MP 7,1 Waiting for Mac Pro 7,1 A1991 (no more)

[Mago]

it doesn't really make sense to switch to AMD and then to ARM within 5 years later.

Seems you are from the few still ones not aware that switching from Intel CPU to AMD CPU its like switching RAM from Crucial to Micron, FYI both AMD and Intel CPU from software perspective are the same thing, few very niche optimizations very quick to update in software and actually dont prevent code to run.

Changes from Intel to AMD are the same as when Apple changes from a Intel core Generation to another one newer: new motherboards/chipset.

Thunderbolt 3 only requires Bios support if you run on an Intel TB controller (as Titan Ridge), Intel released Thunderbolt IP and several 3rd party promised compatible controller chipset (Likely targeting both AMD/Intel/ARM motherboards).

 

[koyoot]

The only way I can see Apple making computers on ARM is by... giving iOS desktop-class capabilities. File system, ability to connect to any peripherals, etc.

Seems you are from the few still ones not aware that switching from Intel CPU to AMD CPU its like switching RAM from Crucial to Micron, FYI both AMD and Intel CPU from software perspective are the same thing, few very niche optimizations very quick to update in software and actually dont prevent code to run.

Changes from Intel to AMD are the same as when Apple changes from a Intel core Generation to another one newer: new motherboards/chipset.

Thunderbolt 3 only requires Bios support if you run on an Intel TB controller (as Titan Ridge), Intel released Thunderbolt IP and several 3rd party promised compatible controller chipset (Likely targeting both AMD/Intel/ARM motherboards).

THERE IS NO SWITCH FROM INTEL TO AMD!

Both are x86-64 architectures with the same f****** ISA's. There are hackintoshes that are running on Ryzen CPUs, without any changes to the OS code, or workarounds.

But, this is a minor issue when you see the negative impact in developers is having Metal, Metal is great, but no one likes it, ask the community everybody wants is Vulkan, and OpenCL 2.x, kicking off from the Apple macOS/iOS ecosystem was a huge mistake, despite solutions to compile Vulkan/OpenCL into Metal, I considered it, but has little sense, since you re not away from new bugs/tuning.

Apple's big error it to believe they are Almighty and Irreplaceable, sometimes hurts to switch, but now I'm very Happy on Ubuntu Linux doing everything I used to do in a Mac, (except Autocad/OmniGraffle ) and all Apps/IDEs I needed have now native Linux/Ubuntu implementations (Pycharm, AndroidDev, CUDA, mROC OpenScad, LibreOffice etc), so i should care on the Mac Pro just to code for Metal ? NO

If you guys want OpenCL, Vulkan, and ease of use of Unix there are... Linux Distros. Elementary OS, Fedora, and latest star of Linux Distros: Solus Project, with Budgie desktop. Heck, even Ubuntu Budgie is quite nice, if you are able to deal with spyware.

In the bonus of Linux, if you want AMD hardware - ROCm. The only place where you can use CUDA replacement for AMD hardware is Linux.

 

[askunk]

Preliminary info on the Vega VII suggest it to be the same MI60 with video output, no PCIe4 no Infinity Link, while seems VegaVII to have a custom xGMI bridge inferior to the one on MI60, or at least coming later as the MI60 market gets cold (driven by FP64 performance, a small niche)

Not entirely correct. Radeon 7 supports PCI-e 4.

What is curious is how they market it as a gaming card, while it's much more than that. 16GB of VRAM are barely used by any game, but come very useful in technical work.

 

[AidenShaw]

Both are x86-64 architectures with the same f****** ISA's.

Trying running Intel's x64 AVX-512, SGX and TSX instructions on a red CPU

 

[koyoot]

Not entirely correct. Radeon 7 supports PCI-e 4.

What is curious is how they market it as a gaming card, while it's much more than that. 16GB of VRAM are barely used by any game, but come very useful in technical work.

Its branding is separate from any previous branding of any AMD GPU.

The manufacturing costs are low enough that AMD can price it at $700 , and market as general purpose GPU, for gamers, and for professionals. Which is quite interesting approach.

Trying running Intel's x64 AVX-512, SGX and TSX instructions on a red CPU

Does OSX use those instructions in any way, shape or form?

No.

P.S. Last time I have heard is that SGX is pretty much useless feature, that Intel tries to sell.

 

[AidenShaw]

Does OSX use those instructions in any way, shape or form?

No.

P.S. Last time I have heard is that SGX is pretty much useless feature, that Intel tries to sell.

So you admit that Intel and AMD are different architectures with different ISAs. Good.

Perhaps you should look into the security research about using SGX (and AMD's incompatible SEV) enclaves for securing encryption and other security features. (Apple's T2 chip also implements secure enclaves.)

