Welcome to our community

Be a part of something great, join today!

Register Log in
  • Hey all, just changed over the backend after 15 years I figured time to give it a bit of an update, its probably gonna be a bit weird for most of you and i am sure there is a few bugs to work out but it should kinda work the same as before... hopefully :)

RED + NVIDIA Announce Realtime 8K R3D GPU Decoding

Blair S. Paulsen said:
AFAIK, nVidia placed their bet on CUDA and backed it up with tools for programmers that sped dev cycles.
Click to expand...

nVidia developed CUDA as a way to extend the capabilities of its GPUs. As the GPUs evolved so did CUDA... but it's what essentially created the market for GPU compute.

Apple appears committed to Metal2 and not interested in reliance on CUDA since they don't control it. Then there's the Apple vs nVidia feud...
Click to expand...

The irony here is that Apple developed OpenCL and submitted the spec to the standards board.

FWIW, unless Apple can exercise tight control over their ecosystem they're just a PC vendor with design chops and high prices.
Click to expand...

That's precisely what Apple is, and has been since the switchover to x86.

It would seem to me that supporting both AMD and nVidia GPU tech would be manageable enough to be worth the coding resources required, it's been SOP for years. That said, Apple standardizing on AMD could increase optimization, stability and simplify OS updates.
Click to expand...

At the expense of performance. AMD GPUs aren't likely to be competitive with nVidia's for a while. Even the Vega is only barely competitive with Pascal, let alone Turing.

nVidia is of course taking advantage of that; the prices stay fairly high and the Turing line has tensor and RT cores that are taking up space that could have been dedicated to more CUDA cores for example.

It's a gamble... but it will probably pay dividends, since AMD and Intel will have to include tensor and RT cores on their next generation GPUs now.

What would it take to get resource hungry applications/processes, built to leverage CUDA, rewritten to offer similar performance per watt/dollar on AMD/Metal2?
Click to expand...

AMD's GPUs are less powerful. Period. So...

Would Apple have to subsidize the coding expense? Would a company like Adobe cooperate even if Apple ate some of the coding costs?
Click to expand...

Apple would have to subsidize AMD's Radeon Technologies Group's R&D.

FWIW - Apple, please fully support nVidia GPUs and CUDA.
Click to expand...

If it's going to offer competitive performance, Apple will need to re-embrace nVIdia GPUs while Intel and AMD play catchup. And AMD doesn't have the bux.
 
Rakesh Malik said:
Apple would have to
Click to expand...

I won't speak for Apple, but part of their general efforts which are effecting generational audiences is to likely move to proprietary hardware in the form of CPU and GPU as well as other parts of the ecosystem. There are added benefits to controlling the whole wheel in terms of optimization and hardware that isn't tied to other company's development cycles for instance. This has led to a lot of their deals with various vendors. I think at this point, particularly in the last 3 years, it's been hitting professionals from hard to medium hard.

But something is certainly changing in their world while this development is going on. I think they are discovering slowly that they can support the mass audience of professionals while developing a fully enclosed ecosystem as that might effect each other down the line. The trick there is on the GPU side, Nvidia from a shear tech perspective is notably ahead in a few categories than anybody else. CPU-wise, it wouldn't be impossible to design better CPUs than either Intel or AMD can offer for sure. Especially with the trajectory of both lines.

Until that future unravels we can just ask for Nvidia driver support and likely some rapid attention/integration when needed to FCP.
 
Phil Holland said:
I won't speak for Apple, but part of their general efforts which are effecting generational audiences is to likely move to proprietary hardware in the form of CPU and GPU as well as other parts of the ecosystem. There are added benefits to controlling the whole wheel in terms of optimization and hardware that isn't tied to other company's development cycles for instance. This has led to a lot of their deals with various vendors. I think at this point, particularly in the last 3 years, it's been hitting professionals from hard to medium hard.
Click to expand...

It definitely seems like most of Apple's efforts have been aimed toward streamlining its mobile device lineups above all else, and going in-house for those processors was a logical thing to do there.

