So, Intel can disarm any threat by cleaning up some legacy bits by emulation, and the 90% won't happen.
Sounds good.
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
Sounds good.
However you choose to cut this cloth the processors are not suitable for the desktop machines we have. Period.
Forgive me if I am wrong as I am only trying to interpret the situation with my limited knowledge;
It was my understanding that in order to migrate from x86 to ARM or anything else with a different architecture, all the existing OS and applications that used to run on Macs will then need to either be re-written or emulated? Somebody can manage an emulation akin to Rosetta with insignificant performance hit to maintain backward compatibility. Newer apps can then be cross compiled (eg. universal library). And at a certain point, when the newer architecture manage to provide considerable benefits in other areas such as form factor, energy efficiency etc, then this is quite likely Apple's logical path way out regardless of what the rest of the world runs on. The immediate sacrifice are people who boot into systems other than macOS or developers who rely on x86 runtime libraries, but it looks like both MS and Intel have been looking into these areas themselves.
It was my understanding that in order to migrate from x86 to ARM or anything else with a different architecture, all the existing OS and applications that used to run on Macs will then need to either be re-written or emulated? Somebody can manage an emulation akin to Rosetta with insignificant performance hit to maintain backward compatibility. Newer apps can then be cross compiled (eg. universal library). And at a certain point, when the newer architecture manage to provide considerable benefits in other areas such as form factor, energy efficiency etc, then this is quite likely Apple's logical path way out regardless of what the rest of the world runs on. The immediate sacrifice are people who boot into systems other than macOS or developers who rely on x86 runtime libraries, but it looks like both MS and Intel have been looking into these areas themselves.
They are. Low-power desktops. Like It has been stated, Clock-for-clock, Hurricane trounces Skylake. If your application world would be designed for only low-power hardware, Intel would already be going down, because ARM, would eat them for breakfast, and Apple would eat both Intel and ARM every single day, for years. Its all about IPC in Low-power hardware, and here Apple is the undisputed champion. And they will be for years.
The CPUs are not suitable for computers like Mac Pro, or even iMac. At this time. In future, they CAN be more than suitable, because thats how they evolve. They are outgrowing mobile world, right now.
Bare in mind that, Intel knows this, and they have 3 threats at this moment: ARM, Apple, AMD. AMD brought extremely capable michroarchitecture, that is scalable from low-power to high-power. ARM has CPU designs that are extremely capable in low-power world. Apple right now has low-power designs, but they are evolving.
Its extremely un-wise to say that Apple will never bring their own custom designed chips to their computers, especially when they can offer much better value(which they already do!), than their competition.
Why do you continue to cherry-pick bits of minutiae that seem to make something look good - when in fact the item that you quote is irrelevant by itself?
So what if a dual-core/dual-thread A10 at 2.3 GHz is 20% better at instructions per cycle - when you're comparing it to a quad-core/octo-thread Skylake at nearly 4 GHz?
The Skylake will destroy the Hurricane on every compute performance metric. Single thread and especially multi-thread.
And I think most of the people following the MP6,1's fade to oblivion would like to see a new new Mac Pro that has a 1000 watt power supply, not one that has been compromised into working with a 25 watt power supply.
Last edited:
@koyoot Aiden is rigth, there is no way to compare Skylake 1:1 with A10 unless Apple releases a Mac with A10, futher you A10 maybe more efficient on Instructions Per Watt, but still half way to reach full IPC, since it dont support SMT , also its single thread results don't include backround tasks, I follow ARM evolution, and while it may reach Intel IPC at some point next 2-3 years, this is not now. its easy to beleive on marketing "desktop class" cpu (a 2005 desktop they meant), go to HPC where each clock each watt and each byte of processed data account huge, you'll see the landscape how it is: ARM architecture still half way to intel in IPC but burns much less than 1/4 the power and solving huge highly parallelized problems this is a Win for ARM over Intel but actually you need 48 ARM cores to beat a quadcore xeon in compute power (and not common ARM core, I mean Kyro Hurricane or cortex A73) .
Also you shuld look at earlier Snapdragon 835 benchmarks, easy beats the A10, still not ready to compete with full dektop cpu (except if you consider a desktop those cheap PC Stick)...
Also you shuld look at earlier Snapdragon 835 benchmarks, easy beats the A10, still not ready to compete with full dektop cpu (except if you consider a desktop those cheap PC Stick)...
Last edited:
As much as I want apple to release a new Mac Pro, I honestly wouldn't be surprise if all we get next year are refreshed iMacs.
Hope I'm wrong but the past few keynotes have been so disappointing, especially the hyped "hello again" event back in October.
If the Mac Pro doesn't come next year it's as good as dead.
Hope I'm wrong but the past few keynotes have been so disappointing, especially the hyped "hello again" event back in October.
If the Mac Pro doesn't come next year it's as good as dead.
Somewhat related, but, dont HP/Dell get most of their profits through their business machines aka workstations & servers?
IPC is all about Single thread performance without HT/SMT, or whatever it is called.
Also about snapdragon: Is it superscalar architecture?
