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

[Imperial Good]

for ryzen users even ambient temps matter ... 5 degrees means 100mhz less turbo boost (ryzen is extremely temp sensitive)

This mostly applies if the CPU is running at quite a high temperature already. Otherwise 5 degrees might be 50 or 25 MHz only. It may also not matter too much given that X4 loads a few cores so will be pushing down the boost frequency significantly from that. This may also vary from core to core as some cores are faster than others. I do agree that it is kind of a mess for consistent benchmarks.

Mem clocks on Intel should not matter that much ..

also it would probably brought no further improvements, since it's Intel ...

It really depends on workload. As some people have posted in this thread, it can make a significant difference. Generally as core performance goes up, memory performance has to as well in order to keep the core continuously fed with data. It will make a more significant difference with Zen2 but it still does matter for Intel, especially with their modern high core count high clock speed parts. This is one of the reasons why traditionally high core count CPUs used 3 or more channel memory configurations, with modern server CPUs like Zen2 based EPYC using as may as 8 memory channels to feed their 64 cores.

This is also one of the driving factors behind migrating to use DDR5. With current dual channel DDR4 consumer CPUs the memory performance is increasingly becoming a limit to their performance requiring the use of larger CPU caches, higher transfer rates and tighter timing to work around. DDR5 should buy another major stepping stone for CPU performance by doubling channel numbers and overall doubling transfer speed per DIMM.

[spookywatcher]

I have updated the original post to include all posted benches thus far for ease of looking.
I'll try to keep updating it as more benches are posted to keep it simple and easy to look up diff cpu's

Hoping to get more!

[exogenesis]

Good job with sorting out standardising with save-games.

Here's my benchmarks comparing
Intel-i7 2700K, 4.5Ghz, GTX 980, 1600Mhz-16GByte (= 9yr old CPU, 5yr old GPU)
Intel-i9 9900K, 5.0Ghz, RTX2070, 2600Mhz-32GByte

Found graphic settings (low to ultra) made no difference whatsoever.
Didn't mess with AA, set to default (triple buffer).

Found graphic resolution did make a difference, not sure why if it's CPU bound, so compared 3840x1600 with 1920x1080.

Also showing zoomed-out map (with nothing extra showing), & SETA make a big difference.

Emptyness / Empire game saves gave FPS (as measured by Steam):

Code: Select all

		i7-2700K+GTX980	i9-9900K+RTX2070

3840x1600	 60 / 12	104 / 20
zoomed out map	 57 /  7	100 / 14
SETA & map	  - /  2	  - /  6


1920x1080	101 / 11	134 / 18
zoomed out map	 72 /  6	120 / 13
SETA & map	  - /  2	  - /  5

NB : GTX980 has 'only' 4GByte memory & at 3840x1600 was getting warning (red explanation mark) from X4 saying something like "your graphic card may not cope'.

[pref]

Let me stress that the dense empire save does not provide a meaningful value as it probably kills the render thread too (2k+ station modules plus all the traders), still has like 600 turrets plus the ones on nearby ships in high attention so is not a typical game situation in any way.

What would be interesting is to see how a medium empire affects performance regardless of the currently rendered scene. Imo the best comparison is to check the 2 empty saves - the young gun and the empire one.
Because the OP's question was more about how the game handles cpu and background tasks. And this would be the meaningful comparison as obviously fps will depend on what you do in high attention and you won't sit and battle inside a stupid plex like that. I just posted that to satisfy my own curiosity seems like that was a mistake.
One of the 2 stations exceeds the max plot limit by like 100%, it's really not how the game is supposed to be used in any way.

Thanks for the input, interesting results regardless.

[spookywatcher]

Let me stress that the dense empire save does not provide a meaningful value as it probably kills the render thread too (2k+ station modules plus all the traders), still has like 600 turrets plus the ones on nearby ships in high attention so is not a typical game situation in any way.

What would be interesting is to see how a medium empire affects performance regardless of the currently rendered scene. Imo the best comparison is to check the 2 empty saves - the young gun and the empire one.
Because the OP's question was more about how the game handles cpu and background tasks. And this would be the meaningful comparison as obviously fps will depend on what you do in high attention and you won't sit and battle inside a stupid plex like that. I just posted that to satisfy my own curiosity seems like that was a mistake.
One of the 2 stations exceeds the max plot limit by like 100%, it's really not how the game is supposed to be used in any way.

Thanks for the input, interesting results regardless.

I'll put it in the original post and I'll update my 4.6 ghz 2800 mhz with it. And any new cpu posts that include I'll include it as well. Good Thinking!

