And the Bulldozer die size is……

[halofanman]

the articles say that instead of using an IBM cell processor, future consoles might uses bulldozer APU variants

[Glow9]

No, you made a single threaded comparison then proceeded to say Bulldozer's multi threading should help.

I can see what you mean I didn't speak too clearly, considering my comment is in an optomistic tone not sure why these retards chimed in with stupid comments .
So looking at the shrink in a 8 core processor one would presume that these would be efficient. Meaning disable 6 of the 8 cores theoretically these should be able to compete no? Or are these cores essentially just another die shrink with 2 added cores?

[BeepBeep2]

I can see what you mean I didn't speak too clearly, considering my comment is in an optomistic tone not sure why these retards chimed in with stupid comments .
So looking at the shrink in a 8 core processor one would presume that these would be efficient. Meaning disable 6 of the 8 cores theoretically these should be able to compete no? Or are these cores essentially just another die shrink with 2 added cores?

I'm not sure if I said this before or not, but Deneb's "Stars" core is pretty close to Bloomfield on a per mm2 performance standpoint, ignoring the rest of the die.

If AMD can make size scale = performance they will surely have a competitive single thread CPU imo.

[zanzabar]

the articles say that instead of using an IBM cell processor, future consoles might uses bulldozer APU variants

what i was getting at was, for the BD to be good on desktops it will need to work with programs made for the lowest common system that that program can be used by, so even though it can be nice and complex with many thread and complex instructions it will only be measured on a desktop platform by how fast is it doing things that a p4 dose. were on a console or server its more on how much power it actually has

[DeathReborn]

thats not really the same as a desktop. the current consoles are on PPC with the ps3 having a full cell chip and the 360 having a 3 cores from a cell in some sort of cell and classic PPC hybrid that they got ibm to steel from the sony/ibm cell project. consoles get coded for what they have while desktops rarely get code path update for cashe use, instructions and now threading, so while its hard to find a game that uses more than 2 cores most games use atleast 3 as thats what the 360 has when they are on consoles and since the consoles are underpowered in comparison so they need to be more optimized just to get the measly 30fps 720p or lower that console games run at.

Xbox 360 Xenon CPU has 6 Thread's (2 per core). I'm not sure if the PS3's Cell SPE units count as a thread each or it's just the 2 from the PPE that are used.

[STEvil]

Xbox 360 CPU is a full core per each two-threaded core, nothing to do with cell at all.

[zanzabar]

Xbox 360 Xenon CPU has 6 Thread's (2 per core). I'm not sure if the PS3's Cell SPE units count as a thread each or it's just the 2 from the PPE that are used.

my point was that console code is for that platform so its measured as what its power is when optimized were desktop parts are rated as when used with code for the LCD of the platform so can it be executed on/with a p4 or dx9

Xbox 360 CPU is a full core per each two-threaded core, nothing to do with cell at all.

its a bastard byproduct of the old PPC core and the cell using a HT based interconnect to link the cores and l2 cashe, but it dose not pertain to my example other than its in existence and the platform its in is optimized for that chip, not how it works.

[halofanman]

http://translate.google.fr/translate...t-chips-23.htm

[TESKATLIPOKA]

AliG I never though AMD will have better performance in single core at the same frequency.
Its unlikely if you look at the size of a module and SB core. SB core is still bigger by 10%.
SB core to core performance against Thuban is better by 35% but the size is +70%. So I am speculating BD will be slower by 5-10% at best but with turbo 10% higher the difference will be zero or BD will be a little faster.

[MBomberZ1]

that is not worded right right at all.

what you just said is the same as "AMD designed bulldozer to be slower in game benchmarks than sandy bridge", which is not correct. first of all, we don't know if BD is slower or not. second, we don't know AMD's design decisions.

if something sucks, you can't say it was designed that way. that is just crazy.

What happened to you??? You used to be a troll monster! Anyways I for one appreciate reading a positive post by you, thanks!

There is still some waiting to do peeps. From the looks of the initial image, I don't see how everyone can assume that these "blank" areas are not being used. Look at the quality of the image, as if it was drawn instead of a closeup photo of the die. Its obviously for general reference to the layout, not a detailed transistor map.

[AliG]

AliG I never though AMD will have better performance in single core at the same frequency.
Its unlikely if you look at the size of a module and SB core. SB core is still bigger by 10%.
SB core to core performance against Thuban is better by 35% but the size is +70%. So I am speculating BD will be slower by 5-10% at best but with turbo 10% higher the difference will be zero or BD will be a little faster.

Sigh, you still don't understand what my point is. That's fine, I'm not losing any sleep over this

[bamtan2]

So I am speculating BD will be slower by 5-10% at best but with turbo 10% higher the difference will be zero or BD will be a little faster.

Yours is an educated guess because AMD is choosing the BD speeds KNOWING how fast SB is. How long has BD been in development while the competition has been out? Exactly...

[zalbard]

AMD is pulling a FERMI....
Talk and talk and no product untill way too late... I remember how we were discussing if it would arrive before Gulftown on this forum, and many others, now we are wondering if it will arrive before or after SNB-E....

The die size of GTX 480 is 529 mm². Nowhere near being close.
And even if you look at Intel chips it does not appear to be quite as huge.



Keep in mind that this is an 8 core CPU we are talking about here.

P.S. Thuban's die size is 346 mm²...
So much panic over nothing...

[Manicdan]

^thuban was the last of 45nm after they got yields to their peak
starting a new node with a chip of 300mm2 is semi risky, but there will be plenty of quads to sell and they will be in plenty of demand for gamers. if the price is really 190$ for a quad and 310$ for a octo, thats not bad of a margin on something half broken, lol.

[IntelUser]

Guys, Sandy Bridge EP is EXACTLY 400mm2. It was revealed by Intel at ISSCC.

[saaya]

wow, no wonder amd changed their contract with glofo to only pay for functional dies and not for wafers

[highoctane]

wow, no wonder amd changed their contract with glofo to only pay for functional dies and not for wafers

Didn't GF replace or fire the chap that authorized this, can't remember his name now...

[eXa]

Just wondering.... With a fairly large die and the modules spread apart. would that benefit cooling anything? does the cache produce much heat btw? It looks very ideal atleast if its mainly the modules that produce heat....

[kaktus1907]

Just wondering.... does the cache produce much heat btw? It looks very ideal atleast if its mainly the modules that produce heat....

cool area is l2 caches and red ones are integer or fpu idk

[eXa]

Thank you

[qcmadness]

315mm^2... in line with previous estimations

not very bad considering it has 16MB+ total cache

[kuroikenshi]

hummm so its 3 weeks before lunch uh!

[Particle]

Three or four weeks--yeah. We're getting very close now.

[kuroikenshi]

jah jah!!

[Lanek]

If all is good and rumors true, yes second week ( or just start of the 3th week ) of september .