KSP1 Computer Building/Buying Megathread

[pmborg]

30 minutes ago, dave1904 said:

 

You missed the point. I do not care what a benchmark score says. If it was not for KSP the 10700k would last for the next 10 years. Its a good CPU but when I got it 1 year ago I did not really believe the amd rumors and did not think we would be having a single core IPC war. Something that we gamers needed. I really only care about single core real world KSP performance. Your 5900x is a great CPU no doubt but I would pay more for a cpu with 4 cores with less userbenchmark scores and better KSP performance. The only reason I care about the 12900k is because of single core performance. KSP does not care if it has 4 cores or 20 and if KSP does not care neither do I   Problem is I do not know how KSP deals with adler lakes architecture. It might actually end up being worse. I do not know enough about these things and the 2 core types in adler lake is making it even more complicated. 

If you want to know my score here it is.  https://www.userbenchmark.com/UserRun/47652692 

you can see in that Benchmark link the HW description that was on another system that I have, that now I use as a server, but this one is faster.

Yes true, what is good in AMD 5950X is that AMD optimized the microcode of 99% of the instructions (so 4.6Ghz = 5.1.Ghz 11900 generation) and AMD have a special turbo when work in single core which boosts from 4.8Ghz to 5Ghz, like KSP and disable the other cores...

With 12900K Intel also optimized the microcode to allow most of the instructions to run in one cpu/cycle time also and that made it really fast.

With my current system I can pull 185FPS on KSP and 130/140FPS in flight.

But 12900k is faster in single core about 19% than AMD 5950X, I am curious to see the results of your upgrade, nevertheless don't forgot about graphic card.

For example from 1070-TI to 2080-TI the FPS just double, is like passing from 70FPSto 140FPS in KSP.

In resume cpu is very important it can handle differences in 30/40% of FPS in KSP, but graphic card have a bigger impact on FPS, once can deliver 100% in 2080-TI or 150% using 3080-TI in relation with 1070-TI.

 

In resume the 12900K is the new bests choice up today, but lets see about the rest of the system in order to check if it has a good performance also.

Dont forget that the performance is the sum: GPU+CPU+BOARD+M2+RAM with this order of impact on KSP.

But in your case your upgrade in FPS will be about 11% from my calculations you will pass from Gaming 215% to Gaming 239% which is quit fast!

 

 

[dave1904]

14 hours ago, pmborg said:

you can see in that Benchmark link the HW description that was on another system that I have, that now I use as a server, but this one is faster.

Yes true, what is good in AMD 5950X is that AMD optimized the microcode of 99% of the instructions (so 4.6Ghz = 5.1.Ghz 11900 generation) and AMD have a special turbo when work in single core which boosts from 4.8Ghz to 5Ghz, like KSP and disable the other cores...

With 12900K Intel also optimized the microcode to allow most of the instructions to run in one cpu/cycle time also and that made it really fast.

With my current system I can pull 185FPS on KSP and 130/140FPS in flight.

But 12900k is faster in single core about 19% than AMD 5950X, I am curious to see the results of your upgrade, nevertheless don't forgot about graphic card.

For example from 1070-TI to 2080-TI the FPS just double, is like passing from 70FPSto 140FPS in KSP.

In resume cpu is very important it can handle differences in 30/40% of FPS in KSP, but graphic card have a bigger impact on FPS, once can deliver 100% in 2080-TI or 150% using 3080-TI in relation with 1070-TI.

 

In resume the 12900K is the new bests choice up today, but lets see about the rest of the system in order to check if it has a good performance also.

Dont forget that the performance is the sum: GPU+CPU+BOARD+M2+RAM with this order of impact on KSP.

But in your case your upgrade in FPS will be about 11% from my calculations you will pass from Gaming 215% to Gaming 239% which is quit fast!

 

 

Nothing needs to be upgraded on my pc. 10700k was a bad choice. 3080 is overkill for ksp. Its basically in sleep mode. Only reason I got it is because my 970 was a bottleneck and I got the 3080 rog strix a week after launch for retail price. It is great value for the money. I am leaving everything in my system the way it is. I doubt ddr5 memory is worth and can save alot by using my existing ram. Even if the 12900K is twice ad good as my 10700k it will always be the bottleneck. 

I'm also considering a 12600k. Way more for the money but I think they will be harder to get.

By the looks of it no-one on forums has benchmarked adler Lake for ksp yet so I will make sure I'll get the data here as soon as I get it. Would love ddr5 but it's just to expensive. I doubt the chip shortage is going to end soon anyways so ddr4 it is.

