i'm pretty sure apple is much better at multitasking than all of us combined ; )
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Will there be a new Mac Pro in 2015?
- Thread starter Anto38x
- Start date
- Sort by reaction score
- Status
Got lost for a sec. You should have just quoted Cook instead of quoting the guy quoting Cook.
This quoting is getting out of hand... I'm quoting you quoting Xteec quoting the guy quoting Cook
Yes, Mantle/Vulcan/Metal is a software technology, that is not dependent on AMD hardware. Sure it has been optimized for Radeon, but it can be used with any other GPU tech. Perhaps AMD agreed to help Apple to build Metal in agreement where Apple becomes Radeon only house for - let's say - five years. And it started 2013 with nMP.
But it could be that next gen Apple TV is based on AMD FinFet 16nm APU processors. That are faster than PS4, have +50% better perf/watt and is cheap. Now that they have tvOS, I'm sure it is not CPU dependent and it will go to a different direction than iOS. That's why they separated it from iOS.
Jim Keller was in charge of Macroscalar Out-Of-Order architecture, that is well known as Cyclone cores in Apple AX chips. He's not a GPU architect, he is in charge of the team that designs also AMD Zen architecture.
The situation is pretty simple. Microsoft started designing DirectX12 many months before the Mantle has been even conceptualized. Then one of DICE engineers wanted to get more control over the GPU, because thats what they thought the API lacked. Went with this idea of low-level API to everybody - Microsoft, Intel, Nvidia, nobody was interested. Then came to AMD. AMD was not interested in the first place also, but went experimenting with it... and they were blown away with the idea. They have started shaping the form of Mantle and within few months they got fully functioning API. Why they did this? Because it was first attempt to get everything from a GPU, all its capabilities both for graphics, and compute. It was first attempt to API that is designed to work for games and professional, computational applications. It was first attempt that allowed VR(ask Nvidia - do they have anything for VR?). The low-level API ticked the all brackets that vendors, developers asked for years. AMD gave to everyone. Because they thought it would be most beneficial not for AMD but for the WHOLE industry to give it out. It is right now everywhere. Linux, Android, OS X, iOS, Windows. People are wondering what will happen with OpenGL especially on OSX.. The problem is: OpenGLNext is Vulkan. And Vulkan is direct rip-off from Mantle driver. The code lines for Vulkan differ from Mantle by 3 first letters of the command line. And Metal is already based on Mantle driver. So there is nothing to worry about.
The situation is pretty simple. Microsoft started designing DirectX12 many months before the Mantle has been even conceptualized. Then one of DICE engineers wanted to get more control over the GPU, because thats what they thought the API lacked. Went with this idea of low-level API to everybody - Microsoft, Intel, Nvidia, nobody was interested. Then came to AMD. AMD was not interested in the first place also, but went experimenting with it... and they were blown away with the idea. They have started shaping the form of Mantle and within few months they got fully functioning API. Why they did this? Because it was first attempt to get everything from a GPU, all its capabilities both for graphics, and compute. It was first attempt to API that is designed to work for games and professional, computational applications. It was first attempt that allowed VR(ask Nvidia - do they have anything for VR?). The low-level API ticked the all brackets that vendors, developers asked for years. AMD gave to everyone. Because they thought it would be most beneficial not for AMD but for the WHOLE industry to give it out. It is right now everywhere. Linux, Android, OS X, iOS, Windows. People are wondering what will happen with OpenGL especially on OSX.. The problem is: OpenGLNext is Vulkan. And Vulkan is direct rip-off from Mantle driver. The code lines for Vulkan differ from Mantle by 3 first letters of the command line. And Metal is already based on Mantle driver. So there is nothing to worry about.
But MS and Sony have been using low level versions of their APIs in consoles for years and the last time we saw it on desktops was with 3DFX Voodoo API(especially under DOS)???? So how was this a new idea that some DICE engineer shopped around? Can I read about this somewhere? The main issue with that story is how low level API is presented as a new idea, but it's been known about for many years. Whenever someone in the gamer community asked why consoles are more efficient the answer for years has alway been about low level drivers and APIs to keep the consoles more efficient than desktop systems.
Last edited:
Maybe we don't see Pro product launch / update this fall, but as mentioned here, iMac's are overdue. And so are external monitors. It could mean new form factor for both. As mentioned here before, separate 5k monitors are nearly as expensive as the cheapest iMac 5k.