 

[koyoot]

So you admit that Intel and AMD are different architectures with different ISAs. Good.

Nope. AVX 512 will be supported by AMD. TGX and SGX are Intel's proprietary features, that tries to sell, and fails miserably, because both are useless apart from few, boarder use cases.

ISA is the same for both AMD and Intel. No matter what way you will spin this.

 

[AidenShaw]

ISA is the same for both AMD and Intel. No matter what way you will spin this.

My spin is that if AMD can't execute an instruction that Intel can execute - by definition the Instruction Set Architectures are different.

Bada bing, Bada boom.

Does OSX use those instructions in any way, shape or form?

Isn't that a bit sad, that Apple OSX doesn't use modern 256-bit and 512-bit instructions?

Perhaps you'd be happy if Apple OSX limited itself to the 16-bit instruction set?

And what about apps that have been optimized to use 512-bit instructions? Tell them to use Windows or Linux?

 

[GrumpyCoder]

Seems you are from the few still ones not aware that switching from Intel CPU to AMD CPU its like switching RAM from Crucial to Micron, FYI both AMD and Intel CPU from software perspective are the same thing, few very niche optimizations very quick to update in software and actually dont prevent code to run.

Good thing you figured this out, maybe I should tell all my students at the university I’m teaching in computer science.
Again, it doesn’t make sense to switch to AMD. But while you’re at it, could you explain a little bit in detail how to get AMD CPUs going with the T2 chip? If you need to dive in, I’ve done hardware design and also wrote operating systems, I’ll do my best to understand your explanation.

Both are x86-64 architectures with the same f****** ISA's. There are hackintoshes that are running on Ryzen CPUs, without any changes to the OS code, or workarounds.

Instructions from AMD and Intel differ, at least parts of it. Any link for Ryzen running Mojave? The latest I know of is the HighSierraV3 installer, otherwise it doesn’t even boot for installation. Do they have iMessage/FaceTime running now or is it still not working?

 

[askunk]

The manufacturing costs are low enough that AMD can price it at $700 , and market as general purpose GPU, for gamers, and for professionals. Which is quite interesting approach.

Actually, the point I was making is that they explicitly presented the card as a gaming card, adding just a few comments on GPGPU work. Pretty much the opposite of what they did with Vega FE, where the card was marketed as a general purpose product, capable also of gaming.

 

[DearthnVader]

Instructions from AMD and Intel differ, at least parts of it. Any link for Ryzen running Mojave? The latest I know of is the HighSierraV3 installer, otherwise it doesn’t even boot for installation. Do they have iMessage/FaceTime running now or is it still not working?

I've got a Ryzen running Mojave, no big deal, just tell it it's a Penryn.

 

[AidenShaw]

I've got a Ryzen running Mojave, no big deal, just tell it it's a Penryn.

Yep - no big deal.

Just tell Apple OSX that it's a CPU from 2007, and ignore twelve years of CPU development.

 

[koyoot]

My spin is that if AMD can't execute an instruction that Intel can execute - by definition the Instruction Set Architectures are different.

If that would be 100% correct, macOS in Hackintoshes would not run on AMD hardware.

Bada bing, bada boom .

And what about apps that have been optimized to use 512-bit instructions? Tell them to use Windows or Linux?

Yep. That is what I actually would do.

Yep - no big deal.

Just tell Apple OSX that it's a CPU from 2007, and ignore twelve years of CPU development.

Well, you can tell software to do anything. Like Intel did with their own compiler, when detecting AMD deviceID, to gimp performance on AMD's CPUs, few years back, which resulted in trial.

Instructions from AMD and Intel differ, at least parts of it. Any link for Ryzen running Mojave? The latest I know of is the HighSierraV3 installer, otherwise it doesn’t even boot for installation. Do they have iMessage/FaceTime running now or is it still not working?

There is plenty of YouTube videos on the matter from last few months, and weeks.

 

[Mago]

Good thing you figured this out, maybe I should tell all my students at the university I’m teaching in computer science.
Again, it doesn’t make sense to switch to AMD. But while you’re at it, could you explain a little bit in detail how to get AMD CPUs going with the T2 chip? If you need to dive in, I’ve done hardware design and also wrote operating systems,

I was a CS teacher too, I've done HW design too, even from time on time I do ASIC Design, My Degree Thesis originally was about my Own OS, I foresee to write something like CP/M running on a paRisc workstation then some industry offered me to work on some specific simulator (finite element analysis google about what it is, maybe you have an stackoverflow section for this, I never had SO when I was a undergraduate intern, many guys now deserves their degrees to StackOverflow and ^C^V)

Beyond your "elevated" (sarcastic) language, I read few things that deserves a bit of educated illustration:

T2 chip has nothing to do with the ISA architecture it runs, itself just another peripheral controller (one that manages SSD storage, Sec. Enclave, Encryption, transcoding etc), like a complement to the MOBOś main chipset (C232, Z399 etc), the Bios should include some "driver" instructions (x86-emt64 likely) to communicate with it, nothins specially tied to an specific main CPU vendor, unless you have fantasies about super-apple engineering figuring this chip does some strange code-injection/hacking into the Intel cpus bringin it super-powers no other system have...OK
[doublepost=1547174867][/doublepost]

Just tell Apple OSX that it's a CPU from 2007, and ignore twelve years of CPU development.

