MP All Models Is Apple ****ing themselves over the ThreadRipper 3 yet?

[goMac]

It has nothing to do with expertise. Based on latest ARM servers benchmarks, 16 core CPU, without Hyper Threading is consuming up to 80W under load. While being slower than equally power efficient, but 8 Core CPU, at the same tasks. Call it lack of software optimization, call it hardware performance. Whatever.

Except again, that's not Apple's ARM CPU. I know it's a frustrating grey area because Apple doesn't ship a big ARM CPU to compare. But ARM performance is not blanket. Different ARM CPUs are optimized very different.

Heck, some ARM CPUs don't even support out of order execution which can be a real problem.

ARM CPUs are hitting Diminishing Returns field when scaled up. Its not up to Apple to fix this.

Its up to ARM to fix this.

Ehhhhhhh.

It really depends what we're talking about.

Most ARM CPU cores are tremendously dumb. It's getting better, but the cores themselves have traded low complexity for power efficiency. Apple has been building more complex ARM cores that may actually not be as efficient, but scale better.

Apple doesn't usually focus on continuous performance for power efficiency. They believe that if the work is cleared as fast as possible, the CPU can go back to idle, and that is better for performance. That leads to designs that are more comparable to desktop CPUs.

Is that at all related to what you're talking about? Not sure. I'm working with generalities here still.

Instruction length. And you know that perfectly well. They may score brilliantly in Geekbench, which executes simplest instructions.

But when you will put the load on those CPUs, the picture may be completely different.

I didn't really want to get into it because it's proprietary information that I can't share, which is kind of unfair to the thread, but I spend a lot of time benchmarking "real world" code on Intel vs ARM, and if there is some sort of hidden performance trap on Apple's ARM CPUs... I'm not seeing it. The only thing I generally see, which is borne out in public benchmarks, is that Intel is still the master of vector instructions.

The only thing to really watch for with Apple CPU benchmarks is they optimize the CPU to spike performance in bursts. But on an iPad that's far less common, and Intel is doing the same thing anyway. Don't know about AMD.

I also think that if Intel was at 7 nm, we might be in a different place right now. Intel is very good at optimizing what they do, and they mostly seem to just be getting hurt by not being able to scale the actual chip beyond 14 nm. I think Intel is probably better at design than both AMD and Apple, but they just can't make up that process gap.
[automerge]1575334934[/automerge]

People need to stop being hung up on die size and benchmarks. Die size can be good for reducing power draw but not the end all be all to silicon performance. Transistor density, lithography methods (e.g. self aligned quad patterning), and instruction sets (e.g. various AVX sets) all matter as well as die size. Benchmarks can be useful as a proxy to performance but processor X being 3% faster than processor Y per core in raw number crunching doesn't necessarily equate to better performance for specific tasks. Unless you have tied your identity to the manufacturer of a specific component in your computer it just shouldn't matter this much. The question should be will this do the job I need it to in a manner that's effective for me.

It's also worth bringing up that Apple's CPUs are a cluster of processors. We've just been talking about the ARM cores, but there are tons of co-processors on the package that could be helpful.

Doesn't apply quite as much to the Mac Pro in since Apple has an entire board to work with. And stuff like Afterburner. But part of the strength of Apple's chipset is they can add dedicated co-processors for specific tasks.

There's a lot of specific tasks (like Metal, or transcoding, or machine learning) that have little or nothing at all to do with the ARM cores.

But... also could be interesting to see a fusion of Apple's custom task specific cores with AMD's custom designs.

 

[Pro7913]

AMD is bad, Nvidia 12nm process is better than their 7nm

That's really irony since both AMD and Nvidia outsourced their GPU to TSMC. Yes, TSMC makes GPU for both AMD and Nvidia. What are you talking about?

Intel is bad, they're stuck on 10nm process, AMD is good and already on 7nm.

In terms of transmitter per milimeter, TSMC 7nm = Intel 10nm. You have no idea what you are saying.

