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

[Blitz4]

The CPU will likely become memory latency bottlenecked at those frequencies. Even though monolithic Intel CPUs generally maintain low memory latencies.

Yes. I'm solely looking at the dense empire performance. It's interesting that the 7900x3D performed worse than Intel silicon considering Intel has a smaller cache.

A lot of the people reporting performance issues either are expecting X4 to do the impossible, more than 60 FPS perfectly stable at max settings with whatever randomness their universe throws in front of them, or they are using older gaming orientated CPUs which no longer cut it performance wise, such as quad core Intel processors from the 7th generation or older.

max settings. It's not clear to me, it should be clear in the settings of the game, precisely which settings negatively impact the cpu. Consider I'm using GeForce Now and can't test that..

Also this benchmark doesn't look at frame latency. I don't know if that's a thing as I never looked at it. It's an issue in a game like Elden Ring at launch, when even if a user got steady 60 fps, the frametime graph could have wild spikes, if even only for one frame, causing perceptible visual and input latency. (Digital Foundry did job showing this issue.) With GeForce Now I'm extremely limited in how I can test and tailor the in-game settings to better that performance/quality balance. There's 3 different FPS number's I can use, it's not clear what those 3 fps number mean and not given any other tools or even know exactly what cpu I'm using, how much % of each vcore on that cpu, etc.

I have a 3900x, but it was temporarily being used for gaming until zen 4 came out and I kind of fried the motherboard doing so. It's a server with server ram, so I'm putting it back to work. However zen 4 came out and I'm not impressed, like look at the 7900x3D performance compared to the 13900k for dense empire. Now with GeForce Now being so good for so many games, I can't see paying for high-end hardware ever again.

[Vheissu]

Do you have any thoughts on the 7700x cpu's performance in X4? Or does anybody in the thread have a 7600x?

They should perform well. The 6 core 7600X might show some weakness in collision heavy scenes but overall it should play decently.

Thanks, much appreciated.

I got a 7600x system almost ready to switch on, but i didnt upgrade for x4 just cos i was worried how long my system would last being about 10-11 years old except the gpu and a few drives.

Still dont think its right that to get anything playable u need a new cpu for this game, they really should impose hard limits for some things to reduce the strain on the cpu for older systems which should be able to play fine but cant.

Unfortunately I don't there's a realistic cap on how much cpu load simulation heavy games can cause. In any case, I somehow always end up playing poorly optimized cpu bottle-necked games so the processor will get good use outside of X4.

[Skeeter]

The thing about all the new 7000 series amd cpus is they all have about the same single core performance scores on stuff i mean the 7600x can trade blows with most and even beat alot cos of single thread being so strong.

[Imperial Good]

I have a 3900x, but it was temporarily being used for gaming until zen 4 came out and I kind of fried the motherboard doing so. It's a server with server ram, so I'm putting it back to work. However zen 4 came out and I'm not impressed, like look at the 7900x3D performance compared to the 13900k for dense empire. Now with GeForce Now being so good for so many games, I can't see paying for high-end hardware ever again.

Been using my 3900X for gaming for years now. No issues at all and certainly the motherboard has not fried.

The dense empire results should largely be ignored due to how many variables are not being kept constant during tests. People get massive differences in frame rate between runs, and newer versions of X4 average significantly lower frame rate than the versions of X4 the tests were made for, in a way that is not representative of actual performance in the current version.

The 7900X3D is likely not ideal due to its hybrid architecture. A 7800X3D will likely perform better and give more consistent results, at least until schedulers and games start to be aware of "higher cache, lower frequency" cores.

As far as I am aware GeForce Now uses server CPUs? These usually perform worse than consumer CPUs in single thread workloads due to slower memory and lower boost clocks. Some aspects, such as memory bandwidth, should be better than consumer CPUs due to the high memory channel count, but I am unsure if this will compensate for the increased latency or worse memory timings. Although server CPUs have a lot of cache between all the cores, I do not think they get the same benefit the X3D CPUs do as the cache is more spread out or further away from the cores.

The thing about all the new 7000 series amd cpus is they all have about the same single core performance scores on stuff i mean the 7600x can trade blows with most and even beat alot cos of single thread being so strong.

