AMD CPUs, SoC Rumors and Speculations Temp. thread 2

[con635]

Guys,Will we see a Carrizo or Excavator based 28nm Next-gen Athlon X4 succeeding the 860K ?

There is rumor of one, 'Bristol Ridge'
http://www.kitguru.net/components/cpu/anton-shilov/amd-preps-bristol-ridge-apus-carrizo-for-desktops/

 

[juanrga]

^Yea Fury X is a bit high in price for what it is but it is selling I guess, if you consider a "out of stock" selling as I have yet to see it in stock at Newegg. Hell Fury is out of stock. That or production is just low.

Leaks before launch mentioned low production due to yields issues

http://www.dvhardware.net/article62651.html

and this was confirmed recently

http://www.kitguru.net/components/graphic-cards/anton-shilov/amds-partners-cannot-get-enough-radeon-r9-fury-graphics-cards-report/

 

[jimmysmitty]

^Yea Fury X is a bit high in price for what it is but it is selling I guess, if you consider a "out of stock" selling as I have yet to see it in stock at Newegg. Hell Fury is out of stock. That or production is just low.

Leaks before launch mentioned low production due to yields issues

http://www.dvhardware.net/article62651.html

and this was confirmed recently

http://www.kitguru.net/components/graphic-cards/anton-shilov/amds-partners-cannot-get-enough-radeon-r9-fury-graphics-cards-report/

The biggest downside I see is there wont be a price war. If they can't produce enough then they cant lower the price so it and the GTX 980Ti will stay where it is.

On a side note, I have found a few for sale. They are on listing sites for $900 bucks. Looks like some of the few to actually get one are inflating the price to profit.

 

[gamerk316]

^^ Supply and Demand at work.

 

[jimmysmitty]

Anyone who pays more than the MSRP for a Fury/Fury X is an idiot in my book especially when for the same or slightly better performance they can get a GTX 980.

But there are diehard fanboys who will think AMD is making money from that when it is the reseller who is.

 

[bharathbigb]

Guys,Will we see a Carrizo or Excavator based 28nm Next-gen Athlon X4 succeeding the 860K ?

There is rumor of one, 'Bristol Ridge'
http://www.kitguru.net/components/cpu/anton-shilov/amd-preps-bristol-ridge-apus-carrizo-for-desktops/

Thanks.

 

[8350rocks]

Spoiler

The comparison is for HPC designs...and POWER9, KNH and VOLTA are all going to be aimed squarely at HPC performance. They are going to do nothing in consumer HEDT, and while POWER9 will make a huge splash in commercial servers, etc. too...I doubt the other stuff has near the impact of POWER9 because of the established clientele IBM has in server market shares. NVidia's Volta will have zero impact on anything AMD does, outside of HPC designs, and KNH will likely make a small splash in the x86 server market as a co-processor...but it will be a very small %, of an already small %, of the world's server markets (x86 server is relatively small comparatively).

And precisely this is the problem for AMD. Zen has been designed explicitly for servers/HPC. It is the main element in AMD strategy to recover market share in the server/HPC market, but customers are choosing CPUs from IBM and Intel instead. Therefore if Zen is being rejected in the same market for which has been designed I smell another Bulldozer.

The same about GCN2, it is explicitly designed for HPC/server and will be used in the HPC APU scheduled for 2017, but consumers are choosing hardware from Nvidia and Intel instead.

Note: The KNH CPU is not a co-processor.

1.) GCN2 is not aimed specifically at servers/HPC...it never was. It will be aimed a bit more at having legitimately powerful workstation class cards on the top end. However, to that end workstation != HPC/Server.

2.) HPCs are a large cluster of hardware assembled one time, then maintained. Even if Intel sells 1000 KNH chips to somebody for a HPC...that is worth less than 3-4 large cluster server sales to financial firms/educational institutions/government entities.

EDIT: The goal is margin + quantity. Not single design wins.

First, the cards used in workstations have the same compute technology (e.g. GCN 1.0) than the cards used in server/HPC clusters. In fact AMD advertises its W9100 card as "supercomputing-class", because the technology inside is the same

http://www.amd.com/en-us/products/graphics/workstation/firepro-3d/9100

Second, there is something named the halo effect. Having your best designs on top supercomputers helps to sales of the rest of products.

