Bay 12 Games Forum

Please login or register.

Login with username, password and session length
Advanced search  
Pages: [1] 2 3 4

Author Topic: DF CPU Benchmarking  (Read 10989 times)

Unknown72

  • Bay Watcher
  • Undead Enthusiast, at your service.
    • View Profile
DF CPU Benchmarking
« on: October 31, 2019, 10:14:19 pm »
Okay so slight change but the idea is to compare eachother's world generation times.

In advanced parameters use Large Island

World Seed: DFcnxOnt0aje0Djfemcv
History Seed: 37CElgQge4H5d1vPxS5E
Name Seed: LWyz4Q5SaXxG5kpIHDPE
Creature Seed: x1SfNwLLbwhW4BZrrTZY

Keep all the other seeds random and set End Year to 250. I used a stopwatch to get more exact times.

My computer specs are:

RAM: 12gb
CPU: AMD FX-4300 Quad-Core Processer 3.80 GHz
Graphics Card: GTX 1050ti


Stable World Generation Time: 2 Minutes to generate a stable world
History Generation Time to 250: 16 Minutes
« Last Edit: November 01, 2019, 04:43:20 pm by Unknown72 »
Logged
Quote from: CaptainArchmage on November 15, 2019, 12:40:35 am
You guys ought to fix Unknown72's turn to some point later in the order. Maybe after Sanctume. Normally that's what happens when someone fesses up to having real life going up like a storm of explosive diarrhoea blasted into a fan.

@me on Discord: Multi#0897

Fleeting Frames

  • Bay Watcher
  • Spooky cart at distance
    • View Profile
Re: World Generation Time Comparisons
« Reply #1 on: October 31, 2019, 11:14:53 pm »
You really should use seeds when doing this, and also check whether the final result matches. Depending on how history goes, even same geography world can generate to year 10000 in a minute or slow down around year 100.

Currently, you're not guaranteed to get same results across different computers (even from same OS to same OS).
Logged

Unknown72

  • Bay Watcher
  • Undead Enthusiast, at your service.
    • View Profile
Re: World Generation Time Comparisons
« Reply #2 on: November 01, 2019, 01:13:21 am »
Oh I didn't think about that. I can gen 2 more and time them and give seeds for 'em, tomorrow.
Logged
Quote from: CaptainArchmage on November 15, 2019, 12:40:35 am
You guys ought to fix Unknown72's turn to some point later in the order. Maybe after Sanctume. Normally that's what happens when someone fesses up to having real life going up like a storm of explosive diarrhoea blasted into a fan.

@me on Discord: Multi#0897

Nordlicht

  • Bay Watcher
  • Wuschelig & Wunderlich
    • View Profile
Re: World Generation Time Comparisons
« Reply #3 on: November 01, 2019, 09:02:12 am »
Maybe with posting the hardware specs too this could turn into some kind of DF CPU-Benchmark. Is it possible to gen worlds from command line and have the game automatically close without saving after it is done?
Logged

Unknown72

  • Bay Watcher
  • Undead Enthusiast, at your service.
    • View Profile
Re: World Generation Time Comparisons
« Reply #4 on: November 01, 2019, 03:59:17 pm »
Oh that could be cool too! I don't know if you can do that from command line though.
Logged
Quote from: CaptainArchmage on November 15, 2019, 12:40:35 am
You guys ought to fix Unknown72's turn to some point later in the order. Maybe after Sanctume. Normally that's what happens when someone fesses up to having real life going up like a storm of explosive diarrhoea blasted into a fan.

@me on Discord: Multi#0897

Unknown72

  • Bay Watcher
  • Undead Enthusiast, at your service.
    • View Profile
Re: DF CPU Benchmarking
« Reply #5 on: November 01, 2019, 04:44:04 pm »
Alright, edited the OP with seeds, updated times and format, and computer specs (the ones that I think matter to DF at least).
Logged
Quote from: CaptainArchmage on November 15, 2019, 12:40:35 am
You guys ought to fix Unknown72's turn to some point later in the order. Maybe after Sanctume. Normally that's what happens when someone fesses up to having real life going up like a storm of explosive diarrhoea blasted into a fan.

@me on Discord: Multi#0897

Strik3r

  • Bay Watcher
  • Persistently work-in-progress.
    • View Profile
Re: DF CPU Benchmarking
« Reply #6 on: November 01, 2019, 05:41:24 pm »
Quote from: Unknown72 on October 31, 2019, 10:14:19 pm
Okay so slight change but the idea is to compare eachother's world generation times.

In advanced parameters use Large Island

World Seed: DFcnxOnt0aje0Djfemcv
History Seed: 37CElgQge4H5d1vPxS5E
Name Seed: LWyz4Q5SaXxG5kpIHDPE
Creature Seed: x1SfNwLLbwhW4BZrrTZY

Keep all the other seeds random and set End Year to 250. I used a stopwatch to get more exact times.

