Home Trending Latest threads New posts Developer diaries

Performance Megathread

Jump to latest Follow Reply
  • We have updated our Community Code of Conduct. Please read through the new rules for the forum that are an integral part of Paradox Interactive’s User Agreement.
Stellaris 51274 Bug Reports 31977 Suggestions 19766 Tech Support 2984 Multiplayer 383 User Mods 4696 Stellaris AAR (After Action Reports) Console edition 1237
Show only dev responses
ThePHD said:
I will be going from an 8700K @ 4.4ghz to a 5800X3D. I will post some results when I get everything set up later. The 5800X3D is known for being the best CPU for Paradox games. It won't be an extensive benchmark but I hope to get a general idea of how much of a performance gain going from an older CPU to a modern one will give someone. Most tests seem to put the 5800X3D against other modern CPUs.
Click to expand...
This is the extremely tl;dr version. The game still slows down but there was a very substantial performance increase. Makes me excited to see what future 3D V-Cache CPUs can do.

3.52

39% faster mid game (2300+)
55% faster end game (2400+)

Overall this is about in line with what i've read about the 5800X3D. I also did my tests with both high and low frequency ram. It barely changed the results at all. I got maybe .5% better results using the faster ram. I might post more details if I have the time. I will also be testing the CPU in other Paradox titles.

edit : End game lag is almost non existent in EU4 and HOI4. It's crazy seeing the end game just fly by with minimal slowdown. The Stellaris end game is by far the worst in terms of how sluggish it feels to play.
 
Last edited:
  • 5Like
  • 1
Reactions:
kane70 said:
a few new postings in the last days here, so I'll add my experience:
I don't know why, but in the 3.5 version (latest non-beta) of the game it is now in the year 2571 almost unplayable.
Okay - I play with a big galaxy, and all extensions are on. But no mods.

Contentwise: the awakened empires are done, and the end game crisis also. Oh, and I play as toxoid on the biological path, all species in my empire are fully genetically optimized.
View attachment 892173

I understand that a Ryzen 3700X and a Radeon RX 6700XT with 16 GB RAM is not high-end. But anyhow, I am frustrated. I'm not done with this galaxy, and I have to abandon it just because of performance. Not cool.
Click to expand...
Do you have pop growth scale to zero? Please post a save
 
ThePHD said:
This is the extremely tl;dr version. The game still slows down but there was a very substantial performance increase. Makes me excited to see what future 3D V-Cache CPUs can do.

3.52

39% faster mid game (2300+)
55% faster end game (2400+)

Overall this is about in line with what i've read about the 5800X3D. I also did my tests with both high and low frequency ram. It barely changed the results at all. I got maybe .5% better results using the faster ram. I might post more details if I have the time. I will also be testing the CPU in other Paradox titles.

edit : End game lag is almost non existent in EU4 and HOI4. It's crazy seeing the end game just fly by with minimal slowdown. The Stellaris end game is by far the worst in terms of how sluggish it feels to play.
Click to expand...
Thanks for this!

I'm curious about the 13900k performance in paradox games, it seems to do relatively well in other simulation games, although the benchmarks I saw still had the 5800x3d crush it in Factorio. But Dwarf Fortress and a couple others, it actually came out above the 3d.

Another thing I was curious about was whether faster memory (DDR5) would cause the gap between 3d cache and non-3d cache CPUs to narrow. Obviously you didn't test DDR5 with the 5800x3d, but with such a minor increase that you saw from memory speed, it seems like the cache will still reign supreme.

I guess I'm still gonna wait on the 7____X3D. Ugh. Really want to just get the upgrade over with.
 
Logros said:
Thanks for this!

I'm curious about the 13900k performance in paradox games, it seems to do relatively well in other simulation games, although the benchmarks I saw still had the 5800x3d crush it in Factorio. But Dwarf Fortress and a couple others, it actually came out above the 3d.

Another thing I was curious about was whether faster memory (DDR5) would cause the gap between 3d cache and non-3d cache CPUs to narrow. Obviously you didn't test DDR5 with the 5800x3d, but with such a minor increase that you saw from memory speed, it seems like the cache will still reign supreme.

