Is the new Mac Pro a failure?

[koyoot]

Oculus won't be working on the nMP. #60

It will be working on the cMP. So that's an entire genre of professional media work that the nMP wont be doing.

I wonder if these guys dusted off their old cMPs or just ditched the nMPs for PCs to do their work:
http://www.thedailybeast.com/articl...-oculus-rift-virtual-reality-movie-henry.html

Netkas is wrong about it. Dual GPU setup will not work on crossfire cards, but with LiquidVR, if you have dual GPU setup two cards will work on two displays. One GPU - one Display.

Even if you will add latency on AMD GPUs together for Crossfire you will end up in 38 ns range, whereas on single GPU setup with Nvidia card you are on 57 ns.

But again, Dual GPU setup is not exactly crossfired under VR.

 

[ZombiePhysicist]

Netkas is wrong about it. Dual GPU setup will not work on crossfire cards, but with LiquidVR, if you have dual GPU setup two cards will work on two displays. One GPU - one Display.

Even if you will add latency on AMD GPUs together for Crossfire you will end up in 38 ns range, whereas on single GPU setup with Nvidia card you are on 57 ns.

But again, Dual GPU setup is not exactly crossfired under VR.

...um... yea.... First, I'm going to trust netkas over others on this issue, because, you know, he knows this stuff pretty well... actually, probably better than almost anyone.

Second, even if you can jerry rig the nMP to some how run this and might be able to squeeze out subobtimal performance and get things working on an nMP, it clearly will be suboptimal to a solution on a cMP. I dont think a lot of professionals will be pining away to do work on a suboptimal rig when they want way lower latency. They won't be doing work on it. They'll get a real system capable of doing that work well.

Realistically they wont be getting old cMPs either (despite them actually being capable of doing the work)--well except some freelance pros that still have some cMPs will be better situated to do this work than those that made the jump to nMPs.

The reality (no pun intended) is all this work is going to PCs and windows, and apple is completely out of an entire category of professional media production. And we have the nMP to thank for that.

 

[996085]

depends on if you're simply talking about the various components inside the machine or the whole of the machine itself..

cmp is simple sheet metal construction.. nmp is using manufacturing processes we've never seen in computer construction and to tolerances typically reserved for the aerospace or medical device industries.

this writer at this link:
http://atomicdelights.com/blog/how-apple-makes-the-mac-pro

gives a little more of a breakdown of what we see in this video:

just a guess but when you bought a cmp, the housing (minus any components etc) probably cost you less than $500..
nmp is probably costing nearer $1500..

if the value you find when purchasing a computer is simply limited to the actual components included and you find little to no value in the construction of the entire machine then yeah, you're definitely going to feel ripped-off if buying a nmp..
they could of put the same components from nmp ,and more of them, into a cmp box for a cheaper cost than the same components along with nmp housing/core..
further still, those same components could go into a plastic or cheaper sheet metal box (ie- certain PCs) for an even cheaper cost than cmp.

I was specifically addressing this point of yours:

"...this math assumes the cost of internal connections/bays/larger enclosure/etc is free.."

Spin it any way you like but the cMP offered more functionality than the nMP for a similar price.
[doublepost=1452546080][/doublepost]

If my 2009 cMP dies, I'm definitely not getting another one even if I do love it to bits! It can't even do 4K properly!

You need to quit thinking of the cMP in its actual implementation and start thinking of it as a cMP form factor with updated components.
[doublepost=1452546299][/doublepost]

if that's how you see it then hey, that's how you see it..

There is no other way to see it unless you place a premium on small size. And that only works if you don't connect anything to it. Once you connect something to it the small size is lost. In fact one might argue the larger tower saves more space once you factor in an external device.

 

[flat five]

I was specifically addressing this specific point of yours:

"...this math assumes the cost of internal connections/bays/larger enclosure/etc is free.."

Spin it any way you like but the cMP offered more functionality than the nMP for a similar price.

it's not spinning.

make the nmp 8x the size in order to equal the size of cmp (and allow for hd housing) and the thing is going to cost like $5000 entry.
[doublepost=1452546563][/doublepost]

You need to quit thinking of the cMP in its actual implementation and start thinking of it as a cMP form factor with updated components.

or maybe you need to quit thinking cmp is not a done deal..
it's gone. it's not coming back. you get that, right?
it's time to move on.
if you want a classic box tower, quit looking to apple for a solution.. you'll save yourself a lot of heartache/headache/aches in general.

