[AT] AMD Reorganizes Business Units - no more high performance x86 cores?

[mrmt]

But let's look at other examples. What do you think cost the most to develop- Netburst, or Banias? Prescott or Yonah? Bobcat or Bulldozer? Throwing money and staff at a problem in no way guarantees you success. There are dozens of ways that a project can fail, and lack of resources is just one of them.

You can certainly not be able to reach your project targets even if given the right amount of resources, but you are sure to NOT reach your project goals if you do not have the right amount of resources. Given AMD budget limitations, whatever projects they are developing now should have a much narrower scope than previous projects, otherwise they'll be ensuring that they will fail again.

On top of that, AMD track record speaks for itself. K12 will either have a limited scope or be a failure.

 

[Homeles]

Then how did ARM succeed?

To continue my comment...

I don't think AMD is doomed, but they're definitely in a tough position. This forum is far too black and white most of the time. You either get arguments that AMD's going to die any day now, an argument that's been said for years, or you get the argument that Jim Keller is computer chip Jesus, who's risen from the dead to save AMD from Satan-owned Intel corporation. As funny as the latter may sound... I think it's a very apt metaphor. People that argue on AMD's half and often do nothing else tend to see Intel as the epitome of corporate evil, and AMD as some saint-like figure. It's honestly a lot like a religion (for better or for worse -- not trying to fire any shots here), as CHADBOGA's signature cleverly points out. On the other side, there's the people that have just become so jaded by AMD's consistent failure to deliver, that they're dismissing any and all chance of AMD making a comeback.

In the end, all businesses fail. Even Intel will go the way of the dodo eventually. AMD isn't going to be around someday, and people need to be less vehemently defensive about that fact. And on the other hand, others need to realize that some people hold their hardware so close to their hearts that you have to tread carefully. It's amazing how worked up people, including myself, get worked up over this. If we weren't essentially anonymous... I wouldn't be surprised if we'd see hate crimes happening. Throw us all together in one room... and there would be some hospital visits at the very least. It's only a matter of time before someone bombs an aborti-- err, SemiCo headquarters.

Classic strawman, nothing more to say.
Rule #2
http://www.craigbellamy.net/2013/06/24/10-commandments-of-rathional-debate/

I'm well acquainted with logical fallacies. There's a difference between a straw man, and not being clear enough in your writing to where your audience fails to understand what you're saying. When you take a look at what you were disputing, I think it's hard not to take your comment the way I did. So I've apparently misunderstood you -- why don't you elaborate on what you meant?

 

[NostaSeronx]

K12 and its sister x86 core will be on par or greater than Project Denver.

 

[SlickR12345]

It probably doesn't mean anything. I'm serious, at a lot of these conferences and most of the press statements and stuff are just utter crap of nothing.

Just put some cool sounding bit pieces, repeat several words like efficiency and performance over and over again and call it a conference/whatever.

It literally probably just mean nothing.

 

[Arachnotronic]

Then how did ARM succeed?

Does ARM sell processors? No, it's an IP company with larger, more powerful, and better funded companies implementing their IP (or building custom cores on top of the ARM ISA) that have succeeded.

 

[jpiniero]

As long as the x86 part doesn't spread themselves too thin, I wouldn't be too concerned about R&D spending. Now the WSA...

K12 and its sister x86 core will be on par or greater than Project Denver.

Now that's being awfully optimistic.

 

[NostaSeronx]

Now that's being awfully optimistic.

There is currently a two year gap between Tegra Logan64 and the K12 APU. So, it isn't really that optimistic. It is literally taking two years for AMD to get on par with Nvidia.

 

[ShintaiDK]

As long as the x86 part doesn't spread themselves too thin, I wouldn't be too concerned about R&D spending. Now the WSA...

AMD R&D budget today havent been this low for over 10 years if we add inflation. And as a note the R&D budget went up 50% when they bought ATI. That budget now has to cover x86, ARM and GPUs.

 

[NostaSeronx]

AMD's R&D covers;
x86 CPUs
x86 APUs
ARM CPUs
ARM APUs
Discrete GPUs
Integrated GPUs
Northbridge Fabric
Southbridge Fabric
Exascale Fabric
High Bandwidth Memory
Chipsets
Porting to next generation nodes
Semi-custom SoCs
Future instruction set architectures + extensions
etc.

