Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.

koyoot

macrumors 603
Jun 5, 2012
5,939
1,853
We don't have to wait on this at all. If FP16 is #1 in our minds, then we'll watch it to find out what mistakes Oxide made.

Polaris is a disaster in progress, each announcement puts it lower on the food chain. Good luck trying to distract us with minor questions about Pascal.
Sorry Aiden but the only one not understanding something is you. Polaris is not a failure as you are trying to convince us all. But hey, you can believe in this, if you want to ;)

Oxide did not made any mistakes. Read carefully again what has been said from the beginning in that posts. But we have to wait for the end of investigation to conclude anything.
 

shaunp

Cancelled
Nov 5, 2010
1,811
1,395
I'd say TB 3 with USB-C connectors is a given. Lots of PC's now have them and companies like LaCie (who are fairly Mac focused) have announced TB3 storage http://www.lacie.com/gb/en/products/raid/12big-thunderbolt-3/#features

I think all the pieces are now there for an update - TB3 for all that external storage, E5-V4 Xeon's + X99 chipset, USB 3.1 (USB-C) and updated GPU. I can't see a reason for them not to do it now. If this is the case then we should expect 3 year update cycles on the nMP? A bit slow, but it wouldn't be the end of the world.
 

ManuelGomes

macrumors 68000
Dec 4, 2014
1,617
354
Aveiro, Portugal
Single GPU is a no go, it would unbalanced the thing.
I'd say there is no way they'll change anything radically on the nMP, this is the way they wanted it to be, and so it will stay, only my opinion of course. I wouldn't change it myself :)

dec, I didn't say you're supposed to OC the nMP, much on the contrary, read again. Apple doesn't want you to surely. I wouldn't do it anyway, it's not a bragging rig or gaming machine. And, as you say, no power7thermal headroom. Only if indeed the PSU is to get upgraded.

Why are we starting the AMD/NVidia war again? Haven't we had enough closed threads?
Pascal is great, indeed. Some compromises had to be made, to be first to market? I don't know, we'll see. Polaris sucks? Let's wait and see. Was there foul play on any side with Ashes? Again, let's wait.
AMD shouldn't have compared setups if the conditions weren't exactly the same, that's a given. They'll suffer for it if it becomes clear. I sure hope they really are getting their act together and acting righteously.
They're both targeting different markets at the moment so they can be civil.
 
  • Like
Reactions: AdamSeen

tomvos

macrumors 6502
Jul 7, 2005
345
119
In the Nexus.
Single GPU is a no go, it would unbalanced the thing.

How exactly would a single GPU unbalance the nMP? Consider situations when only one GPU is active driving the display and no OpenCL is done on the other GPU. I don't see how a nMP with only one GPU would be unbalanced. Unless applications are explicitly designed for multiple GPUs, running unbalanced is the normal state of computing.
 
Last edited:

flat five

macrumors 603
Feb 6, 2007
5,580
2,657
newyorkcity
How exactly would a single GPU unbalance the nMP? Consider situations when only one GPU is active driving the display and no OpenCL is done on the other GPU.

i think there's more to it than "one display + no openCL = only one GPU in use"



[0075] • It should be noted that various computational components can have duty cycles that generate copious amounts of thermal waste heat. For example, SSD module 420 can be disposed on only GPU riser board 414. SSD 420 can cause substantially more heat to be emitted into GPU airflow region 630 than GPU airflow region 632, given otherwise similar operating parameters of both GPU riser boards 412, 414. In such a case, balancing operations can be implemented to offload GPU operations from GPU riser board 414 to GPU riser board 412, thereby balancing heat dissipation across airflow regions 630 and 632.

