Intel moves up debut of billion-transistor chip

[hkh]

Intel wants to make a billion transistor chip by 2005 thats next year that must be a dual core chip.

http://www.eet.com/semi/news/showArticle.jhtml;jsessionid=J4U2C3EDILRWUQSNDBGCKHY?articleID=23900776

 

[SLee]

Should be about 1.75 billion transistors, code-name Montecito and is the first Itanium processor on Intel's 90nm process. It is dual-core and has multithreading capabilities. However, the vast majority of the transistors comes from the 24MB L3 cache and the amount of transistors dedicated to the cores is quite small.

 

[Captain Newbie]

Well, that rules out me having to buy a new oven, all I need is one of those

Good lord, 1.75B transistors... Should be interesting.

 

[bored-sarcasm]

oven? just get one of those easy bake ones.

 

[imgod2u]

I thought Montecito was to debut with 9MB of L3 cache. Tukwila would debut with a ton of L3 cache (may be 24MB but I don't recall) as well as SMT and multicore.

 

[SLee]

I think they changed some of the code names around; the upcoming 9MB L3 cache model remains Madison while the 2005 processor is now Montecito. Here's a picture of the die:
http://www.vr-zone.com.sg/?i=920&s=1

 

[sandman001]

Uhmm..it seems to me this thing is gonna be frikken huge, and very hot.

 

[Sjaak]

well, 1.75B transistors at 65nm will hardly be any bigger then a 200M transistors at 180nm

 

[SLee]

Uhmm..it seems to me this thing is gonna be frikken huge, and very hot.

Roughly 600mm^2, but much of its cache and with Intel's manufacturing capabilities, the cost to manufacture it is probably $100-$150. It should also fit within the existing Itanium 2 thermal envelope, around 110W TDP at release which is reasonable in comparison to competing Power 5s and SPARCs.

 

[BeerHunter]

All the wrong moves.

What they need to do is bring Dothan to the desktop, Add HT, 1000Mhz bus it and they will have a A64 killer. Then dual core it, instead of the gawd aweful presshot as now planned (can you say 200W chip!!?!), up Mhz assuming it will scale in the 2.4-2.8 Ghz realm and it will put the nail in the coffin of the appretice.

BTW have they ever sold one of these itaniums?

 

[SLee]

Why are you bringing up desktop products with the next major Itanium 2 revision, which is strictly targeted at the high-end server market. And Intel sold over 100K in 2003.

 

[ozzlo]

that's 100k of em selling for bout $2000 each which is bout 200,000,000 dollars if my math is right.... it usually never is.

 

[Strages]

lmao... that math is correct btw

 

[Sjaak]

thats one helluva amount of money for 100k chips. And the sales are rising fast.

If it turns out as intel wants it now, we will have to survive the next year on prescott, the two years thereafter on Dothan (dual core), and then the Itanium 3 will be ready for desktop (full 64 bit)

 

[Foxie3a]

All the wrong moves.

What they need to do is bring Dothan to the desktop, Add HT, 1000Mhz bus it and they will have a A64 killer. Then dual core it, instead of the gawd aweful presshot as now planned (can you say 200W chip!!?!), up Mhz assuming it will scale in the 2.4-2.8 Ghz realm and it will put the nail in the coffin of the appretice.

BTW have they ever sold one of these itaniums?

Yes, they've sold 2 of those itaniums...

And its harder than it sounds to bring the Dothan to the desktop. Basically its an "AMD"*GASP* Chip. It does more per clock.

Why can't AMD take their nice little A64, and jack it upto 3.4ghz? That's what they should do... And that's what your saying, but of Intel.

I'm sure that Intel is trying to do that, but its hard, and they dont want to put that out on the market, and have it fail.... Besides, they dont want to kill AMD.. its fun to have a small(I emphasize small) competition.

 

[Sjaak]

Why can't AMD take their nice little A64, and jack it upto 3.4ghz? That's what they should do... And that's what your saying, but of Intel.

Thats impossible, technology-wise. If they could, they would. If intel could make a 5Ghz prescot, they would.


(yes i know, people clocked way beyond 5Ghz, but that wasnt quite with standard conditions )

 

[OC Noob]

All the wrong moves.

What they need to do is bring Dothan to the desktop, Add HT, 1000Mhz bus it and they will have a A64 killer. Then dual core it, instead of the gawd aweful presshot as now planned (can you say 200W chip!!?!), up Mhz assuming it will scale in the 2.4-2.8 Ghz realm and it will put the nail in the coffin of the appretice.

BTW have they ever sold one of these itaniums?

Other people have addressed the desktop vs server aspect, so this comment doesn't pertain to that.

As far as bringing Dothan to desktop, I believe that is now the plan and I can't remember, but I think it will be dual core sooner or later too.


ps Dothan is not "an AMD chip". People seem to forget, pre-P4 chips did more work per cycle. Its not like they reverse engineered an AMD chip. There was (luckly for Intel) atleast 1 smart person working for Intel in the mobile chip division that initiated a chip that is clock for clock more efficient then hte PIII, P4, Athlon or A64 and is an amazing chip for its purposes.

I think the technology to keep Prescotts pipe full is pretty amazing so if they can take the best of Dothan and the P4 it should work out nicely.

 

[Quailane]

Why bother with ram, when you can get it all on the l2 or l3 cache? What are the prices on these things going to look like? I hope note like one of Bill Gate's paychecks.

 

[MetalStorm]

Why bother with ram, when you can get it all on the l2 or l3 cache? What are the prices on these things going to look like? I hope note like one of Bill Gate's paychecks.

That is the idea behind having such huge amounts of cache, however the servers they'll be used in will have huge amounts of memory as they are most likly going to be working on huge data sets anyway.