But something is certainly changing in their world while this development is going on. I think they are discovering slowly that they can support the mass audience of professionals while developing a fully enclosed ecosystem as that might effect each other down the line. The trick there is on the GPU side, Nvidia from a shear tech perspective is notably ahead in a few categories than anybody else. CPU-wise, it wouldn't be impossible to design better CPUs than either Intel or AMD can offer for sure. Especially with the trajectory of both lines.
Click to expand...

Designing a better CPU than what Intel and AMD can come up with is hard, but I agree not impossible; the catch is making it profitable. AMD and Intel aren't exactly slouches, and now that AMD has lit the fire under Intel's booty (again) things are starting to look interesting once more. :)
 
Rakesh Malik said:
At the expense of performance. AMD GPUs aren't likely to be competitive with nVidia's for a while. Even the Vega is only barely competitive with Pascal, let alone Turing.
Click to expand...

https://www.pugetsystems.com/pic_disp.php?id=50954&
$400 RX vega 64 is on the same level as the $1200 RTX2080ti.

https://www.pugetsystems.com/pic_disp.php?id=50955&
$400 RX vega 64 is on the same level as the $700 RTX2080.

https://www.pugetsystems.com/pic_disp.php?id=50957&
$400 RX vega 64 is on the same level as the $700 RTX2080.

That looks pretty competitive to me.

Rakesh Malik said:
nVidia is of course taking advantage of that; the prices stay fairly high and the Turing line has tensor and RT cores that are taking up space that could have been dedicated to more CUDA cores for example.

It's a gamble... but it will probably pay dividends, since AMD and Intel will have to include tensor and RT cores on their next generation GPUs now.
Click to expand...

AMD just included it on it's VEGA20 but won't include it on it's upcoming NAVI RX 3080 (or what ever it's name may be).
RX 3080 is a 150Watt GPU with VEGA64+15% performance for around $300, also pretty competitive.


Rakesh Malik said:
AMD's GPUs are less powerful. Period. So...
Click to expand...
The figures from PugetSystems indicate otherwise.



Rakesh Malik said:
Apple would have to subsidize AMD's Radeon Technologies Group's R&D.



If it's going to offer competitive performance, Apple will need to re-embrace nVIdia GPUs while Intel and AMD play catchup. And AMD doesn't have the bux.
Click to expand...

When a company needs Apple for their survival than they are doomed (Look what happened to Imagination).

AMD is perfectly capable to compete with every other hardware vendor in the IT business and it has one big advantage, one-stop-shopping.

And looking at the new posibilities with TSMC's 7nm: same performance with 0.5x the power or with the same power >25% more speed (TSMC states: same performance with 0.4x the power or with the same power >35% more speed) this would mean that the succesor of the TR2950x will have a base clock of around 3.6GHz and a boost of around 4.5 GHz all-core (32c/64tr) with a TDP of around 250 Watt(at 4.5 GHz all-core) and a price point of around $1.000.

Apple doesn't need to re-embrace NVidia and they will leave AMD as soon as they can.
 
Last edited:
Once the genie is out of the bottle, it probably won't be long before real-time decompression is done via other APIs/GPUs. One of the things that I found interesting (borederline aggravating) in that video presentation is that it took two nVidia engineers one weekend (e.g. two days) to get the shit up and running at 8k/24fps, while RED engineers only got 25%-33% that performance over ~8-10 years of development :/

So, yeah, unless they've got an agreement in place, AMD, Intel, Apple et al, could hopefully port it over to OpenCL, GL, Metal, etc. Hell, even RED said OpenCL wouldn't be too far behind (no word on if it'll run half as good as the CUDA version though).
 
Mike P. said:
This is the video presentation I was referring to (re: two nVidia engineers, two days for 8k/24fps)
Click to expand...
... with beautiful CUDA code and some nice new RTX Graphic Cards.
 
Yes the real question remains is why is this limited to CUDA?...
 
andrewhake said:
Yes the real question remains is why is this limited to CUDA?...
Click to expand...

