Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
It seems like the rumors about the 7.1 are going nowhere! With "Vega-Vega", "Polaris-Polaris", things have always been headed in the wrong direction. So much speculation that I have decided to provide "insider" information to help.
1. The nMP is a UFO. An Unidentified Future Object.
2. You must follow the path of other UFO's which of course starts at Roswell!
3. In 1947 there was The Roswell New Mexico Project. This is when they began work on the 6.1
4. For those that were there and remember, the sign on building 6 read "The Roswell New Mexico Project" In the summer of 1948 the name was changed to "The New Mexico Project" and later NMP and finally nMP!
5. Since most of them were mathematicians they knew 6.1 read as "Six and One. This of course represented the six engineers and the one supervisor working on the project.
With this insight I suggest everyone uses the internet to research before posting nonsense in this thread!
AppleCrop2.png
 
https://www.techpowerup.com/230360/first-intel-processor-with-amd-radeon-graphics-within-2017

The deal will go on. Future products from AMD will be manufactured by Intel fabs, also. First product - AMD Navi GPU.

Intel will use them also, in the CPUs.

Also guys, what do you think about this idea:

Current form factor Mac Pro. 3 16 core APU with 4096 GCN core chip each, and 16 GB of HBM2.

Total: 48 cores, with 48 GB HBM2, and 12288 GCN cores. All packed in 500W TDP(Cause each APU would have 150W TDP).

Each CPU has 64 PCIe Lanes.
 
  • Like
Reactions: Mago
According to some of the people who make Octane Render, they said they had a meeting with AMD and that AMD is dropping Open CL support. They also complain about some things I don't get. I don't know what Otoy's problem is with AMD's GPGPU stack and drivers, and I have no idea what "Boltzmann" is.

https://render.otoy.com/forum/viewtopic.php?f=30&t=59213#p303235

In fact, we just had a meeting with AMD at our office last week, and told them (because they need to hear it from a partner ) to stop wasting resources on stuff like this and focus instead on fixing their GPGPU stack and drivers, so we can finally ship V3 on AMD (on windows/Linux at least).

Instead, we were told (and this was news to us, so maybe to others as well) they are now dropping OpenCL! They also told us they are NOT fixing the binary driver issue they said would be fixed in two months (been 4) so we could finally ship a stable build of V3 on AMD.

The solution for us (and all other GPU renderers hoping to work on AMD) is to switch to Boltzmann, which is going to be ready with a new SDK 'very soon'. We tried it last year and it was usefuless, hopefully second times the charm. That also won't support Mac, but maybe Apple will surprise is all and ship with NVIDIA again.

We will support Mac by or before 3.1 with headless rendering until a better option becomes viable.

I know Otoy's been trying to get their product on AMD hardware for some time. I have no idea how reliable they are as a source for this type of thing. They haven't shied away from complaining about AMD before. They've been quite direct (and perhaps a little dramatic) about issues with AMD in the past, only to turn around and say everything's OK again a few months later. The only thing I take away from this is that they don't complain about nVidia this way.

Don't take this as gospel. Take it as more gossip and/or use it to inform any hardware purchase decisions if Octane Render is a must for your workflow. If you spend all day in Final Cut Pro, this probably will have little effect on your decision to stay with Apple or not.
 
According to some of the people who make Octane Render, they said they had a meeting with AMD and that AMD is dropping Open CL support. They also complain about some things I don't get. I don't know what Otoy's problem is with AMD's GPGPU stack and drivers, and I have no idea what "Boltzmann" is.

https://render.otoy.com/forum/viewtopic.php?f=30&t=59213#p303235



I know Otoy's been trying to get their product on AMD hardware for some time. I have no idea how reliable they are as a source for this type of thing. They haven't shied away from complaining about AMD before. They've been quite direct (and perhaps a little dramatic) about issues with AMD in the past, only to turn around and say everything's OK again a few months later. The only thing I take away from this is that they don't complain about nVidia this way.

