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 MacPro predictions? Hopes? Fears?

If you were Apple ... what would need to go inside the MMP to get 1,000,000 people to want to buy one rather than a 18 core iMac Pro?

Do you think Apple will create something that sits on your desk ... or will need to dissipate 1Kw?
Even with Apple turning up the money hose in the R&D department, there a lot of fundamental tech that will be the same as the iMac Pro.

To get the world to sit up and take the plunge - I think Apple will need to lower the cost of ownership and find a way to scale their offering past the currently perceived bottlenecks.

The most scalable change could be NVMe over fabric:
+) The ability to have a boat load of GPUs connected at a distance (GPU -> PCIe -> NVMe -> MMP)
+) Decoupling the (relative expensive) CPU to a dedicated NVMe over fabric.
+) Providing an order of magnitude higher Data Bandwidth.

https://www.nextplatform.com/2018/03/20/a-reference-architecture-for-nvme-over-fabrics/
http://searchstorage.techtarget.com...brics-Nonvolatile-Memory-Express-over-Fabrics

NVMe at WWDC .... Not sure this is ready for prime time.
With mass uncertainties and Merger and Acquisition - it might be too early for Apple to commit.
With Intel burying their 10nm CPU offerings in the Sand (so they can provide h/w meltdown fixes in the 10nm cores) .. I don't see Apple wanting to release anything new that depends on 'old' cores at WWDC either.

My guesses for WWDC this year :
+) ML : AR : NLP : <-> Watson
+) Siri <-> Opening up intension processing
+) Pushing MacOs frameworks iOSward (and Removal of 'unused' NS frameworks)


AJ
 
AJ, who are you talking too?

1KW, whatever works for sales. 1kw, is not much for a high end system. When you come to 3D workflows it's worse. 4 or 10 top end GPUs fed Bya top end 4 chip cpu system. How much will that be? The truth is to reduce power consumption, you need more customisation, and.more efficient other things. ARM, a Red media processing module (yes, this might be how you get GPUs too). Yes, an Apple Red deal might be possible too. Once you apply Red's technology to processing anything video codec, you get a much larger sales base making the module cheaper. The fact that it would presumably work best on Red codecs, is a bonus. You would make such a system with the GPU in it (practicality/power consumption permitting).
 
Wayne,

After ‘designing themselves into a thermal corner’ with the 2013 nMP, I was speculating into the ether, on how Apple might create a high Compute system without necessitating a 2kW+ heater (and the accompanying noise pollution) within arms reach of its user.

If Apple moves to a high speed (and low latency) fabric using something like optical connectors into a CPU coupled ASIC - they just might be able to create a small footprint 500w CPU Brain that replaces the nMP, and could then leverage cheaper flash, JBODs, and remote GPUs to create a system that can scale from casual to very high end.

+) 1 TB/sec (aggregated) fabric would be sufficient until 2019
+) A 4 TB/sec fabric might give such a system ‘legs’ for the lifetime of the CPU Brain.

AJ
 
With the iMacPro shipping, I would encourage Apple to release the mMP when they can offer a serious step forward. I think Anthony makes some crucial points about technologies that would help the modular architecture reach it's potential.

Cheers - #19
 
Antony Newman said:
Do you think Apple will create something that sits on your desk ... or will need to dissipate 1Kw?
Click to expand...

Who cares about power consumption of a workstation anyway?

I mean, seriously - we shoot on 50000+ cameras, with glas worth a new truck, let alone all other gear, plus spending a ton on computers, and now we are talking about a few bucks of our electricity bill (tax deductible anyway)?

The extra electricity is probably worth under 1%, of what we spend annually, to turn around our businesses.
 
Blair S. Paulsen said:
With the iMacPro shipping, I would encourage Apple to release the mMP when they can offer a serious step forward. I think Anthony makes some crucial points about technologies that would help the modular architecture reach it's potential.

Cheers - #19
Click to expand...
Lol! They're my points I've been posting about in the why don't they do a new Mac pro/technology threafs for years. I even just posted one about adci then to get the size/power down. And this thing about energy consumption being negative, the amount of energy used is proportional to the amount of work done. So if you don't need it that second,byih don't have to use it, and the machine can run cooler with less power. However, there is not enough processing power for everything, so expandable modules with high energy envelopes that aren't going to risk burning the place then are a must. It's all positively practical. You can get by with the maintain, good, if you need more, add appropriate more and multiple modules. Keep going till you hit what you need, and dr with the cooling consequences (like an fireproof rack enclose that will vent 20kw of computing power going up.in smoke (I would actually compartmentalize each blade so you might only loose one, and have a power down/cut out system, with fire suppression (but Apple doesn't work for me. So they probably aren't thinking of it). I so love industry level computing, just my sort of mindset).
 