[splashboom]

I see that single core performance is what you care about, would you say it the same about KSP2?

For benchmarks I found the ones from GamersNexus and LTT pretty informative.

[dave1904]

6 hours ago, splashboom said:

I see that single core performance is what you care about, would you say it the same about KSP2?

For benchmarks I found the ones from GamersNexus and LTT pretty informative.

I watch their content all the time. I like gamersnexus but he obviously cannot test all games and especially not ksp with 100 mods. I might be wrong but I do have the feeling single core performance will still be important for KSP2. If I could build 10000 part space stations in ksp I would. Same goes for KSP2. I will build craft until the fps tanks. 

[Vanamonde]

@dave1904, there's a whole thread full of hardware advice, and your post has been merged into it. Lots of info to be found here. 

[dave1904]

1 hour ago, Vanamonde said:

@dave1904, there's a whole thread full of hardware advice, and your post has been merged into it. Lots of info to be found here. 

thanks alot. Still managed to loose myself  

[Elthy]

Keep far away from Userbenchmark, that website is the worst benchmark ive ever heard of, as they are extreme Intel-fanboys and will adjust parameters until their beloved Intel is ahead again.

We realy need a good benchmark for KSP. When i upgraded from an i5 3570k@4,2Ghz to an R5 3600 i made my own benchmark for comparsion: I made a new installation of KSP, with all graphics turned down and delta-time to the setting that slows down time instead of loosing accuracy (forgot which direction to pull the slider). There i created a creative save where i put a massive Saturn5 replica i pulled from the internet on the launchpad. Each benchmark would be to launch that rocket straight up (with SAS) and measuring how much real-time it takes for one minute of ingame-time to pass.

The results:
i5: 5:06
R5: 3:58
with overclocked RAM: 3:46

[dave1904]

7 hours ago, Elthy said:

Keep far away from Userbenchmark, that website is the worst benchmark ive ever heard of, as they are extreme Intel-fanboys and will adjust parameters until their beloved Intel is ahead again.

We realy need a good benchmark for KSP. When i upgraded from an i5 3570k@4,2Ghz to an R5 3600 i made my own benchmark for comparsion: I made a new installation of KSP, with all graphics turned down and delta-time to the setting that slows down time instead of loosing accuracy (forgot which direction to pull the slider). There i created a creative save where i put a massive Saturn5 replica i pulled from the internet on the launchpad. Each benchmark would be to launch that rocket straight up (with SAS) and measuring how much real-time it takes for one minute of ingame-time to pass.

The results:
i5: 5:06
R5: 3:58
with overclocked RAM: 3:46

I know that. I just uploaded my score because I knew the 3080 would hammer the 2080ti That is the only reason I have a high score. His cpu is a good bit faster especially single core

[quangdinh]

my pc specs: 

intel core i5

8gb of ram

windows 10 20h2

sorry for being inactive:))

[JxEq]

yes bruh, i have 4gb ram and it doesn't lag at all even with 30  mods

[Mahnarch]

You've been on the forums over a year and you haven't downloaded the game, yet?

 

Yeah. You should be good, PC-wise, as long as you don't have a bunch of background programs running at the same time whilst trying to stream and facetime your mom on her birthday all at the same time.

[Vanamonde]

There's a whole thread full of hardware advice, @quangdinh, and your question has been merged into it to keep the info all in one place. Enjoy the game.  

[quangdinh]

On 11/19/2021 at 11:41 PM, Mahnarch said:

You've been on the forums over a year and you haven't downloaded the game, yet?

 

Yeah. You should be good, PC-wise, as long as you don't have a bunch of background programs running at the same time whilst trying to stream and facetime your mom on her birthday all at the same time.

nah
i have played it since 2018
back then i used to play old versions (0.10.1, 0.13.3, 0.22, 0.90, 1.0, 1.0.5, 1.2.2) bc my pc was weak back then

now i doubled the ram (from 4gb to 8 gb) so i can play 1.12.2 smoothly

Edited November 23, 2021 by quangdinh

[quangdinh]

On 11/20/2021 at 2:32 AM, Vanamonde said:

There's a whole thread full of hardware advice, @quangdinh, and your question has been merged into it to keep the info all in one place. Enjoy the game.  

tysm

[Dogywall]

Im getting a PC soon but im just not sure what will work Amazing with ksp. I would like to download mods and stuff but im not sure exacly what to get. if you could tell, i dont know alot about PC's

[AtomicTech]

At least an 10th gen i5 with minimum 16 GB's of RAM. 

