K8L not til 2008?????

[Duvie]

Originally posted by: starwars7

Originally posted by: Duvie
wait for Intel to release its 2nd qtr numbers.....

When will they be releasing those numbers?

i believe 7/19.....

http://www.intel.com/intel/finance/bus_outlook.htm

here is the forecast before the actual announcement...Keep in mind many dont even hit forecast...

 

[zsdersw]

Originally posted by: Stoneburner
i was reading about intel planning 32 core cpu's by 2010 and i laughed. IS intel really this stupid where they just find a single path and continue down it until they hit a brick wall? SO the ghz race is now the core race?

INTEL is SOL.

Oh yes, because *you're* the authority on where things are going and the validity of Intel's plans.

Try again, idiot.

 

[Duvie]

Originally posted by: zsdersw

Originally posted by: Stoneburner
i was reading about intel planning 32 core cpu's by 2010 and i laughed. IS intel really this stupid where they just find a single path and continue down it until they hit a brick wall? SO the ghz race is now the core race?

INTEL is SOL.

Oh yes, because *you're* the authority on where things are going and the validity of Intel's plans.

Try again, idiot.

In the end it will all be a moot point if the software canno ttake advantage f it...multitasking aside we need more apps able to take advanatge of 2 cores now let alone 32....


I think multicore will be the next big thing but I think 32 core is not likely...We will just be quad core in 2007-2008...I doubt we hit oct-core or 8 cores until end of 2009 at the earliest....32 not likely...

remember it will take several reduction in micron size to even make that profitable....How fast do you think intel is going to retool about 3 to 4 more times???

 

[dmens]

fyi, keifer is not for general computing.

as for this fluff about server vs desktop, or memory support for K8L, that is all fluff. K8L is a major core revision, so only the core tapeout date matters for market release. Everything else is irrelevant.

 

[Stoneburner]

Originally posted by: zsdersw

Originally posted by: Stoneburner
i was reading about intel planning 32 core cpu's by 2010 and i laughed. IS intel really this stupid where they just find a single path and continue down it until they hit a brick wall? SO the ghz race is now the core race?

INTEL is SOL.

Oh yes, because *you're* the authority on where things are going and the validity of Intel's plans.

Try again, idiot.

IF only authorities are allowed to bark then this forum needs to be done away with. I can't imagine a more assinine non response. Feel free to call me an idiot though, as you are clearly as authority on the matter as per federal rules of evidence.

 

[zsdersw]

IF only authorities are allowed to bark then this forum needs to be done away with. I can't imagine a more assinine non response. Feel free to call me an idiot though, as you are clearly as authority on the matter as per federal rules of evidence.

It's not a question of "allowing" you to "bark".. it's simply a question of the idiocy of the remark. "Intel is SOL" is about as idiotic of a remark as "AMD is SOL". *Anyone* looking at the Intel-shuffled tea leaves or the AMD-shuffled tea leaves and proclaiming either is SOL is pure idiocy.

You'll note that I never said anything about Intel's 32-core initiatives.. I'm simply saying how ridiculous it is to come to the conclusion you have at this time.. and, in particular, how foolish it is to say Intel is SOL.

Feel free to make yourself look like a fool. Expect to have people call 'em as they see 'em, though.



Go troll somewhere else for a week.
AnandTech Moderator

 

[Stoneburner]

Well i was intending to be ironic based off my reply regarding AMD. I was also referending intel's penchant for going to the extreme based on a single idea. Based off the article i read and intel's past history I think it was a fair extrapolation. In fact it struck with me with a powerful sense of deja vu because those articles about Intel reaching 10 ghz by 2008 or so.


Anyway, we need more clarification as to AMD's release schedule before we can judge what happens but it really looks like both companies are repeating their past mistakes, which is why I said both were SOL. Forgive my attempts at wryness.

 

[Griswold]

Originally posted by: dmens

as for this fluff about server vs desktop, or memory support for K8L, that is all fluff. K8L is a major core revision, so only the core tapeout date matters for market release. Everything else is irrelevant.

