-
Theoretical MacPro Replacement
I’ll start with this link posted by David Lawrence as point of reference.
https://www.marco.org/2011/11/02/scaling-down-the-mac-proI’ve posted elsewhere what my theoretical replacement would be but I think it warrants a new thread given the linked article and various comments people have made.
The new box would be:
Rack Mount in size (eg 19″) and a few RU in height (depends on cooling logistics).
SSD boot and one additional Hard Drive for internal storage and no further internal drive expansion.
One 16 lane PCIe for GPU and possibly a second 16 lane slot for a 2nd GPU or other demanding PCIe card and no other slots.
No internal optical drive.
Either 3 or 4 Thunderbolt ports.
The box would look like an overside MacMini.There’s been debate over i7 desktop processors vs Xeon for multiple CPUs. This (as well as other things) can impact the power supply and cooling.
I think people are overlooking a key solution as far as ways to expand CPU power.
Clustering
Compressor and QMaster are designed for this.If Apple designed a box with a lower entry price point in the aforementioned form factor, it would open it to a lot more people who need something more powerful than a Mini and more expandable than the iMac (which maxes at two Thunderbolt ports, a GPU that can’t really be replaced and a monitor sans choice). This box would be attractive to many more people than the current MacPro. One could either rack mount them or stack them. If you needed more CPU power you’d buy another box and Cluster them. I’m also mulling the possibility that the Cluster would be connected by Thunderbolt given the greater bandwidth compared to Gigabit Ethernet. The price for each box should probably target the $1800 to $3000 price range depending on the CPU/GPU it comes with.
Granted this may not be the solution for everybody but the above “modular” system would cover a lot of the higher end needs with probably higher sales and margins for Apple.
We can’t overlook how Apple has attempted to make built in clustering possible with QMaster. Of course such a setup would allow the boxes to work as separate lowered powered workstation. Imagine two NLEs but clustered when doing encoding or rendering for advanced compositing and graphics works.
-
32 Replies
-
Your simply ingoring the fact again that only a handful of encoding options can be clustered. The rest cannot! Plus compressor isn’t everything. For a lot of jobs Compressor is unusable and other applications need ONE beefy machine with not clustering options at all.
-
Sounds good, now if we could only find someone to make them.
Herb Sevush
Zebra Productions
—————————
nothin’ attached to nothin’
“Deciding the spine is the process of editing” F. Bieberkopf -
Don’t speak in generalities. Be specific. QMaster isn’t just about Compressor though. It was also designed for use with the now defunct Shake and it may include other functions as Lion expands. Clustering is common with a lot of high end graphics work.
-
It already does right now. You can use Qmaster to send jobs from After Effects or Nuke to a cluster.
But you need a second license of the Applications and all the plug-ins you are using. Plus you need some sort of shared storage where everything can render to. So if the current Macpros don’t sell because of price, your clustered solution would cost three times the amount at least.MVC encoding cannot be done under OSX at all. Under bootcamp, your cheapest option is 6000 dollars and cannot be clustered, 90 minutes take approx 20 hours to encode. Next stop: Cinevision which allows distributed processing at roughly 40.000 dollars starting point plus additional licenses. Then there’s Cinemacraft at 60.000 dollars.
Plus theres the whole 3D realm I cannot really comment on because it is not my field but given the prices of some high-end cuda cards I doubt anybody on that area is looking forward to your vision of future power computing. -
Clustering for any serious work requires significant networking and shared storage infrastructure. Building a cluster of lower-performance, lower-cost nodes to replace a higher-performance, higher-cost workstation may be penny-wise and pound-foolish.
That’s not to say that i7 doesn’t make a nice and cost-efficient render node. Rather, a render cluster doesn’t necessarily remove the need for a high-speed workstation on an artist’s desk. There’s a lot of time overhead associated with submitting and collecting network renders, so when time is money, saving a couple thousand dollars on the artist’s desktop doesn’t always pay. A powerful, local workstation for previews and a cluster for bigger test renders and for final renders is ideal.
I like Qmaster quite a bit for its ease in establishing a cluster, but it’s far from ideal as a production render manager. Its control for third-party renderers is somewhat naive, lacking any sort of render monitoring or dynamic reassignment. Since Qmaster has been mostly untouched since its initial release, I’m not optimistic that it’s very high on Apple’s development priority list.
Walter Soyka
Principal & Designer at Keen Live
Motion Graphics, Widescreen Events, Presentation Design, and Consulting
RenderBreak Blog – What I’m thinking when my workstation’s thinking
Creative Cow Forum Host: Live & Stage Events -
[Herb Sevush] “Sounds good, now if we could only find someone to make them.”
