Another noobie with lots of questions..
Posted: Thu Nov 10, 2005 1:42 pm
Hello all,
I do a LOT of simulation work involving trading in financial markets (I trade for myself) which results in runs taking as long as 3 days (yep three) on a water cooled dual Xeon 2.8 with 2 gig of memory and nothing else special. The system is 3 years old. The simulations deal with an artificial intelligence program that does thousands of runs of mostly floating point computations. The disk IO is miminmal and the video needs are minimal as well. The application is single threaded for the foreseable future - its a third party program.
I'm looking for a way to dramatically shorten the time these runs take and have begun realizing that just upgrading my processor(s) will not have a very dramatic effect - ie cut the times by 1/2 or greater. I say this purely on the basis that after researching this I've noticed that while my system will run Super PI 1m in 56 seconds that the best overclocking results can get that down to 20 seconds or 3 times quicker and that seems to be an extreme result.
Now to my questions:
1. Am I correct in my statement above about not seeing a dramatic increase in speed by just upgrading/buying the latest and greatest processor currently available?
2. What would the hottest avail system without any overclocking run Super PI 1m?
3. Is it feasible to create a reliable\stable system that could cut my run times substantially?
4. If the answer to #3 is yes then approximately (in terms of Super PI) what kind of increase would I be looking at and approx how much $$ would it take?
5. What processor manufacturer would be the most appropriate for this?
Thank you very much for any and all input on this!
wingnut
I do a LOT of simulation work involving trading in financial markets (I trade for myself) which results in runs taking as long as 3 days (yep three) on a water cooled dual Xeon 2.8 with 2 gig of memory and nothing else special. The system is 3 years old. The simulations deal with an artificial intelligence program that does thousands of runs of mostly floating point computations. The disk IO is miminmal and the video needs are minimal as well. The application is single threaded for the foreseable future - its a third party program.
I'm looking for a way to dramatically shorten the time these runs take and have begun realizing that just upgrading my processor(s) will not have a very dramatic effect - ie cut the times by 1/2 or greater. I say this purely on the basis that after researching this I've noticed that while my system will run Super PI 1m in 56 seconds that the best overclocking results can get that down to 20 seconds or 3 times quicker and that seems to be an extreme result.
Now to my questions:
1. Am I correct in my statement above about not seeing a dramatic increase in speed by just upgrading/buying the latest and greatest processor currently available?
2. What would the hottest avail system without any overclocking run Super PI 1m?
3. Is it feasible to create a reliable\stable system that could cut my run times substantially?
4. If the answer to #3 is yes then approximately (in terms of Super PI) what kind of increase would I be looking at and approx how much $$ would it take?
5. What processor manufacturer would be the most appropriate for this?
Thank you very much for any and all input on this!
wingnut