AMD Vega 56 and Vega 64 GPUs destined for iMac Pro detailed in Linux driver
The GPUs that will be in the iMac Pro -- the Vega 56 and Vega 64 -- have been detailed by an AMD-provided driver update for Linux, with the cards able to utilize much as twice as a much data in each register as previous cards when 32 bits of precision aren't needed.
The data was collected from the flagship graphic chipset's drivers for Linux's Direct Rendering Manager, and collated by enthusiast site Wccftech and others.
The Vega Pro 56 has 56 compute units, with 3584 stream processors, and 8GB of HBM2 RAM pushing 400 GB/s of data. While single-precision (FP32) calculations are 11 tflop roughly analagous to the Nvidia GTX 1080ti , half-precision 16-bit calculations (FP16) such as that used for image and graphic processing, ray tracing, artificial intelligence, and game rendering hit a peak of 22 tflops.
A configurable upgrade for the iMac Pro, the Vega Pro 64, has 64 compute units, with 4096 stream processors, and 16GB of HBM2 RAM -- but no declared bandwidth as of yet. The gleaned information concludes that the Vega Pro 64 has FP32 single-precision calculations at 13 tflop, with FP16 at 25 tflop.
Pricing on either chipset has not yet been announced, or can be determined from the drivers, obviously.
Pre-orders for the RX Vega Frontier card have started for either $1200 for an air-cooled version or $1800 for a water-cooled one, but how accurate these prices are is not yet known. The card intended for workstations, with a similar market segment to the iMac Pro, and is scheduled to ship on June 27, according to AMD CEO Lisa Su.
The official launch of the line beyond marketing information, or information gleaned from previously accurate driver updates is expected at the SIGGRAPH trade show in July.
The advantages of FP16 for the consumer aren't quite clear, or easily comparable, to Nvidia cards as of yet. The Vega card with its new architecture doubles FP16 performance, giving calculations that rely on it a big boost over older generations of Radeon cards, and Nvidia ones.
Some modern Nvidia cards have equivalent FP16 performance to FP32, and some have notably poorer FP16 performance than FP32. The new Nvidia 1080ti card, often used in builds attempting to match the iMac Pro by enthusiasts, delivers less than a teraflop of FP16 performance.
The Nintendo Switch and PlayStation 4 Pro operating systems take full advantage of FP16 calculations to boost performance when possible, as does Metal when developers choose to implement it.
Geometry and heavy scientific calculations are better performed on FP32, and the even more complex FP64, for the higher precision allowed. Dynamic lighting and image editing will see a boost with FP16, as will gaming, and machine learning tasks.
Real-world tasks will be a mix of FP16 and FP32. Actual task-related benchmarking after iMac Pro release will ultimately tell the tale.
However, the possibility remains that the numbers aren't accurate. But, given the announced RX Vega Frontier cards, it seems fairly likely to be correct.
The data was collected from the flagship graphic chipset's drivers for Linux's Direct Rendering Manager, and collated by enthusiast site Wccftech and others.
The Vega Pro 56 has 56 compute units, with 3584 stream processors, and 8GB of HBM2 RAM pushing 400 GB/s of data. While single-precision (FP32) calculations are 11 tflop roughly analagous to the Nvidia GTX 1080ti , half-precision 16-bit calculations (FP16) such as that used for image and graphic processing, ray tracing, artificial intelligence, and game rendering hit a peak of 22 tflops.
Actual task-related benchmarking after iMac Pro release will ultimately tell the tale.
A configurable upgrade for the iMac Pro, the Vega Pro 64, has 64 compute units, with 4096 stream processors, and 16GB of HBM2 RAM -- but no declared bandwidth as of yet. The gleaned information concludes that the Vega Pro 64 has FP32 single-precision calculations at 13 tflop, with FP16 at 25 tflop.
Pricing on either chipset has not yet been announced, or can be determined from the drivers, obviously.
Not just for the iMac Pro
A a pair of Vega-powered PCI-E cards were also announced in conjunction with the Vega Pro 56 and Pro 64 chipsets. The Radeon RX Vega Frontier edition is in essence a Radeon Vega Pro 64 in a PCi-E card, and the RX Vega gaming card has very few specifications known at this time. It is unclear if macOS will see drivers for either card.Pre-orders for the RX Vega Frontier card have started for either $1200 for an air-cooled version or $1800 for a water-cooled one, but how accurate these prices are is not yet known. The card intended for workstations, with a similar market segment to the iMac Pro, and is scheduled to ship on June 27, according to AMD CEO Lisa Su.
The official launch of the line beyond marketing information, or information gleaned from previously accurate driver updates is expected at the SIGGRAPH trade show in July.
