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

Two to Tango

Two to Tango

Optimizing software is the crux of the biscuit. In the first few months of the nMP's life the real stars are likely to be Apple programs like FCP-X and of course OSX Mavericks.

It will be interesting to see how companies like Adobe behave in terms of allocating engineering resources to coding for the new topology. Perhaps the "canary in the coal mine" leading up to NAB is ColorFront's Express Dailies product on the nMP. Their engineering is as good as I've seen for genuine professional solutions running on commodity hardware.

For the last 3 years or so nVidia/CUDA has been the hook up for leveraging GPU power for pushing pixels. Sophisticated OpenCL/GL implementations need to play some serious catch up or we may continue to see poor benchmarks for many programs on the nMP.

Bruce asks a good question about how many resources software vendors are willing to commit to seriously supporting the nMP. Apple could make it twice as good for half the price and unless you only use Apple apps it's not going to get far without that support.

Cheers - #19
 
Blair S. Paulsen said:
Bruce asks a good question about how many resources software vendors are willing to commit to seriously supporting the nMP. Apple could make it twice as good for half the price and unless you only use Apple apps it's not going to get far without that support.
Click to expand...

That's kind of a non-issue, since Blackmagic/DaVinci, Autodesk, and Colorfront have already committed to the platform. From what I've already seen and know, Open CL has, if anything, a better upside and future roadmap than any single sourced approach like Cuda, even though that platform has hardware support directly from its owner. The fact is that the industry represented here, while small, is the target for a number of software publishers. And Apple is a rather significant, if not the somewhat dominant, platform in that industry. I don't see that changing any time soon.
 
Jeff, sorry if I missed the post but did you order?

PCG
 
Bruce Allen said:
Jeff, imagine for a minute that Apple put you - Jeff Kilgroe - in charge of Mac Pro hardware decisions. Would you have recommended that they do what they have done?

EG put two old-ish, mediocre AMD cards in their spiffy cylinder with expensive Xeon chip? Instead of putting, say, a single 6GB Titan in them as standard (maybe with option to add another 1 or 2 Titans)?

That would give great OpenGL performance and decent compute performance with existing apps, plus the option for amazing compute performance for those who actually need it.
Click to expand...

In keeping with the cylinder design, I would have designed something a bit bigger and more modular. Perhaps something based around a square central chimney to add in 4 boards instead of 3. Could have had one primary CPU board and then 3 more that could have been user-spec'd. Mix and match -- could have been a quad-CPU system or one CPU plus 3 GPUs or dual CPU + dual GPU, or 1 CPU + 2 GPUs + additional storage, and so on….

For GPUs, I'm not sure. Part of me wants to say that I would have loaded it with dual Titans. The other part of me wants to agree with Apple and make a push to OpenCL while cutting CUDA out of the equation. I agree there are more powerful GPUs, but when you look at the overall specs of the D700/W9000, I don't think they made a bad choice. It may have been the best choice due to a number of factors we're not aware of. All the problems with those GPUs up to now have been software problems, not hardware. I'm not placing a whole lot of stock in anything, OpenGL benchmarks or otherwise, in regards to software running on them before the software is better tuned and drivers mature a bit. We already have an EFI update for the new Mac Pro to address some GPU performance issues. OSX 10.9.2 has both updated OpenGL and OpenCL modules to better support the Mac Pro. I'm not seeing any benchmarks yet running the 10.9.2 developer seed.

For GPUs, I think I would have gone with the R9's if it were up to me. CUDA is both a blessing and curse to nVidia. Sure, that's where the best performance is at, at this moment. CUDA is proprietary and is a big turn-off to many out there. While Apple does have a somewhat walled garden in a sense, it's not so simple. They are pushing a vertical market in terms of their hardware combined with their OS tuned for that hardware. Everything else they like to keep pretty much open and do a rather good job of that.

In pure OpenCL computational sense, as it currently stands, two D700's (assuming a 12.5% under-clock of a W9000) are greater than 1.5 Titans. And for what seems to be about the same price as 1.5 SuperClocked Titans. In PC land, I've seen OpenCL FP32 benches that put the W9000 at 3.1 Tflops and a SuperClocked (7% over) Titan right at the rated (for a standard Titan) 4.5 Tflops. I'm betting Apple manages to coax at least the 3.1 out of the D700's and get close to the rated 3.5. If they manage that, for something right about 7 Tflops combined, at what amounts to a $600 upgraded GPU option, I think they will have done well.

I'm not going to sit here and complain about what the system could have been or should have been. This system serves a market and I think it will do quite well with their target audience. Yes, they could have put in newer, bigger, better GPUs. Yes, they could have put in dual CPUs. So what? They could have put in 4 CPUs. They could have put in 4 GPUs, they could have built in 8 drive bays and made hardware-accelerated RAID-6 built of 1TB SSDs a standard configuration. They didn't. They built a system to serve the segment of this industry or other industries with similar hardware needs. Considering the needs of a very large and diverse professional market, I think they did quite well. And by the time this system is due for an upgrade in the next 18 to 24 months, most are not going to want to swap out GPUs or the CPU board or whatever, but will probably opt for a whole new system. This is the last iteration of Xeon for the "Romley" platform. Next up in Q2 2015 we get get the C8xx chipsets, Haswell (possibly skip to Broadwell) core based Xeons, PCIe v4.0, USB 3.5 and Thunderbolt 3.