Don't take this as gospel. Take it as more gossip and/or use it to inform any hardware purchase decisions if Octane Render is a must for your workflow. If you spend all day in Final Cut Pro, this probably will have little effect on your decision to stay with Apple or not.
Botlzmann is CUDA compiler. If you have CUDA application, you just put CUDA code to HIP, which is other name for Boltzmann, and it will automatically compile everything you need for AMD GPUs.

Apple does not care about it because they have Metal. Its developers, who have to care about supporting Metal.

AMD is working on their own API, without any abstraction, just like CUDA, and it is based on Mantle. First glimpse for it is the aforementioned ProRender.

P.S. Otoy claimed they have been able to use CUDA on AMD GPUs. What happened from that moment?
 
what do you think about this idea:

Current form factor Mac Pro. 3 16 core APU with 4096 GCN core chip each, and 16 GB of HBM2.
Fine.

What about a Power8 Mac Pro loaded with 2 Tesla GPU? and 700W TDP (full copper/heatpipe thermal core).

I think could be cheaper for apple to comeback to power architecture than build a new one on ARM,
 
Fine.

What about a Power8 Mac Pro loaded with 2 Tesla GPU? and 700W TDP (full copper/heatpipe thermal core).

I think could be cheaper for apple to comeback to power architecture than build a new one on ARM,
I would have to see how powerful are Power CPUs right now. Have not seen them for awhile.
 
Botlzmann is CUDA compiler. If you have CUDA application, you just put CUDA code to HIP, which is other name for Boltzmann, and it will automatically compile everything you need for AMD GPUs.

Apple does not care about it because they have Metal. Its developers, who have to care about supporting Metal.

AMD is working on their own API, without any abstraction, just like CUDA, and it is based on Mantle. First glimpse for it is the aforementioned ProRender.

P.S. Otoy claimed they have been able to use CUDA on AMD GPUs. What happened from that moment?
Thanks for the explanation.

As for Otoy and AMD, they bounce back and forth with their optimism. At one point they said AMD was off the table, then they came back and said something about being able to run Cuda using some in house solution, then I think they came back again and said it was dead.

I think they are trying really hard to get on AMD. They have some super smart people with impressive tech at their company, and the want to expand their market. But I keep hearing about road blocks outside of their control.

If anyone can get a Cuda engine on AMD it is them. But I can't think of any incentive for them to share their solution with any other company looking to move Cuda code to AMD unless it's open source or they are feeling generous.
 
Remember that only four years ago (or so) Apple built a brand new FACTORY in Texas to manufacture the nMP. You don't go building new factories to build radically new engineered hardware aiming to shut down the whole line in 5 years.

I think this delay, along with all mac delays (as has been stated numerous times) simply has to do with the disintegration of Moore's law. There simply hasn't been significant enough changes in hardware to justify a new model. Another Mac Pro will definitely come out some day. The Mac Mini on the other hand... that model looks doomed.
 
Remember that only four years ago (or so) Apple built a brand new FACTORY in Texas to manufacture the nMP. You don't go building new factories to build radically new engineered hardware aiming to shut down the whole line in 5 years.

I think this delay, along with all mac delays (as has been stated numerous times) simply has to do with the disintegration of Moore's law. There simply hasn't been significant enough changes in hardware to justify a new model. Another Mac Pro will definitely come out some day. The Mac Mini on the other hand... that model looks doomed.
What are you on?

Apple didn't build a factory in Texas, they have a contract with a manufacturer in Texas (http://iphone.appleinsider.com/arti...pertise-during-visit-to-texas-mac-pro-factory). Additionally there have been several appropriate candidates for updates to the Mac Pro CPU and GPU chips.
 
If anyone can get a Cuda engine on AMD it is them. But I can't think of any incentive for them to share their solution with any other company looking to move Cuda code to AMD unless it's open source or they are feeling generous.
If anyone would want to share how to run CUDA code on AMD GPUs, he would be immediately sued by Nvidia, and that is presumably why Otoy has resigned from trying to run CUDA code on AMD GPUs, or even sharing it.