You have missed the point or completely not checked the link, or you are unaware typically for you. what is being discussed. ONE and the same benchmark(Geekbench) running on two different cores: A10, and Skylake CPU downclocked to 2.34 GHz.
What is compared? Every single factor that shows SINGLE THREAD performance of the CPUs.
And in this factor, A10 is already faster than Skylake hardware. THIS is IPC, which some of you try to downplay, or tried to downplay before, when it came to one particular CPU vendor. Simpler it cannot be put.
Performance per core. Where is the place where Intel is getting back? On Memory bandwitdh! Performance of Apple CORES is higher than best what Intel can offer.
Last edited:
It would be nice if you included links for your information.
When I goto the PrimateLabs page for the MSI MS-7970 (i7-6700K), it shows 5057 (single) and 20827 (multi) for N-Body. Your table shows 2757.
Why is the number at PrimateLabs almost twice as high as yours?
Because as, have been said before/if you would pay attention, both CPUs, the A10, and Skylake core was clocked at 2.34 GHz, to get best possible comparison between the cores.
Going to stick with my 5,1 for now; I did not buy the nMP due to lack of expansion capabilities similar to cMP.
What am I missing out on? Thunderbolt, USB-C on my tower? Doesn't bother me as I do neither video or photo or other intensive CPU tasks as others do on this thread (unless you count ripping CD/DVD's
_
I made the mistake of not reading up on the 2013 Retina MBP's when they came out- I thought they would be similar to the previous line of MBP, only to find out after I purchased for my wife the internals could not be upgraded (at the time). Of course now they can be upgraded but it took some time for it to be figured out. My biggest complaint is the "closed" mentality Apple is going towards- I'm sorry but if I am spending serious cash on a piece of equipment I want to have the option of modifying it, as I have on my current system.
I looked at the nMP last week when visiting the Apple store- still not impressed, and having to pay 8K for a system that will be outdated almost as soon as purchased is not my cup of tea. Learned my lesson on the iPhone 1.
What am I missing out on? Thunderbolt, USB-C on my tower? Doesn't bother me as I do neither video or photo or other intensive CPU tasks as others do on this thread (unless you count ripping CD/DVD's
_I made the mistake of not reading up on the 2013 Retina MBP's when they came out- I thought they would be similar to the previous line of MBP, only to find out after I purchased for my wife the internals could not be upgraded (at the time). Of course now they can be upgraded but it took some time for it to be figured out. My biggest complaint is the "closed" mentality Apple is going towards- I'm sorry but if I am spending serious cash on a piece of equipment I want to have the option of modifying it, as I have on my current system.
I looked at the nMP last week when visiting the Apple store- still not impressed, and having to pay 8K for a system that will be outdated almost as soon as purchased is not my cup of tea. Learned my lesson on the iPhone 1.
Cheese on bread!!! We’re not talking about performance per watt or physical size or weight or anything else. Simply, are these processors desktop class.IPC is all about Single thread performance without HT/SMT, or whatever it is called.
Also about snapdragon: Is it superscalar architecture?
You have missed the point or completely not checked the link, or you are unaware typically for you. what is being discussed. ONE and the same benchmark(Geekbench) running on two different cores: A10, and Skylake CPU downclocked to 2.34 GHz.
What is compared? Every single factor that shows SINGLE THREAD performance of the CPUs.
And in this factor, A10 is already faster than Skylake hardware. THIS is IPC, which some of you try to downplay, or tried to downplay before, when it came to one particular CPU vendor. Simpler it cannot be put.
Performance per core. Where is the place where Intel is getting back? On Memory bandwitdh! Performance of Apple CORES is higher than best what Intel can offer.
One major thing tells me they are not. They are not in Apple desktops.
What you are saying is that they would be if……..No if’s no buts. Simply are they? Not they would be if you looked at thsi cost, or they would be if you did B or considered A.
Are they suitable to put into a regular desktop computer right now?
No.
Yes. It is just you are refusing to see it. Mac Mini, or even base iMac - the CPUs are more than enough for them.
It's been over 1000 days with no update and no mention. It's already dead; resurrection is unlikely.
I feel like people have been saying "if it doesn't get updated in xyz" for the past year and a half. We've inevitably hit an arbitrary "deadline" and the xyz gets moved up.
Further, if it does some how get resurrected, would buying it even be a wise decision? Why "sign up" to be put into this kind of cycle again? All evidence strongly suggests Apple has no commitment to the headless workstation market. Why invest is a product stream that the creators themselves don't truly care about?
very interesting comparison between A10 and skylake! I guess it would be easy to reach higher clockspeeds with active cooling. also, adding more memory channels on the A10 to increase memory throughput can't be that complicated...
Recent Geekbench tests on Intel's latest desktop i7-7700K show it delivers 6675 single-core performance and 20950 multi-core performance: https://browser.primatelabs.com/v4/cpu/1366503
If there is some ARM CPU that can beat this single-core performance (much less multi-core), I'd like to see it.