So here's another (more or less) sophisticated guess... what if future iMac's and external monitors are going to be one device? It would solve the 5k problem, because they'd work as an external GPU for MacBook and other TB2/3 devices. And when ever you feel that your iMac is not powerful enough, you buy new Mac mini.. or MP. And it'll work with Apple Pen. Comes with new stand so you can turn the display as drawing board.
Mac would be artsy again!
Last edited:
Did you have had ANY low-level API on desktops since that Voodoo one? Was it that vast used on ALL major compute platforms that is right now? Was it one solution that was adopted on all major platforms: consoles, PCs, Macs, ultra mobile market? Thats what is new here.
Ok I'll accept the story that a nameless DICE engineer went around pitching the new idea of low level APIs to companies who already knew about and used low level APIs and that's 'new' ...because desktop.
I also don't think that we'll see it before Q1'16 now, not everything is ready.
Feb/Mac is more likely.
By that time we should also get the new keyboard (with SF font and no eject button, power instead) and mouse.
And maybe TB3D?
Indeed the best Fiji chips seem to be scarce already for the retail cards (Nano, X)
Feb/Mac is more likely.
By that time we should also get the new keyboard (with SF font and no eject button, power instead) and mouse.
And maybe TB3D?
Indeed the best Fiji chips seem to be scarce already for the retail cards (Nano, X)
http://www.forbes.com/sites/davealt...evelop-console-style-api-for-radeon-graphics/
P.S. Im not saying that idea of Low-Level API is new. New is just the vast adoption of it regardless of platform and hardware you have.
The idea for it came from one person in DICE. Its easy to find out from who, now. Ill give you a hint, Soy. His surname ends with "...son"Forbes said:
P.S. Im not saying that idea of Low-Level API is new. New is just the vast adoption of it regardless of platform and hardware you have.
Last edited:
And for those who put much faith in AMD Zen plateform, Jim Keller, the guy in charge of the project just left AMD... Again. http://hardforum.com/showthread.php?t=1875937
Don't you think that Zen is already done at this stage, with design of Zen v3 or even Zen v4?
http://semiaccurate.com/forums/showpost.php?p=234879&postcount=45
http://semiaccurate.com/forums/showpost.php?p=234879&postcount=45
OK guys, somewhat not so great news. Q1'16 for Xeon E5 1600 (up to 8 cores) and 2600 (up to 22 cores) v4 Broadwell.
Let's hope it's really Q1'16 and no more hiccups.
I guess no nMP in 2015 for sure now. Should we change the title of the thread? Something in the lines of "Will there be a new Mac Pro in Q1'16?"?
Q1 for general availability could mean the nMP coming out a bit later, unless Apple is already doing the work.
Q4'15 seems to only see E3 1200 v5 and Xeon D.
Q2'16 will see the remaining E5 4600 v4 and E7 x800 v4 Xeons, as well as Q3 the Phi x200.
Let's hope it's really Q1'16 and no more hiccups.
I guess no nMP in 2015 for sure now. Should we change the title of the thread? Something in the lines of "Will there be a new Mac Pro in Q1'16?"?
Q1 for general availability could mean the nMP coming out a bit later, unless Apple is already doing the work.
Q4'15 seems to only see E3 1200 v5 and Xeon D.
Q2'16 will see the remaining E5 4600 v4 and E7 x800 v4 Xeons, as well as Q3 the Phi x200.
Yep, that seems definitive now. /Intel Server Release Dates 2016:
http://www.cpu-world.com/Releases/Server_CPU_releases_(2016).html
But if Apple incorporates some of the major advances in SSD/GPU/IO tech since 2013/Q4, then an upgrade to nPM(7) in 2016/Q1 would certainly be significant. They are still offering the refurbished nMP(6) at just 15% of list, but that really ought to be more like -40%.
If only Intel had updated C610 to something like C615 with PCIe 3 instead of PCIe2, that would be awesome.
We would still have the DMI problem since the CPUs most likely weren't updated either. But it would be somewhat better.
At least, with all the delay, I guess the CPUs should come out quite fine tuned, no additional erratas or revisions/steppings. Maybe they did some magic too.
Highly unlikely though, before dec comes down on me
Still, TB3 is still MIA and by then it should be available massively.
Also, at the time El Cap should be all ironed out.
That leaves the GPUs and what Apple will go with...
We would still have the DMI problem since the CPUs most likely weren't updated either. But it would be somewhat better.
At least, with all the delay, I guess the CPUs should come out quite fine tuned, no additional erratas or revisions/steppings. Maybe they did some magic too.
Highly unlikely though, before dec comes down on me
Still, TB3 is still MIA and by then it should be available massively.