 

[koyoot]

First, I'm going to trust netkas over others on this issue, because, you know, he knows this stuff pretty well.

Second, even if you can jerry rig the nMP to some how run this and might be able to squeeze out subobtimal performance and get things working on an nMP, it clearly will be suboptimal to a solution on a cMP. I dont think a lot of professionals will be pining away to do work on a suboptimal rig when they want way lower latency. They won't be doing work on it. They'll get a real system capable of doing that work well.

Realistically they wont be getting old cMPs either (despite them actually being capable of doing the work)--well except some freelance pros that still have some cMPs will be better situated to do this work than those that made the jump to nMPs.

The reality (no pun intended) is all this work is going to PCs and windows, and apple is completely out of an entire category of professional media production. And we have the nMP to thank for that.

I know this stuff also pretty well, never spoke about it, because I don't give a S*** about VR.

I will ask you, and everyone else who knows VR. What is the reason why Nvidia has big lag in VR on their hardware, that not allows good enough user experience in the end? There are 3 reasons for that, as a hint for you.

Secondly. I will ask directly you. If you want the lowest latency for VR, which hardware you will pick? AMD or Nvidia?

 

[ManuelGomes]

Some people here do get a kick out of things coming up that the nMP "can't do", right?
Well, why don't we just wait and see, shall we? Instead of starting off with "it won't", wait on it. It's not like tomorrow you'll start using the tech, still green as we all know. Let it go mainstream. By then, maybe even it's optimized to the point where it will run. Or, likely, when it gets mainstream the nnMP will be here (for sure) and will be a great machine for it. Complain about the current nMP? Well, cMP also needs newer cards to get there, right?
Haven't used it yet, but for most setups (and note the "most" word - should have bolded it) it should be a poor experience still, maybe some time from now it will be awesome. Not everyone has expensive/brutal GPUs to go with it.

Apple might be working on it, they have been buying some startups that could hint that, as well as other things of interest in the near future. Who knows what's to come?

 

[ZombiePhysicist]

I know this stuff also pretty well, never spoke about it, because I don't give a S*** about VR.

I will ask you, and everyone else who knows VR. What is the reason why Nvidia has big lag in VR on their hardware, that not allows good enough user experience in the end? There are 3 reasons for that, as a hint for you.

Secondly. I will ask directly you. If you want the lowest latency for VR, which hardware you will pick? AMD or Nvidia?

I wish I had netkas's skillset here to debate you. But I dont. And I wont debate you on his behalf. He can if he wants.

What I do know is netkas almost single handledly has made made a zillion cards work on the cMP and contributed more to the mac community than most here. He's helped hundreds if not thousands of people and runs a community to help them with crazy technical work (which I have to think may be painfully thankless oftentimes). His technical expertise is backed years of work and contributions to this space. I'm going to guess here and say, significantly more than you might have, but am all ears to hear how your technical prowess is greater than his.

So I'm going to go and rely on his statements on this over yours, unless you have some spiffy arguments/proof to the contrary. Hint. Just saying he's wrong is not at all compelling to me.

If you pursue arguments, please use small words and use pictures because I'm totally daft in this space (as well as in many others). About all I *think* I understand about the subject is you want latency down to 20ms or less visually so that people dont perceive lag and that the VR is compellingly responsive enough to fool the brain 'well enough' to think 'hey, this seems pretty realistic'. As for how that perceptual lag threshold works it's way down to the guts of video card tech, I freely admit I'm beyond clueless and as such, relying on the biggest guru (likely outside of folks working at apple) on the topic, Netkas.
[doublepost=1452547561][/doublepost]

Some people here do get a kick out of things coming up that the nMP "can't do", right?
Well, why don't we just wait and see, shall we? Instead of starting off with "it won't", wait on it. It's not like tomorrow you'll start using the tech, still green as we all know. Let it go mainstream. By then, maybe even it's optimized to the point where it will run. Or, likely, when it gets mainstream the nnMP will be here (for sure) and will be a great machine for it. Complain about the current nMP? Well, cMP also needs newer cards to get there, right?
Haven't used it yet, but for most setups (and note the "most" word - should have bolded it) it should be a poor experience still, maybe some time from now it will be awesome. Not everyone has expensive/brutal GPUs to go with it.

