CPUs, how they are made and the way they work.

There are some things that I didn't find on the net but I am interested in.

One would be the CPU manufacturing process. I'm very curious about the way the CPU are made and, in particular, the way AMD or intel
rate the CPUs. How they came to the conclusion that a processor is 1400Mhz and not 1600Mhz? For example I belive that AMD makes 1400 Athlon CPUs but, because of the marketing reasons some of them are sold as 1200 or 1333 CPUs.

The other thing is the way a CPU works. Lets say in a 3D game. I noticed in some 3D games that even if someone changed the CPU from an Athlon 750 to an Athlon 1200Mhz the game would not run faster cause the videocard is the bottleneck, BUT, cause there is a but, the explosions and the secvences where many objects are moving the new 1200CPU will make a difference.

I thing this could be a more informative discution that the usual benches and scores that are floating everywhere on the net.

Do you have any links?

I know about the Athlon XPs and I've read about that Athlon 2000+ on tomshardware
In my example the main actors were some Athlon (thunderbird core) and the question was deeper: How came AMD to the conclusion the an Athlon (thunderbird) is a 1333 Mhz CPU and not 1400Mhz or 1200Mhz? They obviosly test the CPUs some way but how?
Or about AthlonXPs: Maybe, I say, just maybe all the good CPUs AMD makes are in fact AthlonXP 1900+ (1600Mhz) but, for marketing reason some of them are sold as Athlon XPs 1800+ or 1700+
As for the second answer, can you post a link about the subject?
Thank you for answering!

Good examples of paper launches could be P3 Copermine 1133 and P3 Katmai 600Mhz. When P3 Katmai was released, everybody could see that intel bumped the Mhz figure by good old "method" -> rising the CPU voltage a little. Interesting...
As for Durons, I think they are so overclockable cause AMD pulls the hand break. I mean, when AMD released Duron 700 there were higher clocked Athlon avaible...The only reason AMD didn't release, let's say, Duron 1000 at that time was not to geopatize the Athlon's sells (like Celeron A/Pentium2 case)...
THX, interestin reading, bbaeyens!

probably this is the reason why lower speed chips overclock well, almost upto the fastest processor in the cost range.

you said this yeild has reached upto 80% with 12 inch wafers, last time I took stock it even 50% was considered decent. And AMD Austin fab was suffering with 1% yeild with their K6 chips. 1% is ludicourously low (I cant imagine there were 100+ K6s even on a 6~8 inch wafer and only one worked!), might be a roumor.

I guess they recycle the silicon lost in these non-working die. And make a processor from every working die be it over the range of as much as 300 MHz or more.

girish

<font color=red>No system is fool-proof. Fools are Ingenious!</font color=red>

Ok a few things.


1% yield is highly unlikely, any fab making 1% yield is losing millions of dollars.


I never said amd was at 80% and they dont use 12 inch wafers, I did say, I had heard they were hitting 90% and they use 8 inch wafers(at dresden). Which for logic is VERY good.


finally, they cannot recycle the silicon, if it fails it is scrapped, no semiconductor company can recycle a finished die that failed.

HOWEVER.

Some companies buy defective die's (i have no idea why) but they usually pay like 1 cent a die, but since its trash anyways its a recoup for the company. I dont think intel or amd sell dead die like that.(would be a good way to let the competition know your yield issues) but dram and flash manufactueres do that kind of deal rather often.

"The Cash Left In My Pocket,The BEST Benchmark"
No Overclock+stock hsf=GOOD!