Performance improvement introduced from Penryn to Skylake related to the instructions set, are all about integer-FP matrix, I dont see a lot of OS-related task being improved thru Matrix FP operations, maybe some crypto gets improved from AVX-512, not big deal for Zen which implements AVX2 (aka AVX256), you cant ignore CPU development when you code cryptographic, Matrix, or transcoding applications, but bare OS management dont account big on neither AVX extension.

About AVX-512, it improves FP/Integer Matrix, really but who cares? this functionality is now taken to the GPUs where makes sense to run such complex high parallelism operations. I don't see many customers taking care of AVX-512 in a current active projects.

 

[Biped]

all this noise about instruction and feature sets, architecture, what runs and what doesn't.

not one mention of compilers

it's almost like you all work entirely in assembly

just like the good old days

you all know how to make an old fart shed a tear in nostalgic bliss

 

[mattspace]

THERE IS NO SWITCH FROM INTEL TO AMD!

Both are x86-64 architectures with the same f****** ISA's. There are hackintoshes that are running on Ryzen CPUs, without any changes to the OS code, or workarounds.

See you can say that, but I have friends working as contractors on a major big budget 3D animated film at the moment, who had an AMD workstation built for this job, and it simply didn't work properly with the mainstream animation package they work in. They lost days chasing down the cause, and it turns out that a lot of features in this package simply don't work properly on AMD processors - they went back to an intel machine, slowdowns disappeared.

You can call them the same, you can blame app vendors for not optimising for one brand, or for committing to proprietary aspects of another brand, but the fact remains for the person actually deploying and using the machine in their practice, AMD and Intel based machines are not the same, or even reliably functionally equivalent.

 

[Mago]

it's almost like you all work entirely in assembly

C++ on GCC/Clang LLVM Python Golang Kotlin TypeScript VeriLog, accidentally some java/asm

PS I'm now in love with Go and Kotlin, I learn Swift, but it is very like Kotlin with some Apple's opionate.

 

[filmak]

Xeon = Higher Cost, Lower Performance.

Intel's only advantage is single core performance (and that should disappear with the release of the next gen Ryzen CPUs). But it isn't 2010 anymore; most programs are multi-threaded.

AMD's CPUs would be perfect for Apple - throw the Ryzen 3000 with APUs into the Mac Mini, Ryzen 5 and 7 series in the Imac, and Ryzen 9 in the Mac Pro.

What about the portables?
Are Amd's offerings better, in power efficiency, performance and heat, than intel's? because, I think, that if they move to AMD they will have to move all of their products, not only the desktops.

 

[4tzer]

- I'm sure apple had enough time to develop an OS that would accept AMD CPU completely.
- also, why not both Intel and AMD support in one OS? MacPro as AMD and future iMacs could be a smart move, integrated graphics... open cl.
- there is this guy who got a working tb3 on AMD, so that myth is bused
- there was this promise that tb3 would be available to other manufacturers by 2018, but intel didn't deliver yet. Could have been a promise to Apple to make it widely available on AMD platform.
- AMD's CPU solution is more future proof.
- Apple takes a long time developing the 7,1; As if they are waiting for a development. When i got my first Mac Pro (into quad) it would have cost the same to make a windows equivalent server grade quality. Ever since prices went up, repairability went down, i hope they come to their senses. There are still people who buy apple blindly, but a lot of them got a product (last mac pro) that had real problems. Apple really shouldn't repeat that.

 

[askunk]

Technically, I do agree completely. Apple has developed a CISC compatible macOS for over 6 years along the RISC version, compiling some bios and drivers for AMD chips is a cheap piece of cake for Apple.

Again, I think the problem is the fact that Apple will very soon start to move to proprietary ARM, which is a wise choice on so many aspects, especially given by how fast the Ax chips are developing. The transition will very likely be painful (all transition have been, more or less) and expensive and will require some reshuffling in the company. Cook is (should be, actually) an expert of the supply chain and I am pretty sure they won't change the present workflow when they will have to do it again soon (or already have started to).

