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

[Kpjoslee]

Too bad. It is leaked information and it always true. First one was 16 core and 2nd one was 32 core.

https://www.funkykit.com/news/amd-ryzen-9-3800-and-threadripper-3990wx-black-edition-spec-leaked/
https://www.reddit.com/r/Amd/comments/a3b0sp/amd_zen_2_ryzen_threadripper_processor_rumors_wow/

Also, Threadripper will use EPYC ROME core configuration which is already shown before. Total 8 chips of 7nm 8 cores will make 64 cores. Truely possible.

Those leaked information were already dismissed long time ago. Lot of applications that demands more than 32 cores also demands high memory throughput. Reason why 2990wx occupies the niche that is right in between 16 core threadripper and 32 core EPYC configuration.

Again. Why do you fail tounderstand that you post Q1 2018 market share, when AMD was outselling Intel 2:1 from September 2018(I specifically said that AMD was outselling Intel 2:1 for past 4 months)? It means: Q4 2018.

Why do you lack basic logic, and understanding of what has been said, and what is posted?

This is data from ONE European retailer. European Amazon, sites like OC.uk were experiencing similar situations, because of Price hike Intel CPUs experienced EVERYWHERE, apart from United States.

For past 6 months the best selling CPU on Amazon.com is... Ryzen 5 2600. The only place where Intel is getting significant market share in sales is American market, because it has not been affected by price hikes. All of this has been discussed by guys from Motley Fool and Seeking Alpha, and widely analysed.

So. Wasn't AMD outselling Intel 2:1 for past 4 months?

P.S. Yes. AMD is very well trusted supplier. Why? Because it is not affected by manufacturing problems, that gobble your fabs, so that you cannot push enough CPUs to feed the demand, so they end up costing 3 times as much as MSRP.

Intel's YoY growth on client sales was larger than AMD, on top of Intel's revenue already being 10 times larger than AMD just on that particular segment. Retail CPU market makes up really small percentage of overall client PC market, and using one particular vendor's numbers which makes up tiny percentage of that retail market isn't really good data to support your argument.

 

[AidenShaw]

Why do you lack basic logic, and understanding of what has been said, and what is posted?

Thank you for so clearly supporting my rebuttal.

I don't think that HP/HPE/Dell/Lenovo/SuperMicro and all the other x64 vendors are buying CPUs from Amazon EU, so I don't think that citing a few European DIY stores can in any way support a statement that AMD is 2:1 to Intel - unless (which you don't) clearly say that you're looking at a small slice of the DIY market.

The DIY market doesn't look for "trusted supplier". It wants "bang for the Euro" right now, and doesn't care if the supplier goes bankrupt tomorrow.

Goodbye for now - I don't want to get involved in pointless bickering. Come back when you have something new to offer.

 

[koyoot]

Goodbye for now - I don't want to get involved in pointless bickering. Come back when you have something new to offer.

And what do YOU have to offer here on this forum, if you are not interested in buying Apple Mac Pro?

 

[AidenShaw]

And what do YOU have to offer here on this forum, if you are not interested in buying Apple Mac Pro?

 

[mavericks7913]

"Kpjoslee, post: 27069243, member: 125317"]Those leaked information were already dismissed long time ago. Lot of applications that demands more than 32 cores also demands high memory throughput. Reason why 2990wx occupies the niche that is right in between 16 core threadripper and 32 core EPYC configuration."

It's not dismissed and It's not even released yet lol. Also, not all people demands require high memory throughput. Oh yeah how much is it to buy 32 cores threadripper?

 

[Kpjoslee]

"Kpjoslee, post: 27069243, member: 125317"]Those leaked information were already dismissed long time ago. Lot of applications that demands more than 32 cores also demands high memory throughput. Reason why 2990wx occupies the niche that is right in between 16 core threadripper and 32 core EPYC configuration."

It's not dismissed and It's not even released yet lol. Also, not all people demands require high memory throughput. Oh yeah how much is it to buy 32 cores threadripper?

You are free to believe if there is going to be 5.0Ghz base clock 64-core threadripper when current 12nm threadripper sits on 3.0 base with 4.2 turbo with 3.4 all core turbo non-overclocked on 32 cores. Those estimates shoots way beyond even the best case scenario improvements expected from TSMC 7nm vs current 12nm TSMC/GF.

For tasks that doesn't require high memory throughput, practicality of using beyond 32 cores is questionable, since lot of those tasks doesn't really scale well up to 32 cores. Also, even if EPYC is moving up to 64 cores, it doesn't guarantee that TR will move up to 64 because they will still be using existing platform with quad channel.

