Welcome to our community

Be a part of something great, join today!

  • 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 :)

Something Special... ( REDCINE-X )

Gotta agree with Jarred, 4GB on a Kepler is the minimum I would even try.

IAC, can you confirm that you are using OpenCL vs CUDA for GPU debayering?

I hate to ask, as I know you were aiming at IBC for the rollout, but this is the one spec that would be really helpful to know sooner than later.

Cheers - #19
 
Jarred, do you have a booth at IBC "revealing" the new software or do you team up with someone else? :)

Great seeing you at NAB!
 
Hope all you Wintel pimps get this out of your system before the MacPro actually ships. If Apple and ATI are willing spend the resources to code the living snot out of the GPU<->CPU interface so they can really leverage dual FirePros with 12GB of VRAM each on 22nm dies with a wide path to the mobo it might take a Godzilla PC build to smoke it. Apple has traditionally gotten more of the potential performance out of hardware than anyone else due to their tight control of the end to end ecosystem. They don't have to bloat their OS to accommodate all the component permutations Windows has to.

Even with the 2011 socket and decent CPUs there are still bottlenecks and kludges in play when running Windows on Intel silicon that choke performance, the dead weight of legacy code. With Mavericks on the next gen chips, well integrated with beaucoup GPU resources efficiently utilized (might take an OS rev or two to get there) I expect a new MacPro with top of the line ATI graphics boards to scream - particularly when using apps that fully exploit OpenCL.

Cheers - #19
 
Hope all you Wintel pimps get this out of your system before the MacPro actually ships. If Apple and ATI are willing spend the resources to code the living snot out of the GPU<->CPU interface so they can really leverage dual FirePros with 12GB of VRAM each on 22nm dies with a wide path to the mobo it might take a Godzilla PC build to smoke it. Apple has traditionally gotten more of the potential performance out of hardware than anyone else due to their tight control of the end to end ecosystem. They don't have to bloat their OS to accommodate all the component permutations Windows has to.

Even with the 2011 socket and decent CPUs there are still bottlenecks and kludges in play when running Windows on Intel silicon that choke performance, the dead weight of legacy code. With Mavericks on the next gen chips, well integrated with beaucoup GPU resources efficiently utilized (might take an OS rev or two to get there) I expect a new MacPro with top of the line ATI graphics boards to scream - particularly when using apps that fully exploit OpenCL.

Cheers - #19

The new Firepros in the Mac Pro don't have 12GB each, they have 6GB VRAM each... i think you are thinking of the new Nvidia Quadro K6000 cards ( only for PC's ) that have 12GB VRAM on each card.
 
Are any of our hacks - Jeff? :) - able to tell how upgradeable the new Mac Pro will be?

The RAM is shown to be swappable of course, but the flash storage, CPU and GPU are held in place with simple torx screws which scream "unscrew me!"
Especially the GPU makes me wonder. It seems the card is fixed but that you could exchange the processor itself?

As for Thunderbolt: I guess it's the same as adding a ROCKET in a Thunderbolt chassis, but would "any" extension GPU make a difference for a tool like the new REDCINE-X build, given it's got some "BAM"?
 
The new Firepros in the Mac Pro don't have 12GB each, they have 6GB VRAM each... i think you are thinking of the new Nvidia Quadro K6000 cards ( only for PC's ) that have 12GB VRAM on each card.

Hehehe, this sounds that you do have more insight than you're letting us know ?

Or is this general knowledge to what they are putting inside the black tube that I missed?
 
The new Firepros in the Mac Pro don't have 12GB each, they have 6GB VRAM each... i think you are thinking of the new Nvidia Quadro K6000 cards ( only for PC's ) that have 12GB VRAM on each card.

OK, so what can we do about this. OS options are Apple or Linux. We wont use Windows of any sorts, not up to the job in our view. Can we but the black drain pipe and retro fir it.
 
The new Firepros in the Mac Pro don't have 12GB each, they have 6GB VRAM each... i think you are thinking of the new Nvidia Quadro K6000 cards ( only for PC's ) that have 12GB VRAM on each card.

My mistake, thanks for the note. 2x6 for 12 total is still a nice bit of VRAM on hand, especially if the utilization of the 2 cards in tandem is well implemented.

IAC, looks like my OpenCL query is premature, I'll await your posts from IBC. Congrats to you and the RedTeam for figuring out how to decode on commodity GPUs, that's going to make a lot of people happy.

Cheers - #19
 
This is indeed special, a development many of us have been waiting for years. Looking forward to it and hoping a SDK is released soon. Obviously I have a lot of questions, waiting for more information around IBC time.

Hehehe, this sounds that you do have more insight than you're letting us know ?