Apple might be working on it, they have been buying some startups that could hint that, as well as other things of interest in the near future. Who knows what's to come?

right, but for some people with cMPs they CAN use it. Right now. And the question posed was if there were things that could be done on the cMP that couldnt on the nMP. Asked and answered.

You're raising another point, that in the future, maybe it will be. Great if so. But that's another question all together. But in the mean time, cMP users dont have to wait, and the entire professional media field of VR is basically absent on the mac platform, IMO, because of the nMP.

 

[cerberusss]

The 2013 nMP lacked a 10Gbit Ethernet port

Yeah, I did wonder about that...

They must've obviously thought about it, and decided not to do it. But why not?

Perhaps it isn't used enough? I've actually never seen a 10 Gbit switch, and I just checked the price; the cheapest unmanaged 8-port 10 Gbit switch is from Dlink and costs about $825.

 

[koyoot]

I wish I had netkas's skillset here to debate you. But I dont. And I wont debate you on his behalf. He can if he wants.

I am not questioning his knowledge. He is absolutely right about latency with crossfire. But Crossfire will not be working on VR. It will be even possible to have AMD and Nvidia setup working on VR. Like I said: One GPU for one display.

I asked simple question. Why does Nvidia GPU require 57 ns latency for VR?

Dual GPU setups for VR already been discussed by Valve and work for AMD:

Here is first argument what is happening with Nvidia GPUs:

Secondly: You need as direct control by application of the hardware as you can.

Thirdly: Kepler and Maxwell does not have Hardware schedulers. The last architecture from Nvidia that had such thing was Fermi. And was inefficient, unfortunately. Thats why Nvidia went with software scheduling. To save power.

Next thing is that for scheduling Nvidia went with software, for Kepler and Maxwell. For VR, like I said before, you want as direct control by application over the hardware as you can. With Maxwell and Kepler you have software that controls both the application and hardware. That creates lag. What can Nvidia do in this case? If they are able to optimize the data paths - great. But from hardware point - nothing.

AMD on the other hand have Hardware schedulers in their hardware, at least on Tonga and Fiji chips.

 

[flat five]

But in the mean time, cMP users dont have to wait, and the entire professional media field of VR is basically absent on the mac platform, IMO, because of the nMP.

rift is, as of now, windows only..
here's your workflow on cmp using this plugin in fcpx (ie- currently, people don't edit with the things on their head.. only for playback)
http://uploadvr.com/virtual-reality...ut-pro-enable-you-to-make-live-action-vr-now/

so you edit in final cut.. reboot into windows.. playback.. boot back to osx to edit some more.. then windows.. etc.
meh.

that aside, you can most definitely produce VR clips for rift using nmp (or imac or mbp etc) with apple software.

you're looking at virgin technology here.. very very few of the 'entire professional media field' is going to have anything to do with VR in production realm.. the technology is not ready for that yet.. it's more for playing & experimenting for now.

 

[koyoot]

One more thing, lets not argue about VR, because this is future tech, and a LOT can change over next few weeks for... everybody. AMD, Nvidia, Rift, MS, Sony, whoever is planning to step into the game.

It is first time in few years that future is really unpredictable for any type of technology.

 

[bogg]

Thirdly: Kepler and Maxwell does not have Hardware schedulers. The last architecture from Nvidia that had such thing was Fermi. And was inefficient, unfortunately. Thats why Nvidia went with software scheduling. To save power.

Software schedulers for a GPU?
Not the way I'd do it...
[doublepost=1452548896][/doublepost]

you're looking at virgin technology here.. very very few of the 'entire professional media field' is going to have anything to do with VR in production realm.. the technology is not ready for that yet.. it's more for playing & experimenting for now.