 

[CHADBOGA]

i3 uses the same cores as i5, i7, and Xeon (well, except for having Turbo disabled ...). In this thread, people have only talked about the K12 *core*, not products that will use this core in varying numbers, like you seem to be suggesting. This was not a very effective anti-AMD jab.

inf64's comment was unclear as to whether he thinks AMD are going to try and make a FX replacement or not.

I interpreted his remarks as suggesting that the K12 would be a successor to Kaveri, thus targeting the i3 range, as Kaveri doesn't target i5's & i7's.

Asking people for clarification isn't making an anti-AMD jab. 🙄

 

[Tsavo]

AMD R&D budget today havent been this low for over 10 years if we add inflation. And as a note the R&D budget went up 50% when they bought ATI. That budget now has to cover x86, ARM and GPUs.

That's scary. R&D is fuel for the tech fire. No fuel, no fire.

 

[NTMBK]

AMD R&D budget today havent been this low for over 10 years if we add inflation. And as a note the R&D budget went up 50% when they bought ATI. That budget now has to cover x86, ARM and GPUs.

10 years ago AMD was still paying for fab R&D. This is now handled entirely by GloFo and TSMC.

There is definitely a risk of them spreading themselves too thin. But they seem to be on the right track- designing a single shared platform for their ARM and x86 products makes a lot of sense. We don't know how deep the "shared lineage" will go, but I am hoping that it will at least go as deep as the uncore, L3$ and iGPU, with only the CPU cores changing between x86 and ARM models. Just my speculation of course. The only information we have is a quote from Jim Keller that they will be

compatible at the pin level and inside

but this could of course mean many different things.

 

[ShintaiDK]

10 years ago AMD was still paying for fab R&D. This is now handled entirely by GloFo and TSMC.

There is definitely a risk of them spreading themselves too thin. But they seem to be on the right track- designing a single shared platform for their ARM and x86 products makes a lot of sense. We don't know how deep the "shared lineage" will go, but I am hoping that it will at least go as deep as the uncore, L3$ and iGPU, with only the CPU cores changing between x86 and ARM models. Just my speculation of course. The only information we have is a quote from Jim Keller that they will be



but this could of course mean many different things.

10 years ago AMD didnt pay for GPU, chipset or ARM. And IC design costs have only gone up since.

It seems there is still an illusion that they will be able to create everything with minimal R&D.

 

[NTMBK]

10 years ago AMD didnt pay for GPU, chipset or ARM. And IC design costs have only gone up since.

It seems there is still an illusion that they will be able to create everything with minimal R&D.

You must have missed the news about chipsets: http://techreport.com/news/26591/report-amd-to-outsource-chipset-design-to-asmedia

But no, I don't think there is an illusion. Of course AMD have an uphill struggle ahead of them. At least they seem to be doing a few smart things to lower R&D costs (sharing tech between ARM and x86, using more automated tools, making more portable CPU designs), but that clearly doesn't guarantee anything.

 

[ShintaiDK]

You must have missed the news about chipsets: http://techreport.com/news/26591/report-amd-to-outsource-chipset-design-to-asmedia

But no, I don't think there is an illusion. Of course AMD have an uphill struggle ahead of them. At least they seem to be doing a few smart things to lower R&D costs (sharing tech between ARM and x86, using more automated tools, making more portable CPU designs), but that clearly doesn't guarantee anything.

I didnt miss it. Currently its a rumour. But assuming its true. Then its just another example that every stone needs to be turned for their R&D budget. And I assume the R&D budget will be lowered as well with any outsourcing.

 

[pTmdfx]

Of course AMD have an uphill struggle ahead of them. At least they seem to be doing a few smart things to lower R&D costs

What about semi-custom NRE cost paid by the customers? This should not be part of the net revenue according to my acct. knowledge, and may have been hidden behind the R&D expense figure on the sheets. There are also third party funds on different research projects, by the way.

 

[pTmdfx]

plus AMD is now with a much smaller R&D budget than they had with at the time they were developing their small core line.

Growing your small core without growing your R&D budget significantly sounds a viable approach. Then perhaps share some of the R&D between the two cores? Sounds nice. There are always things that can be shared, like researches on architectures, implementation and techniques of the process technologies, etc. Particularly, we are talking about two built-from-sketch cores started together, alongside the migration towards standard process technologies.

Now it is true that a core larger in microarchitecture must drive higher the design cost, but let's not forget how much additional cost it adds depends on more than the size but also how the core is worked out. In Bulldozer's case, the cost was likely related to the bleeding-edge node and also the manpower of the (relatively) highly custom design. If one can grow Cats at a low budget alongside a bleeding-edge, cash-burning architecture, I don't see why it is impossible to build something bigger in architectural size while maintaining a affordable budget either.