 

[mavericks7913]

You are free to believe if there is going to be 5.0Ghz base clock 64-core threadripper when current 12nm threadripper sits on 3.0 base with 4.2 turbo with 3.4 all core turbo non-overclocked on 32 cores. Those estimates shoots way beyond even the best case scenario improvements expected from TSMC 7nm vs current 12nm TSMC/GF.

For tasks that doesn't require high memory throughput, practicality of using beyond 32 cores is questionable, since lot of those tasks doesn't really scale well up to 32 cores. Also, even if EPYC is moving up to 64 cores, it doesn't guarantee that TR will move up to 64 because they will still be using existing platform with quad channel.

What you dont know is that EPYC and Threadripper share same hardware parts. They just need to disable or kill cores to make 32 cores or activate all cores to use all of them. This is how they use Ryzen, EPYC, and Threadripper. EPYC has more specialized features for a workstation that Threadripper does not have. Also, EPYC's clock speed is lower in order to keep the stability.

Since EPYC Rome with 64 cores is revealed, it is possible to see Threadripper with 64 cores. If EPYC can, then Threadripper can. I know that Threadripper has only quad channel but it's only less than $2000 so why not? We dont know if Threadripper will have higher clock speed but 64 cores Threadripper is not 100% impossible.

 

[wallysb01]

Compared to more than two decades for Intel?

That's what "trusted supplier" is based on, not a flash in the pan.

This conversation is killing me. AMD has closed the gap, but hasn't jumped out to particularly meaningful lead. Why would a computer maker, specifically Apple, who cares so much about selling an entire package, abandon a stable and improving (albeit slower than we'd all like) platform?

Again. Why do you fail tounderstand that you post Q1 2018 market share, when AMD was outselling Intel 2:1 from September 2018(I specifically said that AMD was outselling Intel 2:1 for past 4 months)? It means: Q4 2018.

Ugh, those numbers were just released. 2:1? LOL. Sure maybe for the DIY enthusiasts buying CPUs directly. That wouldn't surprise me. Those types are tinkerers. They like the new shiny object even if there is relatively little reason to use it. When it comes to computer venders, is still like 7:1 intel....


Look, Most of us want the CPU market to become more competitive and faster evolving again. But you're kind off of the rails here Koyoot. Intel has stumbled, no doubt, but they are far from knocked out. They still hold advantages in the platform, single core performance, AVX-capable workloads, and simply more developers are tuning their code to be optimized in Intel platforms. The legacy matters as well. The next 2 years are going to matter a lot for both companies. The next Ryzen may not live up to expectations, as was extremely common for AMD for about a decade. Intel may get back on track after learning a few things from the lumps they took on their 10nm process. The next generation from Intel will be them just kind of getting back on their feet. Its the one after that which will really give us a sense how they are going to do moving forward.

 

[mavericks7913]

This conversation is killing me. AMD has closed the gap, but hasn't jumped out to particularly meaningful lead. Why would a computer maker, specifically Apple, who cares so much about selling an entire package, abandon a stable and improving (albeit slower than we'd all like) platform?



Ugh, those numbers were just released. 2:1? LOL. Sure maybe for the DIY enthusiasts buying CPUs directly. That would surprise me. Those types are tinkerers. They like the new shiny object even if there is relatively little reason to use it. When it comes to computer venders, is still like 7:1 intel....


Look, Most of us want the CPU market to become more competitive and faster evolving again. But you're kind of off the rails here Koyoot. Intel has stumbled, no doubt, but they are far from knocked out. They still hold advantages in the platform, single core performance, AVX-capable workloads, and simply more developers are tuning their code to be optimized in Intel platforms. The legacy matters as well. The next 2 years are going to matter a lot for both companies. The next Ryzen may not live up to expectations, as was extremely common for AMD for about a decade. Intel may get back on track after learning a few things from the lumps they took on their 10nm process. The next generation from Intel will be them just kind of getting back on their feet. Its the one after that which will really give us a sense how they are going to do moving forward.

Well, until Intel solve CPU gate, I wouldn't trust them. Also, it is a fact that Intel is falling because of AMD Ryzen. AMD improved a lot of things within 3 years while Intel didnt since Sandy Bridge by keep using 4 cores for a long time.