Explaining FP16 and FP32
The Nvidia GTX 1080ti has around 11 tflop of FP32 performance, but it is short on FP16 cores with one per streaming multiprocessor -- and FP32 cores can't do FP16 calculations. Previous testing with eGPUs have listed cards performances when doing FP32 calculations, for simplicity.The advantages of FP16 for the consumer aren't quite clear, or easily comparable, to Nvidia cards as of yet. The Vega card with its new architecture doubles FP16 performance, giving calculations that rely on it a big boost over older generations of Radeon cards, and Nvidia ones.
Some modern Nvidia cards have equivalent FP16 performance to FP32, and some have notably poorer FP16 performance than FP32. The new Nvidia 1080ti card, often used in builds attempting to match the iMac Pro by enthusiasts, delivers less than a teraflop of FP16 performance.
The Nintendo Switch and PlayStation 4 Pro operating systems take full advantage of FP16 calculations to boost performance when possible, as does Metal when developers choose to implement it.
Geometry and heavy scientific calculations are better performed on FP32, and the even more complex FP64, for the higher precision allowed. Dynamic lighting and image editing will see a boost with FP16, as will gaming, and machine learning tasks.
Real-world tasks will be a mix of FP16 and FP32. Actual task-related benchmarking after iMac Pro release will ultimately tell the tale.
Grain of salt
Previous versions of cards, including the Nvidia 10-series and Radeon 500-series cards had specifications leak from vendor-supplied drivers for Linux.However, the possibility remains that the numbers aren't accurate. But, given the announced RX Vega Frontier cards, it seems fairly likely to be correct.
Comments
Hopefully we will see a Mac Mini replacement with these cards. Either that or a Mac Pro more focused on the desktop workstation market.
It ties in with LLVM/Clang/ with AMD's /ROCm/VR/AR and Apple's own OpenCL stack/AR/VR and both Machine Learning interests, to much more on the Server end with Naples.
https://github.com/RadeonOpenCompute
http://gpuopen.com/
http://radeon.com/en-us/
Add in an external GPU stack of Radeon Instinct GPGPUs for Vega and Apple becomes the leader in many fields necessary for a complete Workstation environment with Apple's platform.
http://instinct.radeon.com/en-us/
As for what I was talking about that is the rumored Mini replacement, one variant which must support an discreet chip video card for Apple to really demonstrate that they are serious about AR, ML and other technologies demanding high performance compute. So yeah I'm expecting that one variant of the Mini will cost in the order of $1200 to $1500 dollars. A good portion of that cost going to cover the video chip. By the way this may or may not mean that the GPU chip is integrated into the mother board. The other possibility would be a cut down Mac Pro with one video card and a more desktop oriented CPU chip. A Mac Pro lite if you will.
I'm only thinking this way due to eh heavy emphasis Apple has placed on AR, ML and other advanced technologies that really need the compute resources a discreet chip offers. From my stand point it looks like Apple is all in with respect to Machine Learning as that is being worked into a lot of Apples technologies.
The problem is the Mini will become more or less useless if it doesn't support Apples AR and ML technologies in the future. While integrated GPU's get better every year they are still a long ways from being highly optimized for these technologies. As for Vega I see that as a requirement in a top end Mini replacement the lower end models can have anything from integrated GPU's to some midrange chip.
As you note the current Mini's are pretty pathetic. I think Apple realizes this and is actually in a mad rush to overhaul the entire Mac desktop lineup. The fact is the Mini isn't really capable of supporting the very technologies they debuted at WWDC this year. Obviously that has to be address or they will look pretty stupid come the time for a shipping High Sierra. In other words the Mini will get a complete overhaul, possibly into a completely different looking machine, in time to support the new tech in High Sierra.
As for Mini demand, contrary to popular opinion here it is a popular machine for people with low end professional requirements. The Mini is extremely popular with app developers for example, I've seen it used as an embedded controller is some really high end instrumentation also. One thing the Mini has had going for it is that it is a very stable platform that remains compact over the years. Of course adding a discreet GPU capability will have a negative impact on its usefulness as a stable platform. I really think that people underestimate the good qualities that the Mini has that keeps it selling in a very very down market. My problem with it is that I need more out of a desktop but I don't need the hit to the pocket that a Mac Pro causes.
Tim Cook wouldn't be the first executive to screw up understanding customer needs. I live across the river from the Old Kodak Park and can attest to just how badly a companies leadership can screw up not understanding markets and customer needs. Right now Apple is living off iPhone but that can go as quickly as it came, it would be nice to see Apple produce computers that its customers actually want and have a need for. Oh an maybe put a bit of effort into marketing those computers as complete systems, the MacPro shows that they have loss all concept of what a pro system is.
Now the question is, as we wait for Apple new desktops to debut, is does Apple get it. That is do they understand what people want these day out of a desktop machine, a small workstation or a high performance workstation. My feeling here is that Mac Pro would have been an ideal small workstation computer given a better pricing structure and a greater focus on what many pro's outside of content creation need. In other words the Mac Pro is a good machine targeted at the wrong people and maybe just a bit overboard for the people that could put it to use.
