MP All Models 3rd gen Ryzen... Still not possible?

[deconstruct60]

Rest assured, Apple will have multiple prototypes running Ryzen, ARM and perhaps their own CPU designs. Just look at their history;
....]

I don't think their own SoC/CPU designs are in the "perhaps" category. Those are probably more likely than the others. iOS shares a common foundation so it is to a very large extent what they are doing anyway. Perhaps not porting some of the older drivers ( Firewire) but the core macOS and some of the mainstream bundled apps.

Ryzen (AMD) isn't some "Land Man on the Moon" project either. AMD is just as x86_64 99+% user space the same as Intel's implementation. There is only some narrow set of low level privilege mode and mild gaps on AVX implementations to deal with. intel and AMD have cross licensing on x86 instructions so there is no huge gulf here at all.

[ Apple chasing the NUMA impacts the first two iterations of Zen had? Probably not. Deep compiler optimizations on Apple's LLVM fork? Probably not. Working enough to run test? Apple can 'hackintosh' probably at least as good as folks on the outside. ]

3rd party ARM solutions? Errrr, that is the one that is a bit doubtful. Qualcomm , Samsung , Mediatek , Hisilocon/Huawei ... I'm not sure why they'd spend tons of effort helping Apple scope their processors as their are deep competitors at this point. The sever ones Ampere / Marvell/Cavium ThunderX2 in Apple Data center perhaps, but the track record of those to bet the Mac desktop farm on seems to be a stretch.

Apple putting prototype N1 reference board through the paces ( https://www.anandtech.com/show/13959/arm-announces-neoverse-n1-platform ) maybe. But highly likely lower priority than the other two. It is aimed at a different workload that higher end Mac systems are. Probably not going to significantly gap both Intel and AMD over next few years.


that video about Apple moving to Intel (from PPC) to chase performance per watt has more weight on the laptop part of the Mac line up versus the desktop one. The new Mac Pro weighs in a 1.5KW power supply. That isn't modest sipping of power. ( I'm sure over time Apple would like to do "more" with the same 1.5K budget because can't get any more out of a normal plug. But points to different parts of the line up aimed at significantly different targets. )
The overall Mac market is relatively small to 'split' into subdivisions, but Apple may go that course ( especially if they can just 'reuse' pragmatically already 'paid for' iPadOS SoC. ). Chasing the ever thinner laptop with that.

 

[orph]

Apple has a history of hissy fits over not getting the 'love' they want from their partners.
IBM do not make the PPC in to what apple wants and they swap to intel
apple has fight with ATI and drops ATI for nvidia
nvidia and apple have a fight and swap to ATI
they have a fight over there iphone things and buy the modem department from intel

who remembers the-
'Steve Jobs stated during his keynote presentation that the Power Mac G5 would reach 3 GHz "within 12 months."'

there must have been some shouting matches when 3GHZ never happened, if you want to piss of apple that's how you do it.
https://en.wikipedia.org/wiki/Power_Mac_G5

you don't play to apple music and they get mad

back when apple was swapping from PPC to intel they looked at AMD and went with intel in the end.
https://www.theverge.com/apple/2011/11/18/2572451/did-apple-have-an-amd-powered-macbook-air
(there is more info if you look for the AMD/intel thing)
https://blogs.wsj.com/digits/2011/08/10/intel-sets-300-million-fund-to-spur-ultrabooks/

look at what happened after that AMD CPUs died for almost 10 years and intel was stable with improvements each year.
bulldozer CPUs must have made apple vary happy with the option the went with

AMD has not got the laptop CPUs that apple wants never mind all the zen 2 bugs

so in 5 years time if AMD are still doing well maybe apple will look at them & if they can pull out some CPUs that can relay do what's needed in laptops in a way intel cant

and that's just ignoring the pull apple must have at intel, apple is able to influence intel to get what they want and so far intel has not done anything to bad

now the one thing intel has done is not hit it's road map but at the same time i'm sure they give apple a sweat deal on pricing

Ps i dont think apple cares about swapping laptops to ARM, they can sell you an ipad.
why sell a laptop when they can sell an ipad pro maxed out for almost $2K with only a 12' display and a 1TB SSD
there minting it on ipads compared to laptops

 

[ssgbryan]

At less than 10% marketshare, I doubt Apple has all that much in the way of "pull" at Intel.

