NS2 and CPU Speed scaling test
Cr4zyb4st4rd
United Kingdom Join Date: 2012-08-09 Member: 155200Members, NS2 Map Tester, Reinforced - Supporter, Reinforced - Silver, Reinforced - Gold, Reinforced - Diamond, Reinforced - Shadow
United Kingdom Join Date: 2012-08-09 Member: 155200Members, NS2 Map Tester, Reinforced - Supporter, Reinforced - Silver, Reinforced - Gold, Reinforced - Diamond, Reinforced - Shadow
So a while back I made a thread comparing NS2 and CPU core scaling This test only ever went over CPU Core count and not CPU SPEEEEEED, which at the time was only tested upto 4.25Ghz. I've been using a 4.7GHz as my daily clock for some time, so I figured why not see how NS2 scales over speed too.
Now this test isn't 100% identical to the previous, this time my system is
A ) running lightly faster
B ) I introduced 18 bots into the testing to give a real "worst case scenario"
C ) Run on low settings to avoid all possible GPU limitations
D ) CPU Multiplier was changed within Windows so I could keep the same NS2 run loaded for each tested speed
Specs:
CPU: i7-5960x @ 4.7GHz (NS2.exe affinity set to 4 cores only)
GPU: SLI GTX 980
RAM: 32GB DDR4 @ 2666Mhz
OS+Game SSD: 512GB Samsung SM951
OS: Windows 8.1 64bit
Settings used:
Vanilla NS2, no mods
How to reproduce:
Results:
So this one is pretty clear cut, the faster your CPU, the better NS2 will perform, no question about it.
Just a look at the improvements on average FPS.
The 3.0 -> 3.5Ghz FPS jump being larger than the rest is most likely due to the GPU clocking up more. It was clocking down when things were being heavily CPU limited at lower speeds. This is a speed jump of ~125mhz from 1126 to 1252mhz on the GPU.
Remember this test is not "real-world" it's probably not even "worst-case" because that would be a beacon during base rush on wooza where even I see ~30fps at times. This is an estimated average look at where your CPU can fit in.
How does my CPU compare?
How do you find out where you stack up? There's a great website/program https://www.cpubenchmark.net/ and PassMark PerformanceTest. I've run each test on the same clock speeds and core counts as I did for NS2, this gives me a overall CPU score, you can use the site above or download the software and test your own system to see.
Tested as a 4 Core i7 (all i7's are hyper threaded, so tests on their site will be 4c+4ht and not necessarily as good in performance inside NS2 which really doesn't utilize past 4 threads, see my original test)
Speed - CPU Mark Score -- "if it had i7 HyperThreading"
So look your CPU up on CPUBenchmark and see how it's score compares to the above numbers to see roughly where your CPU is for comparison. Do note, while a 4790k etc has a higher score, it will be using 4cores and 4 hyperthreads on their official benchmarks, but it's performance of 4 cores at stock speeds would be less than the above scores. So you could have a higher score and lower performance for that reason.
Edit: A little note on the CPUBenchmark score. Golden told me his 3770k on the chart should outperform my CPU because it has a score of 9591 and if he overclocked to 4.3 it would have a higher score. That's due to the 3770k being tested on their end with 4 cores and 4 hyperthreads.
We're trying to compare NS2 performance and not overall CPU performance, which is why it's limited to a 4core test.
So for actual comparisons to a real world i7 you should be able to add 30% additional to the above scores to get "i7 results" i5's don't have HT and would compare much more realistically to the original numbers. I've added this to the above chart for you.
Now this test isn't 100% identical to the previous, this time my system is
A ) running lightly faster
B ) I introduced 18 bots into the testing to give a real "worst case scenario"
C ) Run on low settings to avoid all possible GPU limitations
D ) CPU Multiplier was changed within Windows so I could keep the same NS2 run loaded for each tested speed
Specs:
CPU: i7-5960x @ 4.7GHz (NS2.exe affinity set to 4 cores only)
GPU: SLI GTX 980
RAM: 32GB DDR4 @ 2666Mhz
OS+Game SSD: 512GB Samsung SM951
OS: Windows 8.1 64bit
Settings used:
Vanilla NS2, no mods
How to reproduce:
Originally copied from the last thread, but slightly modified for how this test was done
Fairly simple, start training on summit
in console enter the following:
scenload xl
damage 0
addbots 9 2
addbots 9 1
freezebots
If you don't set the damage multi you will die.
The bot commands add 9 bots to each team
Freezing the bots means they stay in the same place so not to affect the framerate each run
Then I started all my benchmarks the second I start moving from the center of marine base, walk with axe to reactor and then running with rifle to the end, should take 2mins pretty much everytime, not 100% the same every time but closer than pub testing and better than an empty map.
