Computer deflation

[jasong]

(Couldn't come up with a better title)

As an obsessive compulsive math nut, I decided to enumerate my DC "problem."

First, you can get approximately twice as much power for the same amount of money approximately every 18 months. That means, overall, costs go down approximately 3.75% a month for the same output, but usually only if you stick to the "average" computer. In my opinion, if you buy significantly older, or top-of-the-line product(yes, I know they're different areas), you're paying more for less power.

Secondly, assuming I keep my job, I like to buy new hardware every 3-6 months. I'll wait if I think something good is about to come out, but basically I'm looking for maximum output for the next two years after whenever I'm looking at the situation.

The actual reduction average is 3.78%, so let's go ahead and overanalyze, and see what that gets us. The left column will be how many months have elapsed in a perfect, cost going down by half for some throughput every 18 months, world. The second column is the cost in that same world, or, actually a proportion of the very first cost. The third column is output times the cost percentage. The fourth column is the column which tells how much bang for buck we get at a point two years after month zero, which will tell us if waiting will get us more throughput for our buck in this imaginary world. I haven't bothered to factor in electricity(my dad pays for electricity and I probably won't be able to buy enough computers for him to complain.)

0 1 24 1
1 .9622 22.1306 1.0393
2 .9258 20.3676 1.0801
3 .8908 18.7068 1.1194
4 .8572 17.144 1.1666
5 .8248 15.6712 1.2124
6 .7936 14.2848 1.2601

So, in this perfect world, I'm better off waiting 6 months to spend the same amount of money if the goal is maximum crunching for a time 2 years in the future.

Comments? Flames? Psychiatrist recommendations? ;)

[tnerual]

if you take elicricity in account it's even better to wait up to 8 or 9 month ... but can you wait ?

the second part is that you can sell the first computer after 6 month then buy the new one for 6 month and so on ...

there is a lot of variable.

but you are true, low speed, last generation processor is the best bang for bucks now (conroe e6300 - e6400 and amd x2 3800+, 4200+)

[axn]

Don't forget the upcoming Core 2 Quad - with it, you could wait a whole year and still catch up in double quick time :idea:

@jasong: By your logic, you'll never buy anything. Every 6 months, you'll reevaluate your position, and then decide that to wait for another 6 months. :D

[Helix_Von_Smelix]

Hi AXN ..... first post:thumbs:
Will have to ask BoK if that is the longest time since joining to first post

jasong>>>>> you have to make a purchase on day one, or no output!!

[jasong]

Don't forget the upcoming Core 2 Quad - with it, you could wait a whole year and still catch up in double quick time :idea:

@jasong: By your logic, you'll never buy anything. Every 6 months, you'll reevaluate your position, and then decide that to wait for another 6 months. :D
I understand what you're saying, but I DID say I was committed to buying new hardware at least every 3-6 months.

Somewhere on the Internet, someone made a Java web page that helped you decide when to buy hardware if your goal was a certain amount of processing by a certain date. Quite fun to play with.

[jasong]

jasong>>>>> you have to make a purchase on day one, or no output!!
Even if you factor in that fact, there's still the possibility that the technology will improve enough to where you're better off waiting, even with a fixed date as a goal.

My math was based on a date 2 years in the future. I assumed technology would advance at a Moore's law rate, which meant that, in this case, waiting six months, even though that cut down the processing time, meant more processing when that date finally came. Of course, one of the problems I didn't factor in was the fact that computers have specific costs and the amount of money spent is normally limited. This would cause a bit of granularity in the equation.

But it's fun to do simple thought experiments, and in some cases(Einstein, for example) it can be very beneficial.

[Helix_Von_Smelix]

and you need to take into account that WU's DO get harder over time. SETI is a prime example