This is especially the case for games like X4 where the CPUs are memory access limited anyway. Their effective clock speed is well below the reliable boost clock speed of the CPUs. This is why the X3D variants usually perform significantly better as the reduction in cache misses allow the CPU to maintain a significantly higher effective clock speed.

Due to the changes in collision/physics in 6.00 there may be performance differences between 6 and 8 core CPUs now in situations that can leverage the Jolt physics engine's scalability. If this difference is meaningful I do not know.

[Skeeter]

nvm

[LandogarX4]

[Blitz4]

The dense empire results should largely be ignored due to how many variables are not being kept constant during tests. People get massive differences in frame rate between runs, and newer versions of X4 average significantly lower frame rate than the versions of X4 the tests were made for, in a way that is not representative of actual performance in the current version.

As far as I am aware GeForce Now uses server CPUs? These usually perform worse than consumer CPUs in single thread workloads due to slower memory and lower boost clocks. Some aspects, such as memory bandwidth, should be better than consumer CPUs due to the high memory channel count, but I am unsure if this will compensate for the increased latency or worse memory timings. Although server CPUs have a lot of cache between all the cores, I do not think they get the same benefit the X3D CPUs do as the cache is more spread out or further away from the cores.

I'm somewhat starting to get that.

Let me tell you. The game looks phenomenal on a 4080 max settings. Yes they use, here: https://www.eurogamer.net/digitalfoundr ... lternative

I spend a few hours trying to guess. I've no clue really. I imagine it's a custom chipset, but maybe it's not. 2U server with 2 nodes. Containing 2x mobo's, 2x AMD Threadripper Pro 3995WX, 128GB DDR, 4x 4080's on their side, 4x PSU's all distributed equally between the two nodes. I can't find one chipset with dual cpu's. It's blow me away if it was, 8x 4080's in a 2U, I don't know if that's possible. Even without dual CPU, words won't describe it. I imagine it's like GN & Wendell demo'd: https://www.youtube.com/watch?v=GrOsiCZ9MVE

I tested a lot of virtualization. No question they're slower, even playing locally. But the tech today, I say for a demanding 3D game with lots of data being pushed around is somewhere between 10-30% slower than equivalent hardware. However, for your less demanding 2D games, it's 0-10% slower. In fact doing virtualization like that at home wasn't as easy until Nvidia unlocked their cards to allow it. Nvidia holds all the power and knows far more than they're showing. They've been selling this service for years to the Enterprise. Level 1 has a lot of vids on it if you're curious.

You can see in that first link, they show the fps differences best they can.

Using a stupid powerful GPU like that, but a cpu that performs like a 5600x. Which settings when turned up increase CPU usage? I could test locally over the course of several days, I'm curious if you have any advice?

[Imperial Good]

I spend a few hours trying to guess. I've no clue really. I imagine it's a custom chipset, but maybe it's not. 2U server with 2 nodes. Containing 2x mobo's, 2x AMD Threadripper Pro 3995WX, 128GB DDR, 4x 4080's on their side, 4x PSU's all distributed equally between the two nodes. I can't find one chipset with dual cpu's. It's blow me away if it was, 8x 4080's in a 2U, I don't know if that's possible. Even without dual CPU, words won't describe it. I imagine it's like GN & Wendell demo'd: https://www.youtube.com/watch?v=GrOsiCZ9MVE

The RTX 4080 tier seems to use some unspecified AMD CPU type which is high performance. My guess is that it could be an undisclosed Zen4 threadripper, or given the size of Nvidia it could even be a custom built CPU.

Using a stupid powerful GPU like that, but a cpu that performs like a 5600x. Which settings when turned up increase CPU usage? I could test locally over the course of several days, I'm curious if you have any advice?

I think draw distance? Maybe volumetric fog?

If you want a ridiculously CPU demanding situation then after allying with all factions you can turn on the allied order overlay. This is notorious for causing single digit or lower frame rates with the map open.