Having said that, I believe as many on this forum that it is a managerial/contractual/logistical issue, not at all a technical issue. If they wanted to switch to AMD chips, they could do it in a day. I am pretty sure there are many Ryzen/Threadripper machines running macOS in some lab in Cupertino, as well as we all imagine they have some concept laptop with ARM chips. Whether and when they will appear on the market it's up to their vision.

Steve wanted to have the most powerful computers on the planet in the catalogue and was trying to bring them to a decreasing price trend. Tim does not seem at all to be following that philosophy, therefore cheap and fast do not seem to be critical parameters, unfortunately.

I am appalled... this is the last few months I give to Apple before I will be forced to face my own transition. (32 years of macOS... geeze... I'm scared to move to Windoze...) Fast, cheap, slow, expensive, modular, split, monoblock... I don't give a damn anymore, Apple give us a bloody computer.

 

[ssgbryan]

What about the portables?
Are Amd's offerings better, in power efficiency, performance and heat, than intel's? because, I think, that if they move to AMD they will have to move all of their products, not only the desktops.

Same thing - AMD is supplying every major PC company with CPUs (with Vega APUs) for both their laptops and chromebooks.

AMD's Keynote presentation Wednesday covered it all.

 

[koyoot]

What about the portables?
Are Amd's offerings better, in power efficiency, performance and heat, than intel's? because, I think, that if they move to AMD they will have to move all of their products, not only the desktops.

Ryzen U CPUs are 15 and 35W. To achieve the same performance level as one APU, with 4 cores/8 Threads, and Vega 10 GPU, you need to have 4C/8T CPU from Intel, with 15W TDP, like Core i5 8350U, AND dGPU like Nvidia MX150 in the laptop.

So if you go with AMD, you get the same performance, simpler package(because of one SOC, instead of two different), simpler cooling, better efficiency, lower production costs.

 

[filmak]

Same thing - AMD is supplying every major PC company with CPUs (with Vega APUs) for both their laptops and chromebooks.

AMD's Keynote presentation Wednesday covered it all.

Ryzen U CPUs are 15 and 35W. To achieve the same performance level as one APU, with 4 cores/8 Threads, and Vega 10 GPU, you need to have 4C/8T CPU from Intel, with 15W TDP, like Core i5 8350U, AND dGPU like Nvidia MX150 in the laptop.

So if you go with AMD, you get the same performance, simpler package(because of one SOC, instead of two different), simpler cooling, better efficiency, lower production costs.

Thank you both.

 

[koyoot]

Speaking of APUs: https://www.anandtech.com/show/1385...yzen2000?utm_source=twitter&utm_medium=social

APUs will not be chiplet based, but designs are "Mobile first". Which is quite unusual, because Zen1 design was Server first, and APU design was Desktop first.

 

[GrumpyCoder]

- also, why not both Intel and AMD support in one OS?

They're not going to have AMD in a MacPro and Intel in the iMac or any other line. They can transition to whatever they want, it's a question of how much effort goes into it (not just engineering) and how much sense it makes in the end.
So if they plan to switch to AMD, then they will do so for every machine, MacPro, iMac, MacMini, MBP, etc. Probably not all at once, but step by step. If they stick to their usual updates, they'll probably have this done within a year. And then what, switch to ARM a while later? So AMD would be a temporary "fix"? ARM is not there yet for a single core, but multicore is fantastic. If it takes another year or two for Apple to get a desktop ARM ready, it doesn't make sense to switch now. If they do, they've probably given up the idea of switching to ARM at all.

And let's face it, the MacPro is a professional machine. I get it, alot of home users would like such a machine at home, despite the fact they don't need it. I don't need one at home anymore either, but that wouldn't stop me from buying one even if it's just to answer mail and browse the web. From a professional point of view, price is not an issue. What kind of professional doesn't make the money back from more expensive professional components? It doesn't matter if a GPU is $500 or $5k as long as it helps to make money. And it wouldn't matter if the new MacPro starts at $5k or $20k. If it get's the job done and it's not paying for itself, then maybe the one using it doesn't need such a machine or isn't as professional as he thinks (hello some youtubers with 500 subscribers).

You can call them the same, you can blame app vendors for not optimising for one brand, or for committing to proprietary aspects of another brand, but the fact remains for the person actually deploying and using the machine in their practice, AMD and Intel based machines are not the same, or even reliably functionally equivalent.

Of course they're not. Good scientific software will tell you so and fall back to the greatest common denominator to run and that can in some cases lead to what you described. Speaking of, one of my new colleagues worked at CERN before becoming a full time professor in our big data group. Might have to ask him what his experience is on this issue as it probably doesn't get more data intensive than what they do there in cooperations with research institutes.