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

[spookywatcher]

After going back to my 6700k, I pushed it to the limit.

CPU : 6700k @ 5.0 ghz
MEM : 3600 mhz cl 15
Cache Core : 4.7 ghz
Young Gun : 156 fps
Dense Emp : 29 - 30 fps
Dense Emp w/ Empty : 155 fps

So the reason there was NO improvement upgrading to the 10600k is due to X4 being so dependant on a single core/thread and it's mem access bandwith and ultimately IPC. My single core cinebench r20 score for my 6700k is 518 @ 5.0 ghz (same as 9900ks). An overclocked 10900k @ 5.3 ghz is scoring ONLY 539. That's only a 4% uplift compared to my 6700k. In the single core dependant X4 a 4% uplift is only about 5-6 fps. A 10600KF @ 4.8ghz is only scoring 501 according to cpu-monkey (although that looks a bit low it should be around 511?). And looking at my fps results it does look like a direct correlation between cinebench r20 single core performance and X4 performance. And more to the point...Comet Lake (intel 10th gen) does not have any IPC increase over Skylake from 2015! That's because it is Skylake++++ (14nm ++).

My 6700k @ 4.8 ghz is only 2 fps lower than @ 5.0 ghz. So I'm running it at 4.8 ghz w/ volts at 1.32 and temps that never get above 68c even in a 75F room temp.

So...Until AMD or Intel come out with a cpu that substantially clocks higher than 5.0 ghz (say in the 6.0 ghz) or dramatically increase the IPC (maybe Alderlake?) there will be absolutely no benefit to upgrading specifically for X4 that I can tell. Even at 6 ghz (with same IPC) that might only scale to a 15% uplift in X4. Or unless the devs figure out a way to NOT make it so single core crippled, X4 will forever be handicapped. Because I don't think we'll see cpu's in the 6 ghz range for a very long time if at all and we'd have to wait for 20-30% IPC increase in a new Archetecture (Alderlake?) However, there was 10 fps increase for me @ 4.8 ghz just going from 2800 mhz mem speed to 3600 mhz. So if you're rocking slower mem...that might be a worthwhile upgrade depending on price.

Any how...that's it. Not much else I can push and or test. Hope these tests help someone out if they are thinking about upgrading.

[SirNukes]

Ryzen 5900x stock, paired with gskill ripjaw 3600 cl16 32gb (2 dual rank dimms) memory slightly tuned. Testing on version 3.3 SV.

Young Gun : 152-153
Dense Emp : 27-30
Dense Emp w/ Empty : 152-155

There is some variance in the fps counter, so I'm giving the approximate ranges when ignoring short fps dropouts, observed around 5-15 seconds after loading.

When using a 1 minute capture from frameview, repeated 3 times and averaged, the actual average fps is a bit lower than quick eyeballing would suggest. Dense empire is around 27 average in the log. But I think the eyeball numbers above are more comparable to what others are reporting.

The memory tuning (mostly secondaries) gained 3.1% in frameview averages. I also toyed with comparing dimm configurations, where 1 dimm -> 2 dimms same-channel gained 3%, and then going to 2 dimms both-channels gained another 1%. From that, and ryzen memory reviews I have read, I suspect filling out with 4 dimms (4 ranks each channel) might gain another 3%, though that isn't really worth an extra $150.

[Falcrack]

I gotta remember to find this thread after Christmas. I am going to get a RAM upgrade, from 16 GB DDR4 2666 MhZ CL16 to 32 GB DDR4 3600 MhZ CL18.

[spookywatcher]

Ryzen 5900x stock, paired with gskill ripjaw 3600 cl16 32gb (2 dual rank dimms) memory slightly tuned. Testing on version 3.3 SV.

Young Gun : 152-153
Dense Emp : 27-30
Dense Emp w/ Empty : 152-155

There is some variance in the fps counter, so I'm giving the approximate ranges when ignoring short fps dropouts, observed around 5-15 seconds after loading.

When using a 1 minute capture from frameview, repeated 3 times and averaged, the actual average fps is a bit lower than quick eyeballing would suggest. Dense empire is around 27 average in the log. But I think the eyeball numbers above are more comparable to what others are reporting.