My computer specs are:

RAM: 12gb
CPU: AMD FX-4300 Quad-Core Processer 3.80 GHz
Graphics Card: GTX 1050ti


Stable World Generation Time: 2 Minutes to generate a stable world
History Generation Time to 250: 16 Minutes

Ok sure, why not.
Here are my results:
Stable world time: 30 secs
History gen to 250: 15 Minutes.

Specs:
32gb RAM @ 2133Mhz(Don't ask, its crap RAM)
CPU: AMD Threadripper 2920X 12C 24T @ 3.5Ghz(4.1Ghz boost)
GPU: Nvidia GeForce GTX 980ti

Exactly the results i expected.
I think everyone is gonna get similar-ish results, unless some madlad is running a crazy overclocked CPU under LN or in an icebox.

Generated using following seeds:(Should be same as in the OP)
Seed: DFcnxOnt0aje0Djfemcv
History Seed: 37CElgQge4H5d1vPxS5E
Name Seed: LWyz4Q5SaXxG5kpIHDPE
Creature Seed: x1SfNwLLbwhW4BZrrTZY
Logged
NOTICE: If you can't update your profile/signature, stop using a Imgur URL for your profile picture.
Upload it to somewhere else.
☼

wierd

  • Bay Watcher
  • I like to eat small children.
    • View Profile
Re: DF CPU Benchmarking
« Reply #7 on: November 02, 2019, 04:02:21 am »
.....

OK, but this is gonna be a strange one that's sure to raise some eyebrows.
I will run the benchmark on my phone, running DF in Eltech's Exagear Windows Emulator. (Yes, it runs DF. Kinda.)

Give me a minute. :P
Logged

methylatedspirit

  • Bay Watcher
  • it/its
    • View Profile
Re: DF CPU Benchmarking
« Reply #8 on: November 02, 2019, 04:58:53 am »
I forgot to set End Year to 250, so my results are probably a small overestimate.
My results:
Stable world time: 55 seconds
History Gen to 250: 14 minutes, 50 seconds
(Total: 15 minutes, 45 seconds)

Specs:
RAM: 8 GB @ 1600 MHz
CPU: Intel Core i5-4460 @ 3.20 GHz (slight overclock to ~3.3 GHz)
Graphics Card: AMD Radeon R7 200

Seeds used:
World Seed: DFcnxOnt0aje0Djfemcv
History Seed: 37CElgQge4H5d1vPxS5E
Name Seed: LWyz4Q5SaXxG5kIHDPE
Creature Seed: x1SfNwLLbwhW4BZrrTZY

Made a mistake copying the Name Seed (between G5k and IHD there's a missing p). Hope that didn't affect things much.

@wierd, have you tried running the Windows version of Geekbench 4 on that emulator?
Logged

anewaname

  • Bay Watcher
  • The mattock... My choice for problem solving.
    • View Profile
Re: DF CPU Benchmarking
« Reply #9 on: November 02, 2019, 08:45:35 pm »
So this process can be a bit easier...

Step 1)
Instead of typing in values for the seeds, just copy and paste this text into your worldgen file, then start DF and choose TEST LARGE ISLAND when you are choosing the world to generate. It is the vanilla Large Island with specific TITLE, SEED, and END YEAR values added so the tester doesn't need to type in values.
Spoiler (click to show/hide)

Instead of using a timer to get the times, walk away and let it run until the history is fully generated, then get the timestamps from these files (If you open the file's Properties window, you will be able to get the timestamp in minutes and seconds. Use the Accessed time if you see both Accessed and Modified times.):
- gamelog.txt (this file is updated when the world gen process is started)
- data/save/current/region_snapshot-1.dat (this file is updated when the year 1 is completed)
- data/save/current/region_snapshot-250.dat (this file is updated when the year 250 is completed)

Then calculate the difference between the 1st and 2nd times (time to create the world), and the difference between the 2nd and 3rd times (time to generate the history). The 2nd file may not be created until at the end of the first year, but that happens in a moment, even on the slow machine I tested.

I ran it on two computers and have this:

3m 24s creating world
51m  2s history's 250 years

Celeron(R) CPU N3350 @ 1.10GHz (max 2400 MHz, 2 core)
Ideapad 120S-11IAP motherboard (whatever is in that laptop)
2 GB LDDR4-2100

0 min 22s creating world
6 min 46s history's 250 years

i5-9400F @ 2.90 GHz (max 4100 MHz, 6 core)
AsRock Z390 Phantom Gaming 4 motherboard
16 GB DDR4-2999 (base 2400 but using XMP to get to 3000)
Logged
Quote from: dragdeler
There is something to be said about, if the stakes are as high, maybe reconsider your certitudes. One has to be aggressively allistic to feel entitled to be able to trust. But it won't happen to me, my bit doesn't count etc etc... Just saying, after my recent experiences I couldn't trust the public if I wanted to. People got their risk assessment neurons rotten and replaced with game theory. Folks walk around like fat turkeys taunting the world to slaughter them.