Since we dont know when it taped out / will tape out, we can only go by the one official time frame for K8L and that is server parts in 2007. End of story.

 

[Hard Ball]

Originally posted by: Griswold

Originally posted by: dmens

as for this fluff about server vs desktop, or memory support for K8L, that is all fluff. K8L is a major core revision, so only the core tapeout date matters for market release. Everything else is irrelevant.

Since we dont know when it taped out / will tape out, we can only go by the one official time frame for K8L and that is server parts in 2007. End of story.

Right, what we know right now:

-- Server/Workstation K8L's are due mid-year 2007; whether it can be pulled forward, we don't know
-- AM3 socket will be due in 1H08, so desktop K8L's will be due sometime before, or near that time.
-- at least some K8L's will be compatible with AM2 socket. But we don't know whether this will mean K8L will arrive before AM3 or not, aside from some speculations.
-- Mobile uarchitecture will be different than that of K8L, so core updates for Turions will be irrelevant as regard to K8L schedule; which also means that some SFF and HE desktop chips will also be from this new mobile uarchitecture instead of K8L, which may arrive at a different time.

 

[Keysplayr]

Originally posted by: Stoneburner
i was reading about intel planning 32 core cpu's by 2010 and i laughed. IS intel really this stupid where they just find a single path and continue down it until they hit a brick wall? SO the ghz race is now the core race?

INTEL is SOL.

I don't know about 2010 and 32 cores, but in the distant future, I can see there being thousands of cores in a single computer. Each the size of a head of a pin. The future is never really clear, but your comment here was extremely short-sighted.

 

[Hard Ball]

Originally posted by: keysplayr2003

Originally posted by: Stoneburner
i was reading about intel planning 32 core cpu's by 2010 and i laughed. IS intel really this stupid where they just find a single path and continue down it until they hit a brick wall? SO the ghz race is now the core race?

INTEL is SOL.

I don't know about 2010 and 32 cores, but in the distant future, I can see there being thousands of cores in a single computer. Each the size of a head of a pin. The future is never really clear, but your comment here was extremely short-sighted.

On current technology (the technology that's really carried us from the 1970's til now), it's rather unlikely.

The smallest process that's possible on silicon based material, such as SiGe, only would be able to function as semi-conductor down to a process of 5nm or so, given the size of the individual atoms, as well SNR considerations and such.

Beyond 45nm generation that is being worked on right now, you have an additional 6 generations, maybe 7, of progressively smaller processes; each following Moore's law at approximately 70% of the size of the smallest feature on the die from the previous generation. You will probably have a max number of 25 billion transistors for the same die area as on current chips. Given that currently, the average dual core chips have 200 mil transistors, give or take some; if you stipuate that the complexity of each core does not increase over time (which is extremely unlikely), you would have maximum of 250 cores that will be possible for each chip, if the die area were to remain the same as that of today.

So barring a new technology coming out of semiconductor matsci (which is entirely possible, semiconductor industry had surprised us before, just not on this scale), you will be looking at realistically a max of 100 core per processor, given that at least some increase in complexity in the cores will be very likely (used in such area as additional execution units, extension to ISA, etc, etc), or perhaps even less. A chip with thousands of cores at this point is nothing but a pipe dream, unless someone finds a material that would be able to replace silicon compounds entirely.

 

[Stoneburner]

I wonder... with AMD"s plans for specialized coprocessor support, will they use CELL type coprocessors? I heard this theorized in an article on CELL a while back... CELL SPE's combined with a general purpose processor would be much more efficient than multipe general purpose processors no?

 

[DrMrLordX]

Cell is having yield problems out the wazoo. I wouldn't expect anyone else to take that approach, at least not the way Sony and IBM have gone about it.

 

[Stoneburner]

well you dont need an entire cell processor, just a few SPE's no?

 

[DrMrLordX]

The more functioning SPEs, the better.