Third, the lack of top design wins is a good indicator that Zen, K12, and GCN aren't faster/efficient than competence. Therefore AMD will try to compete once again on price, but this round they have the problem of debts and cannot continue being the cheap company. I recall most of people disappointed by pricing of Fury X card, despite I warned during months it was not going to be a cheap card. Something similar will happen with Zen CPU. I am warning more than one year before!

Fourth, about the idea of selling tons of workstation GPUs in future... that will not happen. It is worth mentioning that Nvidia is the indisputable leader in this camp with a market share of 79.4% and that AMD itself in their corporate strategy and long-term plan don't expect share grow in the medium term.

Sure, they advertise the W9100 as such, primarily because it can be used in HPCs. That tells people that it is designed for compute solutions. Does not mean that they are targeting HPCs specifically though.

As for the "halo effect" sure, that meant something when HPCs made the news still back in the 1990s. However, in this day and age, nobody cares what is in a super computer anymore. They care about what CPU has the most raw performance, and what GPU has the best raw performance. To be completely honest, in the US, Perf/Watt is not even relevant mostly, it is only raw performance. In places where electricity is more expensive, Perf/Watt plays a bigger role in other places.

Design wins based on projections are not yet shipped, and are not guaranteed. So, making assumptions regarding future product strength/quality over design wins for silicon that is a year out is putting the cart before the horse. I would not hold my breath for design wins due more than a year from now to pan out without some changes.

While AMD are realistic about WS class GPUs, you should also realize that even marginal growth would make a signficant impact to margins. Which is to say...the same as servers impact on CPU sales. Even so much as 2% total market share growth could bring margins up an entire percentage point.

 

[salgado18]

I believe that AMD doesn't have any design wins because its past CPUs weren't good, so no one will close a deal with them on a new one. Like, if you spend the past 5 years making bad products, how can I thrust you to make the next product something better than the competition?

Maybe that's why they are advertising Zen and Zen+: Zen is meant to be the proof that AMD can design good chips, and Zen+ is the one to win designs.

 

[jimmysmitty]

I believe that AMD doesn't have any design wins because its past CPUs weren't good, so no one will close a deal with them on a new one. Like, if you spend the past 5 years making bad products, how can I thrust you to make the next product something better than the competition?

Maybe that's why they are advertising Zen and Zen+: Zen is meant to be the proof that AMD can design good chips, and Zen+ is the one to win designs.

Proof that AMD can design good chips is K8. Although a lot of the design behind that chip was from an Alpha design.

That said, you don't need a good product to get wins as the companies will look at the theoretical ideas behind it and see if they fit. Remember Intels KNH is not even close to being out yet and could be a flop, I doubt it though as it is still based on the same idea behind KNL, yet it already has wins in the hPC market.

It is possible that what this tells us is that Zen is not what we are hoping for but merely a design that will keep AMD slightly competitive.

 

[con635]

http://www.kitguru.net/components/graphic-cards/anton-shilov/amd-we-have-taped-out-our-first-finfet-products/

Zen? Reliable benchmark leaks by christmas? AM4 leaks soon?

 

[esrever]

If they have only taped out their first finfets now, it looks like artic island is going to be 2016 Q3 or Q4.

 

[Cazalan]

When the analyst asked if Intel's 10nm delay would affect them, Su seemed quite excited.

 

[jimmysmitty]

http://www.kitguru.net/components/graphic-cards/anton-shilov/amd-we-have-taped-out-our-first-finfet-products/

Zen? Reliable benchmark leaks by christmas? AM4 leaks soon?

When the analyst asked if Intel's 10nm delay would affect them, Su seemed quite excited.

I am pretty sure she wont be as excited when GloFlo has issues with 10nm and beyond. I think I would rather be in Intels shoes where they have complete control over the process node.

Look at the GPU market. We should have been on 20nm GPUs, in fact Hawaii XT was supposed to be a 20nm part. But because neither GloFLo nor TSMC could get it done and viable they had to stick with a 28nm node which caused a big delay and a part that ran hot and used a lot of power. Now they have to wait for 14nm instead next year.

Intel at least can push slight updates while working to make their process node work without having to rely on someone who may never get it right (20nm).

 

[esrever]

Glofo does have all of IBM's research and tech. I think it will go more smoothly for them. Especially since their biggest customer is still AMD. They need to keep AMD afloat or else they lose a pretty large chunk of their business.

 

[jimmysmitty]

Glofo does have all of IBM's research and tech. I think it will go more smoothly for them. Especially since their biggest customer is still AMD. They need to keep AMD afloat or else they lose a pretty large chunk of their business.

