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

[pref]

Empire @ dense area: 19-20, occasional lag spikes

I'm curious about the lag spikes mentioned by pref in the above (old) post. I can't find any other players mentioning it and if other hardware setups are having similar issues. I have noticed the spikes are a lot more common if I hog some RAM with background processes, so I suspect it is some kind of garbage collection or something (I'm considering getting another 32Gb).
So has anyone had the lag spikes issue and solved (or reduced) it somehow?

It's not the memory, only have 16g but game never uses much more then 10 iirc, always have a couple g free.

What affects performance the most is the turret count, i assume their targeting (los checks, acquisition) consumes lots of CPU resources. Removing the turrets improves FPS and reduces lag spikes considerably.
Later on i built another huge plex right next to this one (they overlap), but removed almost all turrets and FPS got better then on this save by 20-40% as i recall.

What also helped on both spikes and FPS was changing memory timings, that put the fps to 24 i think.

[Imperial Good]

Overall I can say that the upgrade gave me an improvement of about +20% in fps and reduced the loading times by about 30%. I expected a bit more fps but I also read that a Ryzen 5600x does not have 100% performance on older chipsets like the x470.

+20% is what is to be expected moving from the 3600 to a 5600X. The 5600X is faster and in theory beats anything Intel currently offers but it is not an order of magnitude faster.

Tightening your memory timings might help gain a few extra percent. I suspect much of the performance improvements with 4.0 are from reduction in the amount of memory bandwidth used, as such increasing memory speed may no longer yield as much gains as it did before.

[SirNukes]

Tightening your memory timings might help gain a few extra percent. I suspect much of the performance improvements with 4.0 are from reduction in the amount of memory bandwidth used, as such increasing memory speed may no longer yield as much gains as it did before.

X4 doesn't use much bandwidth. Cpu programs generally don't, outside heavily vectorized and multithreaded stuff. Switching dual channel to single channel (halving bandwidth) only lost me 1% performance in the dense empire, and that 1% is consistent with the latency benefits of parallel memory accesses. Though the 4.0 changes could certainly change the impact of memory latency, making it either more or less important (depending on if compute or dram access was reduced more).

[Imperial Good]

X4 doesn't use much bandwidth. Cpu programs generally don't, outside heavily vectorized and multithreaded stuff. Switching dual channel to single channel (halving bandwidth) only lost me 1% performance in the dense empire, and that 1% is consistent with the latency benefits of parallel memory accesses.

This contradicts what some other people have posted, where moving from single to dual channel gave several percent improvement in performance. Additionally people tried various memory clock speeds in older versions of X4 and found that higher clock speed memory gave higher performance in the order of a couple of percent, considerably more than a similar increase in CPU core frequency. Of course higher clock speed memory might also have better timings which might also improve performance.

[Max Bain]

Overall I can say that the upgrade gave me an improvement of about +20% in fps and reduced the loading times by about 30%. I expected a bit more fps but I also read that a Ryzen 5600x does not have 100% performance on older chipsets like the x470.

+20% is what is to be expected moving from the 3600 to a 5600X. The 5600X is faster and in theory beats anything Intel currently offers but it is not an order of magnitude faster.

Tightening your memory timings might help gain a few extra percent. I suspect much of the performance improvements with 4.0 are from reduction in the amount of memory bandwidth used, as such increasing memory speed may no longer yield as much gains as it did before.

Here is my benchmark of my PC specs: https://www.userbenchmark.com/UserRun/37805611

As you can see, my CPU is quite good, even better in single core than the average 5900X (maybe because I set precision boost to level 3?)
https://cpu.userbenchmark.com/AMD-Ryzen ... ating/4087

So I was expecting results like the other 5900x users which are above 30 (with 4.0 beta 4). All new Ryzens should perform similar in X4 because of the single thread performance which is pretty close to each other and X4 will not profit from more than 6 cores anyway.

[Imperial Good]

So I was expecting results like the other 5900x users which are above 30 (with 4.0 beta 4). All new Ryzens should perform similar in X4 because of the single thread performance which is pretty close to each other and X4 will not profit from more than 6 cores anyway.

Some people may be using Linux for their results. The Linux build of X4 has repeatedly been reported to perform better than the Windows one, even when tested on the same hardware by the same person.

