First Steamroller processor core exposure

[Ajay]

Excavator has an ATI high density library. AMD considers it equal to a full node shrink.

And work was being done on Excavator: http://www.linkedin.com/pub/sheng-sun/7/7a4/89

I've seen other AMD profiles list design work done on XV (eXcaVator) but can't find them right now.

--------------------------------------------------------------------

It is possible that this is a color coded print out from the design tools, and parts of the design certainly do look synthesized. It is pretty grainy for a printout, but that could be purposeful.

The FPU doesn't look like 2x256 FMA to me, the bottom section looks like an FMA unit (2x128) with fully fanned out multipliers, but he top section has totally different logic. Granted, I'm very rusty on reading CPU die shots (was decent at doing so up to the P4), so I'm not sure if that fan out is representative of 64b muls or 128b.

Considering work has been done an XV and the fact that SR is late, I'm guessing that some of XV was grafted into SR and what we are seeing is Kaveri 2.0 - and that also jives with Kaveri being only a 2 module unit (and thus likely has a larger modules than PD in terms of actual size). Whether it's a real die shot, or a synthesized and somewhat blurry print out (or screen shot really, that probably makes more sense) is something that would be hard to tell, I would think, if it was done by someone within AMD with the correct skill set.

If it's real, it could signal a larger boost in IPC than we expected from Kaveri as well as higher multi-threaded performance.

In any case, I think it was leaked by AMD marketing to generate some noise to counter the info we are getting on Haswell.

 

[Ajay]

Well if AMD says so then it must be true.

They have a stellar track record of promising the moon and delivering it to date, combined with the fact there is zero conflict of interest on behalf of the AMD employees who convinced management to utter such a statement as you have paraphrased in your post.

Yep, I see nothing wrong with taking this on faith and treating it like the gospel truth. Nothing wrong with that at all. :hmm:

Oh come on, AMD wouldn't lie to us AGAIN! :whiste:

 

[CTho9305]

That die doesn't look real to me in some areas. It's possible the structures were optimized for...something...compared to more traditional structures. I'm not calling fake, but it looks odd to me.

The interface section sticks out to me more than anything else (probably because I'm an I/O designer). I/Os are highly regular structures that share a ton of wires across the lanes of the port(s), which means colinear alignment makes the most sense. The I/Os in this image look like they are grouped in small grids instead of lines, which makes routing congestion much more difficult to overcome.

What are you considering I/O here? The claim is that this is just a dual-core module, which wouldn't have any PHYs, right?

 

[Arachnotronic]

What are you considering I/O here? The claim is that this is just a dual-core module, which wouldn't have any PHYs, right?

I'm just going to outright ask you: is this a die shot of Steamroller?

 

[SiliconWars]

I'm just going to outright ask you: is this a die shot of Steamroller?

No matter what answer you get you'll be no closer to the truth.

 

[CTho9305]

I'm just going to outright ask you: is this a die shot of Steamroller?

I honestly don't know - I quit AMD last year, and I didn't work on the Bulldozer family, so I only know high-level bullet points about those cores. I also know that plans change (sometimes drastically), so my knowledge could easily be outdated.

 

[NTMBK]

I honestly don't know - I quit AMD last year, and I didn't work on the Bulldozer family, so I only know high-level bullet points about those cores. I also know that plans change (sometimes drastically), so my knowledge could easily be outdated.

I don't suppose you have any idea where the XBone and PS4 APUs are being fabbed, do you? It'd be a pretty big deal if AMD managed to get at least one of them fabbed at GloFo.

 

[CTho9305]

I don't suppose you have any idea where the XBone and PS4 APUs are being fabbed, do you? It'd be a pretty big deal if AMD managed to get at least one of them fabbed at GloFo.

Sorry, I couldn't find an official disclosure, and I'm not going to be the one who leaks something.

 

[NTMBK]

Sorry, I couldn't find an official disclosure, and I'm not going to be the one who leaks something.

Fair enough, it was worth a try.

 

[Vesku]

Sorry, I couldn't find an official disclosure, and I'm not going to be the one who leaks something.