Since the dawn of computers, there has always been an elusive dream that some architectural trick or instruction set would unleash dramatically better performance at a given fabrication level than competing products. Time and again this has proven untrue, impractical, or transient. The history of computing is littered with these failed efforts. Well-known examples include the IBM Future Systems Project, Data General Fountainhead Processor, the entire RISC workstation market, and Intel Itanium. The Data General case was even chronicled in a Pulitzer Prize-winning book, "Soul of a New Machine", by Tracy Kidder.
If an ARM CPU ever beats the multicore performance of the *latest* Intel desktop CPU, that would be an amazing feat. However so far they haven't even beat the single core performance.
There is a tendency to look at ARM's progress and project that to the future. But past is not prologue. It is easy to make big gains when you are starting from a low point, but they taper off. That happened to other CPU designs as well, years ago. There is no reason to think a similar performance flattening won't happen with ARM as well -- IF their performance ever approaches the latest Intel desktop CPU.
The PowerPC CPU that once ran Apple Macs increased in performance by a factor of 10 over a few years but it tapered off and you don't see those anymore. The past is not prologue.
I love ability to read in context of people on this forum.
https://forums.macrumors.com/threads/waiting-for-mac-pro-7-1.1975126/page-126#post-24121225
Read this post. Look at the table comparing A10 and 6700K. And bare in mind, both A10 and 6700K were clocked at 2.34 GHz.
No. The reason why Apple cannot put higher core clocks is because they are not able to put higher amperage through the cores.
They can manage the core clocks through laveraging the FO4 delay, or by increasing the "per clock delay" ratio of resources, but not just because of increasing the amount of power coming through the CPU. Intel can do this. Apple cannot because of the nature of architecture(ARM) and Microarchitecture(Macroscalar Apple custom designed cores).
There is nothing in that thread or table that supports your statement that "Apple chips are currently faster in Single core performance than anything Intel has".
When an Apple or ARM CPU in a stock configuration can produce better performance than the stock latest Intel desktop CPU, then and only then will that statement be true.
There is no Apple or ARM-based CPU that can remotely approach the single-core performance of the latest Intel desktop CPU, which is the 4.2Ghz i7-7700K.
On workstation-class machines like the Mac Pro -- what this thread is about -- there is even less hope that an ARM or Apple CPU will ever match the multicore performance of a high-end Xeon CPU.
On the server side, the software compatibility problem is lessened for ARM. Nobody cares what OS a server is running, provided it can deliver the performance and reliability. There have been efforts to push ARM-powered servers but they have mostly been miserable failures thus far: https://gigaom.com/2016/11/09/more-arm-cpus-in-the-datacenter-for-2017/
You say that I have posted that: "Apple chips are currently faster in Single core performance than anything Intel has", yet your quote from my post clearly says: ...In single threaded applications, clock for clock Apple CPU will be faster than latest and greatest offering from Intel...clock for clock, Apple chips are currently faster in Single core performance than anything Intel has.....
This is what you have quoted. Moving goal posts?
It is your problem that you have been not able to read with understanding what has been written. In every single post, from the beginning of this merit about Apple's direction of their hardware, it was all about IPC(Clock-for-clock) direct comparison between Apple A10, and Intel Skylake. It was in the post that I had provided a link to on another forum(semiaccurate.com, which has been posted there by this guy: http://www.fool.com/author/4997/index.aspx?source=iapsitlnk0000003) on previous page. How come people have missed it? How come later tried to spin whole discussion into their faulty understanding of the merit? That table shows that CLOCK FOR CLOCK, if you do not understand it properly, Apple A10 is already faster than the fastest CPU architecture that Intel provides.
IPC is all about Single thread performance without HT/SMT, or whatever it is called.
Also about snapdragon: Is it superscalar architecture?
You have missed the point or completely not checked the link, or you are unaware typically for you. what is being discussed. ONE and the same benchmark(Geekbench) running on two different cores: A10, and Skylake CPU downclocked to 2.34 GHz.
What is compared? Every single factor that shows SINGLE THREAD performance of the CPUs.
And in this factor, A10 is already faster than Skylake hardware. THIS is IPC, which some of you try to downplay, or tried to downplay before, when it came to one particular CPU vendor. Simpler it cannot be put.
Performance per core. Where is the place where Intel is getting back? On Memory bandwitdh! Performance of Apple CORES is higher than best what Intel can offer.
Anandtech did a pretty thorough analysis a year ago comparing the Intel broadwell based MacBook against the last gen A9X and gave the edge to Intel. They are competitive with each other both in performance and efficiency.
If the mac did switch, remember there would also have to be an emulation layer for running legacy x86 apps on the mac. This worked ok for PPC to Intel because Intel was so much faster than PPC, but the switch from x86 to ARM would be a lateral move at best meaning legacy apps would be very slow.
Furthermore, Apple wouldn't switch the mac to ARM unless it could do it across the entire lineup. Intel has chips that range from 5 W to 140 W with the requisite chipsets and motherboard designs to go with them. Thats not something Apple can easily replicate and would be a poor financial investment given the decreasing sales of laptops and desktops worldwide.