Matrix Games Forums

Benchmark test on 1405 Civ.
In general, it does seem as though performance in the newer build is pretty slow in comparison to older ones; this isn't really demonstrated on the average pulse times but certainly is on the duration - over 10 mins to do a 2 min test.

Dimitris suggested you to start a new thread with these performance issues, did you do it?

https://www.matrixgames.com/forums/view ... 8#p5145768

1405 beta to 1436 beta showed a huge improvement from 14min 30s to 5 minutes.

New 1436CIV
"Old" performance 1328.18

Add your results above 1400 version runs here
https://docs.google.com/spreadsheets/d/ ... id=0#gid=0

1463CIV

13950HX(Q1LP) aka 13900k(basically laptop chip with chipset adapter to LGA1700, without plate[Delided]) ~200$
P@5.5 E@4.4 DDR5@6000 XMP


Game needs much more cores
How much AVX2 code optimizations vectorization via #include <immintrin.h> ? @Dimitris
Maybe offload some calculations to GPU via OpenCL or DirectCompute, but that hardAF, that would be big improvement

Images clickable https://imgur.com/a/BXBv0Sb

13950HX(Q1LP)
P@5.7 E@4.3 DDR4@4100
Fully tuned system
Like 20% uplift



(Open image in new tab for details)

Laptop with i9-13980HX and 4090 laptop. No tuning, but cleaned out a bunch of vestigial apps.

First run...


Second run...

Ryzen 5800x3d 1463

add your benches to sheet https://docs.google.com/spreadsheets/d/ ... id=0#gid=0

thewood1 wrote: Mon Sep 30, 2024 6:05 pm Laptop with i9-13980HX and 4090 laptop. No tuning, but cleaned out a bunch of vestigial apps.

Those settings affect benchmark, autosave probably the most
i disabled all of them and run again,
i thought that my OC was really bad for a sec and laptop overtakes me

Navigation I can see. But the others are all graphics and this is a headless run.

I am not sure how I got the 10,153 peak pulse time and the low of 220ms. I tried it multiple times, and I find the benchmark not very reliable, at least on build 1328.18.

Results added to spreadsheet

Build 1500 is certainly up there as one of the fastest builds in recent times.

Test 1 - Desktop PC; specs are listed on the screenshot, apart from the GPU (GTX 1080TI).

Build 1527. Tried with both the original 495 and 509 db. No noticeable difference. I got a few warning messages that the error fuses were activated with both dbs. Will create a Tech Support thread in a bit. But overall, a very significant in performance over build 14XXs.

Ran this again on the latest patch. 2:50 on a 7900X3D.

New laptop, two metrics.

Balanced Power Profile
Ultra Power Profile

Priority set to high via task manager.

Finally I answered my own question: what if I run CMO on a $200 old server?

This sucker have two sockets and all filled with same kind of CPU.

For comparation, my SFF 9700 box runs this at about 0.8x.

Seymour Cray:
"If you were plowing a field, which would you rather use?... Two strong oxen or 1024 chickens?"

Got a new 9800X3D system. Saw up to 46GB memory use which seems crazy. Plenty faster than 5800X3d yay.

Build 1567
Ryzen 9 7900 with 64 GB DDR5-5600