CPU performance tests

[Xanoxis]

If anyone is interested, Zen 3 5600x 3200mhz DDR4 finished replay on 13:50~. Lowest speed was around +2/+3, dropped for a sec to +1 on air fight. In lobby benchmark gives 100~ score. No overclock on cpu overall.

I think similar results can be achieved with newest intel cpus, as I would assume most of it comes from good IPC in recent gen cpus.

[epic-bennis]

+1 in airfights is interesting. Seems as if sup com is aging like wine.

[tatsu]

@xanoxis wow!
i think it's the ram speed that's really helping here.

[TheVVheelboy]

Well it's been said again and again that ram is important for supcom but you people never remember that part.
Also might as well post the old benchmark I did:

"Finished the first run for my standard setup
3.8ghz r5 2600(undervolted) 2666mhz ram 12-13-13-13-44 Lobby score at 197 points avarage after 10 runs.
It took 16:31:22 to finish the replay at avarage from 3 runs.
To make it easier to bench I stop the game at 0:10 mark and then set sim speed to +10. Then start simultaneously the replay and the timer.
I hit the stop on timer the moment replay comes to halt."

[maudlin27]

i5 11600k (no overclocking) with 16gb 3200mhz RAM took about 13m7s to run the replay (pausing the game at the start, setting +10 sim speed, then starting timer and unpausing)

[Xanoxis]

Fast RAM is good addition, but to be fair, I used the same setup for DDR4 on Zen 1700 3.7ghz OC, and lobby benchmark was around 240~ and I was dropping to -1 on somewhat big games and air fights. Big jump in IPC in last two CPU gens is what's helping the most.

I think with Zen 4 and DDR5 in 2022, all cpu issues in FAF will be solved, even cheapest cpu from it will be enough to run everything smoothly. But sadly, everyone would need to get new cpus, and that won't happen for years.

[ZLO]

Intel Core i7 7700
Z270 mobo
Memory Type DDR4
Memory Size 16 GBytes
Channels Dual
Memory Frequency 1069.1 MHz (1:16)
CAS# latency (CL) 15.0
RAS# to CAS# delay (tRCD) 15
RAS# Precharge (tRP) 15
Cycle Time (tRAS) 36
Row Refresh Cycle Time (tRFC) 374
Command Rate (CR) 2T
Uncore Frequency 801.8 MHz

benchmark-replay result 18 minutes

[newhope]

10700 FA cpu score 118:)

[Exselsior]

I’ve got a Ryzen 5600x with a beefy aftermarket cooler and a healthy 64gb of DDR4-3600 ram giving me what I think is the fastest score I’ve seen of 95. When I’m back in town I’ll run the benchmark from earlier in the thread because I’m curious

[epic-bennis]

I run the Intel 11950xt founders edition game.pref mod at 75 score

[tatsu]

run the replay!

[epic-bennis]

That feature is not supported by my setup

[ZLO]

HP Laptop 15s-eq2025ur
Ryzen 3 5300u
integrated GPU
Not very expensive
Not very powerfull
Pretty good for 15 watt
Very very loose default memory timings
No idea what overclocking adjustments are possible
Bios looked very basic.
Replay run time 19:57 (single run)
25-30 FPS during +10 in a replay. Graphix on low except for terrain fidelity on medium.
Forgot to check CPU score and i don't care about it =\

How do i make a freaking spoiler
I did not have time to update AIDA64, that might theoretically be important.
Aida64 mem-test looks pretty horrifying, especially considetring this CPU has not that much cache

AIDA 64 CPU-queen test

CPU-Z bench and other screenshots (results are good, but CPU probably did short term boost during that time)

[brainstormer]

Couldn't run the replay, after the loading I got a black screen with cursor.
???

[Jip]

That means something crashed, likely UI related.

[Deribus]

@deribus said in CPU performance tests:

Intel Core i7-7700HQ with 16GB 2400 MHz DDR4: 21:50

Xeon E3-1245 v3 with 32GB 1600 MHz DDR3: 20:14

Upgraded to i5-12600k with 32GB 3500 MHz DDR4: 11:01

Almost half the time

[prodromos]

@zlo

Any gpu with 1 teraflops single precision or more theoretical performance is enough for supcom graphics.
That goes back a decade or more. For example first cards to achieve this were Radeon 4850 and gtx 470.

Regarding integrated graphics most APUs today do pack about 1 teraflops of performance in their gpu, but I would not lose my sleep over them, because they are usually found in expensive packages (more expensive than some decent gaming laptops). And I would not pay more to get some shiny surfaces and good for nothing substance.

Regarding CPUs, any cpu with single core geekbench score of about 400 is good enough for most 1v1 scenarios, with the exception of 20x20 or bigger maps(which is depressing if you think how much silicon is wasted in modern smartphones). Any cpu with single core geekbench score of about 1000 will get you through almost all scenarios,

except the "epic" games of 12v12 players on 81x81 maps. I guess from this viewpoint SupCom is still not played optimally, though latest Alder Lakes should be good enough for a full blown Seton's without a hitch.

PS. And yes , I am old enough to have seen bans on people trying to play Seton's if their cpu did not have a GB score of 400, which was considered to be awesome, getting bans when the demand became 500 , 600 and so on. All these years no-one acknowledging the fact that they did not actually have a Seton's capable cpu

[StevenDS123]

@jip said in CPU performance tests:

This is a cool thread - perhaps more people can join in and do some standard benchmark test to see how a CPU + RAM configuration works in practice? Especially today a computer to run Supreme Commander decently shouldn't be that expensive.

Edit: I have a super-duper expensive laptop that throttles down. My information wouldn't be too relevant I believe

CPU Benchmark Online” is a processor performance test allowing you to test online your CPU speed my account access and to find out how fast is your processor in comparison to others. You do not need to install kohls charge or configure anything — the test works directly in the browser...

[Cyberdyne_Skynet]

I got an 10700k @5.2 allcore, 4.9 cache, 4x 16gb 3800mhz Ram with cl16

got 89 points in the lobby cpu benchmark

any ideas what the new 12900k is able to hit there? maybe something like 75?

but even this is for sure not enough horsepower for larger maps with thousands of units, or bigger air battles.... xD

[Sprouto]

Actually, map size has almost no impact on performance except when it comes to AI pathfinding, and that's mostly due to the number of pathing points a larger map will have, in general. Map size does consume more memory, up front, which can be a limiting factor for very large games due to the maximum memory available, but performance is pretty much all about TOTAL unit count - more than any other factor.