Didn't mean to slight your contributions Wayne. Several of us have offered ideas for workstation performance delivered in new form factors. I've been a Mac guy for 30 years and watched them time several new hardware introductions to component availability, newer bus standards, etc. As they say, there is always something better coming soon - so at some point you just have to dive in. Determining the optimum critical path from design targets to actual shipping hardware is - like most engineering - about compromises. In the case of the mMP, I was simply joining the chorus that feels like the iMacPro gets you about as far as current components and desktop TDP limits can support.

IMO, the only way to make a mMP with currently shipping commodity components that would blow the doors off the iMacPro is to build up a beast that would be large, expensive, power hungry, etc. That approach is not what Apple has embraced over the years. I'm as ready for the mMP as anyone, but I'd rather wait a few extra months if there's significant opportunity to utilize newer tech with a longer useful lifespan - especially if it means optical coherent fabric or other super fast interconnects that would support robust modular topologies. Like many others, I really want a quiet edit/color suite - but I also don't want to run much of anything besides networking and storage pools over 10GigE. For bomber GPUs, co-processers, PCIe cards and such I'd want something faster and wider - ideally with support for longer cable runs than T3 over copper.

Cheers - #19
 
I was referring to Antony. But yeah, I hear you. I merely referring to another way to do it. The imacpro replaced the trash can in terms of light pro segment. It is not as much power dissipation as they could do, but if you don't want to try things like the screen or start fires it is that compromise in that format. Now decades ago, I wanted to do a tower compitrewith a monitor on the side of it. You could put 20 1080's in something like that, just cut a hole in your desk for the tower to reach the floor and propup your monitor (sounds like a good April 1st joke. An iMac pro rival that is a blade cabinet sticking up through a desk with. Monitor hanging on the front of it (and maybe as cheap as a fully decked out next gen macpro). But it is time for a MAC ULTRA.

But with an unencomberd modular Mac, you have one structure that can be on the desktop or on the floor and scsle to the cabinet. You start with a brain on the desk, and add to that through fabric. My own design was meant to have thousands of lines. Really nifty stuff for the game home market. The desk is able to run 1jw modules, but that's dangerous, so best fit a cabinet. If still too hot, you move it to the floor or out of the room. If you really want to be ambitious, you go up from micro blade modules to half and full width cabinet blades going up to 15A a blade for home use in a fire cabinet. But the reality is you could have hundreds of micro blades running at a kw each, whatever the cabinet can take. What Spllehad been missing out on is the server market. That would allow them to fill out their own server farms, give.tjem the differential profit to do their own hardware a d processors (dominate the market) and to pay for better workstation technology and processing, and better Imacs. Apple becomes an Intel and IBM, all the while feeding the technology and success into their other side of business, as well as adding to their profit (how much is domination of the server, infrastructure and workstation markets worth?). The truth of the matter is they are faulterimg, and 3brn if they were not, they would be saturating their market, and need to look elsewhere for dominance. Buying Nvidia out to run a new seperate computer division (Nvidia remains in charge, but runs commercial based product and games hardware and cards) would be a good idea. Nviduais already exploring professional, processing farmd, supercomputing and cars. Apple could get into the component business this way. Even if you don't buy an Apple product you maybe buying an product with Apple components in it. Even Steve would like that one.

So, there is a path there that scales from small innocent and low power to server farms.

Now, at 1kw, it's possible to make something pretty silent with passive airflow. Liquid oil immersion is one option, but of course they might not want that spilling out in somebody's room.

But the issue is that none of them have enough processing power. OIwould like to have 1000+X more. 10-100 GHz is barely scrapping it. Next decade they think that magnetic computing will deliver that. Why waste $30k+ on a system that isn't adequate. Might as well pick a lower target work space and buy something a lot cheaper. But fur yourself, you don't get it, to do the higher end stuff is going to require extra power and cooling. To get what you want further work may require fans.