I spend a few hours trying to guess. I've no clue really. I imagine it's a custom chipset, but maybe it's not. 2U server with 2 nodes. Containing 2x mobo's, 2x AMD Threadripper Pro 3995WX, 128GB DDR, 4x 4080's on their side, 4x PSU's all distributed equally between the two nodes. I can't find one chipset with dual cpu's. It's blow me away if it was, 8x 4080's in a 2U, I don't know if that's possible. Even without dual CPU, words won't describe it. I imagine it's like GN & Wendell demo'd: https://www.youtube.com/watch?v=GrOsiCZ9MVE

The GeForce Now racks do not contain actual RTX 4080 GPUs. Instead they contain enterprise grade GPU accelerator cards which are then virtualised into something that performs roughly similar to a RTX 4080. I would not be surprised if each GPU accelerator card is shared with 2 instances as they seem to have half the VRAM of comparable enterprise cards. For CPU the obvious choice would be some high core count moderate frequency EPYC CPU that they allocate apart into individual instances, however it is remaining competitive against high performance desktop CPUs like the i9 13900k which makes this choice unlikely. As such it will be either some undisclosed high core count, high frequency, Zen4 threadripper CPU, or possibly even a 3DVcache EPYC which frequency does not matter as much.

The reason they can pack so much hardware into a single rack is due to the high powered cooling solutions often used. Typically they use hundreds of watts of fans to generate very high speed airflow over the heatsinks resulting in the heat sinks operating very efficiently. In these conditions a tiny heat sink stuck on top of the CPU package with some shrouds to direct air flow might out perform a huge noctua heat sink commonly used to cool desktop CPUs. Outside of power efficiency, the main down side to this approach is the horrendous noise generated by all the fans, with data centre staff often being required to wear ear protection when working nearby the racks.

[m4nt4]

So finally I managed to run the 3 savegames:

Sys:
CPU: R5 3600
GPU: RX 6900 XT
Ram: 32 GB DDR4 3000 Cl15

Gameversion 6.20 beta
Fullscreen 720p, low quality.

Empty: 114 fps.
Dense + empty: 63 fps.
Dense: 14 fps.

I am thinking about getting an 5800X3d.

[Imperial Good]

Empty: 114 fps.
Dense + empty: 63 fps.
Dense: 14 fps.

I am thinking about getting an 5800X3d.

5800X3D will make a significant difference with performance. Dense empire results do not reflect practical late game results due to migration issues with the save.

[BigBANGtheory]

Sys:
CPU: AMD R7 7800X3D
GPU: RTX 4090
Ram: 32 GB DDR5 6000 Cl30

Gameversion 6.20 beta
Fullscreen 4k, ultra quality with MSAA x2

Empty: 167 fps
Dense + empty: 127 fps
Dense: 27 fps.

[Skeeter]

So i finally got round to finishing my new pc and sticking x4 on.

So on 1080p as thats my living room tv so not the normal 1440p monitor it will be going on later after a bit more setup first with things then ill move new pc to old pc location.

So erm yeah going from 3570k to 7600x. Pretty good for this game tho i didnt upgrade for it but interesting to try it and see the diff.

So i notice that it looks nice on my big tv in living room, hehe. I was getting about 300fps with no vsync on high preset and no aa in the main menu of cradle of humanity version. Dont have boron or pirate dlc. So i changed it to from high to custom and put the shadows up ultra and ssr to medium and the 2 sliders max from 70 and no cockpit reflection, ssao normal. No aa as for some reason i dont need it on this screen lol, well i could use taa (if it was available) for them bits that flicker a bit when the cam moves on stations bits and pieces. But normal polys were fine i think and aa wouldnt help at all.

So in space i was getting about 150. In stations it can be about 90 or more. A far cry from about 35-60 on stations often middle of that values. In space it could be 75fps with vsync on on 3570k but only if empty space or not much around but could dip for no apparent reason unless somits happening i cant see which meh but its the game aint it. With this its now 130-150fps tho thats at start of terran cadet and just flying about in space. Map is pretty high fps but its only got 3 sectors on for now since new game think i got 90fps or 130fps on map i cant remember too well now even tho was just on it 20 mins ago lol.

Oh and with this new comp i can finally enable SAM that gpu vram full usage thingy with amd stuff. Couldnt use it before with my 6700xt cos of old mb and cpu but now its on with this new setup.