The more posts I read from you, the more down-to-earth and reasonable you seem to be.
I don't understand why everyone is getting so carried away with this VR thing, sure it probably is the next big thing, but it will be years before any huge numbers of the average consumers have the tech in their homes. So there's no chance in hell the "entire professional media field" is doing VR stuff already. The market honestly isn't there and probably won't be for years...

This is an even greater change than the smartphone revolution, and that as well took a few years to penetrate the market completely.

 

[merlinmage]

Actually USB 3 did change (even though it kept the physical form so USB1.1/2.0 devices could be connected to it, but it does have more pins inside that USB1.1 and 2.0 does not.)

Which doesn't prohibit USB 1.1/2 devices to be plugged in USB 3 ports. Which is the whole point of this debate.

 

[ManuelGomes]

Everyone's all over it cause it kinda suits them now.
If we can tap at both GPUs individually in OS X (not XFire or anything), like koyoot says it's one GPU for each eye how the tech works, maybe the D700 will hold it's own, why not? Current res for each eye is low enough, and refresh rate. So why suffer (or beat on it) in anticipation?
By the time anyone has it and it runs properly and smoothly we'll be discussing 2018's nMP.

 

[996085]

it's not spinning.

Yep, it's spinning.
[doublepost=1452550867][/doublepost]

Some people here do get a kick out of things coming up that the nMP "can't do", right?
Well, why don't we just wait and see, shall we? Instead of starting off with "it won't", wait on it. It's not like tomorrow you'll start using the tech, still green as we all know. Let it go mainstream. By then, maybe even it's optimized to the point where it will run. Or, likely, when it gets mainstream the nnMP will be here (for sure) and will be a great machine for it. Complain about the current nMP? Well, cMP also needs newer cards to get there, right?
Haven't used it yet, but for most setups (and note the "most" word - should have bolded it) it should be a poor experience still, maybe some time from now it will be awesome. Not everyone has expensive/brutal GPUs to go with it.

Apple might be working on it, they have been buying some startups that could hint that, as well as other things of interest in the near future. Who knows what's to come?

This isn't a discussion about old technology versus new technology but rather cMP form factor versus nMP form factor.

 

[mcnallym]

I wish I had netkas's skillset here to debate you. But I dont. And I wont debate you on his behalf. He can if he wants.

What I do know is netkas almost single handledly has made made a zillion cards work on the cMP and contributed more to the mac community than most here. He's helped hundreds if not thousands of people and runs a community to help them with crazy technical work (which I have to think may be painfully thankless oftentimes). His technical expertise is backed years of work and contributions to this space. I'm going to guess here and say, significantly more than you might have, but am all ears to hear how your technical prowess is greater than his.

So I'm going to go and rely on his statements on this over yours, unless you have some spiffy arguments/proof to the contrary. Hint. Just saying he's wrong is not at all compelling to me.

If you pursue arguments, please use small words and use pictures because I'm totally daft in this space (as well as in many others). About all I *think* I understand about the subject is you want latency down to 20ms or less visually so that people dont perceive lag and that the VR is compellingly responsive enough to fool the brain 'well enough' to think 'hey, this seems pretty realistic'. As for how that perceptual lag threshold works it's way down to the guts of video card tech, I freely admit I'm beyond clueless and as such, relying on the biggest guru (likely outside of folks working at apple) on the topic, Netkas.
[doublepost=1452547561][/doublepost]

right, but for some people with cMPs they CAN use it. Right now. And the question posed was if there were things that could be done on the cMP that couldnt on the nMP. Asked and answered.

You're raising another point, that in the future, maybe it will be. Great if so. But that's another question all together. But in the mean time, cMP users dont have to wait, and the entire professional media field of VR is basically absent on the mac platform, IMO, because of the nMP.

You might want to cut back on the hyperbole.

We now have a claimed figure of a zillion cards flashed in cMP's. I don't know how many cMPs Apple sold as they don't release the figures, however I won't dispute your figure. ( Is convenient for me )

Now assuming that people have put two cards into a cMP then we are talking half a zillion cMP's out there.

So based on your assertion (wild exaggeration) of the figures then we have half a zillion cMP out there, yet only 228 user ( at time of typing ) that the nMP is a failure.

