Post your FPS & CPU's scores - instructions how to measure included - 5950x + 12900k + now 5800x3D!!

[spookywatcher]

Got some 5800X3D results for everyone:

https://www.youtube.com/watch?v=Dj8L3xHuyUE

Interesting results:

Game: 5.10 with all DLC
CPU: 5800X3D
GPU: 3060ti
Ram: 32 3600 cl 16
Resolution: 3440x1440

Young Gun Empty: 170
Dense Empire: 26
Dense Empire Empty: 96


Not sure why the empire results were so low. I did run with all DLC on, not sure if that gave it more to calculate or not. I did try it in 1080p but that didn't really seem to make a difference. If it's just clock speed though, then why are the Young Gun Empty results higher than even the 12900k?

That's awesome. Thanks for testing! Very much appreciated. Interesting results. I'll add it to first post.

[Imperial Good]

Not sure why the empire results were so low. I did run with all DLC on, not sure if that gave it more to calculate or not. I did try it in 1080p but that didn't really seem to make a difference. If it's just clock speed though, then why are the Young Gun Empty results higher than even the 12900k?

You need to run them with exactly the same settings as the others for the results to be fair. Logically running with a DLC enabled is likely to lower performance over without as the universe has more ships and is more complex to simulate. Ideally the same patch version should be used as otherwise optimisations or regressions could affect the results.

12900k is possibly beating it purely due to the results given being DDR5 memory. Although the cache helps a lot, it just cannot compensate for out right ~40% more memory bandwidth and other DDR5 improvements. Could also be GPU related. Especially for dense empire as you are pushing resolution quite hard.

Still this is good signs for that CPU and I would not be surprised if higher was possible with some tuning, given the spectrum of results in this thread with similar processors.

[Action_Parsnip]

Got some 5800X3D results for everyone:

https://www.youtube.com/watch?v=Dj8L3xHuyUE

Interesting results:

Game: 5.10 with all DLC
CPU: 5800X3D
GPU: 3060ti
Ram: 32 3600 cl 16
Resolution: 3440x1440

Young Gun Empty: 170
Dense Empire: 26
Dense Empire Empty: 96


Not sure why the empire results were so low. I did run with all DLC on, not sure if that gave it more to calculate or not. I did try it in 1080p but that didn't really seem to make a difference. If it's just clock speed though, then why are the Young Gun Empty results higher than even the 12900k?

Ok I've already posted in this thread with my same machine but on an older version of x4 and got results that didn't line up with yours very well.

So tonight testing again with x4 on the same version as yours with nearly all DLC installed:

5600x
3800 16gb ram, tuned main + sub timings,
Rx580

Game version 5.10
All DLC except the new ToA
VRO
Tons of other mods
....all this smashing into a 3.x save game test I get the following:


Young Gun Empty: 111-115
Dense Empire: 11-13
Dense Empire Empty: 48-52

Here's the kicker boys: this is 720p all LOW...

....AND this was after waiting 2 solid minutes after the game had loaded for tests 2 & 3, not moving the ship at all or changing cameras, to let everything catch up and the FPS to stabilize. In both those tests the FPS was all over the place and going over your youtube video carefully you didn't experience *anything like* the hitches I did. This is with a NVMe SSD! It's only a cheapo Sabrent Rocket Q but still, my setup and yours look a class apart. System memory was pretty stretched here but not critically so and just to be sure I had all other programs closed while running these tests.

[Animaga]

Not sure why the empire results were so low. I did run with all DLC on, not sure if that gave it more to calculate or not. I did try it in 1080p but that didn't really seem to make a difference. If it's just clock speed though, then why are the Young Gun Empty results higher than even the 12900k?

You need to run them with exactly the same settings as the others for the results to be fair. Logically running with a DLC enabled is likely to lower performance over without as the universe has more ships and is more complex to simulate. Ideally the same patch version should be used as otherwise optimisations or regressions could affect the results.