Seems the game likes these cpus tho i dunno what its like mid game with map explored and in it a bit, hope the fight that happens earlish on terran cadit plays smooth, i think it will, be interesting to see how combat in this is with a reasonable large fight ur in with a station around. Didnt really test too much im just tryin to get new pc setup with accounts and what not, i put steam on thought id test x4 so dled it and had a quick 20 min sesh. At least now i dont think ill be worried about low fps where freesync aint working, infact i probs need enhanced sync on now since ill be past my highest range of 75 on my monitor when i put it in the other room with the 1440p mon. Cos too many fps. I think my 6700xt is working better since new cpu too since it was losing 30fps ish on average of gaming use cos of that old cpu the 3570k. Now its working fully and has a cpu that can feed it fast enough now its flying.

Its nice to have one of the best cpus for single core ipc performance even tho its the cheaphest kinda one of the 7000 line. Hopefully no more upgrades for another 10 years now.

[Chris2000SP]

Empty: 114 fps.
Dense + empty: 63 fps.
Dense: 14 fps.

I am thinking about getting an 5800X3d.

5800X3D will make a significant difference with performance. Dense empire results do not reflect practical late game results due to migration issues with the save.

My save game started on version 5.10. Do the old Engine has issues against 6.xx ?

I have a ryzen 5800X3D and a radeon 6800 XT. I got a boost in my save when i upgraded my CPU to the 5800X3D. But now i have a lot of ships in the universe and i get low fps.

[Imperial Good]

My save game started on version 5.10. Do the old Engine has issues against 6.xx ?

It is more the case that there is migration code that runs between versions, and the dense empire busy station save is very badly affected by that migration code. After a few minutes and teleporting to a different location, everything should settle down. However, due to RNG differences, this would not produce consistent results for benchmarking.

The 5800X3D should still run X4 perfectly fine. My R9 3900X still runs X4 perfectly fine, late game. How many ships does your universe have? Are you using mods?

If you have performance issues, try the current beta as well. In case it is a known performance issue that has already been fixed for next release.

[Chris2000SP]

My save game started on version 5.10. Do the old Engine has issues against 6.xx ?

It is more the case that there is migration code that runs between versions, and the dense empire busy station save is very badly affected by that migration code. After a few minutes and teleporting to a different location, everything should settle down. However, due to RNG differences, this would not produce consistent results for benchmarking.

The 5800X3D should still run X4 perfectly fine. My R9 3900X still runs X4 perfectly fine, late game. How many ships does your universe have? Are you using mods?

If you have performance issues, try the current beta as well. In case it is a known performance issue that has already been fixed for next release.

I do not know but my file is in it's compressed state 115 MB in size. There was a time where it was smaller. If i decompress it, i get a 964 MB file:

Code: Select all

$ wc -l quicksave.xml
20455964 quicksave.xml

(Lines of Code)

EDIT:
I discovered that in the 6.00 beta was my file a little bit bigger in size than now. Maybe that is what you mean with jumping inside Stations or what? I have snapshots of my save files.

EDIT:
The 5800X3D is actual my best choice. I will not change that. I got more FPS in specific situations in game. Especially if i open the Map. I had on the map with my old CPU bad timings and such.

EDIT:
I make tomorrow some benchmarks with the Mangohud tool.

[Chris2000SP]

I made with Mangohud a plot. First is when i open the Map. The Strait line is fling the ship.

4K Res. ryzen 5800X3D radeon 6800 XT. VSYNC ON.
https://i.postimg.cc/VLGB26nK/X4-2023-0 ... 34-csv.png

Greetings

[Scoob]

For me, going from a 3900X to a 5800X3d DOUBLED my frame rates in many instances. It was always a busy local sector that caused my fps to drop below the self-imposed 60fps limit. By "busy" I mean larger battles, involving dozens of ships and thousands of projectiles flying at one time. The 5800x3d far exceeded my expectations, X4 loves the extra cache.