 

[thisisnotmyname]

It's also worth bringing up that Apple's CPUs are a cluster of processors. We've just been talking about the ARM cores, but there are tons of co-processors on the package that could be helpful.

Doesn't apply quite as much to the Mac Pro in since Apple has an entire board to work with. And stuff like Afterburner. But part of the strength of Apple's chipset is they can add dedicated co-processors for specific tasks.

There's a lot of specific tasks (like Metal, or transcoding, or machine learning) that have little or nothing at all to do with the ARM cores.

But... also could be interesting to see a fusion of Apple's custom task specific cores with AMD's custom designs.

which really for all the hand-wringing about who has the flavor of the moment CPU this is really where the real advancements are going to come from. We've milked Moore's law and all of it's subsequent variants to death. Dedicated purpose driven silicon is where we're headed in which case arguing over which manufacturer is currently x% faster will seem pretty quaint.
[automerge]1575335353[/automerge]

That's really irony since both AMD and Nvidia outsourced their GPU to TSMC. Yes, TSMC makes GPU for both AMD and Nvidia. What are you talking about?



In terms of transmitter per milimeter, TSMC 7nm = Intel 10nm. You have no idea what you are saying.

I'll abandon my plans to moonlight as a standup. Both of those quotes were parody of what you had posted.

 

[koyoot]

I didn't really want to get into it because it's proprietary information that I can't share, which is kind of unfair to the thread, but I spend a lot of time benchmarking "real world" code on Intel vs ARM, and if there is some sort of hidden performance trap on Apple's ARM CPUs... I'm not seeing it. The only thing I generally see, which is borne out in public benchmarks, is that Intel is still the master of vector

Is this code executed on the hardware itself, on ARM CPUs, or on Fixed Function units like Integrated Signal Processor, versus bare metal hardware on Intel?

I have tested quite a lot of this, as well, on different CPUs. Beginning on A53 and ending on A76.

That's really irony since both AMD and Nvidia outsourced their GPU to TSMC. Yes, TSMC makes GPU for both AMD and Nvidia. What are you talking about?



In terms of transmitter per milimeter, TSMC 7nm = Intel 10nm. You have no idea what you are saying.

The efficiency of AMD GPUs mainly comes from much lower efficiency of the process, than expected.

Also Nvidia has like 5 times more silicon design engineers than AMD, so they can optimize the patterning much better than AMD. AMD spends 250 mln per design. Nvidia on Volta GPU spent 3 bln alone.

But also don't expect miracles from Nvidia's Ampere when it goes for Efficiency. It will be gigantic HPC GPU, with more than 250W of TDP. Nvidia has to compete in Data Center with not Only AMD, which started eating into Nvidia's cake, but Also - Intel. And the only reason how they can compete with both clients, who will start offer massive discounts on bundled hardware is by locking them into CUDA, and massively bigger dies, smearing around reticle limit of any process.

 

[Pro7913]

The efficiency of AMD GPUs mainly comes from much lower efficiency of the process, than expected.

Also Nvidia has like 5 times more silicon design engineers than AMD, so they can optimize the patterning much better than AMD. AMD spends 250 mln per design. Nvidia on Volta GPU spent 3 bln alone.

But also don't expect miracles from Nvidia's Ampere when it goes for Efficiency. It will be gigantic HPC GPU, with more than 250W of TDP. Nvidia has to compete in Data Center with not Only AMD, which started eating into Nvidia's cake, but Also - Intel. And the only reason how they can compete with both clients, who will start offer massive discounts on bundled hardware is by locking them into CUDA, and massively bigger dies, smearing around reticle limit of any process.

I know what you are saying.
[automerge]1575335811[/automerge]

which really for all the hand-wringing about who has the flavor of the moment CPU this is really where the real advancements are going to come from. We've milked Moore's law and all of it's subsequent variants to death. Dedicated purpose driven silicon is where we're headed in which case arguing over which manufacturer is currently x% faster will seem pretty quaint.
[automerge]1575335353[/automerge]


I'll abandon my plans to moonlight as a standup. Both of those quotes were parody of what you had posted.

