Welcome to our community

Be a part of something great, join today!

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 :)

New Mac Pro

The whole thing with the 4K projector is that it's laser-powered and even though the government regulators have loosened their regulatory sphincter in regards to using lasers in projectors that will be in home/public areas, it's still a bit of a battle to appease them and RED has taken a "They'll show theirs and we'll show ours!" routine for product launches, so it's not surprising to me that they haven't let the projector loose yet given the beta testing roll-out that their competitors are in with their projectors.
 
After all the news, I'm still happy I went for the Z820 16-core 3,10GHz instead of waiting for the Mac Pro.

I'll throw in a new and more current GPU in there a couple of years down the road and wait for 3,5" large SATA6 SSDs to get cheap. HP just also announced a Thunderbolt 2 PCI card coming early 2014, which is great.

Now, if only there was a 64-bit Quicktime implementation for Windows... That 32-bit server is ridiculous. Rendering 5K R3D files down to H.264 is actually faster than rendering out a simple Prores 422 - all because of lame Quicktime for Windows, the encoder is crippled and not using CPU efficiently. Basically anything in a QT wrapper is sluggish on Windows and I guess that is the way it is going to stay unless Apple come out with an update... but why would they, really?
 
Robert Ruffo New said:
Not quite as it's line skipping, but no editor needs more than that to work. You can render full res at the end.
Click to expand...

No, it is not "line skipping." The nature of wavelets is that the image is sampled at multiple resolutions, allowing the image to be extracted at various resolution levels. What you lose when you extract the lower resolution wavelets is high frequency detail, which is why the image looks progressively softer as you go through the multiples of 2, in other words, 1/2 the full resolution, then 1/4, 1/8, and so forth. Each wavelet extraction is a full resolution image at the level it represents, not "line skipped." Different compression schemes use a different number of levels, but all wavelet based approaches work essentially the same way. There are a lot of resources on the Internet regarding wavelet compression. I would suggest looking at them before making statements that are not accurate.
 
The New Mac Pro is tempting but, with all the new cpu from Intel with lots of core available on windows-based machine, the forthcoming GPU acceleration in Premiere following the update on the Red SDK.
I'm not sure I'll invest in it. But it's tempting.
 
Apple's future.

Apple's future.

Going beyond high end use, is Apple killing itself? I went in to get a local price on the Mac Pro, $3999 for the bottom end version. This us nearly double the $1999US it is worth. I did not get the 12 core price. This is going on for years, but what do you expect from a company that killed the innovative Newton product, and the very good arm based notebook with it, and replaced it with colored plastic. The Imac have been continuing the bland downward spiral for some years. A thin screen that is certainly not a thin screen is not an big innovation, unless it is a small Asian company. No 3D, no touch, no tablet mode, no add others here. A premium brand. Looking at the Mac Pro, it is, nope not much, again, another Asian like innovation. Years of nothing much, now no 4x4k ports despite two AND GPU chipset, no PCI card/s. Maybe tb ports can be used to get up 4x4k.
 
Daniel Kelly Brown said:
As a Premiere editor, I would not invest in this machine simply because of its lack of NVIDIA GPUs.
Click to expand...

I understand fanboyism from being an Nvidia fan myself but CUDA is dead and it won't be gaining any share of the software being made in the future as GPGPU applications have almost entirely moved to OpenCL across the board.
 
the mac pros look pretty awesome to me....of course it will be interesting what the final prices will be but i honestly did not think we would see one starting at 3 and 4000....i have a rocketX on order but i am pretty sure that a configuration that allows me to handle 6K very nicely will come in at about the same price.....so in a way i might actually save some money....but to me it really comes down to how fcpx and resolve will be made for/with this machine...as much as it obviously makes sense that a crazy fast machine decked out will generally perform well, i think that the perfect hardware software combo can be much leaner (on both sides) and much more efficient....i understand that everybody has different needs and wants but this looks pretty perfect to me...and so far at a better price then i expected....
 
Not sure what you mean by 4x4k ports. The new Mac Pro can drive up to three 4K displays simultaneously (two via Thunderbolt and one via HDMI). Do you need to drive more than 3 displays at that resolution from a single computer? Also confused about the prices you mention. The new Mac Pro starts at $2999...not sure where you got the prices you mention:
http://store.apple.com/us/buy-mac/mac-pro
 
Jeff Kilgroe said:
The E5-2600 V2 Ivy Bridge Xeons supply 40 lanes of PCIe v3, as pointed out. The GPUs in the Mac Pro each reside on a PCIe v3 X8 interface, plenty fast enough for any GPU available on the market right now or for the foreseeable future. By the time faster GPUs are available that can saturate more than 8 lanes of PCIe v3, PCIe v4 will be out and so will new Mac Pro systems...
Click to expand...

