1 |
On Wednesday 08 November 2006 19:00, Mark Knecht wrote: |
2 |
|
3 |
> The problem with running the neural network app is that it's a huge |
4 |
> install under Windows. It requires Internet access as it has a |
5 |
> hardware key that has to be validated against the specific machine. |
6 |
> Probably takes 1 hour just to set up. Then, once it's set up it takes |
7 |
> maybe 15 minutes to run a single solution on my older Athlon XP |
8 |
> 1600+. With that as background I'm sure you can understand that I'm |
9 |
> not anxious to do it more than once or twice. |
10 |
|
11 |
OK, scrap that idea |
12 |
|
13 |
> What I was hoping to do was find some basic way of comparing the |
14 |
> BogoMIPS on my old Athlon XP machine with BogoMIPS on some new |
15 |
> machines at the dealer. They haven't had any problems in the past |
16 |
> with me bringing in a LiveCD and booting Linux. If I could do this |
17 |
> then I might estimate that the new machine will run the same speed or |
18 |
> will run 3X the speed when doing these neural network jobs? |
19 |
|
20 |
For this purpose bogomips is meaningless - it's simply a measure of how |
21 |
fast the cpu can execute a very specific and very tight loop and is |
22 |
used for some timing setting or other during kernel initialization. |
23 |
Bears almost no resemblance to real life operation, other than it's |
24 |
safe to say that bogomips goes up as cpu freq goes up |
25 |
|
26 |
[snip] |
27 |
|
28 |
> 5) The current Athlon XP Windows machine is busy running Trading |
29 |
> Solutions but when in Linux I *think* it has a BogoMIPS spec around |
30 |
> 2800. No way to verify that right now. |
31 |
> |
32 |
> It's pretty boring but it seems that you can sort of double the CPU |
33 |
> MHz spec and come pretty close to the BogoMIPS numbers. However that |
34 |
> doesn't take cache size into account so maybe BogoMIPS isn't even the |
35 |
> right thing to be looking at. |
36 |
|
37 |
bogomips is usually about double the cpu speed, but you can't count on |
38 |
that. I would imagine that cache size and fpu performance were |
39 |
significant factors. |
40 |
|
41 |
Let's assume this app of your is floating point intensive (fairly safe |
42 |
assumption), and doesn't use a heck of a lot of RAM or disk (already |
43 |
shown to be true). So now you need to rate the fpu of the various |
44 |
processors and machines around. So I would read reviews of various |
45 |
machines in computer performance mags where they publish sane |
46 |
benchmarks, to get an idea of what would be best |
47 |
|
48 |
alan |
49 |
-- |
50 |
gentoo-user@g.o mailing list |