The memory tuning (mostly secondaries) gained 3.1% in frameview averages. I also toyed with comparing dimm configurations, where 1 dimm -> 2 dimms same-channel gained 3%, and then going to 2 dimms both-channels gained another 1%. From that, and ryzen memory reviews I have read, I suspect filling out with 4 dimms (4 ranks each channel) might gain another 3%, though that isn't really worth an extra $150.

Thanks SirNukes. Going to put your results at the top of the list on first post!

[SirNukes]

After messing around a bunch with the amd overclock options for the 5900x, I couldn't get any significant performance improvement. From reading around, a couple percent benefit might be possible with undervolting, at risk of crashes, but all of the easy options (pbo limits, scaler, max clock boost) do little or cause harm for single threaded stuff. And there is enough run-to-run variance in X4 that 1-2% can't be picked out confidently.

But, I did try out the latest X4 beta and found a big performance bump in the slow test.

X4 4.0 beta 4 (eyeballed fps):
Young Gun : 154-155 (+2 fps)
Dense Emp : 33-36 (+5 fps; just barely see glimpses of 36)
Dense Emp w/ Empty : 153-156 (+1 fps)

In the frameview 1 minute capture, the average went up from 26-27 to 32-33, over a 20% increase.

[Gavrushka]

LOL, feel like the poor relation here. Thought I'd be coming along and posting FPS up with the best, but it wasn't to be.

I7 10700K, not overclocked, 32GB Corsair 3600 RGB PRO CAS 15. MoBo Prime Z490-P with M2 NVME drive.

Young Gun: 147
Dense Empire: 36
Dense Empire (Empty): 149

I've no idea why I'm so much lower than other machines, although I appreciate a 5900x is faster than the I7. There's no component/driver issues, so I guess I just suck!

I'm running the latest X4 version (4.0 Beta 4)

[Falcrack]

Here is my system:
i7-7700K @ 4.49 Ghz
16 GB DDR4 2666 Mhz CL16
GTX 1060 6 GB
Asus Z270a motherboard

My graphics settings are as follows:
Resolution: 1920x1080 (lowest possible for me)
Anti-Aliasing: None
Texture: Low
Shadows: Off
Soft Shadows: Off
SSAO: Off
Glow: Off
Chromatic Aberration: Off
Distortion: Off
LoD Setting: 0
Effect Distance: 0
Radar: Off
Screen Space Reflections: Off
Volumetric Fog: Off

I tested with 3.3:
Young Gun: 127-130
Dense Empire: 22-23
Dense Empire (Empty): 120-127

I tested again with 4.0 beta 4:
Young Gun: 126-130
Dense Empire: 24-25
Dense Empire (Empty): 126-130

[SirNukes]

I7 10700K, not overclocked, 32GB Corsair 3600 RGB PRO CAS 15. MoBo Prime Z490-P with M2 NVME drive.

Young Gun: 147
Dense Empire: 36
Dense Empire (Empty): 149

I've no idea why I'm so much lower than other machines, although I appreciate a 5900x is faster than the I7. There's no component/driver issues, so I guess I just suck!

That is the highest in the most demanding test, at 36. If that's a stable 36, it's pretty great. I can see 36.x pop up a couple times on the 5900x, but it can't hold there.

[Gavrushka]

The PC is a week old, and I still have it on a pedestal, so thanks for the partial reprieve from my humiliation, SirNukes. --- The framerate was stable but at 35-36, not 36, although it did flicker downwards when a ship ploughed into me after a while. I've my processor in 'balanced' mode, but I don't suppose that makes a jot of difference in high processor load situations. Will try again in 'pretty' mode and post a video to see what difference it makes. (Didn't see a jot of difference in my own game around my megastations, regardless of graphic settings, although I'd not changed the SSAO setting when testing.)

*EDIT* just did a video and uploaded. (It's 1.4gb so it'll likely need downloading from OneDrive (ad-free premium account) to watch properly.) - I noted in the original post about how SSAO can hit the processor, but I didn't see a noticeable difference. - I started in low, then went to Ultra (after a minute) then added SSAOx4 at 3 minutes. I did get a couple of stutters here and there, along with an anomaly at one point with the contrail, but the framerate stayed over 30 almost constantly, sometimes going into mid 40s. - Main reason, I guess, is because I play at 1920x1080.

