Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[Maklak]

I'm looking for a semi-standardized way of comparing different computer's suitability to play DF. A mature fort succumbing to FPS death that plays for a season without player intervention with FPS uncapped, and displays either average FPS over time, or total time it took to run that fortress for a season (no player input, but stuff designated tu mine/build/make (with manager) would be ideal, but what to do short of that?

Granted, I could just load the same save on different machines and play it out, but having something standardized in one package would be so much better. It might even become a popular benchmark for single-core speed. 

Now I'm using PCMark2002 (old enough to only use one core, uses some compression algorithm, not terribly well optimized for modern processors), but it is a synthetic benchmark, and not so reliable.

I also use super PI. It takes up 300MB RAM, and performs lots of calculations on one core.

Any thoughts?

[kiffer.geo]

How about a script that puts an independant armoured axedwarf in every tile of the arena and then graph resultant frame rate over time. The higher it starts and the faster it climbs back to it's un capped max the better (obviously).

Then to test path finding we need to set up a maze like structure with lets say 100 dwarfs and 100 levers, each lever is assigned to a dwarf, many of the levers change the lay out of the maze, forcing the dwarf heading towards his assigned lever to repath...
Some levers don't change the maze, they might flip bridges that have 100s of stones dropped onto them, or pump magma up under ice, then take it away again to melt and refreeze the ice over and over again.

Also cats, lots and lots of cats.

Also Happy New Year!

[Maklak]

@Kiffer.geo

That is not really what I meant. I asked about:

1) A standardized way to measure real DF performance that mimics real mature fort (so no crazy maze of levels and floodgates). Such a test seems nonexistent at the moment.

2) Short of 1) some other way to compare speeds of computers. Super PI seems best bet so far.

[jaked122]

@Kiffer.geo

That is not really what I meant. I asked about:

1) A standardized way to measure real DF performance that mimics real mature fort (so no crazy maze of levels and floodgates). Such a test seems nonexistent at the moment.

2) Short of 1) some other way to compare speeds of computers. Super PI seems best bet so far.

I'd say, that that old... what is it, battlemasters(I can't remember the name, but it was used as a test box for submitting rendering methods to improve dwarf fortress) or something with the same engine(not really, but uses same principles), except you can spawn hundreds of units at once. it would probably be a simple way of testing the math which dwarf fortress uses(path-finding, fluid dynamics, and rendering).

[Draco18s]

An AMD Phenom II x6 with a combo liquid nitrogen liquid helium cooled CPU overclocked to 6.5 GHz.

In any case, PassMark is your best bet.  DF eats CPUs for breakfast, so get one rated higher.

[Gzalzi]

In any case, PassMark is your best bet.  DF eats CPUs for breakfast, so get one rated higher.

I'm currently playing on borrowed laptop with a Intel Core i3 370M @ 2.40GHz (Overall Rank: 237). The CPU in my computer is a Intel Core i7 920 @ 2.67GHz (Overall Rank: 39).

Oh god I can't wait to get my computer back so I can run this game a little better (a little because it only uses one core).

Need a way to see the performance of only a single core of the CPU versus others...

Edit: Found a review of an i7 920 using only a single core. Now I need to find one for the i3 370M.

[Kilo24]

Oh god I can't wait to get my computer back so I can run this game a little better (a little because it only uses one core).

Need a way to see the performance of only a single core of the CPU versus others...

The code Baughn helped with does incorporate multi-threading.  Therefore, the non-legacy version does benefit from multiple cores.

[Gzalzi]

The code Baughn helped with does incorporate multi-threading.  Therefore, the non-legacy version does benefit from multiple cores.

Yes, but by how much? From what I've seen, it's isn't much. Are there further plans for multicore support?

[Draco18s]

Are there further plans for multicore support?

No.

[de5me7]

not an expert on testing in anyway, but presumably a good way to create a standarised test, would be to upload a 'standard fort' to teh DFonline data base and get people to download it, run one season and with no intervention, report average fps, and their system specs. But, you would have decide what the standard fort would look like. If it had a pop of 200, some computers would just die trying to load it.

[o_O[WTFace]]

The code Baughn helped with does incorporate multi-threading.  Therefore, the non-legacy version does benefit from multiple cores.

Yes, but by how much? From what I've seen, it's isn't much. Are there further plans for multicore support?

I think much of the graphics stuff is spun off onto another core (or the GPU?)  Also a second core can deal with the random background crap, OS business and allow you to play music in case the dwaven national anthem ever gets old.  The FPS boost is going to be small but probly larger then atomsmashing all that leftover rock. 

[Gzalzi]

Are there further plans for multicore support?

No.

That's what I figured.

Kinda Completely breaks the game.