I guess I'm still gonna wait on the 7____X3D. Ugh. Really want to just get the upgrade over with.
Click to expand...
I honestly wish I had waited. The performance on this CPU is a great upgrade but I feel like the 7950X3D is going to be absurd if the rumors are true. I'll probably still end up getting one at some point.
 
ThePHD said:
I honestly wish I had waited. The performance on this CPU is a great upgrade but I feel like the 7950X3D is going to be absurd if the rumors are true. I'll probably still end up getting one at some point.
Click to expand...
jepp, I also plan to upgrade - but because of bigger compute tasks which I regularly need to execute on my environment. Buying such a monster for games like Stellaris sounds like overkill to me. Such games should run well on a mid-sized CPU, ignoring if it's early, mid, or late game.
 
  • 1
Reactions:
I noticed that the forceful mining drone expansion fleets got stuck in one of my games. After wiping out all the barges, the game sped up significantly. Not sure if this is a known thing.
 
  • 2
Reactions:
kane70 said:
jepp, I also plan to upgrade - but because of bigger compute tasks which I regularly need to execute on my environment. Buying such a monster for games like Stellaris sounds like overkill to me. Such games should run well on a mid-sized CPU, ignoring if it's early, mid, or late game.
Click to expand...
I'd agree if it was just Stellaris - but I play several paradox games, and it's also been a recurring theme in other games I've played for many years. WoW, especially with lots of UI add-ons, benefits a ton from the X3D. HoI4 starts chugging late game, especially with ExpertAI mod. CK3 was super laggy in some late game scenarios - people made a population control mod to delete insignificant characters for it. I can't imagine whats gonna happen with Vic3s pop system :). MOBAs (DotA) which I play gets a ton of value from it.

If it was only one game bottlenecking I'd completely agree. I would say it's a majority of my playtime though that has massive gains from the X3Ds. They benefit alot of more mainstream games like regular FPS' and stuff too, but I think anyone considering a somewhat specialized CPU like these should make sure they're doing it because it greatly benefits a majority of what they play and likely will in the future, for the life of that PC/CPU.

Or if it's just not alot of money to them and they judge it worth it :)
 
Sorry if this has been asked before, but is there any good big save file with lots of pops etc. for some ~recent patch I could use to test and compare late game performance? Search found some save files for older versions.

I am soon upgrading from 3800X to 5800X3D and I'd be somewhat interested to see how much it improves things.
 
rubert said:
Sorry if this has been asked before, but is there any good big save file with lots of pops etc. for some ~recent patch I could use to test and compare late game performance? Search found some save files for older versions.

I am soon upgrading from 3800X to 5800X3D and I'd be somewhat interested to see how much it improves things.
Click to expand...
for some mod test i did i looked into the bug section. of course bugged saves are not perfect but if its not game breaking stuff it is somewhat helpfull.
 
Thanks for the tip. I opted to run observer game myself on 3.5. I'll try to remember to post the save game later during the week when it is bit more developed. Not entirely sure what kind of settings would be the most demanding.

I chose huge galaxy, 2* habitable planets, no crisis, grand admiral, 0.15 pop growth scaling (if I remember right, will need to double check it), xeno-compatibility enabled, 25ish empires, 0.5 tech/tradition cost to speed things up a bit.

edit: I could have reduced the pop limits further but that is around the same setting I normally play in Huge galaxy so I felt it was suitable for my personal testing. And given how many pops empires have managed to create in this observe game I think I might limit it in the future...

1667248287863.png
 
Last edited:
I posted in this thread a while back when I got the 5800x3d and I can play Stellaris again on huge galaxies with lot's of empires without getting annoyed. To me, the 5800x3d was a worthwhile investment. I also have 32g memory as I like mods and when playing past 2400, the game can use a lot of memory especially if you save/load games.
 
See next post for the save game used for the test. Game version 3.5.3 without mods. Settings used in the spoiler below. I considered using the more absurb settings mentioned in the previous dev diary but opted for more "realistic" settings I might actually use for my real game. Plus I don't think I have will to keep the game running to 2700.