Anyhow, video link (as I said, it might be best to download due to its size and bitrate.) https://1drv.ms/u/s!AnoYsE1M89YrgcBscII ... A?e=kWJ0EK

[Roeleveld]

CPU: 5950X (no overclock)
Mem: 3600 DDR4 (XMS enabled)
GPU: 2070Super - All settings turned down, but still at 3440x1440
Version: 3.3 (latest non-beta as per 2020-12-19)

young guns: 169 - 171
emire dense: 26 - 31
empire empty : 167 - 169

[Max Bain]

CPU: Ryzen 3600 (no overclock)
Mem: 3200 DDR4
GPU: 1070 - Ultra settings except SSAO off and VSync off, resolution 2560x1440
Version: 4.0 Beta 4

young guns: 132 - 134
emire dense: 20 - 22
empire empty : 139 - 143

[Buzz2005]

CPU: Ryzen 3600x (PBO on)
Mem: 16gb ddr4 3200
GPU: 1080ti, 2560x1440 SSAO low, SSR off, GLOW low, rest on high and LOD 100%, no AA
Version: 4.0 b4

young guns: 124-125
empire dense: 21-23 ( holy **** that station )
empire empty: 126-128

[spookywatcher]

CPU: 5950X (no overclock)
Mem: 3600 DDR4 (XMS enabled)
GPU: 2070Super - All settings turned down, but still at 3440x1440
Version: 3.3 (latest non-beta as per 2020-12-19)

young guns: 169 - 171
emire dense: 26 - 31
empire empty : 167 - 169

Outstanding! I love that 5950x. Can you report back what your memory timings are what mobo you using?? Thx! You're going to the top of the leader board. And that 5950x is a beast.

and @ maxbain and buzz Remember...trying to see how the cpu's do with zero chance of bottleneck by the gpu. I'm wondering if you both are a little low because of some of your settings. e.g. SSAO should be off. LOD should be default. and textures to low etc etc.

[Max Bain]

and @ maxbain and buzz Remember...trying to see how the cpu's do with zero chance of bottleneck by the gpu. I'm wondering if you both are a little low because of some of your settings. e.g. SSAO should be off. LOD should be default. and textures to low etc etc.

I had SSAO off and LOD default (100) as well as VSync off. Because there was nothing to see on screen (except in the dense save) I am pretty sure that it was not limited by my GPU. Its also absolutely in line with the other 3600 values. Maybe post it with the hint that I had ultra settings except SSAO and VSync off and a high resolution? Running the test is a bit of work for me because of my setup and mods.

Btw. I tested the dense save with SSAO high and VSync triple buffered and the difference was minimal (just 1-2 fps less with 1 the average).

Edit:
@Roeleveld: Thats a good result. You should try 4.0 Beta out. It gives about 20% better fps for me and others.

[spookywatcher]

and @ maxbain and buzz Remember...trying to see how the cpu's do with zero chance of bottleneck by the gpu. I'm wondering if you both are a little low because of some of your settings. e.g. SSAO should be off. LOD should be default. and textures to low etc etc.

I had SSAO off and LOD default (100) as well as VSync off. Because there was nothing to see on screen (except in the dense save) I am pretty sure that it was not limited by my GPU. Its also absolutely in line with the other 3600 values. Maybe post it with the hint that I had ultra settings except SSAO and VSync off and a high resolution?

For sure added. Especially since that is the first 3600 version4.0 report. Much Thx

[SirNukes]

While toying around, I found that turning off the radar can give another 5-10% improvement. Though I assume everyone here has radar turned on.

CPU: 5950X (no overclock)
Mem: 3600 DDR4 (XMS enabled)
GPU: 2070Super - All settings turned down, but still at 3440x1440
Version: 3.3 (latest non-beta as per 2020-12-19)

young guns: 169 - 171
emire dense: 26 - 31
empire empty : 167 - 169

I'm curious what memory that is. In theory the 5950x should only be ~1% ahead of the 5900x, but your empty scene numbers are 10% better than my 5900x with 3600 16-19-19-39 memory. (My old gtx 970 might also be an issue if it's causing excess cpu overhead somehow.)

