Morning blandy,
stay warm with RMBS.
This is from Ace's Hardware, one of the most Pro-AMD anti-Rambus sites on the web:
aceshardware.com
"DDR SDRAM (266 MHz) can not offer twice as much bandwidth as SDRAM (133 MHz) because data is sent at twice the rate, but the addresses still arrive at 133 MHz. The dual rambus channel offers twice (!) as much bandwidth as PC2100 DDR SDRAM (AMD760 chipset) while the paper specs indicate a gap of only 50%."
This is EXACTLY what some of us have been saying for almost a year, and we have been vindicated. The Pentium-4 was designed from the ground up to use RDRAM. Instead of crippling RDRAM with a slow front-side/main memory bus, the P-4 architecture unleashes it.
Where is the so-called "latency" of RDRAM on this platform? It's non-existant. The P-4 destroys the competition in any test of pure memory latency and bandwidth. DDR looks crippled, for precisely the reason that some of US have been pointing out for months--the data lines may be double-pumped, but the ADRESS LINES are NOT.
Pentium-4 opens with a lead that is provided amost entirely by it's platform and RDRAM. As SSE2 is added and this incredible chip scales toward 4GHz over the coming 24 months, that gap is only going to get wider. I predict that AMD will announce it's first RDRAM chipset in July, and ship it's first RDRAM chipset in October 2001.
GAME OVER. RAMBUS WON. Infineon, Micron, and Hyundai had better settle very, very soon, or they will be out of business. SDRAM is seling in the spot market right now for $3.5 X 16Mb--that is BELOW COST for most producers. If they aren't making RDRAM, or if they face treble damages for patent infringement, they will shutting down their fabs and turing their keys over to a little company from Mountain View, CA.
volts |
