Intel, AMD quality control...

[dcdomain]

What's going on here? Intel has all sorts of problems with its products, now AMD? Check out this link: http://www.theregister.co.uk/content/2/15211.html Either both companies are really pushing the limits, or they are just simply turning out products to quickly to try to maintain competitiveness and not testing them thoroughly enough... feel sorry for everyone that ordered the Microns.

 

[Noriaki]

Well the register is hardly a reliable source, but it does make me glad I decided to stay SDRAM for a while yet.

 

[Adul]

I think micron try to jump the gun a bit fast on this one. Then again, they are selling DDR SDRAM.

 

[Dulanic]

Again false info from The Register..... The problem is with the BRAND on motherboard Micron uses specifically NOT the chipset.

 

[pm]

It's a delay to an unreleased product. I'd hardly call that a problem with quality control.

If you start shipping it and then you have problems, that's a problem with quality control. To me, if you have a problem with an unreleased product, then this is a forecasting problem not a quality control issue... the quality control guys are clearly doing their job fine finding bugs (or "erratumnotbugs" in this case). The forecasting guys messed up in guessing that there wouldn't be any issues.

 

[Noriaki]

Patrick has a good point...still I'm glad I wasn't waiting to jump on the DDR bandwagon.

 

[dcdomain]

Hmmm... true true... but I remember when I first got into computers, there weren't that many problems as there is now. Maybe it's because I read up on this stuff more...

I'm glad I stuck with SDRAM too!

 

[pm]

No, I really don't think that there were fewer problems back in "good old days" - at least that's not the impression that I get. The problems were less visible back then to the end-user, but they certainly existed. I hear engineers tell me problems that they had when releasing various "old" CPU's at Intel and I'm fairly certain that things have not gotten worse with time. The "I remember back when we shippped processor XXX" stories seem to indicate that things actually used to be much worse at times in some ways.

I think people just paid less attention to these problems back then. Certainly I don't think people who were outside of the IT culture were paying attention to whether internal preproduction ship dates slipped by a little bit. I don't think most people even knew who manufactured their CPUs back in the 286 and 386 days.

 

[Midnight Rambler]

To add another point, anytime you increase the complexity of a design you invariably also add the potential for more problems. A car assembled from 1700 parts, vs. from 10,000+ parts, can be held in process control much easier/tighter. I would imagine the same holds true for semiconductors as well, so comparing the days of when a chip held a few million (if that) transistors to today when they are comprised of tens of millions is not very valid.

 

[pm]

The designs are more complex, but there's a lot more automation nowadays.

I talk to the "old-timers" and they tell me about how they used to make chip masks
using the equivalent of construction paper, x-acto knives and rulers. They used to check for problems by spreading the mask out layer by layer on the floor of a really big room and would check for design violations on their hands and knees with a ruler. Since there would be a dozen or so layers, this took a long time, was tedious and was error prone because it was tedious. Now all of this is completely automated. The mask is generated by a CAD program and it is checked by another CAD program and the numbers of errors that slip through are very small if there are any at all. There are way more transistors and wires, but there are way better software tools for designing them.

Besides, cars may be more complex nowadays, and there are more parts in them than there used to be and there's more to go wrong, but they last longer (overall life expectancy), they fail less frequently (mean time to failure), they have better fuel economy, better power and less emissions compared to older (and simpler) cars made in the 60's and 70's.

I really think this is a matter of perspective. It was easier to hide problems back in the "good old days" (ie. 1985 or so) because no one outside of the IT depts. at a few big companies were really paying attention. Certainly if the internal ship date on a chipset slipped by a quarter no one outside of an astute BYTE magazine reader would notice. Nowadays this is more visible - people here know more about what the internal roadmaps are, and thus are actually going to notice when that schedule departs from reality. Just because you are aware of it now doesn't mean that it didn't happen in the days before confidential roadmaps were posted at public websites.