12900k is possibly beating it purely due to the results given being DDR5 memory. Although the cache helps a lot, it just cannot compensate for out right ~40% more memory bandwidth and other DDR5 improvements. Could also be GPU related. Especially for dense empire as you are pushing resolution quite hard.

Still this is good signs for that CPU and I would not be surprised if higher was possible with some tuning, given the spectrum of results in this thread with similar processors.

Ok, so I did run some more tests. I tried multiple screen resolutions but nothing changed as far as FPS goes. My GPU was never taxed more than 50% on any of these tests.

So first I went back and disabled the Terran and Avarice expansions (had to leave split because the save needs it):

CPU: 5800X3D
GPU: 3060ti
Ram: 32 3600 cl 16
Resolution: 3440x1440
Version 5.10 only Split DLC

Young Gun Empty: 180
Dense Empire: 28
Dense Empire Empty: 109

So other than the giant station that's a gain of about 10FPS having the two newer DLCs disabled.

Tried to go back to 3.10 to match the save, but that is no longer an option. The earliest version of the game listed was 3.20 so that's what I did:

Version 3.20 with Split DLC

Young Gun Empty: 180 (It does start at 210 but over the course of a couple of minutes slowly drops then bottoms out at 180)
Dense Empire: 30
Dense Empire Empty: 193

Now the Dense Empire Empty save is unique because in version 5.10 where the save is isn't actually empty anymore. You load in looking at multiple ships and an accelerator. In 3.20 it is actually just empty space you are staring at.

[jebbyk]

CPU: ryzen 5600x (undervolted and caped to 4.45Ghz on all cores, this setting outperformes Auto OC a little bit)
RAM: 16GB (2x8 3200mhz, timings: 16,18,18)
GPU: GTX1060 3Gb (all setting are the lowest possible even lod, shadows and radar. 720p + FSR:performance)

5.10 all DLCs

Small addition: I've set a affinity for game process manually to use only logical processes 0,2,4,6,8,10 to restrict use of AMD's "hyperthreading". It gives me additional 5-8% of performance

Young gun: 138-140fps
Empire empty area: 85-100fps
empire dense area: 20-22fps

[spookywatcher]

CPU: ryzen 5600x (undervolted and caped to 4.45Ghz on all cores, this setting outperformes Auto OC a little bit)
RAM: 16GB (2x8 3200mhz, timings: 16,18,18)
GPU: GTX1060 3Gb (all setting are the lowest possible even lod, shadows and radar. 720p + FSR:performance)

5.10 all DLCs

Small addition: I've set a affinity for game process manually to use only logical processes 0,2,4,6,8,10 to restrict use of AMD's "hyperthreading". It gives me additional 5-8% of performance

Young gun: 138-140fps
Empire empty area: 85-100fps
empire dense area: 20-22fps

Thanks for that. I'm going to try that today via the bios and see what happens. Great idea.

[jebbyk]

Thanks for that. I'm going to try that today via the bios and see what happens. Great idea.

I'm using Ryzen master for tuning. I'm lazy to go to BIOS for each change. I think you could do the same with 5800x3d

[spookywatcher]

Thanks for that. I'm going to try that today via the bios and see what happens. Great idea.

I'm using Ryzen master for tuning. I'm lazy to go to BIOS for each change. I think you could do the same with 5800x3d

I'm on linux... ryzen master no worky for my daily driver.

With only a few mins of testing...I can't really tell if it's helping. It looks like clock stretching was lessened and maybe 2-3 more fps...maybe. But not sure about that. I need more testing.

[Valhalla_Awaits]

12900k is possibly beating it purely due to the results given being DDR5 memory. Although the cache helps a lot, it just cannot compensate for out right ~40% more memory bandwidth and other DDR5 improvements. Could also be GPU related. Especially for dense empire as you are pushing resolution quite hard.

That is quite possible, I know going from stock speed DD4 ram, to top of the line XMP overclocked DD4 made a hell of a difference in framerate in busy sectors. DDR4 to DDR5 may be causing a similar result. We'd need either processor to be using the other kind of ram for comparison to tell. I doubt the GPU is making much if any difference even at 4k, as a 2080ti doesn't seem to care if I'm at 1080/1440/4k framerate wise. A 3090 would matter even less. Though admittedly stations are rendering a stupidly high amount of geometry even when not being looked at.