 

[Kpjoslee]

What you dont know is that EPYC and Threadripper share same hardware parts. They just need to disable or kill cores to make 32 cores or activate all cores to use all of them. This is how they use Ryzen, EPYC, and Threadripper. EPYC has more specialized features for a workstation that Threadripper does not have. Also, EPYC's clock speed is lower in order to keep the stability.

Since EPYC Rome with 64 cores is revealed, it is possible to see Threadripper with 64 cores. If EPYC can, then Threadripper can. I know that Threadripper has only quad channel but it's only less than $2000 so why not? We dont know if Threadripper will have higher clock speed but 64 cores Threadripper is not 100% impossible.

It has nothing to do with them sharing same hardware parts. AMD has been using that strategy since they launched Ryzen and first EPYC and they are continuing their strategy and I am very aware of that.
32-cores on x399 platform is already hitting that limits of practical balance between total memory bandwidth vs total number of cores, and lot of rendering tasks and computations that requires usage of more than 32-cores demands more memory bandwidth. While physically it is possible to have 64-core on TR4 socket, I doubt it will happen until they release the new platform designed with 48-64 cores in mind with 6-channels of memory at least.

 

[wallysb01]

Well, until Intel solve CPU gate, I wouldn't trust them. Also, it is a fact that Intel is falling because of AMD Ryzen. AMD improved a lot of things within 3 years while Intel didnt since Sandy Bridge by keep using 4 cores for a long time.

Dude, come on, there has been plenty of progress since Sandy Bridge.

 

[ssgbryan]

Two words:

Trusted Supplier​

Companies value being able to depend on a supplier. AMD may being hitting a home run right now, while Intel has stumbled - but companies want nine innings a game, all season long.

AMD has been in the gutter for so long they are not trusted. A single good product doesn't make AMD trustworthy.

Only the DIY gamers will jump ship so quickly (and the AMD fans who cherry-pick DIY sales from overseas and claim that AMD outsells Intel 2:1). And the 'core queens' who cheerlead for 16/32/64 cores are just as bad - since most of them are probably running games and applications which struggle to scale to 6 or 8 cores. (And yes, again thank you ssgbryan for pointing out one of the problem spaces that scale easily.)

Nobody ever got fired for buying IBM, until they were. Intel hasn't been able to go the distance for a while, which is why our cMPs are still competitive a decade later.

If it was just one good product, I wouldn't be paying attention either. But it is a series of CPUs: Ryzen is trouncing Intel at the low end, Threadripper is doing the same at the mid-range, and Eypc is starting to make a dent at the data center. Early days, but I have a lot more faith in AMD right now than I do in Intel. AMD is delivering on their road map, whereas Intel seems to be falling into the place AMD was at a few years ago.

And that doesn't consider the wholesome, crunchy goodness that the AMD ProRender engine brings to the table. That may not matter to what you do, but those of us that use Blender, Maya, 3DS Max, Solidworks, and the Unreal Engine, it does. The 1st two even run native on OSX.

Btw, a sale is a sale is a sale believe it or not, whether it is overseas (and what constitutes overseas depends on where one lives) or local.

And just where are you getting your probably running games nonsense? We Core Kings know that most games don't use a lot of cores - we have known it for some time. It is why most of the gamers are sticking with Intel, at least for the moment. If the 3rd generation Ryzen series can turbo more cores than Intel, that will mean that another group will start looking for the exits. Doesn't matter if the buyer can use the extra cores or not - it isn't about today's requirements, but tomorrow's requirements.

 

[mavericks7913]

Dude, come on, there has been plenty of progress since Sandy Bridge.

lol, how much better? Not like AMD Ryzen within 3 years lol. Keep using 4 cores till Ryzen show up is the biggest example.

 

[wallysb01]

lol, how much better? Not like AMD Ryzen within 3 years lol. Keep using 4 cores till Ryzen show up is the biggest example.

Well, when the denominator is so low....

And this is the Mac Pro forum. Xeons haven't been stuck on any particular core count. The pricing has been jacked up, but again, we can thank AMD's failures for that. Maybe that will change now, but that doesn't really mean Apple should switch to AMD at the first sign of them pulling even to slightly ahead of intel.

 

[mavericks7913]

Well, when the denominator is so low....

And this is the Mac Pro forum. Xeons haven't been stuck on any particular core count. The pricing has been jacked up, but again, we can thank AMD's failures for that. Maybe that will change now, but that doesn't really mean Apple should switch to AMD at the first sign of them pulling even to slightly ahead of intel.

Well, they already planning to ditch Intel CPU. Also, AMD EPYC is cheaper in these days.