Uhhhhh....... because Nvidia co-sponsored, participated in, and supported the development program? Because CUDA is a very well developed development framework? Because Nvidia hardware is the most advanced in the industry? Because it's available on the most widely used computer platform (although Jared made a very obvious point of encouraging Apple to settle whatever dispute it has with them and support that hardware.......)? Do I need to go on??
 
Misha Engel said:
https://www.pugetsystems.com/pic_disp.php?id=50954&
$400 RX vega 64 is on the same level as the $1200 RTX2080ti.

https://www.pugetsystems.com/pic_disp.php?id=50955&
$400 RX vega 64 is on the same level as the $700 RTX2080.

https://www.pugetsystems.com/pic_disp.php?id=50957&
$400 RX vega 64 is on the same level as the $700 RTX2080.

That looks pretty competitive to me.
Click to expand...

Those aren't exactly convincing graphs to compare true performance to. The 2080 beats the Vega 64 in every benchmark. I agree the price for performance is cheap relative to the Vega, but it's already behind for a flagship card.

From their own site: https://www.pugetsystems.com/recomm...miere-Pro-CC-143/Hardware-Recommendations#GPU
 
I suspect that the reason for the "why only CUDA" comment was really Apple related. Apple is currently the only platform I can think of that does not support the most recent Nvidia technology. That is Apple's decision, not Red's. And although AMD cards might be cheaper, they're not THAT much cheaper, nor are the Nvidia cards that expensive (an RTX 2080 non-ti is about $700...).
 
M Most said:
Uhhhhh....... because Nvidia co-sponsored, participated in, and supported the development program? Because CUDA is a very well developed development framework? Because Nvidia hardware is the most advanced in the industry? Because it's available on the most widely used computer platform (although Jared made a very obvious point of encouraging Apple to settle whatever dispute it has with them and support that hardware.......)? Do I need to go on??
Click to expand...

Yes that's all well and good. But it all just comes down to RED opening up their SDK to allow this to take place on the GPU. Nvidia/CUDA isn't the only platform capable of processing 8K R3Ds on a GPU in this way, it's just the only one being given access to this at the moment. Hopefully it will be opened up beyond CUDA, because competition is good.
 
M Most said:
I suspect that the reason for the "why only CUDA" comment was really Apple related. Apple is currently the only platform I can think of that does not support the most recent Nvidia technology. That is Apple's decision, not Red's. And although AMD cards might be cheaper, they're not THAT much cheaper, nor are the Nvidia cards that expensive (an RTX 2080 non-ti is about $700...).
Click to expand...

It's $800.

And it's Nvidia's decision as well. Imagine if Nvidia refused to make their drivers also work with DirectX? And only supported Optix and CUDA?
 
Andy_Johnson said:
Those aren't exactly convincing graphs to compare true performance to. The 2080 beats the Vega 64 in every benchmark. I agree the price for performance is cheap relative to the Vega, but it's already behind for a flagship card.

From their own site: https://www.pugetsystems.com/recomm...miere-Pro-CC-143/Hardware-Recommendations#GPU
Click to expand...

VEGA is more than a year old (14th August, 2017) and RTX2080 is brand new.

CES 2019 will bring the competitor for the RTX2080 at a whole new price level.

And with Adobe PP it doesn't make much of a difference if you use a $10.000 GV100 or a $400 RX VEGA.
 
andrewhake said:
Yes that's all well and good. But it all just comes down to RED opening up their SDK to allow this to take place on the GPU. Nvidia/CUDA isn't the only platform capable of processing 8K R3Ds on a GPU in this way, it's just the only one being given access to this at the moment. Hopefully it will be opened up beyond CUDA, because competition is good.
Click to expand...

True, but if Apple wasn't in some kind of dispute with Nvidia - and I would venture to say that very few, if any people here know what that dispute is or why it exists - we wouldn't be having this conversation, regardless of the competition or lack of it.
 
You must log in or register to reply here.
Back
Top