And yet, you dont know what you are saying.

 

[DoofenshmirtzEI]

AMD is the new king for just about everything CPU related. Zen 2 changed the industry. Honestly, RIP Intel 2019.

I'm having some serious deja vu here. I swear I've heard this before. I have been around since the days of the 8088, so there's a lot of memories floating around in there.

I feel like we are entering the Athlon days all over again...where Intel was insignificant all of a sudden.

Oh, yeah, there it is.

 

[goMac]

Is this code executed on the hardware itself, on ARM CPUs, or on Fixed Function units like Integrated Signal Processor, versus bare metal hardware on Intel?

Hardware itself. Limited SIMD where (as noted previously) Intel seems to hold their own. Outting myself just slightly a bit more, I spend most my time on Mac and iOS porting of performance sensitive things.

I have tested quite a lot of this, as well, on different CPUs. Beginning on A53 and ending on A76.

Which aren't Apple CPUs.

Apple is one of the few that paid ARM enough money that they can completely ignore the reference designs. And generally they don't follow the reference designs. They'll follow the instruction set spec, but I'm surprised they haven't introduced custom instructions already.

The reference designs ARM puts out seem like they're borrowing more and more from Apple. But I don't spend much time on Cortexes. The time I have spent on Cortex based CPUs has not made me feel as if they are as quick as Apple's, but I've never run hard numbers. This is why I few any benchmarks of server ARM CPUs very skeptically. They're likely using less optimal reference Cortex based designs.

I like Intel. I feel like they ship the most optimal designs. But while Intel is handicapped, the A series hype is real.

I think it's also the same thing holding them back right now against AMD.

 

[DoofenshmirtzEI]

Then you clearly do not understand the scale . We have never before seen AMD APUs, ALL OF THEM, let alone one APU, in Apple kexts. And you clearly forgot most important part: Van Gogh, which is Semi-Custom product. Semi-Custom. If it appears in Apple kexts, and it is semi-custom product it means two things:

1) Apple is testing stuff on this hardware, because they already have samples of it. That is what they need Kexts in the OS for.
2) APPLE ORDERED THE DESIGN OF VAN GOGH! Nobody else could, and it wont appear anywhere else, if it is product for Apple, for which, everything points to. You get it right now, what Van Gogh is, and what is happening?

3) Apple is giving Intel a warning kick in the butt. "Get your act together or else."

 

[danwells]

iPad Photoshop is NOT Photoshop - they have a smallish subset of PS running (which includes some routines they define as the engine). The other BIG issue for still photographers is Lightroom - Adobe's just looking for an excuse to ditch big, powerful Lightroom Classic and force everyone to Lightroom CC. Classic will NEVER make it to ARM, while CC started there - it's "desktopified" LR iPad.

 

[ssgbryan]

Seriously? It's a pipe dream to think you'll see a thread ripper, let alone EPYC, powered Mac Pro anytime in the near future. Is it possible some time years out, sure, anything's possible (including ARM) but you won't see one in 2020 or 2021 or 2022 or ... You'll have to wait until AMD can deliver CPUs that meet power consumption specs for the full line of Apple machines before Apple considers a move and then add another year or two of R&D for Apple to bring it to market.

So, you are saying 2020? AMD is launching their Zen 2 mobile chips at CES.

 

[deconstruct60]

If it’s JUST less work to deal with, is there another OS out there that has taken the “less work” and “higher quality code” route and support 32-bit libraries?

If not dropping the 32-bit libraries that would be another OS out there that has taken the "more work" and "lower quality code" route. Errr, yes there is (although Apple has been screwing up the "higher quality" aspect of late. Can still do dumb things on a more manageable sized code base. )

The part that you carefully snipped out is that Apple development group is "just big enough' ( Apple wants to have that "Small sized company feel" so don't run very overstaffed.). Microsoft is a much larger dev group and they cover more stuff..... Linux if cobble up all the forks, system vendors, and distributions is much bigger group of bodies also.