[Toombstone]

for ryzen users even ambient temps matter ... 5 degrees means 100mhz less turbo boost (ryzen is extremely temp sensitive)

This mostly applies if the CPU is running at quite a high temperature already. Otherwise 5 degrees might be 50 or 25 MHz only. It may also not matter too much given that X4 loads a few cores so will be pushing down the boost frequency significantly from that. This may also vary from core to core as some cores are faster than others. I do agree that it is kind of a mess for consistent benchmarks.

Well actually i find that the boost algorhytms depend on the PBO limits the problem is that the default settings on any board do a terrible job with boost. Offset+PBO (and patience and tweaking) can scrub off ALOT of heat. For example Prime with 192k FFT when the board is on default heats up my ryzen to about 80C (this is after 24h just to clarify) tweaked with limits and offset my max temps are 72C again after 24h so that is a massive difference. Depending on your luck with the CCX or multiple CCX's this can be a big difference for sustained boost clocks. In any game my cpu does not drop below 4125 before the tweaking i was seeing as low as 3950. The problem with AMD chips is result vary so much from cpu to cpu ...

[Panos]

Ram speed & timings play great role in this game.
On a 3900X have seen the difference between 3600C16 (XMP) and 3800C16 with tight timings (TFAW especially) is night and day. 3800C14 is even better.

I found that ryzen in general likes TFAW. Btw gg on the lottery 3800 14 is one good kit

Is a 5 year old Gskil Ripjaws V 3600C16. B-die Samsung. On the Intel systems used it can do 4133C16 on the Ryzen 3900X can do 3800C14-14-14-28 with very tight subtimings.
Also ofc on Ryzen helps that using 1usmus powerplan with CPPC activated as the came hits 2 cores at 4605Mhz at 100% and the others are working between 4300-4400 (CCD1) and 4500-4550 (CCD0).

[Panos]

Good job with sorting out standardising with save-games.

Here's my benchmarks comparing
Intel-i7 2700K, 4.5Ghz, GTX 980, 1600Mhz-16GByte (= 9yr old CPU, 5yr old GPU)
Intel-i9 9900K, 5.0Ghz, RTX2070, 2600Mhz-32GByte

Found graphic settings (low to ultra) made no difference whatsoever.
Didn't mess with AA, set to default (triple buffer).

Found graphic resolution did make a difference, not sure why if it's CPU bound, so compared 3840x1600 with 1920x1080.

Also showing zoomed-out map (with nothing extra showing), & SETA make a big difference.

Emptyness / Empire game saves gave FPS (as measured by Steam):

Code: Select all

		i7-2700K+GTX980	i9-9900K+RTX2070

3840x1600	 60 / 12	104 / 20
zoomed out map	 57 /  7	100 / 14
SETA & map	  - /  2	  - /  6


1920x1080	101 / 11	134 / 18
zoomed out map	 72 /  6	120 / 13
SETA & map	  - /  2	  - /  5

NB : GTX980 has 'only' 4GByte memory & at 3840x1600 was getting warning (red explanation mark) from X4 saying something like "your graphic card may not cope'.

Overclock your RAM. 2600Mhz is really low and is proven that not only AMD but Intel CPUs greatly benefit from fast ram.
Now if you have 2600Mhz kit depending what you have bought you could be able to do at least 3200C16 with tweaking.

[Imperial Good]

Overclock your RAM. 2600Mhz is really low and is proven that not only AMD but Intel CPUs greatly benefit from fast ram.
Now if you have 2600Mhz kit depending what you have bought you could be able to do at least 3200C16 with tweaking.

This is usually not the case since the reason that memory was not sold as a 3200 kit to begin with means that it is very poorly binned or old technology. If the RAM vendor can make the memory run faster within reason they will and will sell it as a faster kit. This is the entire idea behind binning. There is no "lottery" with slower kits as all the fast ones have already been sold as faster ones.

[exogenesis]

Updated my table to include with the 'empire empy sector' save

Emptyness / Empire-empty-sector / Empire-dense game saves FPS :

Code: Select all

		i7-2700K+GTX980	i9-9900K+RTX2070

3840x1600	 60 / 71 / 12	104 / 109 / 20
zoomed out map	 57 / 27 /  7	100 /  57 / 14
SETA & map	  - /  8 /  2	  - /  30 /  6


1920x1080	101 / 99 / 11	134 / 132 / 18
zoomed out map	 72 / 28 /  6	120 /  47 / 13
SETA & map	  - /  8 /  2	  - /  22 /  5

Not sure why some 1920x1080 FPS are lower than the relevant 3840x1600 ones...

The RAM at 2600MHz is a bit of mysterious 'Alienware specialness' for the Area51-m laptop,
nobody seems to be able to figure why they specced it so low - only way to improve is to install faster RAM
(can't really fiddle with it since it's a work laptop).

