Low Yield != slow working chip. You are wrong. Unstable process causes slow binning parts. In the semiconductor business you dont make chips on purpose to be a given speed. You make it, you test it and see where it falls. Certain thicknesses of Poly, Nit, copper/aluminum often determin the actual performance of a chip.
Perhaps you simply misunderstood me when I said "Yield." I'm hardly wrong. Since AMD has made neither head nor tails of the slow clockspeed, yield <i>just may be their problem</i>. I'm not saying that yield is a direct cause of slow chips. That's the silliest thing I've ever heard; we both know that yield is the number of properly-functioning chips from a wafer. You can drop the angry++ high-and-mighty "OMG I WORK IN THE SEMICONDUCTOR INDUSTRY SO I CLEARLY KNOW MORE THAN YOU" act you've learned to love lately.
I was simply implying that AMD may, or probably is, having trouble with their process, and as a result cannot produce chips over 2GHz, or is having trouble doing so economically.
Sorry but I do know that our yield on DDR3 is not quite what it is on DD2 but it still performs quite well (both are the fastest chips on the market). Just an example of good performance/process (should be quite similar to DDR2) but crappy yield compared to a mature part.
0
LeonardoWake up and smell the glaciersEagle River, AlaskaIcrontian
edited October 2007
It has been surmised all over the 'tech press' that AMD is indeed having problems with yield and the SOI process itself. I have not seen AMD make any statement to counter those assertions. AMD desperately needs a miracle or they are consigned the middle and low performance market for years to come. With the new 45nm Intel quads coming out very soon (already being benched at reputable sites), AMD can't even claim superiority in memory efficiency. This is really frustrating.
Yeah, looks like AMD will be sucking hind tit for the foreseeable future. And if the 45nm Intel parts clock anything like some of the benches we are seeing on the ES chips on the net right now, they will have headroom to spare whenever AMD can ramp clock speeds up to close to parity.
0
LeonardoWake up and smell the glaciersEagle River, AlaskaIcrontian
edited October 2007
4.27GHz already achieved on a sample, and it wasn't even extreme cooling. AMD is hurting for sure. Intel production CPUs usually perform at least on par with the engineering samples. In six months, out-of-the-box quads will be 4.0GHz-capable with air cooling that is available right now. The motherboards to do it on will be the same price as competent boards now - as low as $89.
The new 45nm intel parts looks promising, but I remember many of us had high expectations for an easy 4GHz OC when the first 65nm dual core C2D's were launched (judging by the early ES test reports). But in reality, the retail parts just barely made it to 3.6GHz with great cooling in most cases. But judging by how well the current 65nm G0 kentsfield OCs gives me hope
0
LeonardoWake up and smell the glaciersEagle River, AlaskaIcrontian
edited October 2007
many of us had high expectations for an easy 4GHz OC when the first 65nm dual core C2D's were launched
The first generation of 65nm dual core, Presler, was an easy overclock to 4.0GHz. It's easy to overlook that C2D was Intel's third series of dual core-- Smithfield (Netburst), Presler (Netburst), and Core 2 Duo (Core 2).
Perhaps you simply misunderstood me when I said "Yield." I'm hardly wrong. Since AMD has made neither head nor tails of the slow clockspeed, yield <i>just may be their problem</i>. I'm not saying that yield is a direct cause of slow chips. That's the silliest thing I've ever heard; we both know that yield is the number of properly-functioning chips from a wafer. You can drop the angry++ high-and-mighty "OMG I WORK IN THE SEMICONDUCTOR INDUSTRY SO I CLEARLY KNOW MORE THAN YOU" act you've learned to love lately.
I was simply implying that AMD may, or probably is, having trouble with their process, and as a result cannot produce chips over 2GHz, or is having trouble doing so economically.
I'm just curious how much the overall jump in proc speed (and clocking) this makes? I understand that there are more and more multi-core ready applications available on the market, but just how much of a performance increase will you see with a Quad Core clocked at 4ghz over a 2ghz overclock? Are we at the point where it will make a difference for anyone but the extreme number cruncher (accountants and science, here's looking at you)? Will this even matter to someone who's not a huge overclocker, but just wants a little performance boost?
Intels 45nm is leaps and bounds ahead of AMD or even Intels own 65nm process. Expect that as 45nm processes mature for speeds to really crank up with easy voltage, much less than before without the heat going sky high.
Check out this article on how Intel is fixing much of the Gate leakage that has really come about at the end of the Netburst reign and even with C2D. Its a radicle change in how transistors are made. http://www.spectrum.ieee.org/print/5553
For those that don't know how a basic transistor works with Source, Drain and Gate it may be a lil hard to grasp but this is a huge step for semiconductors. A must read for enthusiasts I think.
Comments
AMD doesn't know how to get it over 2 right now.
OR
The amount of successful chips >2 GHz from a wafer is too low to be profitable.
I was simply implying that AMD may, or probably is, having trouble with their process, and as a result cannot produce chips over 2GHz, or is having trouble doing so economically.
Play nice Thrax.
Check out this article on how Intel is fixing much of the Gate leakage that has really come about at the end of the Netburst reign and even with C2D. Its a radicle change in how transistors are made.
http://www.spectrum.ieee.org/print/5553
For those that don't know how a basic transistor works with Source, Drain and Gate it may be a lil hard to grasp but this is a huge step for semiconductors. A must read for enthusiasts I think.