Perhaps someone better then me can do the maths however what percentage of half a zillion is 228. I don't see this convincing Apple to reverse course and go back to a Tower Format.

You might also want to cut back on claiming the cMP can do the VR as well. I have responded to your claims that the cMP can do this as well in the Oculus Rift thread with the d500's, using a post in there from MacVidCards that you may have missed in your eagerness to promote the all-conquering cMP as can do everything.

MacVidCards has tested Windows on his cMP fitted with the X5690 i.e. the best CPU you can fit in there and that doesn't meet the recommended figures for performance from the perspective of the CPU ( which you can't upgrade ) It doesn't meet the Single Threaded CPU performance that Oculus have set for the Recommended Figure.

So we are actually at the point where NEITHER the cMP or nMP can meet the recommended figures. You might want to go back and find another thing that can do on the cMP that cannot on the nMP.

 

[koyoot]

You might want to cut back on the hyperbole.

We now have a claimed figure of a zillion cards flashed in cMP's. I don't know how many cMPs Apple sold as they don't release the figures, however I won't dispute your figure. ( Is convenient for me )

Now assuming that people have put two cards into a cMP then we are talking half a zillion cMP's out there.

So based on your assertion (wild exaggeration) of the figures then we have half a zillion cMP out there, yet only 228 user ( at time of typing ) that the nMP is a failure.

Perhaps someone better then me can do the maths however what percentage of half a zillion is 228. I don't see this convincing Apple to reverse course and go back to a Tower Format.

You might also want to cut back on claiming the cMP can do the VR as well. I have responded to your claims that the cMP can do this as well in the Oculus Rift thread with the d500's, using a post in there from MacVidCards that you may have missed in your eagerness to promote the all-conquering cMP as can do everything.

MacVidCards has tested Windows on his cMP fitted with the X5690 i.e. the best CPU you can fit in there and that doesn't meet the recommended figures for performance from the perspective of the CPU ( which you can't upgrade ) It doesn't meet the Single Threaded CPU performance that Oculus have set for the Recommended Figure.

So we are actually at the point where NEITHER the cMP or nMP can meet the recommended figures. You might want to go back and find another thing that can do on the cMP that cannot on the nMP.

Maybe not, but If Apple would want to update the Mac Pro tower to anything new, it would meet the Rift requirements.

 

[bogg]

Which doesn't prohibit USB 1.1/2 devices to be plugged in USB 3 ports. Which is the whole point of this debate.

Sure, and I admitted as much. But then again, Thunderbolt isn't apples technology, is it really their fault that intel decided to change the connector with TB3?

 

[ZombiePhysicist]

rift is, as of now, windows only..
here's your workflow on cmp using this plugin in fcpx (ie- currently, people don't edit with the things on their head.. only for playback)
http://uploadvr.com/virtual-reality...ut-pro-enable-you-to-make-live-action-vr-now/

so you edit in final cut.. reboot into windows.. playback.. boot back to osx to edit some more.. then windows.. etc.
meh.

that aside, you can most definitely produce VR clips for rift using nmp (or imac or mbp etc) with apple software.

you're looking at virgin technology here.. very very few of the 'entire professional media field' is going to have anything to do with VR in production realm.. the technology is not ready for that yet.. it's more for playing & experimenting for now.

You cant boot into windows to produce the VR because the nMPs cards are suboptimal for the job.

 

[bogg]

You cant boot into windows to produce the VR because the nMPs cards are suboptimal for the job.

Suboptimal doesn't mean unusable, and as Flat Five seems to talk about movies or the likes rather than games (which you would need to compile and thus also more or less develop in a Windows based IDE), remember the system requirements for the Oculus Rift is for the most resource demanding applications, which should be games. Pre-rendered Movies and similar stuff should work just fine on a nMP....

 

[Roykor]

To wrap misconceptions about Apple, VR, and Rift I will say this.

There is a very big possibility that Oculus is not developing for OS X because... it is very likely that Apple is preparing their own VR hardware.

http://www.macworld.co.uk/feature/apple/apples-virtual-reality-vr-rumours-images-3601447/

And there is a word, that AMD provided few months ago for Apple engineering samples of their APUs, or custom designed hardware. What for? Apple TV/Console from Apple that can run VR.