Can you give an insider assessment of why Llano and later products have had such slow ramp ups in notebooks? Kind of odd from an outsider viewpoint that the OEMs seemed fine with pumping out hot and power hungry Turions and Phenoms but actually appeared less interested when AMD finally had a more mobile appropriate offering. I have actually been interested in getting an AMD notebook to try out but the variety has been so poor I couldn't justify not getting an Intel notebook.

Edit: It's not like AMD doesn't have the ability to meet demand, they've been paying to not have product made.

 

[Arzachel]

Can you give an insider assessment of why Llano and later products have had such slow ramp ups in notebooks? Kind of odd from an outsider viewpoint that the OEMs seemed fine with pumping out hot and power hungry Turions and Phenoms but actually appeared less interested when AMD finally had a more mobile appropriate offering. I have actually been interested in getting an AMD notebook to try out but the variety has been so poor I couldn't justify not getting an Intel notebook.

Edit: It's not like AMD doesn't have the ability to meet demand, they've been paying to not have product made.

If you have to ask, he probably can't That said, sticking a CPU and a GPU on the same die, both needing pretty different electrical properties to work optimally, is harder than it sounds and coupled with GloFo's process woes, they're probably not getting that many chips hitting the expected clocks within the power envelope needed to go into laptops at first. More importantly, consumers are willing to pay extra for the same experience to get "Intel Inside", which makes OEMs that already operate on razor thin margins (readretty much all of them) even less interested.

 

[CTho9305]

Can you give an insider assessment of why Llano and later products have had such slow ramp ups in notebooks? Kind of odd from an outsider viewpoint that the OEMs seemed fine with pumping out hot and power hungry Turions and Phenoms but actually appeared less interested when AMD finally had a more mobile appropriate offering. I have actually been interested in getting an AMD notebook to try out but the variety has been so poor I couldn't justify not getting an Intel notebook.

Edit: It's not like AMD doesn't have the ability to meet demand, they've been paying to not have product made.

I don't think I can really give you a useful answer here... I don't have much insight into that side of things (designers can finish working on projects over a year before they hit the shelves). There's enough information out there that your guess is probably about as good as mine in this case.

 

[Lepton87]

CTho9305 your decision to quit AMD was because you didn't want to stay on a sinking ship or was there any other reason?

 

[Idontcare]

CTho9305 your decision to quit AMD was because you didn't want to stay on a sinking ship or was there any other reason?

Surely you can appreciate how answering that question would be a potential career-path killer? You've got to be ready to burn some bridges on your way out to put up that kind of a personal opinion in a public arena

 

[Lepton87]

Surely you can appreciate how answering that question would be a potential career-path killer? You've got to be ready to burn some bridges on your way out to put up that kind of a personal opinion in a public arena

You're right, I didn't think of that. If he wants to answer he can do it thru PM but of course I don't expect him to do that and that's understandable. But saying it was because of other reasons wouldn't be burning bridges. On the other hand as far as I know he didn't disclose his surname like pm did.

 

[Seraphimcaduto]

CTho9305, if I may ask, what did you work on while at AMD? I'm not asking for specifics, I just wanted to get an idea what kinds of things people can do when working in research and development, design, ect... I'm always amazed how much people here know about CPU design and implementation here; This is coming from the chemist scratching his head while looking at the core picture still trying to figure out what the pic was.

 

[Lepton87]

It's really a shame that AMD lost a lot of good engineers such as CTho. With such a huge brain drain I'm not confident(to put it lightly) with their ability to compete with Intel I want to see AMD from the days of K7, K8 when they actually had better products then Intel most of the time. From the end of 1999 to 2006 AMD was actually very competitive, I don't know how they couldn't establish a good position in IT segment in 7 years. One product failure and they are on the brink of bankruptcy. In contrast Intel hardly even felt the failure of an architecture that was P4. It's the same with graphics cards, they make better cards/or with better value most of the time and people still buy their competitor's cards. Is is all due to marketing?

 

[CTho9305]

CTho9305, if I may ask, what did you work on while at AMD? I'm not asking for specifics, I just wanted to get an idea what kinds of things people can do when working in research and development, design, ect... I'm always amazed how much people here know about CPU design and implementation here; This is coming from the chemist scratching his head while looking at the core picture still trying to figure out what the pic was.

