Author: Gregor Overney
Date: 14:33:57 05/14/99
Go up one level in this thread
On May 14, 1999 at 15:58:20, Dann Corbit wrote: >>And how would you put 480 chips in >>one? >16 clusters of 30 to a board. > Yes, on a board but not on a chip. Boards with such a high chip count are very expensive. >>Lets make a some small calculation: DB (1997 version) runs at 200 >>Mnodes/sec (right?). It uses 480 chips and 32 (?) RS/6000 CPUs. Lets just assume >>that those 32 (?) RS/6000 CPUs are just sitting there to diplay an IBM logo. >>Well, 200,000,000 divided by 480 times 15 gives 6.25 Mnodes/sec (why 36?). >Why are you multiplying by 15? 200M / 480 = 416666.6666667 (about 1/2 million >NPS per chip is achieved in total.) Surely there are losses because of >communications, barking up the wrong tree, etc. > According to an earliar post, this DB chip is expected to be 15 times faster than a single chip that has been used in IBM's 1997 version of Deep Blue. I cannot confirm this but others might be able to. The delay of Merced is not very relevant. There is the next generation of 21264 (probably called 21364). There are other designs such as Elbrus E2K. - BTW., Elbrus has received a quite interesting press: "The microprocessor will function 3 to 5 times more quickly than Intel Merced while still running all legacy MS DOS and Windows software. Fabricated in a 0.18-micron process, the chip would run at 1.2GHz and deliver 135 SPECint95 and 350 SPECfp95" 135 SPECint95 would be roughly five times the speed of a currently sold 21264 running at 500 MHz. - Exciting new CPU's are waiting at the horizon making it very difficult for general-purpose designs (especially when those special designs are sold for too many $$$). Gregor
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.