[mattaw]

Just got a 5900x, so tested my 5600x first.

CPU: 5600x Version 5.10 HF2 with all DLC no mods
GPU: 2070 Super
Ram: 64GB 3600 CL 16
Resolution: 1920x1080
Young Gun Empty: 122-127 FPS
Dense Emp: 18-22 FPS
Dense Emp/Empty: 66-73 FPS

Identical system, but 5900x and better cooler

CPU: 5900x Version 5.10 HF2 with all DLC no mods
GPU: 2070 Super
Ram: 64GB 3600 CL 16
Resolution: 1920x1080
Young Gun Empty: 127-133 FPS
Dense Emp: 20-21 FPS
Dense Emp/Empty: 77-82 FPS

[Nutsman]

..... interessting topic
I would say the "Dense Empire" savebench will be the most important for CPUs.
If i catch this right the current Intel platforms for example 12900k or 12700K with DDR5 perform the best way in X4 and push the FPS over 30 in Dense Empire
What about the new 5800X3D with the new AMD Cache feature.... it looks to me we not see this impressive boost to go in the same region as Intel.

Any other conclusion about this ?

[Imperial Good]

What about the new 5800X3D with the new AMD Cache feature.... it looks to me we not see this impressive boost to go in the same region as Intel.

From what I can tell both trade blows with each other pretty well. 12900k does better in Dense Emp but the 5800X3D does better in Young Gun and Dense Emp/Empty.

I would not recommend people buying AM4 and the 5800X3D new given how AM4 is EoL with AM5 scheduled to be released later this year. If you already have a compatible AM4 motherboard and an older Zen2 or earlier CPU it is by far a fantastic upgrade which is on par with the 12900k at a fraction of the power used and cheaper memory, and possibly even cheaper price. Otherwise if you are buying a new motherboard then going with Intel might be the better long term choice as the platform is both DDR5 and PCIe 5.0 compatible.

If AMD has a Zen4 CPU with anywhere close to as much cache Intel really has to be worried as that will absolutely dominate games like these.

[Nutsman]

@Imperial Good
thanks for the hints this makes sense

[IratusAvis]

Yesterday I upgraded from Ryzen 3900X & RTX 2070S to Ryzen 5800X3D & RX 6800 XT and noticed a huge performance gain while playing my own savegames (2560x1440). I will try to deliver my scores later. Had not so much time for playing because I had to build my old stuff into my wife's Sims-4-Machine.

EDIT:

Game: V5.10
CPU: 5800X3D <<<
GPU: 6800 XT (Red Devil)
RAM: 32 GB DDR4-3600 CL16 19-19-39 (DOCP enabled)
Resolution: 2540x1440
Settings: High, no SSAO

DLCs: SV only

Young Gun Empty: 178-185 FPS
Dense Empire: 21-26 FPS
Dense Empire Empty: 114-124 FPS

DLCs: SV, COH, TOA

Young Gun Empty: 167-172 FPS
Dense Empire: 21-26 FPS
Dense Empire Empty: 91-103 FPS

[lmike]

Hi. I have a powerful system also
CPU: 12900K stock
GPU: 6900 XT (Red Devil) + 15% PL in windows tests
RAM: 32 GB DDR5-6200 CL40
OS - Windows 11
All DLCs enabled, v 5.10 beta

This is my results:

Resolution: Windowed 1280x720
Settings: Low preset, no MSAA or upscaling

Dense Empire Empty: first 3 mins about 110-115FPS, then drops to 100-105
Dense Empire: 21-24
Young Gun Empty: 145-147


Resolution: Windowed 3824x2073
Settings: Ultra preset, no MSAA or upscaling

Dense Empire Empty: 100-105, GPU usage ~80%
Dense Empire: 21-24
Young Gun Empty: 145-147

Upd:
I don't liked my results. I thought it is caused by Windows 11 or slow SSD Adata Swordfish 2TB. That's why I tried the same test on Linux and SSD Samsung 970 evo

Windowed borderless 3840x2160
Settings: Low preset, no MSAA or upscaling
All DLCs enabled, v 5.10 beta
OS - Ubuntu 20.04

Dense Empire Empty: 94-105
Dense Empire: 19-25,
Young Gun Empty: 104-117

In best scenario Linux is the same as Windows 11 in terms of perfomance, in worst scenario it is worse.