[Roeleveld]

While toying around, I found that turning off the radar can give another 5-10% improvement. Though I assume everyone here has radar turned on.

CPU: 5950X (no overclock)
Mem: 3600 DDR4 (XMS enabled)
GPU: 2070Super - All settings turned down, but still at 3440x1440
Version: 3.3 (latest non-beta as per 2020-12-19)

young guns: 169 - 171
emire dense: 26 - 31
empire empty : 167 - 169

I'm curious what memory that is. In theory the 5950x should only be ~1% ahead of the 5900x, but your empty scene numbers are 10% better than my 5900x with 3600 16-19-19-39 memory. (My old gtx 970 might also be an issue if it's causing excess cpu overhead somehow.)

Ok, more details:

Hardware:

Memory: G.Skill Trident Z ( F4-3600C16D-32GTZRC )
Timings: 16-19-19-39
Bios claims 3603Mhz (Doubt the extra 3Mhz makes a difference)

Mainboard: Asus Strix X570-E

CPU-cooler: Arctic Freezer II-360

GPU: MSI GeForce RTX 2070 Super Gaming X Trio

More numbers:
CPU runs around 3.9 - 4.1 constantly with the occasional 4.5 added in. (this is while at desktop)

I think the GPU does affect the numbers as well, which is why I added mine.
Am currently in several queues for a RTX3080, when I get that, I can redo the numbers and compare.
And the fact I can keep my CPU and GPU below 65C while gaming also helps.

Also, in case it matters:
OS: Linux (5.10.1 kernel)
NVidia drivers: 455.45.01

[Gavrushka]

I think the GPU does affect the numbers as well, which is why I added mine.
Am currently in several queues for a RTX3080, when I get that, I can redo the numbers and compare.
And the fact I can keep my CPU and GPU below 65C while gaming also helps.

I had some of the poorest FPS scores for modern hardware with young gun & empty empire tests, but I've also a watercooler and ridiculous amount of fans in my case, including a pair of 240mm ones taken off a light aircraft! Temperatures are very low too, no matter how stressed. - I benchmarked all my components, and all performing at least in the top quartile.

The RTX3070 I use is well below the levels you can expect from 3080, but in the test I could pump it all the way up to Ultra and SSAOx4 without affecting the framerate.

One silly question I'll ask (will test it later) is whether running the CPU in 'balanced' power mode (as I do) is likely to impact the framerate at all? I don't see how as it must crank to max power instantly when any given core goes under higher load.

[Roeleveld]

I think the GPU does affect the numbers as well, which is why I added mine.
Am currently in several queues for a RTX3080, when I get that, I can redo the numbers and compare.
And the fact I can keep my CPU and GPU below 65C while gaming also helps.

I had some of the poorest FPS scores for modern hardware with young gun & empty empire tests, but I've also a watercooler and ridiculous amount of fans in my case, including a pair of 240mm ones taken off a light aircraft! Temperatures are very low too, no matter how stressed. - I benchmarked all my components, and all performing at least in the top quartile.

The RTX3070 I use is well below the levels you can expect from 3080, but in the test I could pump it all the way up to Ultra and SSAOx4 without affecting the framerate.

One silly question I'll ask (will test it later) is whether running the CPU in 'balanced' power mode (as I do) is likely to impact the framerate at all? I don't see how as it must crank to max power instantly when any given core goes under higher load.

I use the "performance" scheduler.
The boost clocks might be affected if the cooler isn't already running at higher RPMs, but not sure how that would affect it all.

What might affect it is additional stuff running in the background.

[SirNukes]

Also, in case it matters:
OS: Linux (5.10.1 kernel)

Ah, I bet that's the reason. I'm on modern windows 10, which from what I understand can be around 5-10% slower than linux (or older versions of windows, even older win 10).

Thanks for the details. Now I can stop worrying that I bought poor memory.