Linked by Thom Holwerda on Wed 11th Oct 2006 20:13 UTC
AMD "AMD described its forthcoming quad-core processor, codenamed Barcelona, in a session at today's Microprocessor Forum. Details of the new microarchitecture on which the processor is based (codenamed K8L) have been known for some time now. Still, the event brought some new info, and here are some highlights that I've culled from some of the reporting on it."
Thread beginning with comment 171086
To view parent comment, click here.
To read all comments associated with this story, please click here.
RE: What is wrong with you people?
by RandomGuy on Thu 12th Oct 2006 13:15 UTC in reply to "What is wrong with you people?"
RandomGuy
Member since:
2006-07-30

Just play chess ;-P

Seriously, go is not that sort of game computers are really good at because the board is too large, creating too many possibilities.

Very rough estimate:
chess-board: 8x8=64
go-board: 19x19=361
Now the number of possibilities increases approximately exponential with time "t", meaning:
possibilities(chess)~64^t
possibilities(go)~361^t
Now chess opening theory is about 10 turns (there's more but real chess computers stop relying on theory from this point on).
10 turns means:
possibilities(chess)~64^10=10^18
possibilities(go)~361^10=4*10^25
7 orders of magnitude between them.
Moore's law predicts sufficient computing powers no sooner than 2040. If it remains true.

There are problems where brute force is a suitable strategy but go is probably not one of them.

But you might hope for quantum computing, evaluating all possible choices at the same time.
Now you only need a way to make the machine let you win from time to time :-D

On a side note, I find it really strange to put more and more cores into one machine and at the same time virtualize the machine so that it can be used by several OSs and people at the same time.
I know it makes sense but it feels _really_ strange...

I think I'll wait until quads are somehow mainstream and get a really cheap dual core then for ripping or compiling stuff in the background.

Edited 2006-10-12 13:16

Reply Parent Score: 1

bnolsen Member since:
2006-01-06

Some things to remember about virtualization:

Most operating systems need at most 20GB - 40GB at most for their OS drive. I'm hard pressed nowadays to find a drive below 250GB and really now wouldn't buy one less than 320GB.

So what you get with virtualization is the abilty to save space and power. Instead of 4 boxes with 4 cases, 4 power supplies, 4 raid-1 sets, etc, just reduce it to a single box with 2-4 cores and a single raid-1. Attach some amount of storage to it, logically split it and give to each virtual copy.

This is great for basic legacy Windows office infrastructure which is ridden with server software that refuses to play nicely with other server software (exchange, ms-sql, domain servicer, etc). Basically give each Windows server a virtual sandbox to play in, a sandbox backup being simply a copy of the server image.

Not the way I would suggesting running unix based infrastructure though (software seems to play more nicely with each other, virtualization only needed for security purposes).

Reply Parent Score: 1

smitty Member since:
2005-10-13

Your numbers aren't all that accurate, because in chess certain pieces can only move in certain ways, lowering the total number of possibilities even further. However, the general idea is correct - for as many possible moves a computer has to evaluate in chess, the corresponding number of Go moves is much, much larger. I believe the best Go programs running on the fastest computers are about equivalent to an intermediate player at this time, and it will be a long time before they become as good as an expert.

Edited 2006-10-12 15:25

Reply Parent Score: 1

tuttle Member since:
2006-03-01

RandomGuy: I know about the complexity of go. I am really a lousy go player, but even I can beat go programs when I concentrate. A friend of mine is a very good go player (dan).

A few years ago we tried writing our own go program. As you mentioned, using brute force is quite hopeless. So we tried having a very low search depth and a very advanced position evaluation.

We got it to play good strategically (e.g. making decent opening moves without an opening library, good liveness estimation). But we could never get it to play good combinations.

Reply Parent Score: 1