Doesn't mean that the process node will go smoothly. Just because IBM gets it to work for SRAM /(what their 7nm currently is) does not mean it will work well with AMDs chip designs.

 

[juanrga]

I believe that AMD doesn't have any design wins because its past CPUs weren't good, so no one will close a deal with them on a new one. Like, if you spend the past 5 years making bad products, how can I thrust you to make the next product something better than the competition?

Maybe that's why they are advertising Zen and Zen+: Zen is meant to be the proof that AMD can design good chips, and Zen+ is the one to win designs.

Zen+ is a minor revision of Zen. If Zen is not a game changer then Zen+ will not be.

Large customers don't sign multimillionaire contracts with CPU companies and just say "ups" if the company fails to deliver promises. Those mentioned design wins are associated to contracts guaranteeing performance and power consumption within certain margins; for example the signed contracts guarantees that the computer will hit at least X GFLOPS, and any value above that is a welcomed bonus for the customer.

It seems evident that the numbers that AMD provided were inferior to the numbers provided by competitors. I am convinced that Zen cannot compete neither with Power9 nor with KNH in raw performance.

 

[juanrga]

http://www.kitguru.net/components/graphic-cards/anton-shilov/amd-we-have-taped-out-our-first-finfet-products/

Zen? Reliable benchmark leaks by christmas? AM4 leaks soon?

Zen will be tapped out next year. First performance leaks are expected for April... if there is no delays.

 

[esrever]

Glofo does have all of IBM's research and tech. I think it will go more smoothly for them. Especially since their biggest customer is still AMD. They need to keep AMD afloat or else they lose a pretty large chunk of their business.

Doesn't mean that the process node will go smoothly. Just because IBM gets it to work for SRAM /(what their 7nm currently is) does not mean it will work well with AMDs chip designs.

They currently have Samsung's 14nm process and so far that seems to be going smoothly for Samsung, the fact that AMD has taped out now would probably mean the 14nm+ process is on time too. 10nm will be what might slow them down but I doubt they will fall very far behind tsmc again just because of IBM's R&D that they currently have, they are still partnered with Samsung as far as R&D for furture nodes go too. 7nm and beyond is way too far off to consider right now. Even Intel is likely struggling with 10nm so we don't even know if 7nm will ever be possible for high performance chips.

 

[juanrga]

When the analyst asked if Intel's 10nm delay would affect them, Su seemed quite excited.

And when an analyst said her in her face, that doesn't trust the vision that he shared, then she admitted is not even confident in her own vision:

On the Computing and Graphics side, there are – basically think about it as four pieces to the business. So we have OEM processors, we have channel processors, we have graphics, consumer graphics and then we have professional graphics.

So amongst those four segments from what we see, we are being a bit cautious on Q3, just given we need to see exactly how the OEMs ramp; the Windows 10 launch, we think Windows 10 is a good product, but we need to see how that launches. As we go into holiday, we believe we'll make progress in graphics. We believe we'll make progress in professional graphics. I've said that the channel looks like it's healthier for us and we need to see how the OEM demand looks. But those are the ways I think about Computing and Graphics. So clearly we're not happy with the performance in Q2 but as we look forward, we need to manage the business the way we see it and this is how we see it today.

I can say you now, before Q3 numbers, that W10 will not be the savior that she expect and AMD will continue loosing market share and money in Q3 and Q4...

 

[jimmysmitty]

Glofo does have all of IBM's research and tech. I think it will go more smoothly for them. Especially since their biggest customer is still AMD. They need to keep AMD afloat or else they lose a pretty large chunk of their business.

Doesn't mean that the process node will go smoothly. Just because IBM gets it to work for SRAM /(what their 7nm currently is) does not mean it will work well with AMDs chip designs.

They currently have Samsung's 14nm process and so far that seems to be going smoothly for Samsung, the fact that AMD has taped out now would probably mean the 14nm+ process is on time too. 10nm will be what might slow them down but I doubt they will fall very far behind tsmc again just because of IBM's R&D that they currently have, they are still partnered with Samsung as far as R&D for furture nodes go too. 7nm and beyond is way too far off to consider right now. Even Intel is likely struggling with 10nm so we don't even know if 7nm will ever be possible for high performance chips.