And there are also OS development teams that throw lots more bodies at keeping legacy apps running much longer than Apple does. But that is more work ( keeping the talent and skill knowledge up while closing up quirks and bugs over time). Sometimes the workaround for that is to stuff a comatose snapshot of the OS/libraries into a virtualized container. (e.g, old, supported version of macOS in a VM. )

There’s zero direct OS competition to Apple because there’s no other OS that runs Final Cut Pro X or Logic Pro X. If your entire use case is Open Source Portable libraries, (i.e. macOS is not your preferred OS, you could use anything) then there’s need for Apple to provide one out of a myriad solutions,

I suspect you are not in the Software business. Folks can develop code to be deploy on other systems. Especially if doing web services or just programs ported to multiple platforms. This has little to do directly with "open source libraries" at all. There are tons of closed source apps that run on Linux. If you look at the "back end" of most of the web services you'll find a flavor of Unix there. Not macOS. Nor Windows ( although more than macOS).

Macs are only good for FCPX and LogicX is a giant pile of hooey. Apple doesn't believe that at ( unless drinking gallons of kool-aid). Apple siloing the Mac into smaller and smaller niche is good way to just kill the product line. Apple nuking 32-bit plug-ins for LogicX isn't going to win them fans in a subtantive subset of the music business . (there are some folks in that group that aren't going to move quickly to the new Mac Pro 2019 at all. )

just buy the hardware that has the performance you want and then install the OS of your preference.

But if you strip out 32-bits there won't be a choice. Doing that would be multiples worse that way apple did with the escape key ( and had to backtrack on. )

 

[DoofenshmirtzEI]

Ask people from Data Centers, Servers, and HPC space, why haven't they switched to ARM, if it is so powerful, and power efficient.

AWS has been offering ARM based EC2 instances for about a year now. I don't know much about them except what was in the reInvent talks last year, but their suggestions for use cases are "all open source software that scales well horizontally." AWS rolled their own ARM chip for these, so if you want to benchmark them, you have to spin up an EC2 instance.

 

[Unregistered 4U]

Do you have any proof? That would be a significant departure from the ARM64 standard.

I’ll PM the link to the ArsTechnica article. ”The 64-bit ARM instruction set, also called AArch64, is unique in that it is totally separate from the 32-bit AArch32 instruction set.”

 

[Unregistered 4U]

although Apple has been screwing up the "higher quality"

I agree. It’s interesting, but Inever thought of 64 bit only leading to higher quality code.

Apple doesn't believe that at ( unless drinking gallons of kool-aid).

If you check out the Mac Pro intro video, they spend a lot of time on what you can do with FCPX and Logic Pro X on the new Mac Pro. Sure you COULD run Adobe Premier or AVID on it, BUT you’re likely to find a better performing/cheaper Adobe Premier option on the PC side (unless macOS is a preference, in which case spending more for a lower performing Mac is just “what you do”).

good way to just kill the product line.

I am NOT against this thought. If FCPX, Logic Pro X, and iOS code development is ever made available for an OS OTHER than macOS, I believe that’s the end of macOS.

And, when I said “just buy the hardware that has the performance you want and then install the OS of your preference.” I meant ANY hardware from ANY vendor that’s not Apple could fill the need. If your code can really run anywhere, and you don’t have a strong preference for macOS, what would be the business case for an expensive Mac over Dell, etc.? EVEN if it was a Mac with an AMD CPU, it will never have cost parity with PC’s from other vendors.

 

[koyoot]

AWS has been offering ARM based EC2 instances for about a year now. I don't know much about them except what was in the reInvent talks last year, but their suggestions for use cases are "all open source software that scales well horizontally." AWS rolled their own ARM chip for these, so if you want to benchmark them, you have to spin up an EC2 instance.

I don't have to. Way more competent people, than me, actually have got their hands on ARM CPUs that are in AWS and tested them.

This is exactly what I was talking about. There is a good reason why nobody in Data Centers, in Cloud uses those servers for anything else, than basic stuff. Yes - they are cheaper.

But also they are much less capable.

3) Apple is giving Intel a warning kick in the butt. "Get your act together or else."