Somewhy me perfomance much lower that other 12900k or even 5800X3D. Idk why is it so. System is relatively new, 3 month only, fresh installation.

[Imperial Good]

Somewhy me perfomance much lower that other 12900k or even 5800X3D. Idk why is it so. System is relatively new, 3 month only, fresh installation.

It is unlikely to beat the 5800X3D in many tests as I mentioned earlier. The cache on the 5800X3D is just so good for games like X4, factorio, e.t.c. due to all the cache misses. The only down side to the 5800X3D, and why I cannot recommend it for new builds, is that AM4 is EoL as far as better CPUs and memory go and scheduled to be replaced with AM5 by the end of the year. Intel's platform is just a lot more "future proof" right now with DDR5 and PCIe gen5 for GPUs.

The lower performance than other 12900Ks is likely due to not testing like for like. They were tested without Tides of Avarice DLC and each additional DLC reduces performance. Some people have even pointed out that the DLC invalidates one of the tests (empty space is no longer empty).

Any difference outside of like for like testing would be due to differences in memory timing. DDR5 might have higher memory bandwidth but it also generally has worse latency than many of the reasonably priced DDR4 kits currently available. Due to all the cache misses the CPU experiences (why the 5800X3D is good) memory latency has a huge impact on performance as the CPU has to stall while waiting to read from memory. This is also why high core frequencies have little impact on performance, with people reporting +300 MHz core clock OCs having practically no impact on modern CPUs compared with tightening memory timings a very little.

[spookywatcher]

Hi. I have a powerful system also
CPU: 12900K stock
GPU: 6900 XT (Red Devil) + 15% PL in windows tests
RAM: 32 GB DDR5-6200 CL40
OS - Windows 11
All DLCs enabled, v 5.10 beta

This is my results:

Resolution: Windowed 1280x720
Settings: Low preset, no MSAA or upscaling

Dense Empire Empty: first 3 mins about 110-115FPS, then drops to 100-105
Dense Empire: 21-24
Young Gun Empty: 145-147


Resolution: Windowed 3824x2073
Settings: Ultra preset, no MSAA or upscaling

Dense Empire Empty: 100-105, GPU usage ~80%
Dense Empire: 21-24
Young Gun Empty: 145-147

Upd:
I don't liked my results. I thought it is caused by Windows 11 or slow SSD Adata Swordfish 2TB. That's why I tried the same test on Linux and SSD Samsung 970 evo

Windowed borderless 3840x2160
Settings: Low preset, no MSAA or upscaling
All DLCs enabled, v 5.10 beta
OS - Ubuntu 20.04

Dense Empire Empty: 94-105
Dense Empire: 19-25,
Young Gun Empty: 104-117

In best scenario Linux is the same as Windows 11 in terms of perfomance, in worst scenario it is worse.

Somewhy me perfomance much lower that other 12900k or even 5800X3D. Idk why is it so. System is relatively new, 3 month only, fresh installation.

Those are pretty low scores. I suspect you are running X4 on the E cores. Maybe try disabling them in the bios and try again just off the P cores.