The 14nm FinFET done by Samsung was a consortium deal. That works for them but I am saying that the 10nm will probably be more difficult for GloFlo and co to get to. If a leader in process technology is having issues with 10nm, then companies that have to collectively combine their efforts will as well.

And IBM did show off 7nm but the question is if their solution for materials will be viable as I have heard both good and bad about SIGe. Guess we will have to wait and see what they really have and what Intel decides to use for 7nm and below.

 

[8350rocks]

Glofo does have all of IBM's research and tech. I think it will go more smoothly for them. Especially since their biggest customer is still AMD. They need to keep AMD afloat or else they lose a pretty large chunk of their business.

Doesn't mean that the process node will go smoothly. Just because IBM gets it to work for SRAM /(what their 7nm currently is) does not mean it will work well with AMDs chip designs.

Not to be too loud about it...however...the agreement with GloFo and Samsung allows either one to consume overflow production to assure volume. It is also encompassing for future nodes until further notice...meaning that the Samsung/GloFo 10nm will be just like the arrangement for Samsung/GloFo 14nm.

Essentially...there is a reason that AMD has not specified a specific vendor for their 14nm FinFETs.

 

[Baralis]

http://venturebeat.com/2015/07/16/more-hints-that-amd-is-building-nintendo-nxs-processor/

If true it could help AMD revenue.

 

[jdwii]

http://venturebeat.com/2015/07/16/more-hints-that-amd-is-building-nintendo-nxs-processor/

If true it could help AMD revenue.

This was leaked already but most here are probably 100% sure Amd is making it but NX might not be a hit it might we don't know. Either way I do not expect Amd to make more then 10$ per unit if that, remember how much of a success the Wii was well that didn't really earn Amd a lot of money either.

 

[jdwii]

Spoiler

When the analyst asked if Intel's 10nm delay would affect them, Su seemed quite excited.

And when an analyst said her in her face, that doesn't trust the vision that he shared, then she admitted is not even confident in her own vision:

On the Computing and Graphics side, there are – basically think about it as four pieces to the business. So we have OEM processors, we have channel processors, we have graphics, consumer graphics and then we have professional graphics.

So amongst those four segments from what we see, we are being a bit cautious on Q3, just given we need to see exactly how the OEMs ramp; the Windows 10 launch, we think Windows 10 is a good product, but we need to see how that launches. As we go into holiday, we believe we'll make progress in graphics. We believe we'll make progress in professional graphics. I've said that the channel looks like it's healthier for us and we need to see how the OEM demand looks. But those are the ways I think about Computing and Graphics. So clearly we're not happy with the performance in Q2 but as we look forward, we need to manage the business the way we see it and this is how we see it today.

I can say you now, before Q3 numbers, that W10 will not be the savior that she expect and AMD will continue loosing market share and money in Q3 and Q4...

Zens launch reminds me of K7 in a way before the K7 i think Amd was close to bankruptcy and they bet it all on the CPU design. I will say that compared to then i think Amd is in a better shape with more options to it but then again more doesn't mean better unless you have the money to fund it all.

 

[cdrkf]

Spoiler

When the analyst asked if Intel's 10nm delay would affect them, Su seemed quite excited.

And when an analyst said her in her face, that doesn't trust the vision that he shared, then she admitted is not even confident in her own vision:

On the Computing and Graphics side, there are – basically think about it as four pieces to the business. So we have OEM processors, we have channel processors, we have graphics, consumer graphics and then we have professional graphics.

So amongst those four segments from what we see, we are being a bit cautious on Q3, just given we need to see exactly how the OEMs ramp; the Windows 10 launch, we think Windows 10 is a good product, but we need to see how that launches. As we go into holiday, we believe we'll make progress in graphics. We believe we'll make progress in professional graphics. I've said that the channel looks like it's healthier for us and we need to see how the OEM demand looks. But those are the ways I think about Computing and Graphics. So clearly we're not happy with the performance in Q2 but as we look forward, we need to manage the business the way we see it and this is how we see it today.

I can say you now, before Q3 numbers, that W10 will not be the savior that she expect and AMD will continue loosing market share and money in Q3 and Q4...

Zens launch reminds me of K7 in a way before the K7 i think Amd was close to bankruptcy and they bet it all on the CPU design. I will say that compared to then i think Amd is in a better shape with more options to it but then again more doesn't mean better unless you have the money to fund it all.

K7 worked out pretty well... not total dominance but certainly a decent cpu. I could live with another K7 type success, here's hoping 😛