AM2 Predictions...

[Russ]

the san diego cores have been doing surprisingly well lately, with a lot of single core opterons breaking 3ghz, the kab3e stepping making an introduction, and some interesting results out of the dual cores. amd really needs to work on consistancy. the dies are just too damn big

y did you quote me? And how is it inconsistant to have big dies?

Just as the San Diego and Manchester E cores brought with them improvements, the F cores (AM2) will also bring improvements, it's not just about the DDR2

Better memory controller eh?

Actually Socket F is for servers. There were some pics on here a few months ago.

 

[(cf)Eclipse]

y did you quote me? And how is it inconsistant to have big dies?

because you were talking about oc'ing. i made a parallel point about how 90nm is still getting better.
and bigger die = larger area for defects.

 

[GilmourD]

Actually Socket F is for servers. There were some pics on here a few months ago.

He's talking about Rev.F, not Socket F.

 

[freeloader1969]

As quoted by (cf)Eclipse

because you were talking about oc'ing. i made a parallel point about how 90nm is still getting better and bigger die = larger area for defectsr

Eclipse, would this rule not also apply to 65nm and smaller? You'd think more errors would creep into the smaller process shrinks due to leakage. Isn't leakage the real battle that goes on in a die shrink?

 

[coldpower27]

As quoted by (cf)Eclipse



Eclipse, would this rule not also apply to 65nm and smaller? You'd think more errors would creep into the smaller process shrinks due to leakage. Isn't leakage the real battle that goes on in a die shrink?

Hmmm, from wah twe have witnessed, leakage wasn't much of an issue until 90nm where IBM and Intel both had problems at first.

AMD went slow and had SOI, so it isn't really fair to compare it. Not to mention the lower clockspeeds and stuff.

A larger die is more prone to defects, since it has a larger surface which must be perfect vs a smaller surface. It's just probability a larger person has a larger percentage chance of being hit then a smaller person.

 

[needmorecarnitine]

sounds a lot like DSL, which is already being used

Yes

Did you read the article?

 

[(cf)Eclipse]

no

A larger die is more prone to defects, since it has a larger surface which must be perfect vs a smaller surface. It's just probability a larger person has a larger percentage chance of being hit then a smaller person.

well said. it's the wafer itself that usually contains the defects, not that the defects are introduced by lithography process

also.. there are rumors that the leakage won't be too much worse at 65nm. intel seems to have done a good job, reducing temps a decent amount

 

[Russ]

He's talking about Rev.F, not Socket F.

oops! I think I just kind of imagined a "socket" infront of that F.

Is that new opty setting you mentioned actually better than the initial steppings? Seems pretty hard to beat 3 ghz w/ a minimal voltage increase (or whatever robberbaron ended up at).

90nm gets awesome overclocks, yes. But it seems kind of hard to believe that 65nm won't smoke it. It's not like it's from a new company or something, it's supposed to be an upgrade. And it's a huge difference in "size".

 

[stiltner]

Personally, a core drop is gonna be the big hit. Just a new socket doesn't mean much to me, sure DDR2 will be nice, but ultra low latency, high quality DDR is still keeping pace with the integrated mem controller.

It'll be interesting to see what the DDR2 future holds, I predict some breakthrough performance options from some mfg's will really make the chips shine like we haven't seen before.

Fact remains, things moving forward is always good for the consumer. It means a constantly shifting push downwards on pricing, and a demand for better, bigger, faster, and it all drives itself on a even level.

Other than that, the only thing worth chatting about is when software mfg'ers catch up to the dual core systems now. For the first time it seems in many years, they're playing catch up, as very few apps out there can multi-thread to CPU level, once that happens, you're really gonna see the benefit. Now you're seeing the ability to use 2 apps, single threaded w/o any performance degredation, its gonna take the shift by the software makers this time to catch up with the hardware.

 

[mwarps]

no


well said. it's the wafer itself that usually contains the defects, not that the defects are introduced by lithography process

also.. there are rumors that the leakage won't be too much worse at 65nm. intel seems to have done a good job, reducing temps a decent amount

There are defects in and on the wafer. Your assumption that there are more IN the wafer is correct due to the fact that we Put Them There so they stay away from the surface.

There are several processes that the wafers go through other than litho, which each introduce their own defects and problems. Litho is actually one of the more dirty processes, because it involves placing a coating on the wafer itself. You get a lot of kills from litho. Other nasties include metallization and ion implant. Nothing like getting nailed by an ion at a couple hundred KeV.

 

[perplex]

I think all the AM2's will be faster than their respective Intel comepetitiors, Intel is now in the Mac's and they're still 32bit. I just can't forsee Intel having a Dual Core 64bit competitior.

Allow me..