Print

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

[Jacko13]

Hey all. Have only been playing a week so very much a noob. I have read the maximising FPS page on the wiki and experimented with the unit settings. Am a little surprised at how jerky and slow my dwarf movement is. Any advice appreciated.

Am using the ironhand tlleset latest version of the lazy noob pack and am pretty sure I have eliminated windows 7 and antivirus issues. I have settled on printer mode 2dasync after much experimentation. My tin is running an intel i7 940 with lots of fast ram and I have no disk space issues. GPU is I understand mostly irrelevant but mine is an ati radon 68 series with a gb of ram.

I decided to splash out with an 8x8 embark space and by the time I hit 100 dwarves with no magma, mist generators or anything very clever it is not unplayable but surprisingly slow and jerky.

Any config suggestions from experience with multi core modern processors?

[Couchmonster]

Use reveal and look for high amounts of moving water in the caverns.
This can kill any fps on any machine
So it still can be easily embark related issue.

What do you use in your init display mode?

Standard? 2D?

[Jacko13]

Thanks for suggestion.  After much experimentation I use the 2DASYNC as it seems to have a very small advantage over 2D which was the only other one that looked good.

[blue sam3]

Put DF on two cores (not core 0, since that's the one the system stuff will use the most) and block anything else from using either of those two. That way, DF will do most of its stuff on one, export the graphical stuff to the other and not have anything getting in its way.

[Jacko13]

Good idea... seems to have bought me a noticeable improvement.

[Jacob/Lee]

Can you even do that with DF? If you can, then it won't give you an amazing increase in FPS, unless you play on some huge resolution.

[Wayward Device]

8x8 embark with 100 dwarfs? That's some serious business. In all honesty, if you did that and it was "not unplayable but surprisingly slow and jerky", then you've got some pretty damn good DF performance. I would suggest that at this point any further tinkering would be counterproductive as you've already got things set as they should be.

EDIT: Any real gains past that are only likely to come from Toady-based efficiency improvements the game itself (there have actually been some awesome ones recently. Goodbye clothing !!Clothing Bug!!, you evil motherfucker*. 

*Some of you may be wondering why the clothing bug is on fire. It's because I assume its burning in hell.

[Jacko13]

Setting processor affinity away from core 0 does seem to have some benfit.  Just out of curiosity what is the biggest embark anybody has seen work well with graphics and over a hundred dwarfs on an Intel i5 or i7?

[Flobulon]

Put DF on two cores (not core 0, since that's the one the system stuff will use the most) and block anything else from using either of those two. That way, DF will do most of its stuff on one, export the graphical stuff to the other and not have anything getting in its way.

Wow, that actually helped a lot for my i7 2600k, thank you!

I personally have never gone bigger than a 5x5 embark, and I find even that can get slow once you've got in excess of 20k loose boulders lying around...

[NW_Kohaku]

I would just point out that things like having 20k boulders as a computational bottleneck has essentially nothing to do with your CPU architecture, and everything to do with your computer having to constantly load from RAM vectors that are megabytes long.  (The same going for pathfinding as well.)

The CPU can spin its wheels making bad predictions all it wants, (and oh boy does it ever) but A* style Pathfinding is fundamentally a matter of RAM latency.

My CPU is a 6 or 7 year old dual-core, but I have low-latency RAM, and it basically runs as fast as most other people's computers that are new when it comes to DF.

[Patchouli]

What RAM do you have, NW_Kohaku?

[NW_Kohaku]

Actually... it's been so long I forget.  This computer has held up very well with time, so I haven't had to care.

CPU-Z says it's DDR2 RAM with a CAS Latency of 5 clocks.

[Thief^]

It's probably DDR2-800 or DDR2-1066, if 5 is "low". For DDR2-667 CAS:5 is normal.

DDR3 has much better latency, DDR3-1600 has a CAS of 9 as standard (11.25 ns), equivalent to CAS 4.5 in DDR2-800, and you can get as low as CAS6 (7.50 ns, equivalent to CAS 3 in DDR2-800)

[NW_Kohaku]

Yeah, it was fast for the time I bought it; You can get significantly faster RAM now if you actively look for it, but most people don't.

My RAM would take about 10 nanoseconds to return the first word, but according to Wikipedia, you can get it down to around 7 nanoseconds with more modern RAM.  (~40% faster)

Wikipedia page for information.

Looking at Newegg, there are faster types of RAM out than what Wikipedia lists, but they all have CAS latencies that are high enough to negate that for the purposes of getting small chunks of data. 

[Thief^]

Yeah, that's because increasing transfer speed is comparatively easy, but decreasing latency is really really hard (and actually gets harder as ram capacities go up, so the fastest ram is smaller).