Fairly simple, start training on summit
in console enter the following:
scenload xl
damage 0
addbots 9 2
addbots 9 1
freezebots
If you don't set the damage multi you will die.
The bot commands add 9 bots to each team
Freezing the bots means they stay in the same place so not to affect the framerate each run
Then I started all my benchmarks the second I start moving from the center of marine base, walk with axe to reactor and then running with rifle to the end, should take 2mins pretty much everytime, not 100% the same every time but closer than pub testing and better than an empty map.
Results:
So this one is pretty clear cut, the faster your CPU, the better NS2 will perform, no question about it.
Just a look at the improvements on average FPS.
The 3.0 -> 3.5Ghz FPS jump being larger than the rest is most likely due to the GPU clocking up more. It was clocking down when things were being heavily CPU limited at lower speeds. This is a speed jump of ~125mhz from 1126 to 1252mhz on the GPU.
Remember this test is not "real-world" it's probably not even "worst-case" because that would be a beacon during base rush on wooza where even I see ~30fps at times. This is an estimated average look at where your CPU can fit in.
How does my CPU compare?
How do you find out where you stack up? There's a great website/program https://www.cpubenchmark.net/ and PassMark PerformanceTest. I've run each test on the same clock speeds and core counts as I did for NS2, this gives me a overall CPU score, you can use the site above or download the software and test your own system to see.
Tested as a 4 Core i7 (all i7's are hyper threaded, so tests on their site will be 4c+4ht and not necessarily as good in performance inside NS2 which really doesn't utilize past 4 threads, see my original test)
Speed - CPU Mark Score -- "if it had i7 HyperThreading"
Speed - Only 4 cores -- "if it was i7 with HT" 1.2GHz - 2638 -- 3430 1.5GHz - 3277 -- 4260 2.0GHz - 4362 -- 5670 2.5GHz - 5496 -- 7144 3.0GHz - 6599 -- 8578 3.5GHz - 7759 -- 10086 (i7 4770k) 4.0GHz - 8857 -- 11514 (i7 4790k) 4.5GHz - 9872 -- 12833 4.7GHz - 10326 - 13423
So look your CPU up on CPUBenchmark and see how it's score compares to the above numbers to see roughly where your CPU is for comparison. Do note, while a 4790k etc has a higher score, it will be using 4cores and 4 hyperthreads on their official benchmarks, but it's performance of 4 cores at stock speeds would be less than the above scores. So you could have a higher score and lower performance for that reason.
Edit: A little note on the CPUBenchmark score. Golden told me his 3770k on the chart should outperform my CPU because it has a score of 9591 and if he overclocked to 4.3 it would have a higher score. That's due to the 3770k being tested on their end with 4 cores and 4 hyperthreads.
We're trying to compare NS2 performance and not overall CPU performance, which is why it's limited to a 4core test.
So for actual comparisons to a real world i7 you should be able to add 30% additional to the above scores to get "i7 results" i5's don't have HT and would compare much more realistically to the original numbers. I've added this to the above chart for you.
Comments
If I am not mistaken bots are/?were? quite CPU intensive, so the (realistic) values may be higher by some constant.
I assume you do IF you have the same GPUs as the OP.
How many cores do you have @_INTER_ ? If you have 4cores and 4ht or 6/8 core amd then thats why. More cores will help tbe passmark score but ns2 will only really use 4. If you only have 4 then you are running ns2 and everything else on those 4.
Its not the easiest way to compare but should be good enough for estimations.
How do you make those FPS graphs? I may just do one just to give you some baseline, what even the worst Intel CPU you can get today can handle (Celeron G1610). I will not use bots though. They DO slow down the game by about a third for 16 of them. But I think I even use higher quality settings than that in the OP (at least a "rich infestation").
Unfortunately the comand does not show it correctly in all cases. It might show as GPU bottleneck eventhough it's the CPU thats holding you back. I was told that some time ago in the tech support forum.
Didn't they make you do the p_logall command to figure that out? Regardless, r_stats is useful for most.
CPU: Intel Celeron G1610
GPU: Radeon 280X
Stat Capture: Fraps
Environment: real public play, server Wooza, ~40 players, ~55 ping
Settings:
There was some weird red-plugging at the end, so that data is probably partially invalid.
As I told you: "mostly 60". Average of the graph is 60.5.
At the beginning of this year I was playing ns2 on an intel G1620 with a Nvidia 750ti. You have a better gpu. I also mostly held 60fps, but the interesting part is I had more drops that lasted longer into the 40's range. I wonder if your more powerful gpu is keeping your fps higher, or if it is the performance improvments since then.
Anyway, @_INTER_ I would not consider you with your i7 CPU bottlenecked at all (assume from your 4+4 configuration, or do you have AMD CPU?).