monokakata
macrumors 68020
Didn't Intel license some of the Alpha technology after DEC brought a case against them for infringement? I seem to remember something about that, back in the mid-late 90s.
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.
MP 7,1 Waiting for Mac Pro 7,1 A1991 (no more)
- Thread starter Mago
- Start date
-
- Tags
- macpro 2016
- Sort by reaction score
Peace
Cancelled
Not sure but they just licensed ARM..
Laugh all you want but I see ARM in the future for Macs.
AidenShaw
macrumors P6
Intel bought DEC's semiconductor operations - fabs and all. This included Digital's IP for SMT - which moved to the Pentium 4 and was called Hyper-Threading.
Digital was fab'ing StrongARM - one of the leading ARM chips. This moved to Intel as well. https://en.wikipedia.org/wiki/StrongARM
So Apple will decide to ignore the entire enthusiast and professional camp, and focus its efforts on competing with ChromeBooks? 😱
Last edited:
JimGoshorn
macrumors 6502
You're very ignorant of a modern x86 processor. The legacy "problems" you complain of no longer exist. Today's x86 processors function very differently internally than the processors you have in mind. Aiden informed you as much in post 1716. Perhaps you'd be so good as to detail what issues x86 has that causes problems?
As to why someone doesn't come up with something "better" well...there's no reason to. x86 is a very capable architecture and runs an awful lot of software.
Maybe you're the ignorant one here.
You just got here, with the same attitude of others, know all. Are you really an "IT" guy?!
Maybe you should find another rock...
You just got here, with the same attitude of others, know all. Are you really an "IT" guy?!
Maybe you should find another rock...
What I've understood (and I've been in the business nearly 30 years) is that x86-64 software is still a prisoner of the legacy code and all these SSE, AVX, TXE... etc., are patches, "extensions" that make the old fart fly. Intel Core CPU's are CISC/RISC hybrids and x86-64 is translated to the RISC unit - a unit that is hidden from programmers. If x86 instruction set goes, Intel will lose a lot of customers, because if software has to be rewritten anyway, there are cheaper or more powerful or more efficient options than Intel. Intel wont undermine x86-64 by any means.
CISC/RISC hybrid is not the most efficient way to do things, but it is better than to have two completely different instruction sets on same CPU. x86 is a railroad that software developers have to follow.. to Intel stations. AMD commuter train also travels on same tracks, but has been a bargain, less efficient solution these days.
CISC/RISC hybrid is not the most efficient way to do things, but it is better than to have two completely different instruction sets on same CPU. x86 is a railroad that software developers have to follow.. to Intel stations. AMD commuter train also travels on same tracks, but has been a bargain, less efficient solution these days.
fuchsdh
macrumors 68020
Stephen Hackett did a small piece on the lack of Mac Pro updates: https://512pixels.net/2016/09/mac-pro-1000-days/
We'll see if it gets any traction with the larger Mac pundits.
We'll see if it gets any traction with the larger Mac pundits.
Yay, 1000 days!
x 1000! 😎Cool beer waiting in the fridge... will join the party in the evening!
Right, that's what I've been saying.
And it was just blowing some steam, I'm not expecting x86 to go anywhere soon. Exactly because there's a whole lot of dependency on it.
If only at the time Intel would have started up fresh with a new design. But it was a way to get you hooked, and has prevailed to this day.
I'm not expecting a major turn around here.
At least Apple had the guts to make the switch (for known reasons, but to me Apple adopting x86 is a bit against what I would expect although I can understand the reasons) from PPC to x86.
Maybe they'll come up with something of their own in the future.
[doublepost=1473864931][/doublepost]Skylake-E?! Highly doubtful.
And it was just blowing some steam, I'm not expecting x86 to go anywhere soon. Exactly because there's a whole lot of dependency on it.
If only at the time Intel would have started up fresh with a new design. But it was a way to get you hooked, and has prevailed to this day.
I'm not expecting a major turn around here.
At least Apple had the guts to make the switch (for known reasons, but to me Apple adopting x86 is a bit against what I would expect although I can understand the reasons) from PPC to x86.
Maybe they'll come up with something of their own in the future.
[doublepost=1473864931][/doublepost]Skylake-E?! Highly doubtful.
thing is Skylake-E is the HEDT part, not WS Xeon -EP.
The parts we're interested in are Skylake-W now, and I doubt they'll be here in 1Q'17 even.
The parts we're interested in are Skylake-W now, and I doubt they'll be here in 1Q'17 even.
koyoot
macrumors 603
Well I have written something similar very long time ago, specifically at polish forum myapple.pl, that there will be point in time, when the power consumption of computers will be governed, because of energy efficiency.