Apple appears to buy up the leftovers.....

Remember, the CEO is a bean-counter, not a visionary.

There is an AMD-OSX community, so moving to better CPUs wouldn't be all that hard.

Intel has nothing for the next few years.

 

[deconstruct60]

At less than 10% marketshare, I doubt Apple has all that much in the way of "pull" at Intel.

Apple 'punches' above their <10% overall (worldwide) market classic PC market share weight.

Apple buys higher average selling CPUs than the rest of the folks. Apple buying
entry MBP 13" processors that street price at $320 (i5-8257U) , $320 (i5-8279U) , $395 (i7-9750H)

versus folks buying budget i3 and i5 processors in the sub $150 zone. Apple can buy half as many chips and still hand over the same amount of revenue. Systems like the iMac Pro and (if they can get away with it) the Mac Pro only goose the revenue numbers higher.

Apple is one of Intel's "better" customers if talking about profits. Apple probably could be better and have more sway if they didn't haggling away a decent chunk of those street prices and ask for stuff that most other vendors don't want to buy in bulk.


It is big enough that if Apple stomps off in a scorched Earth , big bang , quick exit that it would show up. I don't think Intel is going to get " Steve'd " ( Jobs might have done that with some of the recent stuff but Apple is in so deep they'd have to throw away a ton of money to do it out of rage. )

Apple appears to buy up the leftovers.....

The cellular modem folks? this is actually the second time Apple left that division in a lurch with a move to Qualcomm. Intel bought Infineon wireless div just about the same time Apple dumped them. ( no one chip for every region solution modem, behind curve on LTE , etc. so left for Qualcomm ). Very similar thing here. After hounding Intel (old Infineon ) to jump through upteen hoops ( pushing out economies of scale break even ) .... Apple bolts. Honestly, Apple should have spent the money to clean that up .

Remember, the CEO is a bean-counter, not a visionary.

Not really a bean counter. ( not finance person. ).

There is an AMD-OSX community, so moving to better CPUs wouldn't be all that hard.

Apple would be looking for vendor support, not community forum positings. Firmware and some other wrap around support areas AMD still needs to do some more work.

Intel has nothing for the next few years.

They have something. There is a decent chance it is enough to mostly tread water until they can turn the corner. AMD mostly aimed at the juicy DATA Center business so don't have as solid a laptop solution space. Intel is going to take lumps. If they don't come off their price points in several areas .... even bigger lumps.

If they Willow Cove / Tiger Lake plugs some of the quirks of Sunny Cove / Ice Lake and Intel can ship that before Q4 '20 in the laptop space they have a decent chance ot turning the corner before 2021.



In terms of the Mac Pro and Apple though it may not matter much. Apple sat through 2014-2016 either snoring or singing "Do worry be happy" about the situation the Mac Pro was in. They could fall right back into that mode for the 2-3 years it takes Intel to turn the corner.

 

[Draeconis]

Apple has a history of hissy fits over not getting the 'love' they want from their partners.
IBM do not make the PPC in to what apple wants and they swap to intel

One thing that always fascinated me; right when Apple was dumping PPC, Sony and Microsoft embrace it to build the Xbox 360 and PS3. In fact it's fairly infamous that early live demos of Xbox 360 software and games actually ran on Apple G5 towers hidden within the display stands. True, the use case wasn't the same, but an interesting historical note nonetheless.