In theory the 5900X will always perform better than a 5600X. It has significantly higher sustained boost clock and unlike with Zen2 this really is sustained under the intended workloads. In fact the maximum frequency of Zen3 processors that is advertised is a "minimum" sustained boost frequency with there being a high chance that actual sustained boost frequencies during stock operation will be even higher depending on temperature and the binning of the processor stock. This is why in most games the 5900X will have a measurable performance advantage over the 5600X of a few percent. This is in stark contrast to the Zen2 CPUs where even under single threaded workloads they were prone to all run at roughly the same frequency well below the advertised boost. For example my 3900X has never really been able to sustain its advertised boost frequency for more than a brief instant before falling 100-200 MHz.

[SirNukes]

This contradicts what some other people have posted, where moving from single to dual channel gave several percent improvement in performance.

I scanned through the thread but only saw one other person give channel benchmark numbers, and in their case they doubled rank count when switching to dual channel. Note that my ~1% result was with the same rank count (4) in both configurations, with separately tested ranking being much more impactful (+3% when doubling ranks in same channel config).

(For those who are curious about ranks: when threads happen to be accessing memory locations that map to the same memory bank, they will fight over which row is open in that bank, suffering from increased average latency due to excessive row switching. Doubling memory ranks will cut such infighting in half by spreading memory accesses across more banks. There are also some benefits regarding secondary/tertiary timings, since spreading traffic across ranks helps maintain stable voltages.)

On my old i7 3770k I used an Intel tool to measured used memory bandwidth, and it only reached around 10 GBps for the dense empire test at around 13 fps (in retrospect this was probably gpu limited). Modern cpus with much larger caches should reduce this down, though I haven't found a comparable AMD tool to check this by measuring used bandwidth on my 5900x.

[Hoborific]

specs:
i7-8700k delidded
32GB of 4x8GB (XMP is 2666 at C16) - two different kits so poor overclock ability, motherboard is t-topology not daisy chain, one kit is micron b-die the other samsung c-die
Windows 10 20H2 (19042.685)
game is on public beta running 4.0 beta 4
Game and OS are installed on a gen3 NVMe (trying to reduce bottlenecks, unlikely to matter in this case)
RTX 3090 (worthless for this test)

for CPU:
Stock = 4.3Ghz, cache 4.1Ghz (auto voltage, default bios)
OC = 5.0Ghz, cache 4.7Ghz (35% improvement, unlocked cache/uncore multiplier set to core multiplier minus 3)

1080p windowed, low preset, 4.0 beta 4 unless specified

all FPS ranges were decided by waiting for the load to stabilize for roughly 10 seconds and then eyeballing my overlay, if the FPS increased but did not remain for longer than a second I did not count it, if it dipped and the FPS had already stabilised, I counted it.

Code: Select all

8700k@stock 2133mhz C15-15-15-36(stock) // Stock CPU Stock Ram, 14.064ns latency
young 123-126
empty 20-21
dense 123-126

8700k@OC - 2133mhz C15-15-15-36(stock) // Overclocked CPU, Stock Ram, 14.064ns latency
young 125-129
dense 23-25
empty 123-126

Code: Select all

8700k@stock - 2666mhz C16-18-18-35(XMP) // XMP 12.00ns ram latency
young 125-128
dense 26-28
empty 126-130

8700k@OC - 2666mhz C16-18-18-35(XMP) // XMP 12.00ns ram latency
young 135-138
dense 28-30 
empty 135-139

I stopped bothering to check stock clocks around this point to save time, stock comparison later on in my post against final memory OC.

Code: Select all

8700k@OC - 3200mhz C16-20-20-40(OC) // 10.00ns ram latency
young 143-145
dense 28-30
empty 141-144

8700k@OC - 3333mhz C16-20-20-40(OC) //9.6ns
young 145-146
dense 29-30 
empty 144-148 

this is where I left my RAM alone and started tightening timings, these aren't tight by many standards but compared to stock on the secondaries and things like tRFC and tREFI I did pick up a noticeable performance, albeit diminishing, total core to ram latency is 52.0ns benchmarked by aida64 (I have not yet adjusted tXP which can significantly improve latency in benchmarks such as aida64)

Code: Select all

8700k@OC - 3400mhz C17-22-22-42 // 10.00ns
young 145-147
dense 30-31
empty 143-145

