[RC5] d.net Pride (Ooops)

Eric Wikman eric at wikman.com
Mon Jul 27 16:14:27 EDT 1998


Sorry, I made an error in my calculation.  I will post my formula so everyone 
can be sure that it is basically correct.  Here is the revised stats (also 
including Bruce Fords new stats he just posted).  

(Daily block increase / One days blocks * Days till 2000 + 1) * Current 
TFLOPS = 

1)
(20638/8682196*523+1)*4421429 = 9.9 TFlops by 2000
(20638/8682196*1254+1)*4421429 = 17.6 by 2002
(20638/8682196*1984+1)*4421429 = 25.2 by 2004

2)
(31397/8682196*523+1)*4421429 = 12.7 by 2000
(31397/8682196*1254+1)*4421429 = 24.4 by 2002
(31397/8682196*1984+1)*4421429 = 36.1 by 2004

I would like to include the stats for the exponential scenario, but am not 
sure which formula to use.  Bruce Ford says the doubling time for method 
three is 168 days.  So:

Convert doubling time of 168 days to % increase per day
 2 = x ^ 168; x = 1.0041343992192 

Once I have that I do not know if I should use a similar formula above like:
((x-1)*523+1)*4421429 = 13.9    or:
x^523*4421429 = 38.2

3) 
13.9 = 2000; 27.3; 40.6 or:
38.2; 780; 15870

4) 2x of 126 days = 1.00551632731
12.7; 30.5; 48.3
78.5; 4380; 242991

If the first scenario is the case then I would like to include it in the email 
posts, but if the seconds scenario is correct and it is predicting to have 
242,991 TFlops by year 2004, then I find that a little ridiculous since IBM, 
Sun, Cray, and Digital will still be trying to put together a 100 TFlop 
system.  

I would appreciate it if someone would check my math, and please be 
polite when you let me know I am wrong :)  

As always, thanks Bruce for the stats.  

Eric Wikman
eric at wikman.com
--
To unsubscribe, send 'unsubscribe rc5' to majordomo at lists.distributed.net
rc5-digest subscribers replace rc5 with rc5-digest



More information about the rc5 mailing list