Intel announces shipment of dual-core processors

[danrok]

From the Front Page:
http://www.aoaforums.com/frontpage/i...d=541&Itemid=2

Any good?

[chrisbard]

I've seen in Sorin's comp how Pentium D 820 works. Is a 90nm, 775, 2.8Ghz, 800FSB, 2x1mb L2 dual core EM64T, XD processor. I haven't had enought time to test it but is a hell of a comp...that is !!!

[cloasters]

A dual-core Xeon sounds multi espensivo. I'm guessing that it costs about 85% of what two single-core Xeons would. Then again, double Proc m'boards are just a tad more expensive than singles. Is Intel releasing a mainboard, oops, a Server mainboard to host this great step forward in marchitecture?

[cloasters]

So Intel is not selling the dual-core "P4" now? It's gotten lackluster reviews in relation to AMD's double-core efforts. Makes you wonder how good Intel's non-Xeon double whammies are.

[Gizmo]

Yes, they are still selling the dual-core P4. These are the dual-core Xeon chips. The P4s are desktop chips, and the Xeons are server chips. Historically, Intel's server family has always lagged behind the desktop family in terms of technology for 'stability' reasons. Or did I misunderstand the question?

[keithwalton]

the 800 series p4's (dual core) are selling well because they are competitively priced. unlike the x2 which in some cases costs more than two chips of the same speed .. intel are also moving towards double core rather than dual core with the 65nm presler cpu which are allready up and running alot cooler than smithfields and will be shipping late this year early next. double core as its being dubbed is different to dual core in that double core is any two fully working dies put on one package -> cheaper than dual core which is one piece of silicon containing two physical cpu's.
amd better have some things up there sleeves as its appearing that 65nm for intel isnt proving to be as much of a headache as 90nm was.

As for the new xeons, they run on the same boards as the noccona's but may require a larger mounting hole for heavier heatsinks.

oh and these chips are by no means the first to dual cores to ship by intel. the smithfields (dual prescott) have been shipping for months now, they were on the shelves before amd made there anouncement

[cloasters]

Thanks for the information, keithwalton! Does this mean that one of the new dual Xeons will run on a two CPU Server m'board if a "dummy" is placed in the spare socket? I'm just asking.

[Gizmo]

That's a good question, too, cloasters. I would think that it would, which would potentially make the two CPU server board into a four CPU jobby. Which would be kinda cool.

[Gizmo]

Originally Posted by keithwalton

the 800 series p4's (dual core) are selling well because they are competitively priced. unlike the x2 which in some cases costs more than two chips of the same speed .. intel are also moving towards double core rather than dual core with the 65nm presler cpu which are allready up and running alot cooler than smithfields and will be shipping late this year early next. double core as its being dubbed is different to dual core in that double core is any two fully working dies put on one package -> cheaper than dual core which is one piece of silicon containing two physical cpu's.
amd better have some things up there sleeves as its appearing that 65nm for intel isnt proving to be as much of a headache as 90nm was.

As for the new xeons, they run on the same boards as the noccona's but may require a larger mounting hole for heavier heatsinks.

oh and these chips are by no means the first to dual cores to ship by intel. the smithfields (dual prescott) have been shipping for months now, they were on the shelves before amd made there anouncement

Hate to disagree with you Keith, but I think you've been smokin' too much of Intel's PR machine.

First, if memory serves, AMD had the dual core Opterons out before Intel had dual core anything, although Intel did get the dual core desktop chips out before AMD.

Second, saying that the new xeons run on the same board but may require larger mounting holes for the heatsink is like saying saying that my Mercury Sable 3.0 Liter engine can be replaced by a larger 3.8 Liter engine by just pulling the old one out and putting the new one in. While technically accurate, it is misleading, because the fact is that it is more practical to simply buy a car with the larger engine if in fact you need the larger engine.