 

[orph]

there's a book about it
'Race for a New Game Machine, The: Creating the Chips Inside the Xbox 360 and the Playstation 3'
https://www.amazon.co.uk/Race-New-Game-Machine-Playstation/dp/0806531010
it's an intresting time

 

[deconstruct60]

One thing that always fascinated me; right when Apple was dumping PPC, Sony and Microsoft embrace it to build the Xbox 360 and PS3.

One key difference was that Sony and MS paid for the R&D for what they wanted built upfront. That tends to make a substantive contribution to extremely expensive stuff getting done. They each wanted something different from the track that IBM had a primary need for ( server and 'maybe' workstation CPUs). Apple also did ( wanted to primarily focus on laptops ), but no big money on the table up front for what would be an even bigger 'fork' on the power consumption front ( Xbox 360 and PS3 at least plug into a wall).

Non IBM PPC was having wins as embedded processor in networking , some autos ( built in mobile generator,... this before electric vehicles a major focus ) , and other "plugged in". Freescale wan't going to upfront Apple's narrow (relative to general market) device either.

During most of the time on PPC Apple didn't directly primarily pay for their own PPC. They spent time building proprietary chipsets ( which didn't grow the PPC ecosystem much either as the CPU vendors had to do that to to a smaller group of buyers. )

Mac all by themselves were particularly big enough to jumpstart forking 2-3 CPU designs each with 2-3 concurrent design pipelines just for the one specific product line. Even less true at the volumes back then, but still pretty much in the same bucket now with the more competitive market out there in this space. [ Again a substantive difference with XboX 360 and PS3 is only need on CPU and only need a new one every 2-3 years. Or just pure "hand me down" usage to extend time (e.g., the iPhone market using last years and year before last as the 'affordable' model. ]

Jumping to Intel was at least an order of magnitude cheaper move. More long term shared R&D costs ( again cheaper ) and at least as much performance. Intel was shifting to focusing on laptops because the general classic PC market was shifting to laptops ( just like Apple's customers were ). So aligned with future direction and cheaper. So Apple bolted. ( And yes it was Steve Jobs looking for cheaper driving that. That Apple went Scrooge McDuck on contractors/component suppliers only after Steve's passing is always a giggle. )

If Apple is looking to mainly prune the bottom 'half' of laptops off for themselves and primarily just need an x86 solution in desktop ( and perhaps 1-2 'large laptops' ) then AMD is pretty well aligned with that path over the next 2-3 years. AMD doesn't have lowered performance laptops as very high priority. And have jumped out on Desktop, Workstation, and "Server" space.


The whole "shared costs on much larger ecosystem" also caught up to the MS and Sony game consoles at this point too. It wouldn't have necessarily paid off for Apple to push the PPC where only they were going. ( the R&D ramp on iPhone/iPad probably looked like a better bet. )

 

[orph]

@deconstruct60 hay thats a relay nice post

Apple got a lot from intel and it's worked out, low power use & easy for devs to port apps
(if you where around in the late 90's and early 2000's you will know how few apps there where on mac back then compared to now, it relay is a new world)

from https://appleinsider.com/articles/1...g-unit-sales-reporting-of-mac-iphone-and-ipad

2005 was the PPC/intel switch over (well the start), just look at how laptops started to outsell desktops over time,

as you mentioned AMD has just not been looking at laptops and intel has a death grip on them i joke intel is still 90% of the market and something like 97% of laptops.

anyway intel has done well for apple, if apple was to start using AMD intel will not be happy which is not something apple will want to deal with.

relay show's apple is a phone company that makes laptops on the side,

 

[deconstruct60]

...

from https://appleinsider.com/articles/1...g-unit-sales-reporting-of-mac-iphone-and-ipad

That is percent revenue not really unit volume (indirectly yet but) . Unit volume makes a huge factor in whether it makes sense for Apple do either do their own so a specific product or pay someone upfront to do all the work ..... or just stick with "shared cost" infrastructure. iPhone is down revenue percentage, but services and iPad are up ( in volume and revenue). iPad just got its own 'OS' also ( so will probably continue tracking up if Apple does a good job with that).