Also, at the time El Cap should be all ironed out.
That leaves the GPUs and what Apple will go with...
DDR4 2400 should also have come down in price, let's hope this time 16GB is the minimum config available.
NVMe SSDs should also be used. Will Apple continue to use Samsung OEM drives or go with Intel's this time around?
It would make them too dependent on Intel I guess.
I could see them going all out Intel wireless/BT radio next year though, since there will be a lot of devices being updated.
NVMe SSDs should also be used. Will Apple continue to use Samsung OEM drives or go with Intel's this time around?
It would make them too dependent on Intel I guess.
I could see them going all out Intel wireless/BT radio next year though, since there will be a lot of devices being updated.
A bit early to call TB3 MIA. Most of the Gen 6 (Skylake) options aren't shipping yet. Middle-end of Nov ...then that would be odd. However, ....
MIA at Apple; but any Mac movement post June is also MIA. Dell, not so much.
Alienware 13 , 15, 17 have Thunderbolt v3 ports......
http://www.alienware.com/landings/laptops.aspx
[ although this initial set is stuck with Haswell/Broadwell and earlier chipset design. HDDs (and not SSDs) probably why not competing for the relatively few PCIe lanes. Dell is promising Skylake models later. ]
Lenovo, not so much.
Lenovo P50/P70 also getting Thunderbolt v3 ports....
http://www.theverge.com/2015/8/11/9129681/lenovo-skylake-xeon-notebooks-p50-p70
but those seem to need a bit more work.... prototypes at IFA earlier in the month.
https://forums.lenovo.com/t5/Lenovo-IFA-2015/ThinkPad-P50-amp-P70-Hands-On/td-p/2164261
It wouldn't be surprising to see TB v3 show up at Microsoft's dog and pony show too in a higher end product.
http://arstechnica.com/gadgets/2015/09/microsofts-big-hardware-event-is-on-october-6th-in-new-york/
Good or bad sign that other folks are rolling out TB v3 before Apple. Good, in that it doesn't have the appearance of an Apple first, Apple only port. It isn't everywhere on general PCs by any means, but there is some movement at the higher price points shared with Apple's Macs.
Perhaps Bad in that Apple is perhaps becoming sloth slow on Mac resource commitments. We'll see if nothing in Oct-Nov comes with TB v3 ports on it then Apple is mailing in their Mac effort. At least the 27" iMac should bump to Gen 6 and chipset with higher synergies with TB v3. Unless there is some kind of power/size problem with TB v3 controller, the MacBook is also ripe for Gen 6 update and move to TB v3.
With respect to the Mac Pro. Yeah, by March 2016 TB v3 limited controller supply shouldn't be a blocking issue at all. GPU wise I don't think anything is going to change from what is out now. Apple would have had more time to polish off their custom GPU card design derived from currently shipping cards. Xeon E5 v4 probably would be 100% bug free by March, but there probably won't be any "show stopper" bugs.
If Xeon E5 v4 comes in March '16 (and doesn't ramp until April) then E5 v5 has a more than decent chance to slide into 2017 (very early that Q1 but the following year none-the-less).
What should happen before March '16 is software updates to get more out the Mac Pro 2013 hardware already out there. That's has been somewhat slow.
Last edited:
Just BTW apple.com (not the mobile version) now says Mac Pro supports up to three 5K displays (under tech specs). Wonder if that means something is coming or if it's a huge mistake.
My point exactly. Apple is lagging once again.
TB3 doesn't require Skylake so designs are possible.
Although I wouldn't see any current Mac with TB3 yet, better wait for the Skylake refreshed models of rMBP, rMP, rMBA, iMac and nMP with the Broadwell Xeons.
HBM2 seems to be going well, with final testing Q4 and mass production Q1'16. We could see boards early Q2, maybe in time for nMP.
I'd say Skylake Xeons will be out only by mid 2017 at best, but then again maybe not, since most of the work I done with Broadwell.
El Cap 10.11.1 public beta out, anyone scanning for hidden info?
TB3 doesn't require Skylake so designs are possible.
Although I wouldn't see any current Mac with TB3 yet, better wait for the Skylake refreshed models of rMBP, rMP, rMBA, iMac and nMP with the Broadwell Xeons.
HBM2 seems to be going well, with final testing Q4 and mass production Q1'16. We could see boards early Q2, maybe in time for nMP.
I'd say Skylake Xeons will be out only by mid 2017 at best, but then again maybe not, since most of the work I done with Broadwell.
El Cap 10.11.1 public beta out, anyone scanning for hidden info?
- Status