522471-aa4ce5b9a93f6717664d1e98f4be173b.jpg.png

http://appft1.uspto.gov/netacgi/nph-Parser?Sect1=PTO1&Sect2=HITOFF&d=PG01&p=1&u=/netahtml/PTO/srchnum.html&r=1&f=G&l=50&s1="20140362522".PGNR.&OS=DN/20140362522&RS=DN/20140362522
 

ManuelGomes

macrumors 68000
Dec 4, 2014
1,617
354
Aveiro, Portugal
You need to look at it through Apple's eyes, not how you think it should be.
Apple is all about harmony. OK, forget about only one SSD on a GPU, that was... whatever.
But would you really see them doing a nMP with a vacant side of the thermal core? Heat build up on only 2 sides of the core? No way. The inside needs to be properly populated, and the dual GPU setup is their vision of the future of computing.
Of course you can wish for it, but I don't think it will ever happen. Same for the i7 CPUs, and the non-ECC mem, single GbE, etc.
Only a coupe of things are not mirrored in the nMP: the SSD and the HDMI port, the rest is perfectly designed. Ah, ad the oddly chosen 12GB in the entry level model, having only 3 mem slots populated, worst decision ever.

flat already stated the rest.. :)
 

Mago

macrumors 68030
Original poster
Aug 16, 2011
2,789
912
Beyond the Thunderdome
First MP 7,1 image leaks ...
Mac_Pro_10.jpg

[doublepost=1465135269][/doublepost]LOL... I can't resist that
[doublepost=1465135518][/doublepost]A Sinle GPU may not unbalance the mac pro if apple still uses only GPU1 for display, but in case they need both GPU to feed the bunch of 6TB3+4USB-C (12-16 DP1.2 signals) video outputt, they need both.
 

tomvos

macrumors 6502
Jul 7, 2005
345
119
In the Nexus.
[0075] • It should be noted that various computational components can have duty cycles that generate copious amounts of thermal waste heat. For example, SSD module 420 can be disposed on only GPU riser board 414. SSD 420 can cause substantially more heat to be emitted into GPU airflow region 630 than GPU airflow region 632, given otherwise similar operating parameters of both GPU riser boards 412, 414. In such a case, balancing operations can be implemented to offload GPU operations from GPU riser board 414 to GPU riser board 412, thereby balancing heat dissipation across airflow regions 630 and 632.

This sounds reasonable. However, offloading GPU operations would require an expensive context switch. All the data and state from one GPU has to be copied to the second GPU. You don't get this for free. Most likely it would result in slight frame rate stutter since it will be difficult to keep a steady frame rate during this offloading process.
And as far as I know, there is no automatic load balancing between the two GPUs based on their temperature at the moment?
 

pat500000

Suspended
Jun 3, 2015
8,523
7,515
Let's wait and see if they will announce it. If not, we can go to back cMP vs nMP 2013 or talk about no more Mac Pro in the future.
 

rockyromero

macrumors 6502
Jul 11, 2015
468
147
First MP 7,1 image leaks ...
Mac_Pro_10.jpg

[doublepost=1465135269][/doublepost] I can't resist that

Neither can I.

I do need to have 512 GB of memory, 4 TB of storage, and upward and downward compatible with all my Apple devices.

Specifically, I want Siri to work very nicely with all the commands.

And I want to have the ability to do VR/AR.

This year.

When I think of anything else I'll post it.

 

smelly cat

macrumors regular
Apr 17, 2010
133
55
I also agree that it's unlikely Apple will allow a single GPU option. However, as far as the vast majority of users are concerned, I think making the system a dual socket Xeon with a single GPU would have been better than the current single socket CPU with dual GPU config.

For all the noise that's been made about GPGPU computing, it still hasn't really replaced pure CPU muscle in my view. The second GPU in the Mac Pro seems to often sit as a case warmer in the system.
 

Peace

Cancelled
Apr 1, 2005
19,546
4,557
Space The Only Frontier
I also agree that it's unlikely Apple will allow a single GPU option. However, as far as the vast majority of users are concerned, I think making the system a dual socket Xeon with a single GPU would have been better than the current single socket CPU with dual GPU config.

For all the noise that's been made about GPGPU computing, it still hasn't really replaced pure CPU muscle in my view. The second GPU in the Mac Pro seems to often sit as a case warmer in the system.