The FPS I'm most happy with are the 'zoomed-out-map + SETA' being 20 to 30 on the 9900K,
was getting frustated with about 3 FPS on the 2700K when trying to get fleet movements done.

Yes, the 2700K PC RAM is DDR3.

[Panos]

Updated my table to include with the 'empire empy sector' save

Emptyness / Empire-empty-sector / Empire-dense game saves FPS :

Code: Select all

		i7-2700K+GTX980	i9-9900K+RTX2070

3840x1600	 60 / 71 / 12	104 / 109 / 20
zoomed out map	 57 / 27 /  7	100 /  57 / 14
SETA & map	  - /  8 /  2	  - /  30 /  6


1920x1080	101 / 99 / 11	134 / 132 / 18
zoomed out map	 72 / 28 /  6	120 /  47 / 13
SETA & map	  - /  8 /  2	  - /  22 /  5

Not sure why some 1920x1080 FPS are lower than the relevant 3840x1600 ones...

The RAM at 2600MHz is a bit of mysterious 'Alienware specialness' for the Area51-m laptop,
nobody seems to be able to figure why they specced it so low - only way to improve is to install faster RAM
(can't really fiddle with it since it's a work laptop).

The FPS I'm most happy with are the 'zoomed-out-map + SETA' being 20 to 30 on the 9900K,
was getting frustated with about 3 FPS on the 2700K when trying to get fleet movements done.

Yes, the 2700K PC RAM is DDR3.

Argh laptop, thats why

[spookywatcher]

Updated my table to include with the 'empire empy sector' save

Emptyness / Empire-empty-sector / Empire-dense game saves FPS :

Code: Select all

		i7-2700K+GTX980	i9-9900K+RTX2070

3840x1600	 60 / 71 / 12	104 / 109 / 20
zoomed out map	 57 / 27 /  7	100 /  57 / 14
SETA & map	  - /  8 /  2	  - /  30 /  6


1920x1080	101 / 99 / 11	134 / 132 / 18
zoomed out map	 72 / 28 /  6	120 /  47 / 13
SETA & map	  - /  8 /  2	  - /  22 /  5

Not sure why some 1920x1080 FPS are lower than the relevant 3840x1600 ones...

The RAM at 2600MHz is a bit of mysterious 'Alienware specialness' for the Area51-m laptop,
nobody seems to be able to figure why they specced it so low - only way to improve is to install faster RAM
(can't really fiddle with it since it's a work laptop).

The FPS I'm most happy with are the 'zoomed-out-map + SETA' being 20 to 30 on the 9900K,
was getting frustated with about 3 FPS on the 2700K when trying to get fleet movements done.

Yes, the 2700K PC RAM is DDR3.

The lower ones you are talking about really look margin of error for the most part.

And oh man...Can I just say Thank You for letting us know it was a laptop. LOL
Seriously...I was wondering why it was so nerfed.
I'll update the OP. Much appreciated!

I'll try to do my 6700k @ 4.8 w/ core cache overclocked to 4.6 and my mem to 3000 cl 14 in the morning. They don't like it that much overclocked and it needs to be about 59 degs in the game room for the overclocks to be stable.

[pref]

I'll update the OP. Much appreciated!

Just checked OP, i have all FPS numbers for proper mem timing:
Empty empire: 142 fps
Empty young gun 139-141
Dense empire: 17-20 fps

Btw mem is a Kingston DDR4-2400 KHX3333C16D4/8GX
I just changed the timing preset to performance from normal. No manual changes.

I consistently get higher fps with the empire save. Imo this means background load is very well hidden and the engine is pretty solid from this aspect.

There could be another test for AI battles btw, we could release a 100 ships on a core sector and save when the command is issued or when the fight is already on.

edit: CAS latency 16

[spookywatcher]

Been away for a while, but I whipped my cpu like a rented mule.
Got it to run at 4.8 and 4.9 ghz with 3000 mhz cl 14 and with 4.6 ghz cache core (uncore). Just couldn't get 5.0ghz to run. Really wish I could have.
Anyhow...here are the results:

CPU : 6700k @ 4.8 ghz
MEM : 3000 mhz cl 14
Cache Core : 4.6 ghz
Young Gun : 148 - 150 fps
Dense Emp : 26 - 27 fps
Dense Emp w/ Empty : 146 - 147

CPU : 6700k @ 4.9 ghz
MEM : 3000 mhz cl 14
Cache Core : 4.6 ghz
Young Gun : 149 - 150 fps
Dense Emp : 27 - 28 fps
Dense Emp w/ Empty : 148 - 149 fps

I really do think and agree with Imperial Good that this is showing a memory bottle neck. I might get the 10600k soon and in prep for it I'll be buying some 4000 mhz ram. So I might get the ram first and put that in this system at 4.8 ghz cpu and see what that does.