Mac unit sales have gone up over the time period covered by the graph. iMacs also did very well over that time period. The issue that lots in the this Mac Pro forum like to put their head in the sand over is that the Mac Pro impact decline started before 2012-2013.

https://www.macworld.com/article/1141829/appleq309.html

2006-2008 desktop basically just treaded water. As the iMac moved off of mobile CPU processors and into desktop ones things got better ( even as price point of Mac Pro went up and wasn't a big of a volume contributor. )
When the better performance per watt caught up to the desktop options for the iMac things also got better in a subset of desktop for Apple also.

2005 was the PPC/intel switch over (well the start), just look at how laptops started to outsell desktops over time,

That is a "brought in more money" graph. Not an outsell in numbers one.

as you mentioned AMD has just not been looking at laptops and intel has a death grip on them i joke intel is still 90% of the market and something like 97% of laptops.

AMD is looking at laptops. They are trying with mobile discrete GPUs ( just not particularly cheap ones). They are rolling out the Zen and GPU improvements to APU ( just at the end of the roll out cycle for a generation). Intel was far more vulnerable in the Datacenter space largely because Intel was far more greedy there. New EPYC's surcharge charge for > 1TB of RAM $0.00. Intel's Gen 2 SP processors tax for > 1TB $3K. (and a >2TB tax of $7k). It is really not hard to whip someone like that if have a product that is at about at parity in performance . Intel price umbrella is so high that AMD can price under and still make a more than reasonable profit ( instead of having to 'eat' razor thin margins.)

It is not going to be hard to whip Apple too if they are passing along that > 1TB tax at roughly those magnitudes.

If things keep going smoothly for AMD eventually they'll attack the laptops more vigorously too. It is a tougher market to crack ( Intel's margins are not 'crazy high' so a lower umbrella to hide under) and the system integration work is much higher so the development costs are bigger. (and AMD only has limited budget to invest. And also batting Nvidia on another 'front'. Part of AMD's problems over last decade or so was with trying to go toe-to-toe on everything with Intel and Nvidia at the same time. That was just beyond goofy strategy of basically a non-strategy.. just follow the other folks on thinner margins. ).

anyway intel has done well for apple, if apple was to start using AMD intel will not be happy which is not something apple will want to deal with.

Intel should largely be not happy with Intel. To a substantive degree, this is a shoot themselves in the foot problem. Intel wandered for over a year cooking up ways of charging even higher prices as opposed to executing and delivering better. They threw the "change 1-2 major things at a time" 'tick-tock strategy out the window and tried doing 5-6 major things in one big jump ( because they had super powers and could leap over complexity like a tall building in a single bound. ..... not ). Basically more than a little "fat ,dumb, and/or lazy"

The sales and revenue 'chop' for Intel is coming. Not just from Apple. What Intel can only do now is manage just how deep it goes. Folks there need to stop looking at their stock (or stock option ) prices and start lots more doing.

relay show's apple is a phone company that makes laptops on the side,

Not really. That too is an apparent Apple exec problem of gazing too much at the stock price and not enough on doing something. the Mac business doesn't need to be "just as big as other division X" goal. They do need to be engaged and doing a very good job. Comatose products for years at a time isn't that. 4 years to sort out a keyboard problem isn't that. They didn't have to starve the Mac product line investment to build a deeper Scrooge McDuck money pit to splash around in.

The problem isn't the iphone business. It is trying to slavishly match the margins of the phone business. (which even the iPhone business is going to have problems with over the next couple of years. The whole escalate the price every year dynamic is running out of steam. )

 

[ssgbryan]

Intel is getting smoked on all levels. The only use case for Intel is single threaded gaming performance - and with games moving to 8 cores/16 threads (see upcoming consoles), it is irrelevant.