True. I’ll admit this all my wishful thinking at the moment. At least I think it makes sense given Apple’s business model as well as where I think Thunderbolt itself might be used for.
In fact, as rumors surface about Apple “consolidating their computers” (there have been rumors about an Air like 15″ MacBookPro for example and the MBP line may head in the direction) it seems at least possible that the same thinking might result in a large rack mount sized box using the Mini design. I believe common design elements can cut production costs.
I think when people are debating consumer vs pro direction they’re not seeing something. It’s really more about how they’re trying to commodify their production yet maintain margins. That means products would have a broad, rather than niche focus. That doesn’t mean forsaking power. It does mean that the product must serve multiple targets and purposes. The modular approach allows for that.
The drawback in all this, especially compared to PCs, is that Apple seems to always be driven towards fewer (not worse) choices. It may well mean having to buy two 6 or 8 core boxes rather than having a single 12 or 16 core box for example. Keep in mind Apple would want to drive down the price of a single box to make this work. More sales and fewer choices drives down the cost of component purchase price for Apple. Ultimately to get to the same power it might cost as much but it allows a small business or facility to start with a single box. That’s important in this economy.
One can only hope that Apple is thinking in this direction as opposed to just abandoning high end multi core needs. Clustering a bunch of iMacs is not going to look elegant or practical in many facility environments with a monitor stuck on each. Also the lack of higher end GPU options would be another problem. That’s why a new box that can target a variety of needs makes sense.
Cook, Ive are you listening?
-
Walter, I think you’re assuming everything is going to remain as is. New design means new development motivation.
[Walter Soyka] “There’s a lot of time overhead associated with submitting and collecting network renders, so when time is money, saving a couple thousand dollars on the artist’s desktop doesn’t always pay”
[Walter Soyka] “Since Qmaster has been mostly untouched since its initial release, I’m not optimistic that it’s very high on Apple’s development priority list.”These are based on assumptions that Apple won’t be addressing them along with the hardware changes. I’d think they would by necessity . . . which won’t exist until the design changes happen.
Compressor 4 shows some hints of this just in improved ease of clustering.
-
[Craig Seeman] “The drawback in all this, especially compared to PCs, is that Apple seems to always be driven towards fewer (not worse) choices.”
A blessing and a curse. This is alternately one of my favorite things about Apple and one of my least favorite things.
[Craig Seeman] “It may well mean having to buy two 6 or 8 core boxes rather than having a single 12 or 16 core box for example.”
Developers have really just caught up to multithreading and multiprocessing in the same box, and we’re at the dawn of GPGPU. Clustering trades raw performance for scalability, but adds all kinds of new challenges which I don’t think will be overcome overnight.
A cluster of two half-powered machines is just not a substitute for a single full-powered machine.
Walter Soyka
Principal & Designer at Keen Live
Motion Graphics, Widescreen Events, Presentation Design, and Consulting
RenderBreak Blog – What I’m thinking when my workstation’s thinking
Creative Cow Forum Host: Live & Stage Events -
I like the macmini cluster idea, and I think it makes sense to further streamline Apples product line..but wishful thingking-wise…I wish they just bring back the most beautiful computer ever made next to the iMac G4…the G4 Cube.
Theoretically, I say, design their own 64-bit ARM chip, place in some 48-core version on a cube, that will take like 5% of the space inside. The 95% of the space left will be for a 512GB SSD, 2 GPUs,PSU and a fan.
There, the new pro level Mac Pro. The only ports youll see below are 6 thunderbolt ports, 2 of them optical.
Quobetah
New=Better -
[Craig Seeman] “Walter, I think you’re assuming everything is going to remain as is. New design means new development motivation.”
You’re assuming that nothing will remain as it is, or even on the same trajectory as it is. You’re speculating products and technologies which don’t yet exist, and suggesting them as reasonable solutions to the potentially imminent demise of the Mac Pro.
I’m pointing out today’s real-world challenges inherent in even small-scale cluster computing.
I just don’t see Apple focused on this area, especially since they abandoned enterprise computing. Even if they do develop an amazing new technology, it will take time before the rest of the industry can adapt their products to take advantage of it.
[Craig Seeman] “Compressor 4 shows some hints of this just in improved ease of clustering.”
Have you tried to integrate Qmaster into any workflow, other than to feed Compressor? It’s really quite limited.
Walter Soyka
Principal & Designer at Keen Live
Motion Graphics, Widescreen Events, Presentation Design, and Consulting
RenderBreak Blog – What I’m thinking when my workstation’s thinking
Creative Cow Forum Host: Live & Stage Events
Reply to this Discussion! Login or Sign Up