Third double core is NOT cheaper to manufacture than dual core, unless your yeilds are sucking swamp water. Think about it for a minute. With dual core, you are dealing with two chips, cut from the same wafer in one operation side by side and mounted on the same die. Because of their positions on the wafer, they are already guaranteed to be as closely matched parametrically as it is going to be possible to get; you could test every chip in the wafer and almost certainly not come up with a better matched pair of chips. With double core, you cut every chip up individually, TEST every chip individually (well, ok, you have to do that anyway), MOUNT every chip individually, then try to MATCH chips with similar parametric characteristics and then mount those on a multichip module. Explain how that is cheaper?! It only works out if your process yeild sucketh so badly that you can't stand a reasonable chance of having two side-by-side chips on the same wafer actually work, or so it would seem to me. The only thing that is saving Intel's bacon out of this whole deal is that it's able to get a lot more chips per wafer at 65nM than at 90nM, so that even with a low yield, their manufacturing costs are probably still lower than they were with Prescott.

Fourth, these chips are still pulling MASSIVE amounts of power. Toms measured the power consumption at 252 watts peak load. While that is a definite improvement over two Prescotts, that is still a long ways from what AMD is getting with their dual core chips, and this is with Intel doing some pretty nifty stuff with power management.


Now, having said all that, Intel is definitely making steps in the right direction, and it will be interesting to see what their next gen effort looks like. But right now, we need to be realistic and call this for what it is; nothing more than a stop-gap measure by Intel to at least be able to talk about dual core CPUs and be taken halfway seriously without breaking the bank in the process.

[cloasters]

252 Watts at peak load? Dot's a lot of Watts. Methinks that one heck of a HSF will be needed. Overclocking these puppies will necessitate mondo liquid cooling. It's rare to see "help, I flooded my box!" posts anymore. All the same, I'm not enthused about H20 inside my cases.

[keithwalton]

the 252 watts is the total power used by there test system not just the cpu and this is actually for the smithfield core which had a tdp of 130w i seem to remember. The presler setup used 214w under full load, thats quite a reduction in power, by rough estimation of a guage i have on my psu my main rig uses 175w (2.4 n/w @ 3g + ati 9800xt) it was a close things on the dual core race i remember that. What i do remember was avaliability and the fact both did as good as paper launches.

and yes it would only be cheaper on a yield issue thing at present if anything is wrong with either die the whole thing has to be scrapped. the double core route will allow them to pair die's that are next to each other the same as before but if one of them doesnt work properly then they'll just use the next one along and either bin the dud die or speed grade / cache grade it as a celeron (single core chip)
I think they're trying to make both the single and dual core chips the same which makes sense on a mass production side of things as they can then cater for which way the demands swing for single / double core.

- Regarding the power thing the single core prescott on there test rig used 200w. the new 65nm core (ceder mill) used 166w and has a tdp of 86w they both run at 3.4GHz, where as the duals ran at 3GHz.

Oh and i do have a problem with this page from amd which details the x2
http://www.amd.com/us-en/Processors/..._13041,00.html

they claim "Only AMD64 technology runs today’s 32-bit and tomorrow’s 64-bit software"
err i dont think so at time of launch of the x2 there was plenty of intel chips around with em64t which does 32/64bit as well. maybe not as good but it does it

[Gizmo]

Originally Posted by keithwalton

the 252 watts is the total power used by there test system not just the cpu and this is actually for the smithfield core which had a tdp of 130w i seem to remember.

You are absolutely right; I misread that paragraph.

[Daniel ~]

Originally Posted by cloasters

252 Watts at peak load? Dot's a lot of Watts. Methinks that one heck of a HSF will be needed. Overclocking these puppies will necessitate mondo liquid cooling. It's rare to see "help, I flooded my box!" posts anymore. All the same, I'm not enthused about H20 inside my cases.

Time to let it go George! I've run H20 the whole time you've known me, nary a drop did slip!

[cloasters]

Well, err, umm. Yeah. I admit it, I'm afraid of water. Jest don't seem right inside a PC case. What? So I'm an old fogey. Maybe I'll get sick of screaming fans, they do get old after five or so years.

[Daniel ~]

If I were a lifeguard, you would be my kind of a swimmer! ":O}

[cloasters]

I'm not afraid to swim, just leery of mixing say $800 worth of components with water. Talk about a possible short circuit..