 

[Kpjoslee]

Well, they already planning to ditch Intel CPU. Also, AMD EPYC is cheaper in these days.

I doubt they will switch to AMD just to use them on their desktop line, and AMD doesn't really offer much on mobile products.

 

[mavericks7913]

I doubt they will switch to AMD just to use them on their desktop line, and AMD doesn't really offer much on mobile products.

lol they are trying to ditch mobile Intel CPU.
[doublepost=1549431555][/doublepost]What I want to point out is that Mac computer is not powerful and attractive but expansive. Only Apple software like Final Cut Pro or Logic Pro can take advantages from Mac and Apple has only 5~6 professional software.

This is why we talk about AMD and Intel but their stupid policy and design ruined everything like Mac Pro 2013.

It's quite late to change this unless they change everything for Mac computer. But what is Apple doing?
- Preventing Nvidia to support graphics cards
- No way to expand and upgrade
- Much more expansive
- Limited options.

Is it really the best way to be specialized only for Apple software?

Just my idea but it would be nice to make a Mac Pro motherboard like Apple 1 so that we can upgrade whatever we want.

 

[Kpjoslee]

lol they are trying to ditch mobile Intel CPU.
[doublepost=1549431555][/doublepost]What I want to point out is that Mac computer is not powerful and attractive but expansive. Only Apple software like Final Cut Pro or Logic Pro can take advantages from Mac and Apple has only 5~6 professional software.

This is why we talk about AMD and Intel but their stupid policy and design ruined everything like Mac Pro 2013.

It's quite late to change this unless they change everything for Mac computer. But what is Apple doing?
- Preventing Nvidia to support graphics cards
- No way to expand and upgrade
- Much more expansive
- Limited options.

Is it really the best way to be specialized only for Apple software?

Just my idea but it would be nice to make a Mac Pro motherboard like Apple 1 so that we can upgrade whatever we want.

Well, back in the day when Apple's main bread and butter was Mac line of products, and they tried to do everything to keep feature parity with latest PCs. They even made a deal to get latest CPU and GPUs before other PC manufacturers back then. But now they no longer rely on Mac sales but overwhelming majority of revenue is based on iPhones and iPads and their App store echosystem, and they are back on their old philosophy of keeping everything proprietary within their walled garden.

 

[deconstruct60]

lol they are trying to ditch mobile Intel CPU.
...

are they ?

https://arstechnica.com/gadgets/2018/11/apple-walks-ars-through-the-ipad-pros-a12x-system-on-a-chip/

Once get out of wading through the single thread drag racing benchmarks, the A12x would have to improve by 38% to even get to parity with the MBP 15" 2018. Funny how when AMD x86 solutions were 20-30% behind Intel solution they were a 'dog' and Apple shouldn't touch them with 10 foot pool.... but Apple 30+% behind ... well just around the corner and they're great.


No Thunderbolt (and PCI-e lane provisioning) . No USB 3.1 gen2 , one internal drive , drives 3+ screens ... nope. And that is all with Intel behind on process tech. When Intel moves forward and next gen "big" A series just does small increment 7nm 7++nm , they likely aren't going to make up huge ground just on process gains.

Apple could goose these short duration geekbench score by throwing a couple more low power, 'small cores' in there that can light up for the duration of this kind of benchmark .... but that is mostly just a cheat on substantive , longer duration workloads.

[ The iMac Pro is about 37,129 on other charts in that article. That is 106% improvement just to catch up. Just to put some Mac pro context on this. ]


Dropping Intel in the one port wonder MacBook, sure. Same number of ports as iPad Pro. No Thunderbolt. Maybe drive two monitors. .... a match better match in direction. Apple could make it even thinner. ( whereas should really stop screwing with anorexia with the MBP line up. ). If Apple's objective is to crank out range laptops that consists of two pieces of glass wrapped in aluminum shell ( https://www.macrumors.com/2019/02/04/apple-exploring-new-glass-panel-keyboards/ ) then sure they probably need their ARM SoC. But not sure why they need to take macOS that way. iOS would work too if on that kind of quest. Lots of folks there already happy with typing on zero key travel, glass keyboards. Decoupling the Mac from the OCD objective would. That mania was a contributing facto to walking the Mac Pro into a design corner. Detaching folks primarily fixated on that from the Mac product line would probably result in an improvement.

 

[Biped]

Apple needs to announce something soon, even if just for the sake of this thread. Things are getting goofy and nasty.

 

[Aldaris]

And outright embarrassing.