My main area of expertise is "physical design". I started writing up a much longer reply but I probably shouldn't derail the thread even more . The Wikipedia article looks like a decent overview, but feel free to ask more questions in PMs or another thread.

 

[NTMBK]

From the end of 1999 to 2006 AMD was actually very competitive, I don't know how they couldn't establish a good position in IT segment in 7 years.

Intel bribes, and limited supply.

 

[Vesku]

Intel bribes, and limited supply.

And threats (go below 90% Intel Inside and we start cutting you off buster!), but also AMD either never attempted or failed miserably at creating a close relationship with an OEM. One where they would guarantee supply (supply them first before the rest of the market) and be favorable with pricing in exchange for them having a strong line of AMD products. No AMD version of a mini-Sony or mini-Dell with an agreement for producing 90% or more AMD products.

 

[Lepton87]

Intel bribes, and limited supply.

How queer of Intel to give bribes when AMD already sold all they could. Limited supply was AMD's management fault but the be fair to them could they do something about it? IMHO they should've ported K7 and K8 to TSMC and sell both CPUs produced outside and inside but wasn't the deal between Intel and AMD precluding this and that only changed recently? With this clause AMD was doomed from the start, unless some investor came up to them and lent them money to build new fabs and fast, but that's just fantasy. They would have to become as big as Intel. Impossible.

 

[Vesku]

How queer of Intel to give bribes when AMD already sold all they could. Limited supply was AMD's management fault but the be fair to them could they do something about it? IMHO they should've ported K7 and K8 to TSMC and sell both CPUs produced outside and inside but wasn't the deal between Intel and AMD precluding this and that only changed recently? With this clause AMD was doomed from the start, unless some investor came up to them and lent them money to build new fabs and fast, but that's just fantasy. They would have to become as big as Intel. Impossible.

Even in a supply limited scenario, denying them an increased presence in products from big name companies like Dell and Sony is quite damaging. AMD getting more business from Dell wouldn't mean they say "sorry Dell not enough supply" it would mean there would be less AMD retail boxes available for enthusiasts. Intel understands marketing, they understood the importance which is why they worked and spent cash to prevent it.

 

[Lepton87]

Even in a supply limited scenario, denying them an increased presence in products from big name companies like Dell and Sony is quite damaging. AMD getting more business from Dell wouldn't mean they say "sorry Dell not enough supply" it would mean there would be less AMD retail boxes available for enthusiasts. Intel understands marketing, they understood the importance which is why they worked and spent cash to prevent it.

That makes sense, they probably expected a good CPU after K8 but when K10 released there probably was a party at Intel I always underestimate how much removed I am from the usual computer buyers.

 

[Idontcare]

Even in a supply limited scenario, denying them an increased presence in products from big name companies like Dell and Sony is quite damaging. AMD getting more business from Dell wouldn't mean they say "sorry Dell not enough supply" it would mean there would be less AMD retail boxes available for enthusiasts. Intel understands marketing, they understood the importance which is why they worked and spent cash to prevent it.

While AMD was certainly supply constrained, the lack of OEM demand from the likes of DELL and HP definitely reduced the profits (and ASP) that AMD could extract from their CPUs at the time.

The consumers who could (and did) buy AMD CPUs at the time benefited from Intel's shenanigans because it lowered the prices that they had to pay for their AMD CPUs. Supply vs. Demand.

That much I don't think anyone would contest or debate. However, where we enter the realm of the hypothetical is when we start contemplating (and fantasizing) about what AMD could have done with themselves had they enjoyed an extra $100 ASP for every processor they sold during those few years in which they were the performance champ and were supply constrained.

No question more money would have spilled into R&D. The K10 might have not be such a disappointment, or maybe not have been so late as 65nm might have been developed more aggressively.

We can play the "what if" game all day long but it doesn't change what happened, and it doesn't change the hand AMD's current management was dealt.

 

[AtenRa]

We can play the "what if" game all day long but it doesn't change what happened

Exactly, it is what they are doing NOW the important thing, it reminds me of the Greek economic/politics Crisis (not Crysis 3 ). What happened in the last 30 years is of no immediate concern, we really have to focus on what we can do NOW and the years to come, to change the situation so we will have a better future.