Intel to release first quad core processor in 2007

[SpeedZealot369]

http://www.dailytech.com/article.aspx?newsid=1471


Intel is stepping up it's game, hopefully AMD has similar plans...

 

[Jules]

um k

 

[ahock]

even though AMD can release their 1st quad core ahead of Intel, I dont think they'll be able to ramp it sufficently since it wont be at 65nm.

 

[Duvie]

Originally posted by: ahock
even though AMD can release their 1st quad core ahead of Intel, I dont think they'll be able to ramp it sufficently since it wont be at 65nm.

AMD wont laucnh a quad core until it is 65nm....Doesn't make fiscal sense to do it with the 90nm...

Dont forget ppl....AMD is set to release quad cores in Q4 2006 with the opterons....AMD services the business sector first and has followed this mode for 3 years now...

 

[openwheelformula1]

hopefully qua cores are reasonably priced.

 

[BrownTown]

haha, yeah right you gotta be kidding.

 

[Viditor]

Originally posted by: openwheelformula1
hopefully qua cores are reasonably priced.

It depends on your definition of reasonable...if $1300 and up is reasonable then I'm sure it will be.

Edit: On second thought, they may do the same with Kentsfield that they did with Smithfield...take their poorly binned Conroes and glue them together for a very low speed quad core at a reasonable price (probably around $800-900).

 

[bunnyfubbles]

Once they're mainstream I wouldn't be surprised if the cheapest quadcores are $400-600. Usually "reasonably" priced means doubling the price of half the cores before them

 

[jiffylube1024]

It looks cool and all, but for the moment, dual core hasn't even been mass-adopted yet. So, good for Intel and their profit margins once they release the chip to the "Xtreme" market, but for the most of us, quad core is basically just another "FX" or "Extreme Edition" - cool tech but priced out of our reach.

In two to three years it should be interesting, though..

 

[BrownTown]

Kentsfield will replace the Conroe XE, so that means a minimum of 1000$ for it, but its the same as getting a 2 processor board in one package, so its not as insane as it sounds. Also it should be cheaper to make than a quad core AMD since its really just two allendale chips stuck together, so Intel can just glue two of them together instead of trying to make quad core chips and having crappy yields. This about AMD quad cores, one big die on a new 65nm process, you would expect yields to be terrible given those conditions. 65nm isnt enough for quad core, you need mature yields on 65nm to keep margins up.

 

[Viditor]

Originally posted by: bunnyfubbles
Once they're mainstream I wouldn't be surprised if the cheapest quadcores are $400-600. Usually "reasonably" priced means doubling the price of half the cores before them

That's possible, but I don't think we'll see mainstream desktop quad cores for at least 1-2 years...these will most likely be for people whose budgets allow them to buy the EE chips only.

 

[Viditor]

Originally posted by: BrownTown
Kentsfield will replace the Conroe XE, so that means a minimum of 1000$ for it, but its the same as getting a 2 processor board in one package, so its not as insane as it sounds. Also it should be cheaper to make than a quad core AMD since its really just two allendale chips stuck together, so Intel can just glue two of them together instead of trying to make quad core chips and having crappy yields. This about AMD quad cores, one big die on a new 65nm process, you would expect yields to be terrible given those conditions. 65nm isnt enough for quad core, you need mature yields on 65nm to keep margins up.

AMD is starting their 65nm ramp at mature yields (they announced this at the conference call), that's another reason why they are delaying the release. They had production chips coming out in January, but they are holding back on volume until the APM software is done tweaking the production process to the highest yield...

 

[BrownTown]

well, of course they are gonna say that yields are great and all, but we will never see any numbers, so impossible to tell if its true or not. Certainly it is unlikely that they will not tweak the process slightly as volume production begins to try to improve yields. But even if it is they still have higher costs than Intel due to a single die being employed. Also, if we are to assume that Intel is competant then it is likely they will have at least as good if not better on their process which will be a year old when AMD starts volume production. Don't forget that by the time AMD has transitioned to 65nm Intel will be starting 45nm production and AMD will again be making quad cores that have die sizes four times what the two dies in Intels will be.

 

[hans007]

Originally posted by: Viditor

Originally posted by: BrownTown
Kentsfield will replace the Conroe XE, so that means a minimum of 1000$ for it, but its the same as getting a 2 processor board in one package, so its not as insane as it sounds. Also it should be cheaper to make than a quad core AMD since its really just two allendale chips stuck together, so Intel can just glue two of them together instead of trying to make quad core chips and having crappy yields. This about AMD quad cores, one big die on a new 65nm process, you would expect yields to be terrible given those conditions. 65nm isnt enough for quad core, you need mature yields on 65nm to keep margins up.

AMD is starting their 65nm ramp at mature yields (they announced this at the conference call), that's another reason why they are delaying the release. They had production chips coming out in January, but they are holding back on volume until the APM software is done tweaking the production process to the highest yield...

the general theme for amd regarding new fab process ramps is this. when they are doing great and actually making chips with high yields they release them, and have a huge press release, have lots of demo chips etc.

when they are not doing well, generally they tell everyone it is great, then push back the launch dates from say q1 to h1 and junk like that. as a long time follower of their stock and sometimes holder of it, that is well what i've noticed oh.. sicne about 1999.

i am not even sure they have demoed a chip honestly at 65nm to the public and if they had it working they would be doing it all day long. that is what they normally have done.

i mean i think they could have gotten a little secretive maybe w ho knows. but honestly i think this ramp is not going that great right now.

