[FUDZ] The heart of AMD’s Llano is K8

[crazylocha]

Additionally, isn't the GPU part of CPU+GPU be quite easy to update/refresh/whatever?

Once AMD has their newer plants up and running, in theory, yes. Easier, NOT easy though.

Being tied to TSMC's scheduling (recent nVidia scheduling as example) has pushed several factors back. Including TSMC's decision part way into BB/BD designs to skip a generation of die shrinks. Hence the need for AM3+ skip on some products apparently.

 

[Phynaz]

Once AMD has their newer plants up and running, in theory, yes. Easier, NOT easy though.

Being tied to TSMC's scheduling (recent nVidia scheduling as example) has pushed several factors back. Including TSMC's decision part way into BB/BD designs to skip a generation of die shrinks. Hence the need for AM3+ skip on some products apparently.

AMD no longer has plants. They are always going to be tied to somebody else's scheduling.

I wasn't aware that TSMC was manufacturing Bulldozer though.

 

[zokudu]

AMD still owns a significant portion of Global Foundries the "not owning your own fabs" clause from the Intel agreement only really allowed GF to be financially independent of AMD though AMD will still get preference on Fab space.

 

[Zstream]

I'm a little daft too, you're gonna have to be a little more explicit if you want me to "get" whatever message it is you think you are sending...

Simply that the article has nothing of substance. You being a moderator and highly technical know this. That is all, nothing more. If there is any indication of such cpu being a die shrunk K8, then he could at least put some relative thought into the article.

 

[Idontcare]

Simply that the article has nothing of substance. You being a moderator and highly technical know this. That is all, nothing more. If there is any indication of such cpu being a die shrunk K8, then he could at least put some relative thought into the article.

Z, of course I know this, I was really quite confident my OP would have indicated this much.

The guy is full of BS, my posting it here was just to have conversation about that fact. There really was no other agenda.

Just killing a saturday chatting about the latest absurd comedy spewing forth from Fudz...the thread title [FUDZ]...

 

[Hacp]

Ouch, dems fighten words, I really can't believe this to be true. It makes no sense for AMD to invest resources to shrink the K8 core versus shrinking the K10 core.

Doesn't K8 have less transistors than K10?

 

[Idontcare]

Doesn't K8 have less transistors than K10?

Sure, but its them extra xtors that improved the IPC of the Stars core over K8. This is Llano, not bobcat, I expect IPC to improve in a 45nm->32nm shrink over Phenom II, not take a step back.

FUDZ crafted his article to obviously imply the worst, I find it absurd and have said just as much. The work involved in doing the layout work to shrink K8 versus K10 at this juncture is not going to be materially different, and if shrinking K10 results in a higher performing, higher price-fetching, mobile processor then that is where the ROI would be and that is where the decision makers would have spent their R&D imo.

 

[Wreckage]

Fusion to fight Sandy Bridge with 6-year-old core

Makes me wonder if the GPU will be just as old.

 

[jvroig]

Fudzilla is correct literally because K8 can refer to anything, even Phenom II, if you are in AMD - unless you follow the real nomenclature that AMD follows, such as "Family 10h" for the most current Phenom II line, or cite the actual codenames you wish to refer to ("Hammer", etc)

And inside AMD's dev team, "K10" is Bulldozer.

If people remember, AMD never officially claimed/called anything as K10 or K10.5 as is popular in the media today. It's something the tech world came up with, not AMD, simply because the last "K" nomenclature AMD used was "8".

Of course, although literally correct, what Fudzilla implies is false and can raise heedless hysterics, so yeah, the conclusion pretty much remains the same that Fudz is obviously aiming for page hits with worthless/needless articles.

Anyway, when AMD has been saying since forever than Llano = Phenom II core, I really wonder what got into Fuad's mind to even bother posting this "article", when everything from AMD about Llano has quite specifically mentioned "Phenom II" instead of just the much broader and vague "K8" internal nomenclature.

EDIT:
Just scanned this thread, it seems Soleron already did mention this fact:

Fuad is just wrong. The misunderstanding comes from the fact that Phenom II is internally referred to as K8, same as Phenom I and Athlon 64. K9 was a cancelled design. K10 is Bulldozer. Another internal name for Phenom II's core is Greyhound+ .

I've no idea how I missed this thread and only saw it now that Wreckage bumped it to critique the GPU. Perhaps I was too engaged in the Bulldozer thread.

 

[Daedalus685]

Anyway, when AMD has been saying since forever than Llano = Phenom II core, I really wonder what got into Fuad's mind to even bother posting this "article", when everything from AMD about Llano has quite specifically mentioned "Phenom II" instead of just the much broader and vague "K8" internal nomenclature.

I'm pretty sure this all comes from AMD actually claiming K8 using their own nomenclature mixed with Fuad not exactly being the best investigative reporter in the world..

 

[Keysplayr]

Could this be similar to what Intel did? Abandoned the P4 arch and expanded on the much older PIII architecture into what we have today? Started with banias and Dothan, a modified P6 architecture. They backtracked to go forward and it worked out really well. Maybe AMD can pull off the same, if there is any truth to this at all.

 

[heyheybooboo]

Could this be similar to what Intel did? Abandoned the P4 arch and expanded on the much older PIII architecture into what we have today? Started with banias and Dothan, a modified P6 architecture. They backtracked to go forward and it worked out really well. Maybe AMD can pull off the same, if there is any truth to this at all.

I'm thinkin' the ultimate result is dependent upon the instruction set.

As I understand it, in the past Intel did not have to release new instructions to AMD until they released a processor that utilized them. That changed with the settlement -- that info is purportedly more readily shared.

It is conceivable that the Stars core doesn't bring much more to the table when pitted against a K8 properly charged with SSE4.x, AVX, SSE5, etc., in the overall scheme of the arch.

Toss in advances in the MC, better compiler support, process enhancements, etc., and maybe it just works better moving forward for AMD at this point in time.




--

 

[jvroig]

Could this be similar to what Intel did? Abandoned the P4 arch and expanded on the much older PIII architecture into what we have today?

No, it's still really a new core they are tweaking, Deneb or Propus. For AMD, they are all K8, since K10 is Bulldozer for them. The names "K10" and "K10.5" is a tech media invention, not something that AMD actually uses internally.