Its not a warning, if Intel, and other sources say that Apple is ditching them in favor of... ?

 

[ericinboston]

The question - for Apple - is not: Which one is faster? The question is: How much will it cost us, and how many customers will we get or lose, by switching to AMD?

Are you going to buy a Mac Pro with AMD processor? How many will switch from PC + Threadripper + Linux to Mac + Threadripper? And how many will switch away from Mac Pro if it doesn't come with an Intel processor?

Apple has been working with Intel for over 13 years. Even if AMD is ahead today, for how long?

Although I am late to this thread and have read most of the comments, you really hit the nail on the head. Some more elaboration from my side would be:

-Apple very likely is tied to Intel for some foreseeable future. The Mac Pro needs a refresh so it's either wait it out (how many years?) or offer something now.

-I don't follow every AMD and Intel release, but just because AMD has the fastest ABC chip today, December 3, does not mean that Intel won't have an even faster chip 4 months from now. I know the AMD chip on this thread is fastest by a good margin, but Intel may close or beat that gap a few months from now.

-Let's not forget that the Mac Pro isn't even AVAILABLE YET. Apple announced the refresh 6 months ago (golly!) and you can't even buy the computer let alone start using it. Apple had to announce most of the specs 6 months ago and therefore had to choose the specs long before that. Hence another reason we are seeing Intel chips in the design.

-The Mac Pro is going to be used by different people for different software titles and different use cases and each user will have different expectations. But I would argue that unless you are running the CPU 24x7x365 at full throttle, the CPU is not the bottleneck. The drive is. I would therefore state that if most people are going to use the Mac Pro for heavy, yet not 24x7 full throttle CPU usage, then who cares if the AMD is 10-20% faster? An example of a software title that can run the CPU full throttle for 24x7x365 is BOINC.

-Apple is a company who produces very, very few models of computers and infrequent updates. Apple likes/wants to control as much as possible on the hardware and software side. Therefore, they have to pick a CPU manufacturer and stick with it for quite a bit of time (decade usually) otherwise Apple's going to be spending more time/money on redesigns, testing, etc. and Apple likes a nice profit margin so I don't see any kind of CPU switching or offering different manufacturers at the same time. I also don't see Apple offering more than 1 CPU manufacturer at a time given the volumes (and needs) of Mac sales. Dell and Lenovo and others sell both Intel and AMD to offer choice and to also, in some cases, allure people to a low(er) budget machine.

-As others have mentioned, is Apple going to lose sales due to switching to AMD? Whether it's a valid technical reason or, more likely, customer belief that AMD is inferior. The customer buying these $6000+ Macs don't give a hoot if they think they could have saved $400 to use the AMD chip.

 

[teagls]

-I don't follow every AMD and Intel release, but just because AMD has the fastest ABC chip today, December 3, does not mean that Intel won't have an even faster chip 4 months from now.

-Let's not forget that the Mac Pro isn't even AVAILABLE YET. Apple announced the refresh 6 months ago (golly!) and you can't even buy the computer let alone start using it. Apple had to announce most of the specs 6 months ago and therefore had to choose the specs long before that. Hence another reason we are seeing Intel chips in the design.

-The Mac Pro is going to be used by different people for different software titles and different use cases and each user will have different expectations. But I would argue that unless you are running the CPU 24x7x365 at full throttle, the CPU is not the bottleneck. The drive is. I would therefore state that if most people are going to use the Mac Pro for heavy, yet not 24x7 full throttle CPU usage, then who cares if the AMD is 10-20% faster? An example of a software title that can run the CPU full throttle for 24x7x365 is BOINC.


-As others have mentioned, is Apple going to lose sales due to switching to AMD? Whether it's a valid technical reason or, more likely, customer belief that AMD is inferior. The customer buying these $6000+ Macs don't give a hoot if they think they could have saved $400 to use the AMD chip.

-Intel isn't coming out with anything remotely faster. Chip development and production takes years and analysts have gotten pretty good at estimating what's to come. 4 months is literally impossible.

