Larry - Re: "The reason is that when tested one against another, the AMD-equipped machines should outperform similarly
equipped Intel-based machines. This difference, along with price, will make Intel machines look less appealing
over time.
The crux of his argument - Outperform and Price.
All the testing to date shows that the AMD K6 is a credible device, but in no way outperforms a Pentium Pro and certainly not a Pentium II. In fact, many tests have shown the K6 performance can be matched by a Pentium MMX - and this was BEFORE Intel announced their 233 MHz Pentium MMM (Last week).
Today's DEC announcement concerning the K6 said that DEC would offer a K6 PC that was 5% less in price than a comparably equipped Intel machine. I don't know if they meant Pentium or Pentium II but it sure implied that DEC thought Intel machines were comparable to K6 but were only able to discount the K6 machines by 5%. Now - where is that 25% savings that Jerry Sanders promised everyone?
Regarding 6 layers for the K6 - well, they have a "local interconnect" level that allows polysilicon gates to be strapped/connected to source-drain diffusion. If you count that, the 5 metal layers plus local interconnect equals "6".
Today Compaq announced 300 MHz Pentium II workstations, and NEC has previewed 300 MHz machines. So, it looks like the bragging rights for full blown performance still go to Intel. In fact, AMD's K6 can rarely (if at ALL) be found in 233 MHz speed grades - and these devices have to be goosed up to 3.2 volts just to make them wiggle at that speed. With Intel actually DELIVERING 300 MHz Pentium II devices and "goosed-up" 233 MHz K6 devices nowhere to be found - well, Intel wins again.
Back to Socket 7/Slot 1 - Intel's 233 MHz Pentium still will command the Socket 7 high end - and I would suspect faster chips will be available when the 0.25 micron devices arrive. So, Intel will be selling Pentium MMX/Socket 7 devices head to head with AMD and Cyrix's M(e)2 chip, as well as coralling the REAL HIGH END Pentium II/Slot 1 workstation market.
Speaking of Slot 1 - I saw a second generation (only 1 month after the Pentium II was released) Pentium II/Slot 1 motherboard from CDC of Taiwan at Fry's (in San Jose) last week. This board was the smallest PC board I have yet seen - measuring about 6 1/2 in x 10 in. It contained 17 ICs including the 3 440FX chip set. It was smaller than the ATX boards by at least 25%. So, the overall cost is coming down quickly - this was selling for $329, $40 lower than the $369 board I saw several weeks ago.
As for the Z80 analogy - it was a better chip than the 8080A for TWO reasons - it ran at only 5 volts (8080A required 12V & 5 V) and the Z80 had a built in DRAM refresh controller. Designers did not have to worry about DRAM memory design - the Z80 did all the work for them. Believe me - that was a BIG selling point back then as DRAM meory system design was a BLACK art.
Back to the K6 - It is now two months after the K6 roll out and AMD has only Vobis, Fujitsu/ICL and DEC signed up for K6 usage, as well as some very small shops (CyberSMASH & Frogleap, Laguski Computers). With K6 availability a real question mark, what large user would commit to this chip when his production lines could be shut down by lack of K6 availability? The K6 is sole-sourced - one AMD Fab in Texas.
Intel's Pentium(s) are made in 6 or 7 fabs - in the US and Ireland. A stable, diversified production base exists , allowing Intel to have "mutiple" sources from several countries. This means a lot to companies like Compaq that are consuming Pentium devices at the rate of 1.8 MILLION per quarter.
Paul |