Metal comes here into play, also.

Im writing this because right now, I can.

This is not going to happen Koyoot. Flat Five (expert in a lot of things) shared with us that it will take 5 - 10 years before this all will be mainstream. Sorry m8. I think you'r wasting your time! (sarcasm)

Anyway, interesting! Another VR headset on the line is always welcome. I will catch up the article later. Are there just a few developers working on it? In my network there is none i know of that is developing anything for the Apple VR device. If so, I would have seen some Apple VR hardware laying done somewhere They are all focussed on Rift (which I think will be leading the VR tech market soon) and HTC Five.

 

[ZombiePhysicist]

You might want to cut back on the hyperbole.

We now have a claimed figure of a zillion cards flashed in cMP's. I don't know how many cMPs Apple sold as they don't release the figures, however I won't dispute your figure. ( Is convenient for me )

Now assuming that people have put two cards into a cMP then we are talking half a zillion cMP's out there.

So based on your assertion (wild exaggeration) of the figures then we have half a zillion cMP out there, yet only 228 user ( at time of typing ) that the nMP is a failure.

Perhaps someone better then me can do the maths however what percentage of half a zillion is 228. I don't see this convincing Apple to reverse course and go back to a Tower Format.

You might also want to cut back on claiming the cMP can do the VR as well. I have responded to your claims that the cMP can do this as well in the Oculus Rift thread with the d500's, using a post in there from MacVidCards that you may have missed in your eagerness to promote the all-conquering cMP as can do everything.

MacVidCards has tested Windows on his cMP fitted with the X5690 i.e. the best CPU you can fit in there and that doesn't meet the recommended figures for performance from the perspective of the CPU ( which you can't upgrade ) It doesn't meet the Single Threaded CPU performance that Oculus have set for the Recommended Figure.

So we are actually at the point where NEITHER the cMP or nMP can meet the recommended figures. You might want to go back and find another thing that can do on the cMP that cannot on the nMP.

Well, along with the people at pixar that are pros to busy to be on this thread, those zillion are too busy too. See how that works.

Good point on the processor. I didn't see that post. Actually, I still dont see it, can you share the link (UPDATE: found it here #26). Thanks. Well, if the single thread aint enough on the 3.46ghz, which is pretty close to performance of the 2.7ghz 12 core, I suspect the nMP's processor is also not enough juice.

So I think you're right on the money. Apple doesnt have a professional machine that can address an entire professional media category. Pretty spectacular.

[doublepost=1452551726][/doublepost]

Suboptimal doesn't mean unusable, and as Flat Five seems to talk about movies or the likes rather than games (which you would need to compile and thus also more or less develop in a Windows based IDE), remember the system requirements for the Oculus Rift is for the most resource demanding applications, which should be games. Pre-rendered Movies and similar stuff should work just fine on a nMP....

Sub optimal means professionals will not be choosing the machine for such use.

 

[bogg]

Sub optimal means professionals will not be choosing the machine for such use.

Rebooting to try it out, is in itself suboptimal, so no matter if it was performance wise capable of it, it wouldn't be the prime choice of professionals until OS X supported it natively anyway

 

[koyoot]

This is not going to happen Koyoot. Flat Five (expert in a lot of things) shared with us that it will take 5 - 10 years before this all will be mainstream. Sorry m8. I think you'r wasting your time! (sarcasm)

Anyway, interesting! Another VR headset on the line is always welcome. I will catch up the article later. Are there just a few developers working on it? In my network there is none i know of that is developing anything for the Apple VR device. If so, I would have seen some Apple VR hardware laying done somewhere They are all focussed on Rift (which I think will be leading the VR tech market soon) and HTC Five.

I do not believe Apple VR hardware, and even AppleTV with AMD hardware is anything more than prototype now.

 

[ZombiePhysicist]

Rebooting to try it out, is in itself suboptimal, so no matter if it was performance wise capable of it, it wouldn't be the prime choice of professionals until OS X supported it natively anyway

That's a very fair point. But I'd speculate that needing to reboot is partly a function of the developers not bothering to develop for two reasons. First, smaller market. Second, no hardware could support it. The second pretty much closes the issue out.