There are two solutions for you. A longer blade where the case is long and narrow running along a wall or the back of a desk, venting up into open air. It might be six foot long and 25 cm high narrow case. So 8 or more gpu's one or more mainboards, and a expansion module that could house a drive array. Each (twin GPU though) in its own module. But, what's the point in sweating that out. The other solutions is a split system, like a air conditioner, where an immersion coolant (or air/gas) vents to outside where a heat exchanger extracts the heat to the environment. Now, your room Canberra almost completely silent (except for creaking from heat expansion). Doing this you would stack things to save space, even cube it.
 
I've used the iMac pro to edit 5k RED footage, and if the Mac Pro is anything as smooth, it will be money well spent IMO.
 
Far out, newly formed group, what happened to the previous effort. Measure twice cut once, do they mean measure ten times cut twice at this extended delay. This is not even an April Fools article, but makes the previous one look like it. Time to abandone Apple.

They are designing workflows, when what is needed is flexibility with more power, which could have been achieved even last year. I smell something, and I wonder if it is timed with Intel next generation product launch. It sounds like a tubeisch concept design or waiting for an Intel NG technology which might extend out or flop. It's so far out, they should have invested in magnetic computing and prepared something the size of the Mac Mini with the performance of the modular. It might even be thier own processing solution, which at least wouldn't be too bad. Actually, profiling workflows would be what they need to do for that, to refine it. But best to have done a sturdy modular shape and connector interfaces first and to upgrade to your own solution with minimal fuss afterwards.
 
Oh, I am so bummed out by this. Now investigating a switch to Linux for our main color projects...
 
I'm very buoyed by the revelation of the Workflow group. Apple's in it for the long game. It's not an accident that this information was released right before NAB, and right as a major new point release of FCPX comes out that addresses some difficult-to-solve problems in typically elegant Apple fashion: built-in, easy support for RAW and captioning. Think of how painful those two things can be... and they solved both in the most elegant way possible in a x.x.1 release.

Of course, I've got the horsepower I need right now. If the iMac Pro + a few eGPUs wasn't fast enough for me, I'd be in a world of hurt waiting for that Mac Pro to emerge. But I think when it does it's going to be killer.
 
John Wee said:
MY prediction is some ass clowns are still going to claim they can build their PC cheaper.
Click to expand...

And still be right and not an ass clown. Arm has come to windows desktop already. Apple can't do too much unless the actual circuits technology leap frogs. Which means their 'charging' should make it more expensive again.
 
so I guess according to Mr. Morellini, that when Adobe and Blackmagic Davinci use a HP Z8 for demonstration of their products, they too are ass clowns. Well, it's a good group to belong to. At NAB, Sonnet inadvertently showed the issue with an iMac Pro. They showed an iMac Pro, but with dual Sonnet eGPU boxes running the new AMD WD9100 GPU cards that are faster than the AMD Vega 64. There will constantly be new cards and accessories from all these companies, and if you can't upgrade constantly (without throwing out your iMac every year), then it's an issue. As I have stated earlier in this thread, when the new Mac Pro gets released in 2019, if Apple refuses to allow for THIRD PARTY DEVELOPMENT of hardware (like they are doing now with Nvidia) - then this will be an issue. Can you imagine if RED wants to come out with a new super duper version of the RedRocket-X (or anything else), and Apple says "nope - only a handful of products that we select, like the AMD Vega 64". Well, that could be an issue, and I see all the current nonsense with
macOS 10.13.3 and 10.13.4 and System Integrity Protection, that makes it almost impossible to load third party drivers without disabling SIP in Recovery mode. That is not what Steve Jobs envisioned when he said "for the rest of us". I understand that Apple is in business to make money, but they need to cooperate with other established hardware vendors in our industry. Do they HAVE to do this - of course not - they are Apple, they can do whatever we want. And then us "ass clowns" can choose another solution if we need to incorporate other hardware.

Bob Zelin
 
WWDC comes and goes and not even a mention of hardware. Yeah, it's a developer's conference but Apple always at least teases hardware here. They teased the iMac Pro at WWDC last year. With no mention of the mMP at all, starting to think the update is still a ways off, at least a year.
 
This thread is 2 years old, the "great new MacPro is around the corner" mantra is now 6 years old and in other news there is overcooking/throttling down to avoid overcooking.
 
You must log in or register to reply here.
Back
Top