Kwan Khan said:
Respectfully disagreed
Click to expand...

I'm confused. It looks like you're agreeing with me. :)


PC Greene said:
Jeff, sorry if I missed the post but did you order?
Click to expand...

I did. 8-core, 64GB, 1TB PCIe SSD, D700's. Delivers "January"
 
I'm still surprised apple didn't include an option for an x8 or x16 PCIe lane iPass connection with x2 or x4 TB2, instead of x6 TB2 ports. The new Mac towered would have been instantly compatible and upgradable with existing PCIe hardware.
 
I think Nvidia will likely have GPU options at some point. I believe Apple went with AMD because they got a good deal like Microsoft and Sony did and because AMD and Apple are both deeply committed to OpenCL which Apple has made mandatory for it's ecosystem. Even the iPod uses OpenCL for stuff.
 
Brian Smith said:
Don't know what's going on there, but that video looks bad… nothing like 4k should look.
Click to expand...

I'm not sure if you're being humorous but there is 18 cheesy effects piled onto the 4K image and still manages to playback without dropped frames.
 
Bob Gundu said:
I'm not sure if you're being humorous but there is 18 cheesy effects piled onto the 4K image and still manages to playback without dropped frames.
Click to expand...


This is awesome! Man, can't wait to get my hands on this. Definitely impressive. Hopefully is will work smoothly with 5K. Not expecting to be able to add that many effects with 5K…but smooth playback without a rocket would be snazzy.
 
Bob Gundu said:
I'm not sure if you're being humorous but there is 18 cheesy effects piled onto the 4K image and still manages to playback without dropped frames.
Click to expand...


Haha no kidding, I guess maybe next time they should stated 14+ effects added but not show the effect :)
 
Remember though that this video is not played back in 4k resolution but more like in 1k resolution.
This will probably make the calculations a lot less heavy. Also most of those filters are pretty useless. What we perfectly would need is
- display at 4k. With real 4k media. Doesn't have to be red media (that will only play with gpu sdk or Red Rocket).
- using plugins or filters that we use ourselves. Neat Video comes to mind. Magic Bullet I never use since they are just too slow, I doubt they will be that much faster.
- differences with 2k. How fast is it to export.
/ndreas
 
Bob Gundu said:
I'm not sure if you're being humorous but there is 18 cheesy effects piled onto the 4K image and still manages to playback without dropped frames.
Click to expand...

Oops... I was looking at something else when writing that response.
 
Man, kind of a drag about how its currently working with Premiere. No real point to buy in yet…unless a possible jump ship to FCPX…unless Adobe can release something soon. Whats everyone else's thoughts?
 
Matt Gottshalk said:
My thoughts are, that as a 100% Adobe user, a current or late model 27" Imac is a better value.
Click to expand...

For the time being, yeah. Make no mistake: the entire industry is moving towards multiple CPUs and GPUs, and I'm talking computer industry, not TV/Film. Eventually, anything worth using, will work better on something like the new Mac Pro versus something like the current iMac.

But it'll take some time for the software to catch up. Nature of change, and all that.
 
Bob Gundu said:
I'm not sure if you're being humorous but there is 18 cheesy effects piled onto the 4K image and still manages to playback without dropped frames.
Click to expand...


This machine sounds perfect for me as my timelines are always littered with 18 cheesy effects per clip. ; ) !
 
There's commitment, then there's buying a ring...

There's commitment, then there's buying a ring...

M Most said:
That's kind of a non-issue, since Blackmagic/DaVinci, Autodesk, and Colorfront have already committed to the platform. From what I've already seen and know, Open CL has, if anything, a better upside and future roadmap than any single sourced approach like Cuda, even though that platform has hardware support directly from its owner. The fact is that the industry represented here, while small, is the target for a number of software publishers. And Apple is a rather significant, if not the somewhat dominant, platform in that industry. I don't see that changing any time soon.
Click to expand...

Thanks for the notes Mike, it reinforces my underlying premise that as long as a few key vendors are fully committed to optimizing for the new MacPro platform we should see excellent performance - despite the single CPU, AMD GPU only (so far) formula. As Jeff notes its typically been the drivers, not the hardware, that has put nVidia in front of AMD at similar price points. Deep integration of the CPU and GPU resources, internal data pathway management and other "tuning" made possible by Apple's extensive scope of control constitutes significant leverage in real world performance metrics.

ColorFront, Assimilate, Autodesk, FilmLight, etc are obviously key vendors for many of us, that said, for thousands of content creators the world over, when choosing a platform its mostly about Adobe. Premiere is unlikely to evolve as far down database boulevard as FCP-X, but if Adobe is willing to do extensive coding for OpenCL and 64 bit processing there's every reason to believe their apps will fly on the nMP.

From the Adobe perspective it has to be a bit of a conundrum. In terms of straight editorial, FCP-X has a huge head start over Premiere so catching up will require a lot of coders on the payroll. There's the whole Flash schism. If Adobe does make the nMP more attractive by tuning their apps to run fast on it, they could drive more sales for Apple. If they see Apple as a direct competitor on the software side is that helping the enemy? Ultimately, as Mike suggests, for some software tools ignoring the Mac fans is bad business due to its popularity amongst creatives/image processors. Would have been fun to be a fly on the wall for some of the internal discussions that must have happened at Adobe over the last year or two...

Cheers - #19
 
You must log in or register to reply here.
Back
Top