That's not really true, or better said, it is true in the general case, but not particularly true in lots of niche markets for GPUs, including video post. There are certainly very compute intensive processing steps that challenge current GPUs (noise reducting, wavelet decoding, etc.), but lots of algorithms that are interesting to us, like for instance color transformations, scaling and blending are so simple and so well-matched to the GPU architectures that they don't even come close to being bottlenecked by the actual computing power of the GPU, but rather by the available local memory, and closely related to that, of course, to the bandwidth available in transfering frame buffers between storage, CPU / Memory, GPU and back. This becomes even more evident when you try to scale across more than one GPU, since you now have to coordinate the upload and download of frame buffers between the GPUs as well.

And it's really easy to actually overload even a x8 PCIe 3.0 bus with frame buffer transfers: a debayered full-resolution Dragon-originated stream at 25 fps and in floating point color needs about 5.8 GBps in order to be moved to and from the GPU. You'd think that the theoretical 8 GBps bandwidth available to a x8 PCIe 3.0 device would be enough, but if you actually try to benchmark what you can really achieve, you see that with overhead you land at about 6 GBps (somewhat higher in downloading from the GPU, somewhat lower in uploading to the GPU). That is really cutting it very close, in fact up the frame rate to 30 fps, and you are over the limit. Go wild, and imagine a 60 fps full resolution Dragon stream (HDRx, HFR, 3D, whatever), and you are well above even the theoretical limit of x8 PCIe 3.0, and scratching at the practical limits of a x16 slot.

Sure, you can compromise, and work in 4k, say that you don't need 32 FP buffers, optimize, try to use GPUdirect for DMA transfers between some devices to get closer to the theoretical limits of the bus, but in the end there is nothing like raw bandwidth when you have to work with such streams. And this is equally true for other applications, not just color-correction: a GPU-based compositing system may not have to work with Dragon streams, but it has to work with tens or even hunderds of layers of lower resolution material, which comes out to the same thing, if not worse. Even a GPU-based 3D renderer can become bottlenecked by transfer speeds - one of the reasons why the progress in this field has been rather bumpy and unsatisfying is not just that many algorithms are hard to port well to GPUs, but also the fact that the local memory of current GPUs is not large enough, so you have to constantly shuffle textures to and from system memory, which in many circumstances means that the GPU idles while waiting for the transfers to complete.

So even with today's formats you can get into situations where a x8 link between GPU and the system can become a bottleneck, you don't even have to look at the near future, let alone the forseeable one. That's some pretty non-forward-thinking system architecture in the new Mac Pro - granted, once they decided to go single CPU and put so many TB controllers in it, they didn't have any technical means to do better, but it is very much a compromise that can and probably will be felt in certain scenarios. No doubt not by the majority of the target market, but still.
 
L. Langer said:
I understand fanboyism from being an Nvidia fan myself but CUDA is dead and it won't be gaining any share of the software being made in the future as GPGPU applications have almost entirely moved to OpenCL across the board.
Click to expand...

Talking abuot Fanboys, i think you take the cake!
 
Jimmy Grewal said:
Not sure what you mean by 4x4k ports. The new Mac Pro can drive up to three 4K displays simultaneously (two via Thunderbolt and one via HDMI). Do you need to drive more than 3 displays at that resolution from a single computer? Also confused about the prices you mention. The new Mac Pro starts at $2999...not sure where you got the prices you mention:
http://store.apple.com/us/buy-mac/mac-pro
Click to expand...

Hmm, it doesn't have my country under the name forum panel. Well Jimmy, it is nearly $3999us in my country, a big hike in what was already very pricey. I was wanting 4x4k for 8k table surface touch computing, I long have sourced 8k display tech. Unfortunately my potential engineer decided he might become ceo of a really large company instead.

They seriously need to consider dropping the price of the basic machine to $1999, US and AU immediately. $3999 sounds excessive even for an i7 version. I suspect the future super iMac might be the reason to charge the current price. It is such a toy, if you changed the graphic chips and used a regular haswell and drive, it could be worth $999, are those things (with more memory sad etc) worth the extra nearly $3000+ no!


Thanks


Wayne.
 
Wayne Morellini said:
They seriously need to consider dropping the price of the basic machine to $1999, US and AU immediately. $3999 sounds excessive even for an i7 version. I suspect the future super iMac might be the reason to charge the current price. It is such a toy, if you changed the graphic chips and used a regular haswell and drive, it could be worth $999, are those things (with more memory sad etc) worth the extra nearly $3000+ no!
Click to expand...

If you want to set your own prices on things you buy, there's always Priceline. If you want to buy things from manufacturers, they set the price, not you. Apple is a pretty darned successful company and they aren't going to drop their prices by 1/3 simply because Wayne Morellini wants them to. Nor should they. If you don't like what they're charging, don't buy one. Simple as that. But "demanding" a certain price point that you deem appropriate is a bit ridiculous, especially when you have absolutely no idea what it is actually costing them to design, fabricate, manufacture, package, ship, market, and support the product, along with generating at least some profit. Not that it really matters anyway. It's Apple's product. They can charge whatever they want. People will either buy it at that price or they won't. If you think you can design, fabricate, manufacture, package, market, ship, and support a better product for less and still be profitable, then I suggest you try to do so. But if you can't, you really should stop telling them what they should or should not be doing.
 
You must log in or register to reply here.
Back
Top