-You say the drive is the bottleneck, but AMD offers much better drive performance. More pcie-lanes, pcie v4.

-Apple will lose customers if customers lose money through productivity and are forced to use other systems.

 

[koyoot]

-Let's not forget that the Mac Pro isn't even AVAILABLE YET. Apple announced the refresh 6 months ago (golly!) and you can't even buy the computer let alone start using it. Apple had to announce most of the specs 6 months ago and therefore had to choose the specs long before that. Hence another reason we are seeing Intel chips in the design.

-The Mac Pro is going to be used by different people for different software titles and different use cases and each user will have different expectations. But I would argue that unless you are running the CPU 24x7x365 at full throttle, the CPU is not the bottleneck. The drive is. I would therefore state that if most people are going to use the Mac Pro for heavy, yet not 24x7 full throttle CPU usage, then who cares if the AMD is 10-20% faster? An example of a software title that can run the CPU full throttle for 24x7x365 is BOINC.

-As others have mentioned, is Apple going to lose sales due to switching to AMD? Whether it's a valid technical reason or, more likely, customer belief that AMD is inferior. The customer buying these $6000+ Macs don't give a hoot if they think they could have saved $400 to use the AMD chip.

Intel will not have anything faster than AMD for foreseeable future.

10 nm woes are so serious, that they are porting back 10 nm Architecture: Willow Cove to 14 nm(which will result in two things - lack of efficiency, and low clocks, apart from bloated die sizes) with Rocket Lake.

AMD is the king of computing since now. Get used to this, guys. Zen 3 will fix all of the problems Zen architecture ingherently has: latency and slow L1 cache write. Zen 3 will have over 2000 GB/s write to cache, and you know what that means?

It will solve any bottleneck that Zen 2 had, and every application that was relying on this(almost all of them) will get bump in performance.

Zen 3 will have 15-20% higher performance than Zen 2. What Intel has for next year? Another 14 nm Skylake respin. Yay!

AMD is not going to be 15-20% faster. Current 64 core EPYC CPUs apart from AVX512 are 100% faster than fastest Xeon 28 core CPUs. Guess what will happen with Zen 3?

Customers believe that AMD is inferior brand to Intel?

Well those guys are plainly stupid, if they believe so, especially right now.

-Apple will lose customers if customers lose money through productivity and are forced to use other systems.

It already started happening, in other spaces, than Apple's sandbox/fairyland.

 

[thisisnotmyname]

Intel will not have anything faster than AMD for foreseeable future.

10 nm woes are so serious, that they are porting back 10 nm Architecture: Willow Cove to 14 nm(which will result in two things - lack of efficiency, and low clocks, apart from bloated die sizes) with Rocket Lake.

AMD is the king of computing since now. Get used to this, guys. Zen 3 will fix all of the problems Zen architecture ingherently has: latency and slow L1 cache write. Zen 3 will have over 2000 GB/s write to cache, and you know what that means?

It will solve any bottleneck that Zen 2 had, and every application that was relying on this(almost all of them) will get bump in performance.

Zen 3 will have 15-20% higher performance than Zen 2. What Intel has for next year? Another 14 nm Skylake respin. Yay!

AMD is not going to be 15-20% faster. Current 64 core EPYC CPUs apart from AVX512 are 100% faster than fastest Xeon 28 core CPUs. Guess what will happen with Zen 3?

Customers believe that AMD is inferior brand to Intel?

Well those guys are plainly stupid, if they believe so, especially right now.

DOOOOOOOOOM!

uh huh, just like AMD went out of business during the (many) years that Intel has been on top. Intel will obviously never have a higher benchmark (which benchmarks are the only thing that matters, I know I'll never be happy if I can't tell people on the internet that I purchased something some company made that has a higher number on an arbitrary set of tasks than something other people purchased that some other company made) than AMD ever again despite their massive resources compared to AMD's serious debt. Obviously Intel should just close their doors now, koyoot has declared their business is done.

?