8700k@stock - 3400mhz C17-22-22-42 // 10.00ns
young 139-142
dense 26-27
empty 138-140

up until now the die characteristics of each set did not effect the overall overclock other than loose timings, from here out I have removed one set of ram leaving 2x8gb in dual channel mode (c-die)
the loss of FPS observed here compared to 10ns runs earlier could be explained by t topology layout of motherboard or the loss of 16gb of ram from the system itself

Code: Select all

8700k@OC - 2x8gb 3466mhz C17-21-21-40 // 9.8ns 
young 143-145
dense 28-29
empty 139-141

My steam client didn't like me changing the RAM out and detected that I was on a new computer, this reset my beta branch so for one run I did 3.30

Code: Select all

8700k@OC - 2x8gb 3600mhz C18-21-21-40 //10ns <!3.30 !>
young 138-140
dense 24-25
empty 137-139

8700k@OC - 2x8gb 3600mhz C18-21-21-40 //10ns <!4.0 beta 4!>
young 143-145
dense 28-29
empty 139-142

all memory settings were tested against to ensure they were stable and running well, my final memory stability was checked with Karhu RAM Test to 6400% (which is proposed to be 99.41% certain), all memory tests were done with CPU cache enabled to both ensure my cache/ring ratios were stable. I currently have a fan blowing over the RAM to ensure stability(this was not necessary however I am using a SFF case so the heat soak from my GPU can raise the RAM temperatures to uncomfortable ranges), RAM can be notoriously unstable at temperatures above 45c and I would not recommend just anyone go overclock their RAM without both understanding it from an in-depth guide and having adequate cooling on their case.

It's quite an interesting relationship between the overclocked CPU and the newly overclocked RAM, at a glance I assumed the stronger Single Core Performance (SCP) introduced by overclocking my processor is a significant gain towards the FPS however that is not the likely factor, by overclocking the CPU I was able to increase the cache/ring ratio to match my CPU Overclock at 4.7Ghz (CPU Multiplier minus 3).

The SCP increase appears to not have come in as a factor due to how loosely timed my RAM is, I believe I am near my limit of my processor itself yet still bottlenecked by my IMC, the observable differences in FPS without the CPU overclock in the dense save I would like to attribute to the lower ring ratio used on the CPU which ultimately bottlenecks my system however I am in some scenarios nearing the lows of 90% utilization on some cores which indicates even if I was to replace my RAM I would very soon hit that single core bottleneck.

My final thoughts would be, do not cheap out on your RAM, target under 9ns of latency on the kit you buy. A simple formula to work out latency at a glance is 2000 times CAS latency divided by RAM frequency so 2000 * C17 / 3400mhz is 10.0ns

I will return soon with other comparisons done but for now that is all the information I have to share, if you feel something was done erroneously please reply and I will address it.

Thanks, Hobo.

[Max Bain]

@Hobo: The ram you used in your 2 ram stick test (2x8gb), is it dual ranked ram? Because I read that dual ranked ram is up to 15% faster than single ranked ram. My ram is single ranked, so I ordered a new 2x8 kit so that I can use dual ranked ram because thats the case if you run 4 single ranked ram sticks (rank is not equal to channel).

[Hoborific]

@Hobo: The ram you used in your 2 ram stick test (2x8gb), is it dual ranked ram? Because I read that dual ranked ram is up to 15% faster than single ranked ram. My ram is single ranked, so I ordered a new 2x8 kit so that I can use dual ranked ram because thats the case if you run 4 single ranked ram sticks (rank is not equal to channel).

My sticks are all single rank for four ranks total, they're much more stable having four on my motherboard while overclocking( I assume I bought the one motherboard that's t.topology for Z370 lmao )

I mean other than read/write interleaving from the dual rank sticks? I can't imagine the latency is any better, it should provide more bandwidth however I'm not anywhere near saturating the bandwidth capability of my RAM which is about 50gb at my overclock. I heard a rumor they overclock better but I haven't gone looking for any comparisons yet, I'll look into it, interesting.

[Max Bain]

Whats your radar setting? Is it on high, low or off? I am bit curious why I do have about 15-20% higher fps in low and empty density scneario but lower fps in high density scenario. If you would have radar off it would make sense.