 

[ekwipt]

Intel is getting smoked on all levels. The only use case for Intel is single threaded gaming performance - and with games moving to 8 cores/16 threads (see upcoming consoles), it is irrelevant.

After yesterdays announcement of Epyc, all of the major players are buying AMD for their data centres (Google, Facebook, U.S. Military, Amazon). 7nm, offer up to 64 cores (x2), offer 128 lanes of PCIe 4.0, offer 8 memory channels (2TB of RAM).

Threadripper 3 will most likely be 32 or 64 cores, trounces everything Intel has got for less money (40% less at least).

I can't see Apple staying with Intel after the next Mac Pro. I truly believe they have to go with AMD

2020 redesigned Mac mini with AMD, redesigned iMac with Zen and two different Mac Pros Zen and Threadripper

 

[ssgbryan]

That would be the smart thing to do, but I don't see Tim doing it.

 

[high heaven]

X570 is connected via PCIe 4.0 x4 to a Zen 2 CPU. TB3 is PCIe 3.0 x4.

You can also put a TB3 card on a 3.0 x8 slot connected directly to the CPU.

I really don't think so. Check youtube and you will find a video that someone trying to attach TB3 PCIe slot but it didn't work.
[doublepost=1565388380][/doublepost]

"happens to work" isn't really supported. It works and the vendor backs up that it works is what wins real design bake offs that Apple highly like runs.

"It should work but we really haven't throughly checked it out" is one reason why AMD has been on the sidelines for more than a few system vendors in workstation space.

Threadripper supports ECC RAM.

 

[cube]

I really don't think so. Check youtube and you will find a video that someone trying to attach TB3 PCIe slot but it didn't work.

As I said, the plug and play situation seems to have changed since you had to hack to make it work.

 

[high heaven]

As I said, the plug and play situation seems to have changed since you had to hack to make it work.

And that's not allowed and not official way to use TB3. Am I suppose to risk the stability to hack it? Im talking about official attachment from Apple, not an individual.

 

[cube]

And that's not allowed and not official way to use TB3. Am I suppose to risk the stability to hack it? Im talking about official attachment from Apple, not an individual.

That is before the reports where it just worked. I'm not talking about Macs.

You don't have to pay Intel anymore to use Thunderbolt.

 

[high heaven]

That is before the reports where it just worked. I'm not talking about Macs.

You don't have to pay Intel anymore to use Thunderbolt.

Where is your proof? AMD CPU is still not able to use it officially. And who makes motherboards with TB3 except for Asrock?

Royalty free doesn't mean you can use it right away.

 

[cube]

Where is your proof? AMD CPU is still not able to use it officially. And who makes motherboards with TB3 except for Asrock?

Royalty free doesn't mean you can use it right away.

ASRock is the first to sell TB AMD motherboards, like they were the first to sell 2.5GbE AM4 boards and a TR4 micro-ATX board. They are also the first to sell an AM4 motherboard that takes Intel coolers.

The other brands likely did not bother as instead of adding a TB chip to the board, you can just plug a card.

AMD can use it officially, it just needs to be supported by the board or card makers.

 

[high heaven]

ASRock is the first to sell TB AMD motherboards, like they were the first to sell 2.5GbE AM4 boards and a TR4 micro-ATX board. They are also the first to sell an AM4 motherboard that takes Intel coolers.

The other brands likely did not bother as instead of adding a TB chip to the board, you can just plug a card.

AMD can use it officially, it just needs to be supported by the board or card makers.

I must warn you that you still not able to prove anything.

1. Asrock is well known for making experimental motherboards. Having the first TB3 motherboard for AMD doesn't prove anything at all. "AM4 motherboard that takes Intel coolers." They'd done something similar for a long time. Nothing special. Check ASRock K8 COMBO-Z and K7Upgrade-880 and 754bridge.

2. If Asrock can add TB3, how come others didn't add TB3? Almost all motherboard manufacturers are from Taiwan. Since others didn't even add TB3, it doesn't convince me at all.