 

[carlos700]

Yes, AMD has faster CPUs for workstations and servers, but they are still perceived as lower end budget parts. Just as when they were faster during the Athlon 64 X2 and K6-III days. I have a hard time picturing Apple charging $5999 for a Threadripper or EPYC box versus a Xeon box.

 

[ericinboston]

Intel will not have anything faster than AMD for foreseeable future...

Thanks to you and teagls for the technical dives. I think that while they are enlightening to us techies, what percentage of Mac Pro customers are a)that technical and b)have the ability to care AND switch from Mac if they are not happy with performance? And, how much does the AMD cpu really affect their particular software title and use-case? We (here) all read tech specs all the time, but the real world is much different than just some tool thrashing the CPU and other pieces of the system.

In my eyes, Apple has neglected the true Mac professional for a long time...like 20 years. And I mean true, diehard Mac users working with CAD or movies or music or similar that every minute or hour of time really counts not only for profit but for delivery of some product/service. I think Apple really just doesn't care for that market, especially since Apple no longer has the monopoly in those use cases. Much of the number crunching software runs on multiple platforms and/or better platforms. Therefore, Apple isn't going to directly compete with some computer farm that can render a movie in 1/10th the time of the best Mac. Apple isn't going to compete with some workstation/server than can host 4 CPUs with up to 80 cores, for example. Apple's going to stuff the fastest specs they can WHILE ALSO ensuring their hefty margin. If Apple can't make the margin, something(s) get's dropped.

Yes, the AMDs are blazing fast...but is it a fit for 70%+ of Apple's Mac Pro use cases to warrant a redesign, OS and software testing, new marketing, a new license agreement with AMD (for 10 years I'm sure), allthewhile somehow negatively impacting its relationship with Intel? Will the iMac Pro switch to AMD since the MP is running AMD? What about the Macbook Pros? If these "pro" models do switch and yet Apple keeps Intel for all the non-pro lineup, wouldn't that imply/infer that Intel can't create "pro" cpus? And if the other "pro" models do not switch to AMD, then, is it really a "pro" class machine (in the eyes of the customer)?

Overall I think the answer is not a technical reason.

 

[koyoot]

DOOOOOOOOOM!

uh huh, just like AMD went out of business during the (many) years that Intel has been on top. Intel will obviously never have a higher benchmark (which benchmarks are the only thing that matters, I know I'll never be happy if I can't tell people on the internet that I purchased something some company made that has a higher number on an arbitrary set of tasks than something other people purchased that some other company made) than AMD ever again despite their massive resources compared to AMD's serious debt. Obviously Intel should just close their doors now, koyoot has declared their business is done.

?

Tell me my friend, why writing "A" immediately for some people means "B".

Intel will not have anything faster than AMD for foreseeable Future means, that at best - we will have a tie between them.

And this may be best scenario possible, for us consumers! IF there is a tie between them, it means lower prices! What more do we want?

And no, Intel will not have anything faster than AMD, for foreseeable future .

P.S. If you fail to deliver most important part of any architecture - the process(!), you are failing to deliver products. Intel did this with 10 nm products, and that allowed AMD to catch up. But contrary to Intel, they do not want to get complacent, and will run this race as fast, and as hard as they can. So yes, if Intel will not fix their ****, they are doomed. Well, at least their fabs are.

Thanks to you and teagls for the technical dives. I think that while they are enlightening to us techies, what percentage of Mac Pro customers are a)that technical and b)have the ability to care AND switch from Mac if they are not happy with performance? And, how much does the AMD cpu really affect their particular software title and use-case? We (here) all read tech specs all the time, but the real world is much different than just some tool thrashing the CPU and other pieces of the system.

In my eyes, Apple has neglected the true Mac professional for a long time...like 20 years. And I mean true, diehard Mac users working with CAD or movies or music or similar that every minute or hour of time really counts not only for profit but for delivery of some product/service. I think Apple really just doesn't care for that market, especially since Apple no longer has the monopoly in those use cases. Much of the number crunching software runs on multiple platforms and/or better platforms. Therefore, Apple isn't going to directly compete with some computer farm that can render a movie in 1/10th the time of the best Mac. Apple isn't going to compete with some workstation/server than can host 4 CPUs with up to 80 cores, for example. Apple's going to stuff the fastest specs they can WHILE ALSO ensuring their hefty margin. If Apple can't make the margin, something(s) get's dropped.