Maybe there was point in the design of Mac Pro, after all? 😉
I think this was discussed inside the European Union in 2014, and 2015, but nothing so far has spawned from this discussion. So far...
In europe maybe, but where the source of electricity is hydro based and you have to use it or lose it (it can't be stocked) and is dirt cheap not really. And before you reply, keep in mind that I work for one of the top hydro electricity producer in the world.
koyoot
macrumors 603
I see that it is the matter of Californian regulators. And as far as I know, California is in USA. So it is not only matter of Europe 😉.
Energy efficiency is problematic around the world, but its not the matter we should discuss here. What should be discussed here is the effect we will see around the world, of governed power consumption of computers, and the need for design of computers with efficiency in mind in first place.
Mac Pro 6.1 design was not so pointless after all, as much as it was spun out by people here. Because this is probably the first computer that was designed with this in mind. Others will have to follow, the suit.
And even if all of this governing of PSU's and the design of computers will fail at first, in the end, we will see upper limit for power consumption. Which is truly great thing to see.
One last thing: Was I writing for very long time, that Energy efficiency will be important, or not? 😉
Energy efficiency is problematic around the world, but its not the matter we should discuss here. What should be discussed here is the effect we will see around the world, of governed power consumption of computers, and the need for design of computers with efficiency in mind in first place.
Mac Pro 6.1 design was not so pointless after all, as much as it was spun out by people here. Because this is probably the first computer that was designed with this in mind. Others will have to follow, the suit.
And even if all of this governing of PSU's and the design of computers will fail at first, in the end, we will see upper limit for power consumption. Which is truly great thing to see.
One last thing: Was I writing for very long time, that Energy efficiency will be important, or not? 😉
Yes, you were writing.
We also need some scalability and flexibility in this efficient designs, a way to become more powerful... efficiently.
koyoot
macrumors 603
External expansion.
How can you have flexibility if your environment is constrained by power design in the first place? Whole point of this is reducing the power footprint for each desktop computer.
External expansion here is the only way we can get more power from your computers in upcoming future. Whether we like it or not.
How can you have flexibility if your environment is constrained by power design in the first place? Whole point of this is reducing the power footprint for each desktop computer.
External expansion here is the only way we can get more power from your computers in upcoming future. Whether we like it or not.
Of course, but aren't we far enough from this point?
Even the expected TB3 is not enough in many cases. I think.
Even the expected TB3 is not enough in many cases. I think.
koyoot
macrumors 603
That is not a question to me. I think that for the whole idea of external expansion exploding we need similar technology like thunderbolt from another vendor, so it would create competition.
Because Intel develops TB there is no push for breakthroughs in it. Because I dont think there is similar technology available for mainstream market that can work with mainstream CPUs and technology.
Also, as long as you don't use too many external solutions/ or better none, you 're ok and within the reduced power footprint for each desktop computer.
Isn't this completely untrue when you need multiple external solutions?
Isn't this footprint enlarged then by a large proportion?
[doublepost=1473885624][/doublepost]
Isn't this completely untrue when you need multiple external solutions?
Isn't this footprint enlarged then by a large proportion?
[doublepost=1473885624][/doublepost]
True, and here is where we need Apple's old good and indepent innovation, like when Apple was intel's competition (in innovation not massive sales).
Last edited:
koyoot
macrumors 603
It depends how efficient are those external solutions. Soon NAS'es will be using SSD technology rather than HDD, and they do consume much less power than HDD's.
If you read the techpowerup article you will ask yourself one question. Why do they define the high-end GPU by memory bandwidth? Because GDDR5 memory bandwidth using 512 bit memory bus is consuming 80W of power, at 6000 MHz.
HBM can use 1/4th of this.
GPUs in upcoming months will be much more efficient, than previous generations. Not by raw power, but power from each watt consumed. Look what Nvidia was able to achieve. 5.5 TFLOPs from 100W GPU(Tesla P4). That is something.
As for GPU designs. AMD touts that because of increased costs future designs will be jumping around of scalability idea. What this means is that GPU designs will be perfectly scalable both in terms of compute units/SM units, or the number of GPU dies on single GPU board.
We were talking about efficiency, yes? Imagine this: Single PCB plate inside external GPU enclosure, that is resembling the Mac Pro 6.1 design. On that PCB there is interposer, and 4 GPU dies, that are connected by that interposer. It also plays a role of coherent fabric. Then we have feature called dual link interposer, that connects the HBM memory not only to particular GPU dies, to whom are the HBM chips attached, but also to other GPU dies on that interposer. It is in very complicated way hardware implementation of software idea which is: HSA 2.0 and Unified Memory.