[lmike]

Retested 12900k
Windows 11, same 720p res, same 5.10 beta

DLCs: SV, COH
Dense Empire Empty test is broken. It spawns inside Heart of Achronomy II

Dense Empire Empty: 108-117
Dense Empire: begins with 27-31, after ~2 min its floating 21-24
Young Gun Empty: 155-157


DLCs: SV
Dense Empire Empty test is broken also

Dense Empire Empty: 108-123
Dense Empire: begins with 27-31, after ~2 min its floating 23-25
Young Gun Empty: 160-163

Testing without E cores maybe I'll do later

[Perkel]

It is unlikely to beat the 5800X3D in many tests as I mentioned earlier. The cache on the 5800X3D is just so good for games like X4, factorio, e.t.c. due to all the cache misses. The only down side to the 5800X3D, and why I cannot recommend it for new builds, is that AM4 is EoL as far as better CPUs and memory go and scheduled to be replaced with AM5 by the end of the year. Intel's platform is just a lot more "future proof" right now with DDR5 and PCIe gen5 for GPUs.

It is ****** amazing cpu. In all games i test with some complex simulations like Dwarf Fortress. X4. Kenshi etc. It literally doubles triples or quadruples FPS. Dwarf Fortress went from 14x14 tile embark in range of 1-2 fps with often seconds per frame rather than frames per second into 120ish fps with split second drops to 0 fps every 5-6 seconds suggesting some bottleneck like cache miss with some calculation and going into ram and back. World creation and 250 years of history went from 15-20minutes on 5600X to just 3-4 minutes on 5800X3D. People who build stuff in Fallout 4 say it also effectively nearly doubled FPS compared to 5800X and it stands like 30% above 12900KS.

Imho dense empire and other tests aren't great tests for CPUs because they are more about pipeline between gpu and cpu. Yes good CPU can help with draw calls issue but ultimately you are clogging pipeline between CPU and GPU.

On other hand stuff that is done entirely on CPU like world simulation can give you best example.

I use fires of defeat start. Just start game wait maybe 2-3 minutes and then you have relatively empty map with some ships and some wrecks. Lower res and details and you are limited mostly now by simulation of world rather than gpu or pipeline clogging.

In my test:

5600X - 110fps
5800X3D - 230fps

It literally doubled fps

What that means is that it takes double the size of world economy/ships etc simulation to hit 5600X simulation fps with this GPU.

[Imperial Good]

What that means is that it takes double the size of world economy/ships etc simulation to hit 5600X simulation fps with this GPU.

That is not quite how computers work. There exists a complexity point for each of the 5600X and 5800X3D which is the largest complexity obtainable before significant cache misses start to occur. If complexity is raised beyond this point performance (instructions per cycle) starts to decrease rapidly as cache misses start to increase. Logically for the 5800X3D this complexity point is exactly 3x bigger than the 5600X as it has exactly 3x as much L3 cache. At this point for the 5600X both the 5600X and 5800X3D will be largely frequency, core count or power bottlenecked and so perform very similarly due to the low utilisation of multithreading in X4 and the similar boost frequencies of the CPUs. Any game where the 5800X3D does not show significant performance gains is at or below this point in complexity.

A typical playthrough of X4 on the 5600X is obviously well past this optimal complexity point. However on a 5800X3D it is either significantly less past the complexity point or even before it allowing the CPU to maintain high IPC. The entire performance offering of the 5800X3D is this higher IPC at higher state complexity levels. Anything below such complexity levels will be run better by a 5900X or 5950X as those have higher boost frequencies and more cores.

Now of course this is all about complexity with regard to IPC performance. Complexity is already at best linear with frame time as the more complex something is, the more cycles the CPU needs to process through it all. As such at the complexity limit described for the 5800X3D it will already be running at most at ~1/3 the FPS of the complexity limit of the 5600X. In reality due to underlying data structures it is possible for performance to scale worse than linearly with complexity, as such values like 1/4 or 1/5 of the FPS could be observed.

This logic applies to all cache tiers, not just L3. Each time complexity exceeds one of the cache levels a large IPC decrease occurs as it starts to fall back to the slower, higher level. With this regard memory is little more than another tier of cache with page file being yet beyond that still.

As such for your statement at the very best it could handle roughly twice the complexity. In reality it is likely to fall short of that due to data structure scaling. If the cache size of the 5800X3D is exceeded before this complexity then it will likely fall short significantly due to the massive decrease in IPC that starts to occur.