Yes, the AMDs are blazing fast...but is it a fit for 70%+ of Apple's Mac Pro use cases to warrant a redesign, OS and software testing, new marketing, a new license agreement with AMD (for 10 years I'm sure), allthewhile somehow negatively impacting its relationship with Intel? Will the iMac Pro switch to AMD since the MP is running AMD? What about the Macbook Pros? If these "pro" models do switch and yet Apple keeps Intel for all the non-pro lineup, wouldn't that imply/infer that Intel can't create "pro" cpus? And if the other "pro" models do not switch to AMD, then, is it really a "pro" class machine (in the eyes of the customer)?

Overall I think the answer is not a technical reason.

The excuses are coming, and coming.

Let me give you all, resistant to changes users, advice.

You can build a DIY machine, for the price of Basic Mac Pro, that will be faster than Fully loaded Mac Pro.

Why don't you give it a try, and then ask: will it benefit us, resistant to changes users?

 

[ericinboston]

The excuses are coming, and coming.

Let me give you all, resistant to changes users, advice.

You can build a DIY machine, for the price of Basic Mac Pro, that will be faster than Fully loaded Mac Pro.

Why don't you give it a try, and then ask: will it benefit us, resistant to changes users?

I'm not sure why you quoted me. I am not defending Apple, merely offering reasons (again, non-technical reasons) why Apple is not offering AMD cpus on the MP or frankly any other Mac.

On a side note, I built hundreds of machines back in the late 80s to mid 90s. I stopped because prebuilt machines were just more affordable considering all the time/effort/warranty chasing to DIY and that I was no longer 20 years old with nothing to do but work and go out and party.

There is no doubt that anyone can build a DIY machine with the latest hardware to beat any Mac or Dell or Lenovo or whatever. That's not the point of this thread.

 

[carlos700]

Tell me my friend, why writing "A" immediately for some people means "B".

Intel will not have anything faster than AMD for foreseeable Future means, that at best - we will have a tie between them.

And this may be best scenario possible, for us consumers! IF there is a tie between them, it means lower prices! What more do we want?

And no, Intel will not have anything faster than AMD, for foreseeable future .

P.S. If you fail to deliver most important part of any architecture - the process(!), you are failing to deliver products. Intel did this with 10 nm products, and that allowed AMD to catch up. But contrary to Intel, they do not want to get complacent, and will run this race as fast, and as hard as they can. So yes, if Intel will not fix their ****, they are doomed. Well, at least their fabs are.


The excuses are coming, and coming.

Let me give you all, resistant to changes users, advice.

You can build a DIY machine, for the price of Basic Mac Pro, that will be faster than Fully loaded Mac Pro.

Why don't you give it a try, and then ask: will it benefit us, resistant to changes users?

Considering the 3275M is over $7K, likely not. Unless you are considering an AMD-based system, then likely so.

 

[koyoot]

I'm not sure why you quoted me. I am not defending Apple, merely offering reasons (again, non-technical reasons) why Apple is not offering AMD cpus on the MP or frankly any other Mac.

On a side note, I built hundreds of machines back in the late 80s to mid 90s. I stopped because prebuilt machines were just more affordable considering all the time/effort/warranty chasing to DIY and that I was no longer 20 years old with nothing to do but work and go out and party.

There is no doubt that anyone can build a DIY machine with the latest hardware to beat any Mac or Dell or Lenovo or whatever. That's not the point of this thread.

What I meant was this: you can build a AMD Based system that will be faster than fully loaded Mac Pro, for the price of the basic model.

Considering the 3275M is over $7K, likely not. Unless you are considering an AMD-based system, then likely so.

You can use Threadripper to get more performance, than fully loaded Mac Pro, and pay for it the price of basic model.