The current year is 2331 and I felt the game had mostly stagnated. Galaxy has 37 262 pops - so not the largest game I have heard of but I'd say pretty large.

edit: if people want me to run it even more I can do that as well, though not sure how much worse the game speed would become. Let me know.

Load the game, open console and type observe to switch out to the poor starter MegaCorp.

  • 3.5.3 - no mods.
  • Huge Elliptical
  • 30 AI empires (some were premade one including Imperial Fiefdom)
  • 15 Advanced AI empires
  • 5 Fallen Empires
  • Marauders off
  • Tech/Tradition cost 0.5x
  • Habitable worlds 3.5x
  • Primitive Civilizations 2x
  • Crisis off
  • Mid- and endgame year 2250 and 2300.
  • Grand Admiral
  • Difficulty Adjusted AI modifiers On (I thiink these were bugged in 3.5.3?)
  • AI aggressiveness normal
  • Empire Placement: Clusters
  • Advanced Neighbours: On
  • Hyperlane, Gateway and Wormole: 1x
  • 2 guaranteed habitable worlds
  • Caraveneers, L-Gates, Xeno-Compatibility On
  • Logistic Growth Ceiling 1.5x
  • Growth Required Scaling 0.1x
 
Last edited:
Sorry about double post, but I just finished my really quick and dirty comparison of AMD Ryzen 3800X and 5800X3D in Stellaris. The exact save game used for the tests is attached to this post.
  • 3800X completed 98 days in 70 seconds, 1,4 days/second.
  • 5800X3D completed ~144,25 days in 70 seconds, ~2,06 day/second
  • Increase of 47%
I suspect slightly longer test would have increased the difference a bit as there a noticeable slowdown after unpausing the loaded game.
  • Post above lists the game settings for the save. I progressed the game few months forward so that I could get the end of year to the test too without making it overly long.
  • Game was switched to observer mode using console and galaxy map before starting the test. Camera was not moved after exiting to the galaxy map.
  • Stellaris was restarted after each test.
  • Each test was 70 seconds long. I was planning doing bit longer tests, but I got a brainfart while doing tests on 3800X and ended up with shorter version as I really didn't want switch back to 3800X.
  • Each CPU was tested 5 times.
  • System used was Asus Prime X570-P motherboard, 32 GB of RAM (3200 Mhz, exact timings evade me), Radeon RX6600, Windows 10 21H2. Motherboard BIOS firmware and AMD CHipset drivers were updated to the latest prior to tests.
 

Attachments

  • benchmark.sav
    16,6 MB · Views: 0
  • 3Like
  • 2Love
  • 1
Reactions:
I stopped playing Stellaris back in 2018 if I remember correctly, in part because of late game stuttering lag. It is absolutely fascinating to read this thread as I consider a return to the game. I eagerly await reports of how much the new AMD processor coming in January improves late game performance.
 
Well, you can look at the performance of the 5800x3d and compare that to the new x3d that has a higher clock rate. It will easily outperform the old cpu by at least another 10% to 15% if not more.
The 7800X3D has a reportedly base frequency of 4.8 Ghz vs 3.4 Ghz and a boost frequency of 5 Ghz vs 4.5 Ghz. The new chipset also has faster memory support, a larger L3 cache and PCI 5. Thing is, will you need the extra speed as the 5800x3d is already fast in the endgame and a new setup with the 5800x3d will be rather cheap.
 
  • 1Like
Reactions:
Anyone have any idea as to the relative contributions to lategame lag of pops vs bypasses (particularly hyper relays, L-Gates, and gateways, since these proliferate lategame) vs ship numbers as of 3.6 under default settings?
 
Is anyone planning to get one of the new AMD CPUs when they come out? Specifically the 7900X3D or 7950X3D. I'm curious as to how big of a difference the extra cache and clock speed will make. I was going to get a 7900X3D on release but the 600 dollar price tag is pretty steep. That and having to buy a new MOBO + Ram. I'll probably just hold onto my 5800X3D for awhile but I would also like to see how it compares to a 7900X3D.
 
You must log in or register to reply here.
Top Bottom
Back