[Hoborific]

Whats your radar setting? Is it on high, low or off? I am bit curious why I do have about 15-20% higher fps in low and empty density scneario but lower fps in high density scenario. If you would have radar off it would make sense.

radar's on low, I believe it was automatic from low preset. feel free to post your specs and FPS from the saves and we can speculate on the reason, would love to do this actually.


I looked into the Dual Rank(DR) vs Single Rank(SR) sticks, as far as I am concerned the performance increase is negligible from the rank interleaving you gain from having DR memory, there's nothing wrong with DR necessarily, it might be considered ideal to use DR 2x16gb sticks on a daisy chain motherboard if the frequency is generally low like sub 3000mhz and you're using something older like the first couple zen generations(like the 3600). I can see it making an improvement due to the horrendous latency you would be experiencing on Zen 2 with budget ram, if it were Intel where the latency is generally really good to begin with or the latest 5000 series of AMD where they vastly improved the latency by using a single CCX for all the cores (assuming you have 8 or less) I couldn't see this making a noticeable difference or even being recommended over 4x8gb single rank.

Would I recommend DR sticks? no, would I use them? probably.

Now I have typed all that and looked at comparisons online I realize now you may have just meant dual rank dual channel in which case it definitely can't hurt, it's harder to drive especially on daisy chained motherboards most manufacturers make it that way on AMD so it will limit your overall OC ability for the RAM itself and probably strain the hell out of your IMC but it's not going to have a negative impact using XMP profiles (assuming the two sets both run together, even at the same rated XMP this can not work)

if the question were, should I buy DR 2x16gb or single rank 4x8gb? I would say take the DR for sure.

I would use 4 sticks in dual channel for dual rank (am right now) however if you're buying new RAM it may be worth investing in one good kit of memory for high frequency low latency like a CL14 3600mhz (or CL16 if it's too pricy), if you need more than 16gb of memory I would go for dual rank sticks, especially if your motherboard/platform/IMC doesn't want to work well with 4 sticks like AMD are known to do.(daisy chain)

If you just bought two new sticks in 2x8gb and you already have say 2x8gb and the new ram is the same frequency and timing but dual rank sticks, I would consider returning them as I don't think your motherboard will play very well with 4 sticks of DR/SR, the IMC will hate its life at least and you may not even reach XMP overclocks. I was not aware they even made 8gb DR sticks

[Max Bain]

radar's on low, I believe it was automatic from low preset. feel free to post your specs and FPS from the saves and we can speculate on the reason, would love to do this actually.

I have posted my results just 1 pages before this. In fact I made a comparson with my older Ryzen 3600 vs my new Ryzen 5600x which gave me +20% fps and 30% faster loading times (yes, I benchmarked loading and save times as well).

I looked into the Dual Rank(DR) vs Single Rank(SR) sticks, as far as I am concerned the performance increase is negligible from the rank interleaving you gain from having DR memory, there's nothing wrong with DR necessarily, it might be considered ideal to use DR 2x16gb sticks on a daisy chain motherboard if the frequency is generally low like sub 3000mhz and you're using something older like the first couple zen generations(like the 3600). I can see it making an improvement due to the horrendous latency you would be experiencing on Zen 2 with budget ram, if it were Intel where the latency is generally really good to begin with or the latest 5000 series of AMD where they vastly improved the latency by using a single CCX for all the cores (assuming you have 8 or less) I couldn't see this making a noticeable difference or even being recommended over 4x8gb single rank.

Would I recommend DR sticks? no, would I use them? probably.

Now I have typed all that and looked at comparisons online I realize now you may have just meant dual rank dual channel in which case it definitely can't hurt, it's harder to drive especially on daisy chained motherboards most manufacturers make it that way on AMD so it will limit your overall OC ability for the RAM itself and probably strain the hell out of your IMC but it's not going to have a negative impact using XMP profiles (assuming the two sets both run together, even at the same rated XMP this can not work)

if the question were, should I buy DR 2x16gb or single rank 4x8gb? I would say take the DR for sure.

