Dave:
ITC conference sysnopsis.
Attended ITC in Atlantic City and Pat Gelsinger was keynote speaker. He stated by welcoming himself back to ITC after absence of thirteen years. He explained that in 1987 he presented paper on improving yields on 386 microprocessors. The frequency of the original 386 was 16 mHz but stated that it could have been 12 MHz if yields weren't improved. That was his segueway into the standard Intel stump speech about the Internet driving the future of computers sales.
Highlights were 100 million computer sold in 1999 rising to 180 millions by 2002. Stated that total connected users will by 400 million. He put up an eye opening slide that stated that current server capacity is 4% of what is needed by the internet by 2005.
He got into performance issues; put of slide stating increases in CPU speed, FSB speed, AGP bandwidth, disk access bandwidth and memory bandwidth. He stated that doubling of each subsystem would be in the order of 12-24 months for each of the subsystems. (I was happily relieved to hear that memory bandwidth was doubling every 18-24 months. No other memory that is available or been postulated can meet that forecast except for RDRAM.)
He got into cost issues of computer equipment with slide of average selling price of retail computer for home use was under $900 and dropping at 7% per year. Computer sold through distribution for business use was $1500 and dropping in price just as fast. Focus was that as cost of computers keeps coming down so does the need for costs to test computer systems to decrease. Showed slide with two lines: capital costs for semiconductor manufacturing vs time -- aggressive downwards slope -- second line ATE costs vs time -- flat line indicating no reduction in cost. He predicted that the two costs would be equal in 2005. I never seen hundreds of engineers let out such audible groans. In the next sentence, he gave test equipment profession an ultimatum: Either reduce costs for testing dramatically or computer industry will find another cheaper method of testing their products than using hi speed ATE.
He specifically mentioned Kingston's utter dissatisfaction with ATE testing costs associated with their recent efforts with PC100 RAM. He stated they had previously used in house PC motherboard test solutions for all previous types of RAM up to and including PC66 cost effectively for many years. He used the term dollar per chip rather than pennies per chip. He stated that Kingston is migrating back to PC motherboard based test schemes for RDRAM, since 100% ATE testing is too costly and not 100% effective in insuring system reliability after insertion of RIMM's into motherboards. Don't we know!
In conclusion, he said nothing contradicted Intel's stated goals of implementing RDRAM from hi-end on down to entry level over next two years. In the course of the fortyfive minute speech, RDRAM or RIMMs were mentioned at least 5 or 6 times in a matter of factly manner in front of a audience of primarily of electronic engineering faculty.
I left after the speech to call my broker and bought a ton of RMBS @ 60 and I plan never to sell it again. The exposition opened at 10:30 and I entered shortly there after. It happened that Teradyne was near the door. I looked for their big assed RDRAM testor and sitting in chairs in front of it were people from Viking Components. Talked about RDRAM with Teradyne rep for about 1-2 minutes, (not too much to discuss between two RMBS shareholders LOL). I was introduced to two RMBS employees at show who happened to be in booth. I proceeded to the Advantest booth. Sitting in front of their RDRAM testor were gentlemen from PNY. I picked up product sheet and moved on to Aglient (HP) booth -- now that's a stock I want to acquire a lot of when it trades for first time in NOV; like like stocks that end in "ent" like in Lucent. Standing in front of this humoungous machine which had oversized waffle iron at one end (sorry I call anything I can't understand it's purpose a machine; I get it from Sister Gertrude from Pottstown PA who in sixth grade referred to AV television set as machine). The product engineer, who had a poor command of English language, stated that it was their newest 32 at a time RDRAM testor. I asked him if they have been selling a lot of them. He stated that they hadn't sold any yet but quickly recovered from his slip and said that many of previous units were sold that tested 16 RDRAM at a time and that this was an easy upgrade to existing users. I tried to find the press relations person to get answer I could understand but he couldn't be found as he was wandering through show and wasn't scheduled to be in booth during the show. So I went back to the product engineer and asked whether that device was used primarily by RAM manufacturers or RIMM assemblers. He said that it was needed by both types of operations but declined to give me any customer names or additional details. I have return call into press relations person who is still probably wandering around exposition floor or maybe casino floor.
In the back of the exposition were the people from Schlumberger. They appeared to be very happy and it seems they had a right to be. Their booth was extremely crowded with people. In addition, they didn't have any of their big assed RDRAM test equipment in their exhibit. It seems that if you already sold that equipment to 3 of the 5 largest RAM manufacturers, you don't have to lug it to the trade show. At their booth, I met more RMBS employees, private RMBS investors and Schlumberger employees who were RMBS investors so I had a fairly good and informative time while trying not to get "too_hyped_up" on RMBS. The 6-0-6 three RIMM memory problem was brought up, the fact that some manufacturers were ready to ship and a approx time line of 2 weeks to 2 months to 820 re-release. I was told to expect significant speed increases with CUmine microprocessors with RDRAM.
Still couldn't find anything about Carmel chipset. I found this existing article from Feb 99 IDF at INtel web site.
developer.intel.com
Please go to pg 18, they refer to about 250% fold increase in graphic performance to Carmel based workstation forecast for Sept 1999. More Intel hype; we should see soon.
john |