3. "You can just plug a card" Like I said, where is your proof? Seriously, I don't like comments without sources or proofs which doesn't support your claim.

4. AMD didn't mention anything about supporting TB3 officially. Supporting TB3 for AMD is a big news and yet, I heard nothing about it. Do you even know if TB3 PCIe card works on AM4 motherboards? If you do, sources?

Without related links or sources, there is no way to trust your statement.

 

[cube]

I must warn you that you still not able to prove anything.

1. Asrock is well known for making experimental motherboards. Having the first TB3 motherboard for AMD doesn't prove anything at all. "AM4 motherboard that takes Intel coolers." They'd done something similar for a long time. Nothing special. Check ASRock K8 COMBO-Z and K7Upgrade-880 and 754bridge.

2. If Asrock can add TB3, how come others didn't add TB3? Almost all motherboard manufacturers are from Taiwan. Since others didn't even add TB3, it doesn't convince me at all.

3. "You can just plug a card" Like I said, where is your proof? Seriously, I don't like comments without sources or proofs which doesn't support your claim.

4. AMD didn't mention anything about supporting TB3 officially. Supporting TB3 for AMD is a big news and yet, I heard nothing about it. Do you even know if TB3 PCIe card works on AM4 motherboards? If you do, sources?

Without related links or sources, there is no way to trust your statement.

Go to the ASRock site and see what they advertise (there are limitations).

The TB header in the motherboards is to be connected to their TB card. Only one of their boards already has a TB chip (it shows no caveats).

Gigabyte has a similar card but no AMD motherboards with a header yet.

Without the undocumented header some functions are missing.

So ASRock is the only one with official support for AMD.

TB is not a pressing need on a desktop, that's why others are not desperate about it.

AMD likely does not want to talk about it (no AMD laptops with TB, for example).

 

[deconstruct60]

I really don't think so. Check youtube and you will find a video that someone trying to attach TB3 PCIe slot but it didn't work.

Solid support for Thunderbolt needs to be done at the firmware, boot, and board level. Intel didn't 'ban' AMD CPU solutions. AMD largely didn't want to put in the basic work on their side. The board markers didn't want to spend much extra money carrying AMD's 'water' on the lack of support. AMD was spending money trying to 'kill' Thunderbolt. That the primary reason it isn't present on most AMD boards.

As it got apparent that Thunderbolt was going to merge intro USB. Step 1 USB Type C port . step 2. Intel working to do a 'hand off" to USB-IF, AMD started to get clue their 'kill it' position was untenable. That was highly likely too late to improve their balance in any Ryzen 3 bake off for Apple design. [ spending money trying ti 'kill' thunderbolt and then going to Apple saying buy my stuff isn't going to improve chances. ]


there was a way to do Thudnerbolt with other CPUs if simply just bought Intel TB controllers and implemented to the specs to pass the Thunderbolt certification specs. Much of the gap there was some parties not wanting to hand over money to Intel for either of those steps. That isn't Intel "stopping" the process, that is more so folk not wanting Intel in solely the driver seat. If have a deficient Thunderbol implementation Intel can just cut you off from TB controllers and deviant product is dead. Likewise, money paid for TB controller just makes Intel product better. AMD had a major hang up with that. But that hang up was primarily going to keep them out of Apple Mac products because those fully boungt into Thunderbolt.

[doublepost=1565388380][/doublepost]

Threadripper supports ECC RAM.

[/quote]

The original quote talked about "Desktop AMD" processors supporting ECC. While it is a checkbox feature for Ryzen , most of the board vendors don't implement it. If not implementing it then bugs tend not to get fixed either by AMD or board vendors. ( The is a sparse set of Core i3 that do ECC , in addition to the Xeon E3 / E variants of the mainstream die. Again the board support isn't there in most mainstream boards not aimed at Xeon E3 / E ).

Threadripper was always aimed at Xeon E5 1600 / W space so all the boards normally do flush out the ECC support.