I would use 4 sticks in dual channel for dual rank (am right now) however if you're buying new RAM it may be worth investing in one good kit of memory for high frequency low latency like a CL14 3600mhz (or CL16 if it's too pricy), if you need more than 16gb of memory I would go for dual rank sticks, especially if your motherboard/platform/IMC doesn't want to work well with 4 sticks like AMD are known to do.(daisy chain)

If you just bought two new sticks in 2x8gb and you already have say 2x8gb and the new ram is the same frequency and timing but dual rank sticks, I would consider returning them as I don't think your motherboard will play very well with 4 sticks of DR/SR, the IMC will hate its life at least and you may not even reach XMP overclocks. I was not aware they even made 8gb DR sticks

I am speaking about these sources where people compared dual vs single ranked performance (aso 2 vs 4 sticks):
https://www.techspot.com/article/2140-r ... rformance/

or another source:
https://www.youtube.com/watch?v=-UkGu6A-6sQ

As far as I read in the internet, 4 single ranked sticks will run like 2 dual ranked sticks which will gain some performance. I will test this out and see what it brings and report it back here.

[Hoborific]

I have posted my results just 1 pages before this. In fact I made a comparson with my older Ryzen 3600 vs my new Ryzen 5600x which gave me +20% fps and 30% faster loading times (yes, I benchmarked loading and save times as well).

at 3800mhz@C15 you should not be having problems at all, have you measured your RAM latency yet? is DOCP actually on? is the ram properly running in dual channel mode?

I am speaking about these sources where people compared dual vs single ranked performance (aso 2 vs 4 sticks):
https://www.techspot.com/article/2140-r ... rformance/

or another source:
https://www.youtube.com/watch?v=-UkGu6A-6sQ

As far as I read in the internet, 4 single ranked sticks will run like 2 dual ranked sticks which will gain some performance. I will test this out and see what it brings and report it back here.

ah these articles are talking about single vs dual rank in dual channel not on the dimm, you meant you bought another kit to run dual rank dual channel, as in 4x8gb?

4x8gb single rank on a dual channel motherboard = dual rank
2x16gb dual rank on a dual channel motherboard = dual rank

you may have trouble running 4x8gb at 3800mhz C15 (damn that's some nice RAM if those specs are accurate), usually adding more sticks means you lose top end frequency, for example my 3600mhz ram dropped back to 3400mhz at 4 sticks in dual channel dual rank, my motherboard is t.topology and supposed to support 4 sticks properly as well.

let me know how it goes, interesting to know your results

[Max Bain]

at 3800mhz@C15 you should not be having problems at all, have you measured your RAM latency yet? is DOCP actually on? is the ram properly running in dual channel mode?

In fact I have lowered the frequency to 3600 and CL15. This gave me a better performance. What is DOCP? The ram is running in dual channel mode (read by Hardware Monitor)

bro these articles are talking about single vs dual rank in dual channel not on the dimm, did you mean you bought another kit to run dual rank dual channel, as in 4x8gb?

4x8gb single rank on a dual channel motherboard = dual rank
2x16gb dual rank on a dual channel motherboard = dual rank

you may have trouble running 4x8gb at 3800mhz C15 (damn that's some nice RAM if those specs are accurate), usually adding more sticks means you lose top end frequency, for example my 3600mhz ram dropped back to 3400mhz at 4 sticks in dual channel dual rank, my motherboard is t.topology and supposed to support 4 sticks properly as well.

I think today you only get at least dual channel mode motherboards.
I am talking about the whole system. If you have 2 single ranked rams, as I do, as far as I know (but I am not a pro) you can only run them in single ranked mode. But if you add 2 more and put them into 4 slots you have the dual ranked mode. At least thats what I read in the internet.
If you have 2 dual ranked rams they will always run in dual rank mode.
Its the same what you have listed here:

4x8gb single rank on a dual channel motherboard = dual rank
2x16gb dual rank on a dual channel motherboard = dual rank
but
2x8gb single rank on a dual channel motherboard = single rank (right?)

[Hoborific]

Okay my final results, as I am now clearly at the edge of my hardware's capabilities

4.0 beta 4, 1080p windowed, low preset

Code: Select all

8700k@OC - 3400mhz C16-21-21-39 //9.41ns, 51.2ns total latency(aida64)
young 148-151
dense 31-32
empty 144-149

by closing iCUE software which was monitoring temps and causing system latency, I was able to reduce my overall latency from 50.3(ran twice, used average) to 47.1ns(ran twice, exact result) which is a total decrease of system latency by 6.3%

I realised while doing my last benchmark that my native resolution may be a factor in my scores while running in windowed mode as my desktop is 32:9 1440p, I opted to use fullscreen for the first time in these benchmarks as indicated below, for fairness to the controls I placed on myself previously I have also done the same test twice at my original resolution for comparisons to my previous benchmarks and one at half size resolution (1440p) to show performance uplift.

