I want pics! and code if you've got it and don't mind sharing so other's can try it out.
This idea started when I was over in Memory and someone said that there is no way you can use all of 2GB of memory..
My Original Response:
I run 2 Gig, and in the following Matlab script, anything over about a 900 by 900 mesh kills my machine (X1Steps and X2Steps in the script) - out of memory errors. This is a short thing I did for an Optimization course I am taking in graduate school- Civil Engineering / Structures.
Here's what you get out of that jumbled mess above: Really hard to see the function's optimal value.
Resulting plots (2) with 50 x 50 mesh:
High Resolution, 1500 x 1500 mesh:
Medium Resolution, 900 x 900 mesh:
You are easily able to see the global optimum in a high resolution graph- It's lost in the 50x50..
Memory Space and Performance are critical in some areas.. MATLAB can eat your lunch very quickly.
Processing Power is not the issue all of the time; as you can see with this script the only issue is available memory- the calculation is not very intensive. The script runs in about 3 seconds at 2000x2000, but it's a nightmare to graph, due to the memory requirements.
This idea started when I was over in Memory and someone said that there is no way you can use all of 2GB of memory..
My Original Response:
l337zax said:I don't see exactly why you'd need 4 gigs o' ram. Even if you're doing mathematical computation, 2 gigs would probably be more than enough. I really wouldn't use 4 gigs of ram unless I got into the server side and was running dual opterons.
If you're worried that much about math computation and memory resources, I'd honestly make the jump into a dual processor system, preferably opteron. They'll blow your single socket A64 3500+ out of the water as far as computational speeds go.
I run 2 Gig, and in the following Matlab script, anything over about a 900 by 900 mesh kills my machine (X1Steps and X2Steps in the script) - out of memory errors. This is a short thing I did for an Optimization course I am taking in graduate school- Civil Engineering / Structures.
Code:
% CVEN689.600 - Optimization
% Homework #2
% Problem 1
clear;
clc;
i = 1;
j = 1;
X1Steps = 900;
X2Steps = 900;
X1Min = -10;
X1Max = 10;
X2Min = -10;
X2Max = 10;
X1Increment = (X1Max - X1Min)/X1Steps;
X2Increment = (X2Max - X2Min)/X2Steps;
zValues = zeros(X1Steps,X2Steps);
for i = 1:1:X1Steps
X1i = i * X1Increment + X1Min;
for j = 1:1:X2Steps
X2i = j * X2Increment + X2Min;
zValues(i,j) = 0.5 - ((sin((X1i^2+X2i^2)^0.5)*sin((X1i^2+X2i^2)^0.5))-0.5)/((1+0.001*(X1i^2+X2i^2))^2);
end
end
% Plot 3D Surface Mesh from workspace
Here's what you get out of that jumbled mess above: Really hard to see the function's optimal value.
Resulting plots (2) with 50 x 50 mesh:
High Resolution, 1500 x 1500 mesh:
Medium Resolution, 900 x 900 mesh:
You are easily able to see the global optimum in a high resolution graph- It's lost in the 50x50..
Memory Space and Performance are critical in some areas.. MATLAB can eat your lunch very quickly.
Processing Power is not the issue all of the time; as you can see with this script the only issue is available memory- the calculation is not very intensive. The script runs in about 3 seconds at 2000x2000, but it's a nightmare to graph, due to the memory requirements.