The Threadripper have been behind the curve on Turbo ( and single thread) performance . It is also big socket oriented ( which would have pushed iMac Pro into more corners). Threadripper 3 should narrow the gap to point perhaps not a significant difference but it also is no where to be seen. Mac Pro was grossly late and hooking to an even later CPU would have been bad idea in 2017- early 2018. To 'win' AMD also has to be on time. Right now their ordering on the roll out for Zen generation isn't lined up with much of the Mac product line in desktop pro space in 2018-2019.

Release window sync wise a 2020 iMac Pro might line with Threadripper if AMD had very good samples to hand out in 2013 for Apple to build again. If not then probably they missed that window too. Really have to look at where Intel and AMD were relative to each other two years ago to see where Apple would be moving. Some sexy AMD benchmarks released to public now may have impact on where Apple is for 1-2 years out in future products.

The next Mac Pro should have launched back in the first half of 2019 ( or even better at very end of 2018). AMD's timelines didn't match up well at all.

Apple probably not going to do major board/socket upgrades on less than a two year cycle. For for Mac Pro that is pretty big miss.

 

[Pressure]

... offer 128 lanes of PCIe 4.0 ...

2P will offer 160 lanes of PCIe 4.0 in total and support 4TB of RAM per socket.

 

[deconstruct60]

Go to the ASRock site and see what they advertise (there are limitations).

The TB header in the motherboards is to be connected to their TB card. Only one of their boards already has a TB chip (it shows no caveats).

Gigabyte has a similar card but no AMD motherboards with a header yet.

Without the undocumented header some functions are missing.

I'm suspicious that this may turn into one of the spos where USB4 forks from Thunderbolt certification and standards. That it will deems the quirky subset that is missing as a "pass". If they do , then I don't think Apple is going to the same thing for passing muster on Macs.

Intel added some stuff to the last round of Thunderbolt controllers where it (and associated firmeware) can put the controller into a sanity check mode for the TB network that is present at boot. ( so don't need GPIO for hot plug reconfiguration or dynamic adaptation. ). This has let more 'hacks' work, but those hacks won't pass TB certification as a documented configuration.

THe firmware and GPIO 'work' need to do here isn't some Intel 'lock out' of Thunderbolt but it is work that many razor thin margin vendors don't want to do. And AMD often chasing the "cheapest solution" has been all to willing to go along with that and skip the doing the additional ground work also. AMD is trying to push into more up market so they are slowly turning the corner.

So ASRock is the only one with official support for AMD.

TB is not a pressing need on a desktop, that's why others are not desperate about it.

Not pressing in the general classic desktop market. In the Mac desktiop market that isn't really pragmatic true. There are lots of folks on this board who don't want it to be true, but it is not their product (not the owners, nor directly pay for the R&D work).

AMD likely does not want to talk about it (no AMD laptops with TB, for example).

Where AMD mostly is on the laptop space ( cheapest options ) there probably won't be much pressure even for the next 1-2 years. Intel weaving TB controllers into the CPU package puts them even further behind at the higher end.
As long as AMD is making deeper in roads on high margin server and desktops they aren't going to worry about it too much for next 2 years or so.

If Apple jumped into to doing their own USB4/Thunderbolt discrete chips then could get AMD out of this ditch (if worked with Apple on integrtating them). For Zen AMD has most just licensed ASMedia I/O chip implementation. I think they are going to be restricted in the laptop space for an extended amount of time unless get very substantive outside help. [ They don't have budgets to do everything for everybody so have to pick and choose where to battle. Laptops is more a defensive position where probably won't gain much ground with tons of outside help. ]

 

[deconstruct60]

2P will offer 160 lanes of PCIe 4.0 in total and support 4TB of RAM per socket.

The 4TB of RAM isn't a 1P vs 2P socket thing.