I agree completely . Apple needs to come up with some form of crossfire.
 

Mago

macrumors 68030
Original poster
Aug 16, 2011
2,789
912
Beyond the Thunderdome
Would you be kind enough to cryptically quote the post that lead you to that conclusion?
No way, anonymous "leakers" (or liers) are easy to find (even darknet its an exageration), I'll never expose an actual apple insider, BTW actually he/she didn't leak nothing directly or indirectly but knows every screw on the new trash can.
 

Aldaris

macrumors 68000
Sep 7, 2004
1,791
1,250
Salt Lake
So you're saying there's a new trash can!
[doublepost=1465183964][/doublepost]... I just miss the days when we had some grainy elevator photo of the side of a box listing specs...
[doublepost=1465184712][/doublepost]Mago, back in he previously closed thread, you mentioned target modes over tb3; y'all know more about how the technical sides of things. This is intriguing to me. The conversation over the eventual merge between iOS and OS X/macOS do you think there could be (sooner than later) a docking solution to where you take a retina screen (iPad) and dock it to a larger system and run it like a desktop/workstation Mac, detach and us as an iPad/mobile solution.

Seeing the cross that already exists primarily through iCloud, but being able to have access to your files/projects and being able to work on them depending on your current situation (iPad vs desktop iMac/Mac Pro). Almost like the theory about the external Gpu on an updated display/think of it as modular computing.

(Forgive me for the fumbled post)
 
Last edited:

Mago

macrumors 68030
Original poster
Aug 16, 2011
2,789
912
Beyond the Thunderdome
None for us, unless use it as eGPU
[doublepost=1465214526][/doublepost]
Mago, back in he previously closed thread, you mentioned target modes over tb3; y'all know more about how the technical sides of things. This is intriguing to me.

I only know the "leak" (or actual lie) is about new Macbook Pros having an "improved target mode", current target mode on Mac Mini allows you to use it as an external HDD for other macs, on non-retina iMacs allows you to use your iMac display as external display for another Mac, the new feature for macbooks is this mode will allow you even to use most of the macbook peripherals as the keyboard, touchpad and webcam as if they where just plugged into another Mac, just imagine goin to work off site, you need a Mac Pro for XYZ you dont need to move that 27" TB display 3, instead you pick your mac pro, your macbook and on site you put your macbook on target mode and it works as keyboard, display and touchpad for your mac pro, even powered by the Mac Pro's USB-C PowerDelivery mode (there is where the newly PSU added 100W will go), very practical and needed, Apple should implemented time ago.

Anoter thing on the leak was about and External compute device, similar to an eGPU but only for compute, this maybe not an Apple development at all, but an novelty on Macs.

then the Tread started to talk about HSA (heterogeneous system architecture) this was not an explicit part of the leak, but infered about the possibility of Intel Xeon Phi Knights landing coming to the macOS environment eventually could enable add 72 core to your Mac Applications just plugging an small appliance like a mac-mini.

The conversation over the eventual merge between iOS and OS X/macOS do you think there could be (sooner than later) a docking solution to where you take a retina screen (iPad) and dock it to a larger system and run it like a desktop/workstation Mac, detach and us as an iPad/mobile solution.

Rigth now you have remote desktop (thru iCloud), I dint use it for a long time but it basically converts your iPad on a emulated terminal, not so good to use on slow machines/slow networks, but I use somthing to connect to Linux Workstations over a dedicated gigabit network with an ac1750 class router and was a very like to be in the real desktop, with HEVC and improved cpu/gpu this scenario could become more common, and someday your Mac will only plug to your network then you plug to your mac from your Apple TV or iPad or an Macbook as if you where directly plugged with an std moinitor and input devices, but we still 3-4 years from this day, the biggest challenge is Wireless Network Speed (mostly on clients, most wireless devices dont use more that 1/3 bandwidth allowed by the router) and display compression (as long I know no remote desktop client implements HEVC or similar, most still stuck to MPEG2-like tech.
 
Last edited:
  • Like
Reactions: t0mat0
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.