DontMineYellowSnow

  • Bay Watcher
    • View Profile
Re: DF CPU Benchmarking
« Reply #10 on: November 02, 2019, 09:31:04 pm »
Just ran this:

11 rejected worlds
55 seconds to stable world creation
~8:40 to completion at year 250 (including stable world creation time, so 7:45 spent on history)

I had a couple of browser tabs open and ran the version bundled with the LNP so DFHack, etc. running in the background

Using all the seeds spelled out up top.

Platform:
i7-8700
SSD
16GB DDR4 16-18-18-38 @ ~2880Mhz
ASRock Z390 Phantom Gaming 4 mainboard



anewaname's results seem to suggest that, if the tests were accurate, that extra bump in RAM clock made up for a considerable difference in processing power.  What kind of RAM you got there?  Specs on latency?  Are you using standard XMP to get to 3000Mhz?
« Last Edit: November 02, 2019, 09:44:02 pm by DontMineYellowSnow »
Logged

anewaname

  • Bay Watcher
  • The mattock... My choice for problem solving.
    • View Profile
Re: DF CPU Benchmarking
« Reply #11 on: November 02, 2019, 11:12:38 pm »
The memory is 15-17-17-35, CMK16GX4M2B3000C15. The overclock was the default XMP profile in BIOS, taking it from 2400 to 3000.

There was a post earlier this year that started me on the path to get these parts, where it was concluded that memory speed is the most important component for DF. In the retrospect, it makes sense because a DF fort usually is less than 2 GB so it will exist entirely in memory, and because DF does multiple lookups on arrays for most anything, there is going to be multiple transactions with the memory for each transaction with the CPU.
Logged
Quote from: dragdeler
There is something to be said about, if the stakes are as high, maybe reconsider your certitudes. One has to be aggressively allistic to feel entitled to be able to trust. But it won't happen to me, my bit doesn't count etc etc... Just saying, after my recent experiences I couldn't trust the public if I wanted to. People got their risk assessment neurons rotten and replaced with game theory. Folks walk around like fat turkeys taunting the world to slaughter them.

methylatedspirit

  • Bay Watcher
  • it/its
    • View Profile
Re: DF CPU Benchmarking
« Reply #12 on: November 03, 2019, 12:40:32 am »
I happen to have a laptop, so I ran the test on that too:
Stable world time: 2 min, 30s
History Gen to 250: 27 min, 29s
(Total: 29 min, 59s)

Specs:
RAM: 8 GB @ 1866 MHz
CPU: AMD A9-9420, 2 cores @ 3.40 GHz
GPU: Radeon 520

Seeds used:
(Used anewaname's worldgen config)

Faster RAM doesn't seem to change things if it's being bottlenecked by a crap CPU, like in this laptop.
Logged

DontMineYellowSnow

  • Bay Watcher
    • View Profile
Re: DF CPU Benchmarking
« Reply #13 on: November 03, 2019, 09:39:15 am »
Quote from: anewaname on November 02, 2019, 11:12:38 pm
The memory is 15-17-17-35, CMK16GX4M2B3000C15. The overclock was the default XMP profile in BIOS, taking it from 2400 to 3000.

There was a post earlier this year that started me on the path to get these parts, where it was concluded that memory speed is the most important component for DF. In the retrospect, it makes sense because a DF fort usually is less than 2 GB so it will exist entirely in memory, and because DF does multiple lookups on arrays for most anything, there is going to be multiple transactions with the memory for each transaction with the CPU.

I always thought the CPU cache was king, but Intel won't have a larger cache in their CPUs until at least 2020-2021 anyhow.  I wish someone would come along with one of those fancy 4400Mhz setups on a capable board/CPU to test this further.
Logged

Soadreqm

  • Bay Watcher
  • I'm okay with this. I'm okay with a lot of things.
    • View Profile
Re: DF CPU Benchmarking
« Reply #14 on: November 03, 2019, 07:35:28 pm »
Windows 10 Pro N
CPU: Intel Core i5-9600 @3.70GHz
GPU: GeForce GTX 1060 6GB
RAM: 16GB (2x8GB) DDR4 3200MHz
HDD (WD10EALX)

Stable world time: 0:25
History gen to 250 years: 9:45
Total: 10:10

Eleven rejects like everyone else. Here's the world map.

I then copied my DF folder to my SSD and ran the same worldgen again. So, the same specs except for the hard drive:

Windows 10 Pro N
CPU: Intel Core i5-9600 @3.70GHz
GPU: GeForce GTX 1060 6GB
RAM: 16GB (2x8GB) DDR4 3200MHz
SSD (SH103S3/120G)

Stable world time: 0:26
History gen to 250 years: 9:50
Total: 10:16

Which isn't the result I expected, but there you have it. At least it made the same world.
Logged
Pages: [1] 2 3 4
« previous next »