Van egy
nagyon érdekes thread az RWT-n. Egy Inteles fejlesztő fickó (bizonyos Victor) beszél benne az Intelen belüli történésekről. Rávilágít egy csomó mindenre és megerősít jó néhány máshonnan származó információt (pl. a Bob Colwelltől származó infókat). Néhány részlet (kiemelések tőlem):
Idézet
Intel did well because it has a lot of good engineering talent that can made success happen despite the barriers of management. Intel grew because of the internet boom needed computers like a forest fire needs trees. And I'll give Barrett credit for one thing: he built lots of fabs (it's all he really understood well was how to build a fab, but then he was a fab guy first).
I still recall when Barrett (and Albert Yu) torpedoed the 64bit Willamette and put more resources on Merced, Intel still managed to push forward. When the FP on Willamette was scaled back so as not to compete with Merced the x86 teams pushed on. When Merced was discovered to be *bigger than the reticle* more resources poured in to fix it. When the “Year of the Itanium” (an internal pronouncement) passed with almost no revenue shipments the x86 teams plunged on.
When billions of company dollars were poured into becoming “an internet company” or a "networking company" buying up tiny money losing companies at premium prices the design teams soldiered on.
Pure friggin’ neglect of the CPU business by Barrett allowed clueless VP’s like Louis Burns kill the low frequency, wide issue, high performance Prescott in favor a deep pipelined high frequency part, “because we know how to sell frequency” (words Burns uttered in my presence at a Q&A).
Intel made billions but many billions more were missed, great products were killed, and marketing took over Intel. I’ll give Ottellini credit – he’s taken hold of the CPU side and made it a company priority again. We no longer hear that CPUs are “on the way out like memory was in the 80's.” We hear that great CPUs are important, and resources are being lined up behind winning designs. And he finally put a real technical guy (Pat Gelsinger) in charge of CPU development.
Elég érdekes, nem? Kicsit más a szóhasználat, mint amit máshol olvahattunk (noha például egyezik a Bobbal készült interjúban írtakkal). 64 bit Willamette-t megtorpedózták, ami egyértelműen politikára utal. A "Year of the Itanium" (ami 2001 volt egyébként) értékelhető bevétel nélkül alakult, egyértelműen kiderül, hogy ez nagy csalódás volt az Intel számára, azaz az Itanium bevételek mégsem úgy alakultak/alakulnak mint ahogy szerették volna. Érdekesek a Prescottról írtak, bár ez szerintem elírás, itt a P4-ről beszél szerintem nem a Prescottról. Késöbb is utal erre más hozzászólásokban.
Idézet
I am *not* one of those people who "hate" IA-64 and think it was all a big mistake. I am taking aim at Merced as a project not the whole product line. What I probably failed to convey was that I don't think Merced *had* to be a disaster. The failure of Merced was in how the design was led, not some fundamental flaw in the concept. Merced could have taped out on time if it had been managed competently. The failure of Merced was because of management choices and that chain of failure led directly to Albert Yu and Craig Barrett. Albert made most of the direct mistakes. Craig failed us because he played almost no role in CPU or chipset development.
I also don't want to claim Willamette was the greatest part ever. Hell we made fun of it and I refused to build a system with my free one. A lot of battles were fought over Willamette and those fights were often lost over what the part should look like (the battles over Prescott were really just an extension of that fight). Those battles with upper management are what drove out Randy Steck and Bob Colwell who you will recall “left” Intel at the same time. If you want to know what Willamette should have looked like ask Bob about the P6 Heavy the next time you see him at a conference.
...
Maybe the worst thing was Yu kept pulling resources off other projects to fuel Merced. I’m not saying it should have been cancelled, but other projects were paying the bills. All the software enabling resources went to Merced. EDA resources went to Merced. Validators? Merced. Mask Designers? Merced. I think it was summer of 99 before it finaly taped out. Depending on how you measured it was about 12-18 months late.
Craig Barrett, who rarely talks about CPUs at BUMs, who always talks about how we are transforming from microprocessors to his latest fad, finally makes a CPU the main thrust of the BUM. It was going to be the corporate goal for the year: “Enable Itanium”. The whole focus of the company was making it a success. Posters on the walls, stories on Circuit, speeches, the whole works. Itanium was it. Willamatte? Coppermine? It’s clear they felt it was the swan song of a dieing IA. Money was drying up, all consumers wanted were Celerons, high-end CPUs were dead. We were going to make Itanium a success.
A success? For who? We were desperately trying to fill x86 demand and making hand over fist doing it. Hey I was all for the idea of Itanium. We weren’t going to unseat Sun with x86 (this was before the bust), we needed something like IA-64 to do it. But here I was working crazy hours because of a hiring freeze to make a part which made the company billions, and more people were going to get sucked away for part that wasn’t going to make us a dime and had a horrible reputation in the market. It was clear Itanium was going to really start with McKinley. I guess Sun was “it” that year, so suddenly Barrett wanted to be Sun. Thank goodness he never wanted to be Netscape.
Röhejes, nem? Persze gondolom minden nagyvállalat ugyanilyen hibákat csinál garmadával (személyes tapasztalataim is alátámasztják), mindenesetre érdekes ilyen eseteket olvasni. Az alábbi hozzászólás már a "momo" nevű hozzászólótól származik, aki a gyártástechnológiai oldalon dolgozik.
Idézet
Well, I personally liked Northwood. I think it proved that process arguably defines the success of a design. :^) Ha ha, I know this will truly anger many chip architects. But, hey devices on silicon rules not paper
. By the way, CTM team has not been involved in the enhancements of Prescott, that may tell you something. A whole different story, huh ? Anyways, thanx to CRB, excellent team at D2 will not be involved with mainstream logic device development in the future.
I have too many stories about the bloody feud between PTD and CTM process teams. It is a sad, sad story. Kane and Able simply...
Tudja valaki mi az a PTD és CTM?
Szerkesztette: hvuk 2006. 02. 24. 16:03 -kor
Athlon 64 939 2 GHz Winchester mag, GigaByte NF4 K8N Ultra-9 (passzív chipset), GigaByte X800 (passzív), 2x512 KingMax HC 500 MHz, Hitachi 160 Gb, NEC 3520, Coolink BAT01VS (1040 ford.), Chieftec 420W, Chieftec ház, Samsung 193P TFT monitor
Súgó
A téma zárva.
meg hogy 4milliard.. es honnan vetted a 2.5milliardot 2005-re? nem ezt keressuk napok ota?(forras)
)