On a side note, outa take a look at Gamer's Nexus youtube vid about the 10600k and the incredible gains it got from just getting rid of the memory bottle neck.

[zakaluka]

Overclock your RAM. 2600Mhz is really low and is proven that not only AMD but Intel CPUs greatly benefit from fast ram.
Now if you have 2600Mhz kit depending what you have bought you could be able to do at least 3200C16 with tweaking.

This is usually not the case since the reason that memory was not sold as a 3200 kit to begin with means that it is very poorly binned or old technology. If the RAM vendor can make the memory run faster within reason they will and will sell it as a faster kit. This is the entire idea behind binning. There is no "lottery" with slower kits as all the fast ones have already been sold as faster ones.

I agree with this. You can tune your individual kit to find what it's capable of but it takes a fair bit of experimentation & testing to push your ram to its limits. Also ram (1) is more technically complicated to overclock correctly than cpu, and not everyone can do it (2) doesn't usually give you a proportional performance improvement for overclocking, especially because caching is a thing & has improved so much over the years.

If you manage a 10% boost on all timings you might get like 1-2% performance overall, but you have to be certain your system is still stable. Worth doing if you have lots of time to kill and actually know how to test stability.

[spookywatcher]

I pushed GoldenBoy to 5.0 ghz + 4.6 ghz uncore + 3000 mhz Cl 14

Had to delid him and put some metal on him. But he did it. Rock solid too.

CPU : 6700k @ 5.0 ghz
MEM : 3000 mhz cl 14
Cache Core : 4.6 ghz
Young Gun : 149
Dense Emp : 27 - 28 fps
Dense Emp w/ Empty : 148 - 149 fps

There is absolutely zero difference between this and @ 4.9 and very very little @ 4.8

This appears to be a memory bottle neck.

Next I might get some new sticks and see how high the IMC of this 6700k can take. I was thinking about upgrading to the 10600k @ 5.0 ghz but not so sure there'd be much of an improvement. Just guessing 20 fps or less. Might be waiting for Ryzen refresh or rocket lake later this year.

[spookywatcher]

I downloaded some faster ram. Specifically the Trident Z F4-3600C15D-16GTZ running at CL 15-15-15-35 and Refresh set to 390 all other settings in auto

So here's the numbers:

CPU : 6700k @ 4.8 ghz
MEM : 3600 mhz cl 15
Cache Core : 4.7 ghz
Young Gun : 154 fps
Dense Emp : 28-29 fps
Dense Emp w/ Empty : 153

4-6 fps increase compared to 4.8 ghz w/ 3000 mhz cl 14. So not much. Definitely showing it's partially memory bottleneck. The micro stutter seems much better. But in the end was it worth $140 eh...maybe(?)

Maybe on a cool morning I'll try 5.0 ghz again with the 3600 mhz mem.

[spookywatcher]

Bought the 10600k and paired it with the Trident 3600mhz cl15 sticks

CPU : 10600k @ Stock
MEM : 3600 mhz cl 15
Cache Core : Stock
Young Gun : 147 fps
Dense Emp : 24 fps
Dense Emp w/ Empty : 146 - 147 fps

Extremely Disappointed in this compared to my 6700k @ 4.8 ghz. So far a complete waste of money.

I'll try overclocking the 10600k tomorrow.

[spookywatcher]

eh...decided to try a quick overclock on it. So far I have not won the silicon lottery with the 10600k. Just can't get 5.0 ghz stable yet.

CPU : 10600k @ 4.9 ghz
MEM : 3600 mhz cl 15
Cache Core : 4.8 ghz
Young Gun : 153 fps
Dense Emp : 28 fps
Dense Emp w/ Empty : 150 - 151

I was doing these numbers with my 6700k @ 4.8

So conclusion...don't upgrade.

[spookywatcher]

Got it to 5.0 ghz. Definitely not stable for long in prime but stable enough to run x4

CPU : 10600k @ 5.0 ghz
MEM : 3600 mhz cl 15
Cache Core : 4.8 ghz
Young Gun : 153 fps
Dense Emp : 29 fps
Dense Emp w/ Empty : 150 fps

No benefit at all running at 5.0 ghz

I also pushed my ram as hard as I could with the cpu at 4.9 ghz

CPU : 10600k @ 4.9 ghz
MEM : 3700 mhz cl 15
Cache Core : 4.8 ghz
Young Gun : 153 fps
Dense Emp : 29 fps
Dense Emp w/ Empty : 151 fps


** ETA **

This was such a pathetic upgrade I just reinstalled my 6700k and it's mobo. Confirmed that I was getting the same fps results. And will be selling the 10600k and z490.
Seriously... Just upgrade your memory and run it as fast as you can. 4 generations of intel cpu updates equaled no benefit.