In future, because of increased silicon design costs, we will see GPUs made from multiple smaller GPU dies, instead of single big ones.
And one last part. People complain that Apple does not update the Mac Pro line very often. Well, you better get used to it, because when the regulation will come to fruition, on high end side, only possible jumps in performance will be available with every new node. And the time spans between node jumps will only increase. We have waited 4 years for 14/16 nm node in GPUs. 7 nm will come in 2020, at least! We are 4 years from next generation GPUs.
If you read the techpowerup article you will ask yourself one question. Why do they define the high-end GPU by memory bandwidth? Because GDDR5 memory bandwidth using 512 bit memory bus is consuming 80W of power, at 6000 MHz.
HBM can use 1/4th of this.
GPUs in upcoming months will be much more efficient, than previous generations. Not by raw power, but power from each watt consumed. Look what Nvidia was able to achieve. 5.5 TFLOPs from 100W GPU(Tesla P4). That is something.
As for GPU designs. AMD touts that because of increased costs future designs will be jumping around of scalability idea. What this means is that GPU designs will be perfectly scalable both in terms of compute units/SM units, or the number of GPU dies on single GPU board.
We were talking about efficiency, yes? Imagine this: Single PCB plate inside external GPU enclosure, that is resembling the Mac Pro 6.1 design. On that PCB there is interposer, and 4 GPU dies, that are connected by that interposer. It also plays a role of coherent fabric. Then we have feature called dual link interposer, that connects the HBM memory not only to particular GPU dies, to whom are the HBM chips attached, but also to other GPU dies on that interposer. It is in very complicated way hardware implementation of software idea which is: HSA 2.0 and Unified Memory.
In future, because of increased silicon design costs, we will see GPUs made from multiple smaller GPU dies, instead of single big ones.
And one last part. People complain that Apple does not update the Mac Pro line very often. Well, you better get used to it, because when the regulation will come to fruition, on high end side, only possible jumps in performance will be available with every new node. And the time spans between node jumps will only increase. We have waited 4 years for 14/16 nm node in GPUs. 7 nm will come in 2020, at least! We are 4 years from next generation GPUs.
Last edited:
That's because few people really need anything like it. I'm sure I'm an upper 1% computer user, but I hardly ever use TB.
This whole regulation thing stinks. Companies buying 100s or 1000s of desktop computers for basic office work already are concerned with power usage. As are data centers and clusters. The only person getting hosed here will be those buying home computers. Now we're going to have classes of computers that we have upgrade to in order to justify certain power needs? Holy perverse incentives Batman! So if I want to up the power beyond a certain point, I need X GBs/sec GPUs or Y Watt power supply? And forcing LEDs as the only option or "setting a new default brightness standard (since most consumers never change their monitor brightness)", really? Tough ship if you're on a budget with a monitor or just want some basic thing that would hardly be on anyway. And seriously, "most consumers" don't change the brightness on the display? I want to see that data and where its coming from. And jesus christ, even if true, you don't think more people will start changing the brightness if you just ship monitors with dimmer defaults?
I really, really hate living in California...
koyoot
macrumors 603
I think we should start to read "between the lines". All of this is made for silicon designers to build more efficient hardware. Dimming the monitors, will result in much more efficient monitors, for it to leave the brightness of them on previous levels. Similar thing about hardware. Rather than looking at the boarders, why not just make the hardware more efficient? It will be easier for manufactures and companies to think this way, rather than adjust to the requirements proposed by Californian gov.
Why not design the desktop computer with extremely low power consumption, compared to previous paradigms of desktop computer, and squeeze every single bit from its efficiency in terms of raw power? For this, we need to get the fastest possible CPU, faster and highest amounts of RAM, lock the GPUs to certain power levels, and possibly down clock them to get lower power consumption for not huge amount of raw power lost. And to get more efficiency, we need to get rid of internal expansion, because then we can use really low power design of PSU. Sounds familiar?
Why not design the desktop computer with extremely low power consumption, compared to previous paradigms of desktop computer, and squeeze every single bit from its efficiency in terms of raw power? For this, we need to get the fastest possible CPU, faster and highest amounts of RAM, lock the GPUs to certain power levels, and possibly down clock them to get lower power consumption for not huge amount of raw power lost. And to get more efficiency, we need to get rid of internal expansion, because then we can use really low power design of PSU. Sounds familiar?
The answer to both is: Yes. Something largely ignored by the nMP advocates.