I run mine on an 8th gen (I think) i5 with 12 GB's of RAM. It's not great most of the time.

[Scarecrow71]

I'm running an AMD Ryzen 9 3900 12-Core Processor on Windows 11 with 32 GB of RAM...and mine is a bit laggy at times.  I should have gotten a better processor, but it is what it is.

[jimmymcgoochie]

I refer you to this:

Generally speaking KSP is bottlenecked by single-thread CPU performance, because all physics calculations are done on one thread and when you have several vessels and/or lots of parts, those add up very quickly to create game lag. Most other games don't have this bottleneck though so unless you're only interested in playing KSP (seriously though, don't just play KSP exclusively as you'll burn out sooner or later) don't just chase single-thread performance above everything else.

Having lots of RAM is a good thing for running mods, especially those with parts in them (and doubly so for parts with highly detailed textures/shapes/interiors) and having a good GPU means you can run the game with higher texture settings and even with visual mods; both of these are true for other games too so more RAM and a bigger/faster GPU are both good choices for a gaming PC. Just be aware that there's a global chip shortage so prices will be higher than normal.

[Curveball Anders]

I use an I-5 3570K (oc:ed at 4.1GHz) with 16GB ram a GTX 960 with 4GB and an SSD running Ubuntu 20.04.3 LTS.

Works really nice for me (but note that I don't use any graphical mods).

 

[AtomicTech]

5 minutes ago, Curveball Anders said:

I use an I-5 3570K (oc:ed at 4.1GHz) with 16GB ram a GTX 960 with 4GB and an SSD running Ubuntu 20.04.3 LTS.

Works really nice for me (but note that I don't use any graphical mods).

 

The only graphical mod I use is Waterfall with Stock/KNES Configs. The hit on the GPU is almost non-existant.

[Dogywall]

could i send a link for a pc and u guys tell me what u think.

[Vanamonde]

There's a whole thread full of PC advice, @Dogywall, and your thread has been merged into it to keep it all in one place. 

[Dogywall]

19 minutes ago, Vanamonde said:

There's a whole thread full of PC advice, @Dogywall, and your thread has been merged into it to keep it all in one place. 

thanks

[Nuke]

i do belive i solved my cpu cooler fitment issue. apparently the amd kit that comes with the nh-c14s is not as versatile as the stand alone kit, which allows 90 degree fitment. and it fits well in this orientation, its not conflicting with my front panel usb headers (who the eff designed these connectors? especially the new usbc mobo headers are they supposed to be this loose? and it looks like they based the connectors on the most hated sata connectors. shame! stop designing bad connectors). and it works with all my cabling and my psu mounting option.

im not quite sure it clears the video card by a good enough margin, so i stuck some kapton tape on the end of the pipes. i also have a second set of brackets to work with, i think if i drill out the tapped holes, slot them with a file and find some longer m3 screws, nuts and washers, i could get a few more mm of clearance. the block is bigger than the heat spreader so that shouldnt reduce efficiency.  its not much but it should clear the back plate on some of the newer gpus. 

still getting nowhere on the gpu front, they are a scarce beast it seems. but the other problem is sorted and thats good.

Edited December 2, 2021 by Nuke

[dave1904]

On 11/8/2021 at 9:56 AM, Elthy said:

Keep far away from Userbenchmark, that website is the worst benchmark ive ever heard of, as they are extreme Intel-fanboys and will adjust parameters until their beloved Intel is ahead again.

We realy need a good benchmark for KSP. When i upgraded from an i5 3570k@4,2Ghz to an R5 3600 i made my own benchmark for comparsion: I made a new installation of KSP, with all graphics turned down and delta-time to the setting that slows down time instead of loosing accuracy (forgot which direction to pull the slider). There i created a creative save where i put a massive Saturn5 replica i pulled from the internet on the launchpad. Each benchmark would be to launch that rocket straight up (with SAS) and measuring how much real-time it takes for one minute of ingame-time to pass.

The results:
i5: 5:06
R5: 3:58
with overclocked RAM: 3:46

Do you by any chance still have that craft file? After getting new hardware I always get into a benchmark mood. To be totally honest it is kind of funny the way the 3570K is only 20% slower than a R5 3600 considering the age difference. I get that its KSP but still you have to give some credit to Ivy Bridge.