Or is this general knowledge to what they are putting inside the black tube that I missed?

Yes, it's common knowledge - 2 x FirePro W9000 for the variant showed off in June.

On a separate note, AMD has their next-generation GPUs (Hawaii) out next month, in good time for a stable build. Could be the GPU(s) to get - those will almost certainly compete with Titan for less money.
 
On a separate note, AMD has their next-generation GPUs (Hawaii) out next month, in good time for a stable build. Could be the GPU(s) to get - those will almost certainly compete with Titan for less money.

Thats what makes this so great. We gotta thank the relentless gaming community for their own accelerated interpretation of Moore's law. Next month a faster card from AMD, the month after that a faster card from Nvidia.

One of the reasons this took so long for us to be able to offload to GPU is that the cards just were not fast enough back when we started all this. There was the quadroplex floating around but it was pretty complicated and expensive, Rocket card was cheaper and easier. Now we finally have graphics cards pushing Ghz and more V memory than our entire computers had back in 2005.

A Rocket-X still is going to win in a fist fight. It's insanely fast. As fast as GPU's are, they still don't do decompression very well. But hopefully this will help alot of people speed things up when a Rocket just isn't available.
 
Hi Jarred,

Can we assume we would be able to output 4k via the GPU directly or do we still need a BMD or a Rocket to do it ? GPU debayer with a seiki could be the cheapest 4k workflow for indie, commercial of music video.
 
Can we get any ballpark figures on GPU memory requirements? I'm assuming it's going to be resolution dependent?

Just wondering if I'm going to have to look at upgrading from my two 3GB GTX580s.
 
Not only would you need to provide power from somewhere else, but you would lose the use of one PCIe slot. Both the 5770 and GTX680 are double-width cards and both require use of two power connectors. Mixing GPU platforms isn't a great idea either if you're looking at using both for computational or OpenGL work.

Jeff, as usual, thanks for sharing your rather vast knowledge. Much appreciated.
 
Let's hope this thread doesn't turn into a PC vs Mac discussion as honestly it should be a totally moot point in today's age as both systems/platform/OS perform virtually the same. Anyway, what I want to know more about is if RCX is optimized for Cuda mainly or does it perform just as good on AMD card using Open CL? (Just trying to figure out what video card hardware to spec for our next workstations)
 
WOW ... real range of emotions up in here. As entertaining as it is to read, and it is, let's just take a moment to appreciate that this is a discussion forum. It's a fantastic resource for tech support and exchanging of ideas, not so much for bitching because someone didn't give you exactly what you wanted, when you wanted it ... for FREE! Whether you're a RED lover, RED hater, or indifferent, just remember that if you can do it better? ... then do it, otherwise just watch your manners slightly.

(Step down off soapbox) Preaching over! :)
 
I knew there was a reason all those Rocket cards were hitting Ebay... ;)


Hmm I check daily on eBay and see one or two, how are you finding these?

Im stoked that we have all these options available soon.

Some of us are staying put with 5K and down so any option is a great thing :)

Again, thanks RED!!
 
Are any of our hacks - Jeff? :) - able to tell how upgradeable the new Mac Pro will be?

Very modular, very upgradeable. I have actually got a personal taste of one a few weeks back. ;)

That said, we are still at the mercy of Apple and other hardware vendors in terms of what components will be produced and offered. The most upgradeable system in the world is anything but upgradeable when there are no new components to upgrade to... So at this point all we can do is see what happens. I have a hunch that it will catch on if a lot of people start buying them.

For all of the people who are denouncing the system because a PC will be more powerful... Yeah, sure, a dual processor PC is going to have more power. OTOH, not everyone wants or needs a dual processor monster. This new Mac Pro makes a lot of sense for the majority of workstation customers, IMO. For those who need to be on the bleeding edge in terms of absolute CPU power and GPU options, the PC is indeed where it's at. My request to all the dual processor PC snobs who have been coming out of the woodwork to trash-talk the new Mac Pro is this. If you really think you need the ultimate in PC performance and power, why are you sporting a dual CPU box? You can stack 8 (yes, 8 of them!) Xeon E7's in one system. Currently that's 80 cores, 160 threads. And it will jump to 128 cores, 256 threads in 2014. Chew on that for a bit. Your badass dual Xeon box is only about 25% of what's possible.

I'm not a Mac guy specifically, most of my systems are Windows PC's. Just keep things in perspective. Apple built the new Mac Pro to fit a specific market and I think it's going to do rather well. It's not a one-size-fits-all workstation. The current/previous Mac Pro did a shit job of being a one-size-fits-all workstation. I think the new one is a lot more fitting to their customer base.
 
Back
Top