If want minimized NUMA impact still don't gain any PCI-e lanes in 2P. What AMD added is an option that you can make the NUMA worse between the socket ( under provision the socket to socket Infinity Fabric links ) if want to add more PCI-e links for some corner case reason. Maybe some systems highly skewed toward to specifically CPU package pinned virtualized loads may want to do that, but general case that is extremely probably a bad trade off in single user workstation context. The two relatively very large packages are blowing away room that any additional slots probably would sit in. ( and in that general market devotion to space for spinning disc platters will do even more. ). If running simulations/computations that are large enough to span both processors packages then chopping down the inter-package line bandwidth isn't going to help that workload at all.


128 is plenty for a workstation. The new Mac Pro only has 64 on the CPU. So could flush out 3 more of the x8 slots the Mac Pro has to x16 and still not be in some dire need of more.

More than 128 is really in the extreme 'corner case' zone. Most systems won't come that way. It most cases it will only bring Intel back into the game when actually had a gap.

 

[cube]

Solid support for Thunderbolt needs to be done at the firmware, boot, and board level. Intel didn't 'ban' AMD CPU solutions. AMD largely didn't want to put in the basic work on their side. The board markers didn't want to spend much extra money carrying AMD's 'water' on the lack of support. AMD was spending money trying to 'kill' Thunderbolt. That the primary reason it isn't present on most AMD boards.

As it got apparent that Thunderbolt was going to merge intro USB. Step 1 USB Type C port . step 2. Intel working to do a 'hand off" to USB-IF, AMD started to get clue their 'kill it' position was untenable. That was highly likely too late to improve their balance in any Ryzen 3 bake off for Apple design. [ spending money trying ti 'kill' thunderbolt and then going to Apple saying buy my stuff isn't going to improve chances. ]


there was a way to do Thudnerbolt with other CPUs if simply just bought Intel TB controllers and implemented to the specs to pass the Thunderbolt certification specs. Much of the gap there was some parties not wanting to hand over money to Intel for either of those steps. That isn't Intel "stopping" the process, that is more so folk not wanting Intel in solely the driver seat. If have a deficient Thunderbol implementation Intel can just cut you off from TB controllers and deviant product is dead. Likewise, money paid for TB controller just makes Intel product better. AMD had a major hang up with that. But that hang up was primarily going to keep them out of Apple Mac products because those fully boungt into Thunderbolt.


The original quote talked about "Desktop AMD" processors supporting ECC. While it is a checkbox feature for Ryzen , most of the board vendors don't implement it. If not implementing it then bugs tend not to get fixed either by AMD or board vendors. ( The is a sparse set of Core i3 that do ECC , in addition to the Xeon E3 / E variants of the mainstream die. Again the board support isn't there in most mainstream boards not aimed at Xeon E3 / E ).

Threadripper was always aimed at Xeon E5 1600 / W space so all the boards normally do flush out the ECC support.

The Threadripper have been behind the curve on Turbo ( and single thread) performance . It is also big socket oriented ( which would have pushed iMac Pro into more corners). Threadripper 3 should narrow the gap to point perhaps not a significant difference but it also is no where to be seen. Mac Pro was grossly late and hooking to an even later CPU would have been bad idea in 2017- early 2018. To 'win' AMD also has to be on time. Right now their ordering on the roll out for Zen generation isn't lined up with much of the Mac product line in desktop pro space in 2018-2019.

Release window sync wise a 2020 iMac Pro might line with Threadripper if AMD had very good samples to hand out in 2013 for Apple to build again. If not then probably they missed that window too. Really have to look at where Intel and AMD were relative to each other two years ago to see where Apple would be moving. Some sexy AMD benchmarks released to public now may have impact on where Apple is for 1-2 years out in future products.

The next Mac Pro should have launched back in the first half of 2019 ( or even better at very end of 2018). AMD's timelines didn't match up well at all.

Apple probably not going to do major board/socket upgrades on less than a two year cycle. For for Mac Pro that is pretty big miss.

How did AMD try to kill TB?