concievably if intel sells quad core as just a dual die package, it wouldnt really cost them any more to make those chips than 2 single cores.

of course they'll never charge just 2x as much im thinking. even when amd first came out with dual core, a dual core cost more than twice what 2 singles cost.

if amd releases a quad core, for yield purposes they probably should do it with a dual die like intel does as with 4 cores on one die, the rate of defects for the entire die would be pretty high. i suppose they could just turn off one core, and sell ike a 3 core or something sort of like what ati does with bad gpus.

a quad core on 90nm would be not economically feasable for amd ona single die. most speculation says that the ddr2 controller (and it being quad ported) on the am2 chips will add about 20mm more die, so the X2 AM2 will be 220-225 . i think that is for a 1mb cache chip though.

lets just pretend they make a single piece quad core, with say 512k cache per cpu, it would still probably be around 350mm^2 die. pretty big. about the same as the r580 gpu from ati i think and pretty low yield / expensive to make.

 

[nyker96]

I think AMD has been preping quad for sometime now. Wouldn't be surprised if they release it soon after going 65nm. But the question is architecture will the next gen AMD stuff comparable to Intel's Conroe cores?

 

[BrownTown]

well, I don't know if AMD has explicitly said that their quad core will be one die, but thats what everyone assumes since they haven't used MCM like Intel did with Presler. Also many people assume that the single die is more powerfull due to cache coherency issues. However that is only a sure thing when all the cores are workign on the same data. Of course Intel is hurt by FSB which will have to carry the extra data. However I'm trying to remember if Kentsfiled uses DIB, i'm not sure. IT would help alot though, esp since a split bus will lower bus speeds.

 

[Duvie]

Originally posted by: BrownTown
well, of course they are gonna say that yields are great and all, but we will never see any numbers, so impossible to tell if its true or not. Certainly it is unlikely that they will not tweak the process slightly as volume production begins to try to improve yields. But even if it is they still have higher costs than Intel due to a single die being employed. Also, if we are to assume that Intel is competant then it is likely they will have at least as good if not better on their process which will be a year old when AMD starts volume production. Don't forget that by the time AMD has transitioned to 65nm Intel will be starting 45nm production and AMD will again be making quad cores that have die sizes four times what the two dies in Intels will be.

Well the past has shown us that was not the case with INtels 90nm....Intels implementation of that was horrific....By the time AMD came out with theirs, Intels 90nm was still horrific and not much better. ...AMD took their time and came out with a 90nm that was good the first time...

Being first does not mean better....

If AMD takes their time to have another near flawless implementation then that is fine....

 

[BrownTown]

Well, Intels 65nm process seems to being going along just fine from what I've seen. Obviously we don't get to see the yield numbers so we can't be sure, but things seem to be advancing according to plan. Upcoming price cuts and dual core clockspeed increases point towards a mature 65nm process that is providing better clockspeeds and less power for less cost. What will be more interesting is the generation after 45nm where things are projected to change signifigantly.

 

[Viditor]

Originally posted by: BrownTown
well, of course they are gonna say that yields are great and all, but we will never see any numbers, so impossible to tell if its true or not. Certainly it is unlikely that they will not tweak the process slightly as volume production begins to try to improve yields. But even if it is they still have higher costs than Intel due to a single die being employed. Also, if we are to assume that Intel is competant then it is likely they will have at least as good if not better on their process which will be a year old when AMD starts volume production. Don't forget that by the time AMD has transitioned to 65nm Intel will be starting 45nm production and AMD will again be making quad cores that have die sizes four times what the two dies in Intels will be.

Actually, I would bet long odds that AMD is able to ramp much faster than Intel every single time...and with better yields. The reasons are:

1. APM - this is the software that AMD uses in their Fabs to tweak the process (and no, Intel doesn't have anything like this). What it does is analyse data on yields and binning for every single wafer dynamically while every process is occuring. It can change a process mid-step and can tweak even small portions of each wafer differently. The scary part for the Fab guys is that they have to keep their hands away from the controls and let the computer take over the manufacturing process completely! It's success can be seen in the fact that operating profits for the chip group have been increasing steadily for a long while now.

2. Copy Everything - This is Intel's manufacturing systemology...it makes sense because of their size, but it also makes last-minute tweaks near impossible. The system works by finding the best manufacturing methodology at one Fab, then copying that to all the Fabs. It sounds great and it is less expensive to implement, but you can see how it would be far less dynamic and responsive than an Automated Production Manufacturing system.

APM is AMD's greatest and most innovative design in their history...it got them the Fab of the Year award, and is the reason Chartered, IBM, and Motorola all did such favourable deals with them (they received the process as part of the package). In fact, the 2005 Fab of the Year is the IBM East Fishkill Fab, which AMD helped to modify and where they are working on the 45nm and 32nm process together.

 

[Viditor]

Originally posted by: BrownTown
Well, Intels 65nm process seems to being going along just fine from what I've seen. Obviously we don't get to see the yield numbers so we can't be sure, but things seem to be advancing according to plan. Upcoming price cuts and dual core clockspeed increases point towards a mature 65nm process that is providing better clockspeeds and less power for less cost. What will be more interesting is the generation after 45nm where things are projected to change signifigantly.

We really don't know (at least I don't) how well the 65nm process is going yet. Intel's sales have been somewhat less than stellar (to say the least, remember that they have given an earnings warning already), so it's hard to tell from general data. There have been a few articles in the financials that Intel's inventory is growing more than expected, but that could be from lack of sales rather than a surplus of parts.