For example, I can have a major Xenon invasion taking place (multiple capital ships, dozens of Fighters) vs. my large defence fleet (several battleships (mods), a Carrier and 60+ fighters) and the fps remains a solid 60 throughout. 3900x drops to 30, or below, in the same situation. When I first got to experience these large fleet vs. fleet battles on the 5800X3d, it was a game changer for me. Seeing these battles play out smoothly is amazing. The 7800X3d would be even better, but the 5800X3d was a simple drop-in upgrade for me, the 3900X going into another system - the one I'm typing on now.

For the reasons Imperial Good states, I do feel that this test save is somewhat obsolete now. While I don't have a game with that uber mega station - which is a little nuts, impressive, but still nuts - I do have very mature games with a LOT more going on (more ships, more stations, more conflicts) that were started in newer versions of the game and runs far better. So, while DLC has added more ships, stations and sectors, this older save seems to chug a bit, compared to significantly busier game universes, started more recently.

I get that, when loading an older save, the game is doing "stuff" for a certain amount of time, so performance is erratic at best, after loading this now ancient save. So, it must be allowed to settle for several minutes before taking the benchmark fps. Also, while things do settle down a bit, but the results are still quite sporadic.

[Eyeklops]

Game: 6.10 with all DLC
CPU: 13900KS (EVGA Z790 Dark Kingpin stock settings with 300w limiter)
GPU: RTX 4090 (stock settings)
Ram: 2x24GB DDR5-8000 CL38-49-49-84 CR2 (stock XMP profile)

Resolution: 2560x1440
Graphics Settings: Low

Young Gun Empty: 159~160
Dense Empire: 38~40
Dense Empire Empty: 135~138

I haven't tuned the memory yet. This is with the default XMP profile. The RAM is actually DDR5-8200 but the XMP profile is for 8000.
AIDA64 Cache & Memory Benchmark:
Memory: Read 120.41 GB/s, Write 108.84 GB/s, Copy 108.68 GB/s, Latency 60.6 ns

[lolol]

Do you guys think itd be worth upgrading from a 5800X3D to a 7800X3D/7900X3D or a 13900K for X4? I am somewhat underwhelmed by the performance a 5800X3D/4080 produces in this game.

[Scoob]

Do you guys think itd be worth upgrading from a 5800X3D to a 7800X3D/7900X3D or a 13900K for X4? I am somewhat underwhelmed by the performance a 5800X3D/4080 produces in this game.

I wonder if there's some other factor at play here. I have a 5800X3d and a 3070 with 32GB of RAM and fast M.2 SSD's. Game runs exceptionally well. I have graphical settings mostly maxed out at 1440p - including draw distances - and the only time I see the FPS drop below a (self-imposed) 60fps is when a large battle breaks out in the sector I'm currently in. It's ALL about what's going on near me.

My current save is modded, with many more ships (and ship models) flying around, more battles, more stuff in general. My save is quite mature, approaching eight days, I have multiple busy stations trading, lots of auto-traders and miners, several large fleets moving around, others defending set Xenon incursion areas. The AI Factions themselves have been boosted too, so they're fielding a lot more, and more ships and fleets in general. It's a busy universe. Yet, despite this, the game is performing flawlessly.

Note: regular Faction vs. Faction battles, such as border skirmishes between HOP and PAR, or HOP and ANT/ARG or any faction and Xenon, rarely cause an fps drop (below 60). However, what does are some of my uber murder fleets with multiple battleships (mods) backed up by a Carrier group with dozens of fighters. It's when 100+ Fighters are engaged, with thousands of projectiles in flight at once that slow-downs can occur.

Now, we all have a different perception on what is "smooth". For me 60fps visuals feel very fluid. I will say though that, in some titles, I prefer a higher fps (90, 120) as I can detect the input delay at 60fps. This only applies to titles that really benefit from "twitch" reflexes. X4 is not such a title. While I do dogfight occasionally, or fly myself when neutering a ship prior to boarding, I'm more an Admiral commanding the action, than a Pilot.

Oddly enough, friend has a 12900k, 3090, 32GB DDR5 and fast M.2 SSDs, yet his game can suffer from massive slowdown during longer play sessions. Mine does not. We even tested vs. the exact same save, my game ran perfectly the entire time, he'd lost more than half his fps after an hour. That hour was spend just standing on a Wharf, watching the fps. We're not sure what's going on. I wonder if you're perhaps experiencing some obscure issue too?