CUDA is Nvidia's proprietary API that has zero to none abstraction from hardware. The applications run as close to hardware as possible. Until AMD will bring Mantle-Based CUDA-like offering, we have nothing else with similar level of zero abstraction.

In theory in opensource software it would be impossible to get zero level abstraction on hardware of Different IEMs. We have to find out, if AMD will be able to bring this. Until then, all we have is OpenCL, ProRender, Metal, CUDA.
 
Looks like nmp is "expandable."
Like the egpu and processor upgrade.
Now if it could run the 5k display Apple is selling ;) (*at native retina resolution).

It's nice to know there are some options for playing with the current MP. Question for more of a hobbyist, could you have a 4 core machine, and up the processor to a 12? Or would that be a change out of the entire board? Out of curiosity.
 
  • Like
Reactions: pat500000
Now if it could run the 5k display Apple is selling ;) (*at native retina resolution).

It's nice to know there are some options for playing with the current MP. Question for more of a hobbyist, could you have a 4 core machine, and up the processor to a 12? Or would that be a change out of the entire board? Out of curiosity.
Couldn't Gtx 1080 run 5k? .
If you're the handy person, Like what Daniel stated...get that 4 core and pop a new processor in it.From that video it looks like he didn't switch out the board but replace the processor. This is actually a good thing...based on watching that video.
I'm not a specific tech person compared to others here...but nmp is almost like cmp in a sense.

I mean that egpu may cost alot like bizon box but worth the investment.
 
  • Like
Reactions: Aldaris
It's nice to know there are some options for playing with the current MP. Question for more of a hobbyist, could you have a 4 core machine, and up the processor to a 12? Or would that be a change out of the entire board? Out of curiosity.

Yes. You can do it yourself or OWC offers a turn-key service where you send in your nMP and they upgrade the CPU for you.
 
Couldn't Gtx 1080 run 5k? .
If you're the handy person, Like what Daniel stated...get that 4 core and pop a new processor in it.From that video it looks like he didn't switch out the board but replace the processor. This is actually a good thing...based on watching that video.
I'm not a specific tech person compared to others here...but nmp is almost like cmp in a sense.

I mean that egpu may cost alot like bizon box but worth the investment.
As far as the thunderbolt 3 LG display. Not that I am a fan of it, but were all in the boat for third party displays at the moment, which means we aren't restricted to Thunderbolt or other proprietary connections. Running off an external GPU really seems to be the solution-and from the video and a few weeks ago 9to5mac did a similar video with a different external box (granted if I remember right they ran it through TB3). But this opens to using standard connections without adapters.

Yes. You can do it yourself or OWC offers a turn-key service where you send in your nMP and they upgrade the CPU for you.

Thanks!

Really sounds like the route to go. (I personally am holding my breath till NAB/WWDC, but if nothing by then I think this is the route I'll take).
[doublepost=1486269871][/doublepost]
Again, "its coming",

dBrand tweeted Apple first 2017 Keynoye Leak:https://twitter.com/dbrandSkins/status/828020010853478401/photo/1

"we are getting down to the wire"...

Which seems the keynote with the Airpods sudden success as opening theme...

No details on when, but usually this leaks a month or two weeks earlier.
I'll agree with @pat500000, wrong thread. And also call it out as fake. If there is indeed an upcoming Apple Event, which would be expected in the spring, it would really break with tradition to have anything to do with iPhones which were a summer release at the beginning and have since become a fall release. I can't imagine the PR/marketing for a 6-month window from the iPhone 7 to the 10th anniversary phone (iPhone 8 or whatever).

Seems like more of a stunt to drum up traction and attention if you ask me. I believe the future is Nikolai Tesla-esque wireless charging, but will take more time to get people used to it. If Apple was really looking at making something game changing would be a combination wireless networking/charging mesh, but that would The be years out. Apple would likely start out small rolling out solutions for watches and iPhones, but again that is where the "puck" will be. And the current captain of the ship is probably going to be pulling out spring watch band colors out of his "pipeline" as opposed to anything close as innovative as what I described above, which would also be great if it had LiFi involved too.
 
Last edited:
  • Like
Reactions: pat500000
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.