 

[fuchsdh]

Apple needs to announce something soon, even if just for the sake of this thread. Things are getting goofy and nasty.

If they announce something it'll just be slightly more relevant ********ing, rather than this.

 

[deconstruct60]

Apple needs to announce something soon, even if just for the sake of this thread. Things are getting goofy and nasty.

These forums aren't going to be a primary driver of Apple marketing. There are goofy and nasty future Mac Pro threads all the way back to 2012. Apple announcing something might tap it down for about 2 , maybe 3 , months and it would spring back up again. Frankly, if all Apple does is a sneak peak then it will probably just get worse as whatever ultimate checklist feature Apple has dropped will be proclaimed the end of the world.

i'll second the "embarrassing for Apple" notion though. They have had all the time, even with a standing starting at April 2017 , for a competent team working around full time to get to the engineering verification phase at this point. By the end of this month it would at the 21-22 months point.

The Cisco document associated with this article shows Apple would have had info to do some decent planning.
https://www.anandtech.com/show/1393...ade-lake-cooper-lake-and-ice-lake-for-servers

On the second page of comment on the article
"... his is not an Intel cpu roadmap but a presentation fail. Check out page 23, it clearly says “Processor timeline and availability based on Intel disclosure to Cisco on April 2017. Subject to change.” April 2017!!! ..."

If Intel told Apple in April 2017 that Cooper Lake was coming in late 2019 and Apple picked the Intel W versions of those as a "jump back in" point then that would have been a goof. (Cisco is sliding into 2020 to Coope Lake). Jumping onto the Intel W variants of Cascade Lake would mean once again they'd be jumping back end of the socket cycle but risk of sliding into 2020 would have been too risky. AMD probably had a similarly accurate timeline around April 2017 if they wanted to jump in on that direction. To still be in the relatively grossly incomplete stage at this point is just plain non execution on task. If they picked Cooper Lake it will probably be yet another "dog ate my homework" talk in April where they announce sliding into 2020.

The competitive system builders with competent leadership will be selling new, updated workstations in 2019 and Apple will be selling "dog ate my homework", "just wait, it will be magical" excuses.

 

[Joe The Dragon]

These forums aren't going to be a primary driver of Apple marketing. There are goofy and nasty future Mac Pro threads all the way back to 2012. Apple announcing something might tap it down for about 2 , maybe 3 , months and it would spring back up again. Frankly, if all Apple does is a sneak peak then it will probably just get worse as whatever ultimate checklist feature Apple has dropped will be proclaimed the end of the world.

i'll second the "embarrassing for Apple" notion though. They have had all the time, even with a standing starting at April 2017 , for a competent team working around full time to get to the engineering verification phase at this point. By the end of this month it would at the 21-22 months point.

The Cisco document associated with this article shows Apple would have had info to do some decent planning.
https://www.anandtech.com/show/1393...ade-lake-cooper-lake-and-ice-lake-for-servers

On the second page of comment on the article
"... his is not an Intel cpu roadmap but a presentation fail. Check out page 23, it clearly says “Processor timeline and availability based on Intel disclosure to Cisco on April 2017. Subject to change.” April 2017!!! ..."

If Intel told Apple in April 2017 that Cooper Lake was coming in late 2019 and Apple picked the Intel W versions of those as a "jump back in" point then that would have been a goof. (Cisco is sliding into 2020 to Coope Lake). Jumping onto the Intel W variants of Cascade Lake would mean once again they'd be jumping back end of the socket cycle but risk of sliding into 2020 would have been too risky. AMD probably had a similarly accurate timeline around April 2017 if they wanted to jump in on that direction. To still be in the relatively grossly incomplete stage at this point is just plain non execution on task. If they picked Cooper Lake it will probably be yet another "dog ate my homework" talk in April where they announce sliding into 2020.

The competitive system builders with competent leadership will be selling new, updated workstations in 2019 and Apple will be selling "dog ate my homework", "just wait, it will be magical" excuses.

and why does apple need to be so locked into an socket or even CPU?.

Apple needs to just give an workstation MB with all of the apple stuff in a slot.

 

[AidenShaw]

and why does apple need to be so locked into an socket or even CPU?.

Apple needs to just give an workstation MB with all of the apple stuff in a slot.

Apple's best move would be to re-skin some Z-series systems, and stick a half-eaten apple on them. (The externally visible bits of a Z-series are just attached to the case (some literally snap-on/snap-off, others riveted or screwed).)

Which could have been introduced in May or June 2017....

But there are some egos that couldn't take such an obvious step.