Code: Select all

8700k@5ghz - 3400mhz C16-21-21-39 //9.41ns, 47.1ns total latency (aida64)

4.0 beta 4, 1080p, fullscreen, low preset
benchmark)
young 142-144
dense 35-37
empire 150-153

4.0 beta 4, 1080p, fullscreen, low preset // run#2 to sanity check, young/empire felt inverted
young 142-144
dense 35-37
empire 148-150

Code: Select all

8700k@5ghz - 3400mhz C16-21-21-39 //9.41ns, 47.1ns total latency (aida64)

4.0 beta 4, 1080p, windowed mode, low preset // native desktop resolution
young 150-152
dense 32-33
empire 144-148

4.0 beta 4, 1080p, windowed mode, low preset // half native desktop resolution of 2560x1440
young 143-145
dense 34-36
empire 143-149

perhaps the test itself should specify fullscreen use as to reduce variance between systems as there are a lot of different resolutions available which can apparently make a real impact on the scores.

2x8gb single rank on a dual channel motherboard = single rank (right?)

Yeah (had a chat over discord, sorted out my confusion)

[Hoborific]

CPU: 10900K (no overclock) Version: 4.0 Beta 4
Mem: VENGEANCE® LPX 32GB (2 x 16GB) DDR4 DRAM 3200MHz (XMP enabled)
GPU: EVGA GeForce RTX 3080 FTW3 ULTRA (High Default, no anti aliasing, 5120x1440)
Young Gun : 136 - 139 fps
Dense Emp : 38 - 51 fps
Dense Emp / Empty : Appears to crash beta 4

Could I bother you for some additional information about your RAM, RAM System latency and how you ran the tests? you're the only one who got above 35 fps on dense and I'm wondering how good your system latency is as that's clearly the big factor

[Gavrushka]

@hoborific, your results with an I7 8700 are better than my I7 10700 with decent components to match.

Mind, I'm not overclocked, but core speed is still, in theory, higher than yours.

And one thing I've noticed is the I7 (and I9) seem to have the edge in the Dense Empire scenario.

[pref]

perhaps the test itself should specify fullscreen use as to reduce variance between systems as there are a lot of different resolutions available which can apparently make a real impact on the scores.

The saves were originally created to test how much resources background tasks take away from rendering, was just trying to show that even a relatively bigger empire costs only small amount of frame loss so it's well optimised.
So only the difference between the 2 empty saves matters.

The dense save was added to see which HW can render that on acceptable (30+) fps.

To make a HW test out of this not only resolution but all render speed affecting params should be fixed for the test. If CPU tests are desired then gfx options should be tuned down to eliminate that bottleneck, and for GPUs the other way around (with intensive scenes showcasing as many gfx features as available, and without high amount of background tasks).

[MSterling]

The Linux version of X4 has been known to perform better than the Windows one when people compared it on the same system. There are many possible reasons for this ranging from Linux being inherently faster all the way down to the Windows build possibly being less optimized for the platform. In any case comparing actual frame rates between the two does not accurately represent if the hardware of one system is faster than the other.

X4 is also quite demanding on the GPU. I use a GTX 760 and it really caps the frame rate.

Windows is terrible at forking processes, too, and one reason why Sun made Java (intended to kill Microsoft Windows as a platform) thread rather than fork. This can free up all sorts of resources in new CPU architectures, because a forked process doesn't bother moving anything to the new process (which could be scheduled onto a new core) when it isn't currently being used or changed, but multithreading you can theoretically change any value, dirtying the MLB and data cache and requiring every other thread to be informed of the dirtied registers and memory. Forking can result in a lot of extra memory use, but this is less of a problem when the L3 cache is 32MB... You should therefore also see better speed on the 5000 series and better for the single CCD 5600/5800 because there is no dirty cache at L3 to tell the other CCDs about.