K12 should be an extremely cost efficient chip, not a top notch performance chip.

Why not pick somewhere else but the two extremes on the scale? Is that building something big must always mean to compete to the highest end of the processor spectrum? Why can't it simply aim at nice performance per watt per dollar at an affordable cost while scalable in some sense? Moreover, the tiny truth is that if it does want to be cost efficient to the edge, it needn't build a custom core. Just use the vanilla one and the sun will rise - if you mean the ARM one.

 

[Homeles]

Why not pick somewhere else but the two extremes on the scale? Is that building something big must always mean to compete to the highest end of the processor spectrum? Why can't it simply aim at nice performance per watt per dollar at an affordable cost while scalable in some sense? Moreover, the tiny truth is that if it does want to be cost efficient to the edge, it needn't build a custom core. Just use the vanilla one and the sun will rise.

Well put.

 

[raghu78]

You can certainly not be able to reach your project targets even if given the right amount of resources, but you are sure to NOT reach your project goals if you do not have the right amount of resources. Given AMD budget limitations, whatever projects they are developing now should have a much narrower scope than previous projects, otherwise they'll be ensuring that they will fail again.

On top of that, AMD track record speaks for itself. K12 will either have a limited scope or be a failure.

AMD's track record after 2011 is good. The bulldozer core is a failed design and the blame lies on the architects who worked on it from 2006 - 2011. Bobcat,Jaguar,Puma are excellent examples of AMD's design prowess. GCN too is a good example of AMD's design capabilities. The semi-custom business was built around these two pillars. AMD has the resources to deliver on a efficient high performance core. btw K12 is clearly aimed at extending the performance range of ARM. Jim Keller said the way we built K12 was we built a bigger engine. I am guessing he meant that K12 is a wider machine with massive resources. Its easy to see that since Keller headed the design of Apple's Cyclone core which is a massively wide core.

http://www.anandtech.com/show/7910/apples-cyclone-microarchitecture-detailed

btw this is on a Samsung 28nm gate first process so you can bet on a even wider machine on Samsung 14 LPE/14LPP process. K12 will be designed to scale to 3.5+ Ghz and it will go up against the biggest x86 cores from both Intel and AMD. There are people who are skeptical about AMD and people who are bullish about AMD. There is no problem if you belong to the former category.

 

[ShintaiDK]

K12 will be designed to scale to 3.5+ Ghz and it will go up against the biggest x86 cores from both Intel and AMD. There are people who are skeptical about AMD and people who are bullish about AMD. There is no problem if you belong to the former category.

If something is too good to believe, it usually is. And with AMDs minimalistic and ever reduced R&D budget its even further away to believe.

Since you compare to Apple. Their R&D budget is 4 times greater than AMDs.

You are not getting something you didnt pay for in R&D. There isnt engineers who just magically come up on their own with completed super innovations over night for chip designs.

 

[NTMBK]

. K12 will be designed to scale to 3.5+ Ghz and it will go up against the biggest x86 cores from both Intel and AMD.

It's a bit early to start declaring frequency targets, given that we know almost nothing about it 😵

 

[NTMBK]

Since you compare to Apple. Their R&D budget is 4 times greater than AMDs.

Apple's R&D also covers designing a tablet line, a phone line, 2 laptop lines, 3 desktop lines, two entire operating systems, a music player line, app development for various browsers, music players, etc... Oh, and a CPU. They're not directly comparable in any way. Come on, you're better than this 😉

 

[Homeles]

It's a bit early to start declaring frequency targets, given that we know almost nothing about it 😵

No kidding. People should leave that stuff to Seronx...

 

[Arachnotronic]

It's a bit early to start declaring frequency targets, given that we know almost nothing about it 😵

If you're an AMD permabull, the next chip will always have amazing performance/clock, clock to obscene heights, sip power, and bring untold riches to AMD shareholders 🙂

We're not a bait shop. But we are a travel agency. To that end I would like to reward you with an all expenses paid vacation, on the house.
-ViRGE

 

[AtenRa]

If you're an AMD permabull, the next chip will always have amazing performance/clock, clock to obscene heights, sip power, and bring untold riches to AMD shareholders 🙂

For a moment i thought you were describing Intel's graphics and tablet financials, next year will be the next big thing that never comes 🙄

Good news! You will be joining Intel17 on that vacation.
-ViRGE