In any event I'm trying to drive home the idea that Apple will need to be far more proactive when it comes to performance GPU's if they really want their new AR and ML technologies to take off. This is why I believe the Mac Minis replacement will have support for external GPU's in at east one and most likely two variants. Either that or the Mac Pro becomes Apple midrange desktop and it gets replaced with a real performance workstation. A Mac Pro chassis with a recent Intel APU style chip and one good GPU card would be an awesome machine for many of us. Sell if for $1500 - $2000 and it will be a hot seller
In the end the Mac Pro is an example of Apple being to focused on one subset of its customer base and then not even getting that right. They completely missed that many of us only need one GPU card, a high performance one at that. Mac Pro really makes you question if they really understand their customer base. Mind you I'm one that think many concepts embodied in the Mac Pro really do represent the future of desktop computing. There is no reason for the massive chassis of the past for one. Second the way technology changes these days it doesn't make sense to buy a machine with the thought of upgrading every part in the machine as technology changes too much by the time new hardware makes an upgrade worthwhile. At best a RAM upgrade and maybe a storage upgrade can make sense, but these day people are fooling themselves by over valuing complete machine upgradability.
Seems Apple is trying to expand out of their iPhone dependence, I'm sure they know better than you and I the risk of being overly dependent on the one product, but they're not gods and sometimes expectations from bloggers far exceed the reality that tech progress can deliver.
Of course none of this matters to some, for those with unreasonable expectations if Tim can't pull an actual rabbit out of the proverbial hat then he's a failure and it's because he's not listening to the best ideas, those which are contained within the minds of Joe and Cindy Nobody. The most interesting ideas from those two morons (Joe & Cindy) include 1) more machines in a non-growth or dying market (i.e. desktop PCs), or 2) a better (Pro) machine in an infinitesimally small market. It's a laughable assessment some insist will work but the armchair CEO position is just as easy to perform as armchair Quarterback, when the reality of these positions is something a bit more complex.
Speaking of the iMac Pro, I thought it would only be the Edition Watch that I cannot afford! These iMac Pros are really some beasts!
Apple makes the line between "want" and "need" very blurry. Sometimes the price is what separates the men from the boys.
I'm crawling back to my play-pen now!
https://www.pugetsystems.com/labs/articles/Adobe-After-Effects-CC-2015-Multi-Core-Performance-714/
To do this manually, use the binary called aerender that is inside the base AE folder. Example commands are here:
https://helpx.adobe.com/after-effects/using/automated-rendering-network-rendering.html
Say that you had a project called Animation.ae on the desktop and you had a comp inside it called Overlays. Assume you want to use Best render settings and an Output Module called PNG Sequence. The template names are just the names in the drop-downs. Normally you just do the whole sequence but this has to wait for each frame to finish one after the other. Instead, you split the render into multiple blocks of frames and do them in parallel. The -s and -e below are start and end frames.
For a comparison, first run a sequence in a single pass in a single command-line, change the names and folder locations to match your own setup, you can drag files into the command-line and use tab for auto-complete:
/Applications/After Effects CC/aerender -project ~/Desktop/Animation.ae -comp "Overlays" -RStemplate "Best Settings" -OMtemplate "PNG Sequence" -s 1 -e 100 -output ~/Desktop/animation[####].png
Then try splitting the task into 4 blocks in 4 separate command-lines, just copy/paste into each window and adjust the numbers, it's best to do the changes in a text editor then copy/paste into the command-line:
/Applications/After Effects CC/aerender -project ~/Desktop/Animation.ae -comp "Overlays" -RStemplate "Best Settings" -OMtemplate "PNG Sequence" -s 1 -e 24 -output ~/Desktop/animation[####].png
/Applications/After Effects CC/aerender -project ~/Desktop/Animation.ae -comp "Overlays" -RStemplate "Best Settings" -OMtemplate "PNG Sequence" -s 25 -e 49 -output ~/Desktop/animation[####].png
/Applications/After Effects CC/aerender -project ~/Desktop/Animation.ae -comp "Overlays" -RStemplate "Best Settings" -OMtemplate "PNG Sequence" -s 50 -e 74 -output ~/Desktop/animation[####].png
/Applications/After Effects CC/aerender -project ~/Desktop/Animation.ae -comp "Overlays" -RStemplate "Best Settings" -OMtemplate "PNG Sequence" -s 75 -e 100 -output ~/Desktop/animation[####].png
The latter 4 processes should be much faster than the single pass. You have to be mindful of memory usage because each process allocates its own memory and not all parts of a sequence are equally intensive so choose the splits accordingly, you can experiment with which amount gives the best results. Adobe should really have a UI for aerender to allow you to quit AE and spool up multiple processes for a project.