not architected with similar provisions for modular extensibility.
No single component in the Mac Pro architecture is the "lone hero on a white horse riding in to save the day" (sorry, Kim Jong-un) in the system performance and quality equation. The system as a whole, and the underlying architectural design that enables the quality attributes, including performance, that make the Mac Pro system so spectacular is the closest thing to being a major accomplishment, but in aggregate and not due to any single contributing hardware or software component. Because it's a system, there's a large system team making it happen, not just a few bright people working on the CPU (which itself is also a system with a sizable team).
I also must add that software is a huge piece of the total system, and in many ways, is usually the laggard and weakest link that's holding the rest of the system back from its full potential. Since the very first multiprocessing capable computing hardware found its way into general purpose computing systems there's always been a huge quantity of unrealized computing potential stagnating in these computing systems. This is largely caused by a sizable impedance mismatch between the type of problems that a computing system is capable of solving based on its processing architecture and the fairly primitive means that programmers armed with programming languages, algorithms, and compilers are able to present their problems to these computing systems running on general purpose operating systems. If we had to justify our computer purchases based on our ability to actually consume the full processing capabilities of the computer we wanted to buy, most of us would be hard pressed to justify a Raspberry Pi single board computer.
For the past couple of decades most of us have been carrying around massive sledge hammers and using them to drive in tiny little nails. One reason for having so many tiny little nails is the software and humanware "problem." Yes, compilers, special purpose libraries, and APIs like Metal are getting better at covering up and working around some of our programming, algorithmic, and problem solving shortcomings, e.g., by taking advantage of parallelizing opportunities. However, we're now starting to see with greater frequency a lot more of the "big nail" problems emerging, e.g., AI, AR, ML, big data analytics, etc., that are stressing the limitations imposed by general purpose computing systems running general purpose operating systems with suboptimal (for the problem space) compilers building apps for an underlying system architecture that hasn't changed radically in over 40 years. At some level the Mac Pro and macOS are caught in this as much as any other computing system, but the Mac Pro's modular extensibility should allow it to adapt to future changes in both hardware and software components that are designed to overcome anticipated limitations and position it to take on more big nail problems.
The Apple Mac Pro system team has done their homework and they have built a very nice platform, but they don't know what they don't know. Only time will tell how much productive life they can squeeze out of their system. The fact is that they are playing a game where the goalposts are always moving. Yeah, it sucks to be them. There's nothing they can say or do to change that fact, but they had to take a shot and ship a product that they felt very comfortable with, and the Mac Pro is what they came up with.
And I agree - the easy stuff has mostly been done.
I bought a Black Macbook back in 2007 or thereabouts and I liked that machine but I never liked that it came with the same white power adapter and cable that all other white Macs used. It just didn't feel right, because the two didn't go together well.
I recently bought a Space Grey Magic Trackpad and I was surprised and also very happy to see a black Apple lightning cable in the box. I couldn't even tell you how many white Apple lightning cables I have amassed throughout the years. That was my first ever black one from Apple.
I'm glad that Apple is starting to explore other color options besides white.
Of course, Apple charges a premium for their darker colors and I have paid extra for it each time (Black Macbook, Space Grey Magic TrackPad). I'm not complaining, as long as they keep offering it as an alternative to the white.
I foresee Final Cut Pro X like 3D tool chain being developed for internal projects with release for external use over time.
Renders, Modelers and a. line of Apple designed GPU hardware to support it all as very likely.