Jump to content
Ouiche

Cpu bench in squad

Recommended Posts

Ouiche   

Hi,

 

As a new player, i noticed a lots of questions regarding CPU performance and squad, but few reliable information available. Giving a rough estimate from an FPS counter in the corner of the screen without may not be the most accurate way to assess a CPU performance - it vary a lot between maps, player count, server, and it's actually not displaying everything.

 

How to solve that? Fraps frametime benchmark + frapsbenchviewer. Frametime record each frame time in game. Frapsbenchviewer convert them into a graph. I don't think the bench is affecting the performances, i did not noticed any unusual performance loss.

 

1/ Use the benchmark option in fraps with the "frametime" option (work with the trial version)

2/ Once it's done, drag & drop the file in frapsbenchviewer

3/ Clic on "copy chart as image to clipboard if you want to save the picture.

 

Note :

-change the recording key for something else than "F11", the game freak out when i push that key for some reason. (goes from fullscreen to windowed, sometime the resolution change)

-don't forget to start after spawning at the beginning of the game, and to stop as soon as possible after the game end, for a reliable result. (otherwise, you'll get 200+fps in the menu)

-the graph show the frame time in ms. Lower = better. Spikes going up = bad.

 

It will give you the average, the 1% low and 0.1% low. Average is not everything, the 1% and 0.1% low are very important for the overall fluidity (stuttering / micro lag spike are never good for that).

 

Ex: Chora aas V2 with 72 players (few minutes of running since i join mid-game, lot of combat on the way and around insurgent checkpoint with an SVD, then we won by tickets near mosque)

-Ryzen [email protected]

-Gtx1070

-1080p, everything on max but the blur, 125%.

 

My GPU is never fully utilized due to the CPU bottleneck (usually, ~70-80% max), that graph should display the raw cpu performance without any gpu limit. You can spot the time when i died and had to wait in the menu ("flat" line around 60k). I don't know what's causing the ~40ms frame, it seems to be a recurring event but i don't feel a lag. Could it be the map? I will try with full texture loading next time and see if it change something.

 

It would be interesting to have various CPU tested, for those looking for an update. I'll record / post more benchs if you're interested. (but you can do it too, it's really easy). Of course, two game can't be compared but it can give people an general idea on how a cpu is going to perform in a real game.

 

(clic on the picture for full resolution)

c9005d344f7d711aa12ad04a64419568b8a55302

Edited by Ouiche

Share this post


Link to post
Share on other sites
moppel   

To make this transperant you should show via a log each cpu core and its SMT thread usage at a give frametime point.

 

Thes spikes are reported by many players atm , i also get them , did you use Texture streaming while recording this( with a log file for cpu core usage you could see if thwe main thread jumps to another core at these lag spikes )?

 

 

Share this post


Link to post
Share on other sites
Ouiche   
On 24/04/2017 at 0:28 PM, moppel said:

To make this transperant you should show via a log each cpu core and its SMT thread usage at a give frametime point.

 

Thes spikes are reported by many players atm , i also get them , did you use Texture streaming while recording this( with a log file for cpu core usage you could see if thwe main thread jumps to another core at these lag spikes )?

 

 

 

I used texture streaming yes. I deactivated SMT (amd hyperthreading) to see if it change something (and it's not really noticeable). For the thread swap, i have no idea how to record that accurately. I don't saw core changing the few time i checked.

 

edit : all tests are done in 1080p, ultra, 125%.

 

With texture streaming: Al Basrah - 74 players. Us side, start from FOB, calm until 1/4th of the game,  heavy fighting in the town (north area) until the end (rocket spam, waves of soldiers, building to building fight)

b71a8315de0c22b7590c3147af55cbb609d60a52

 

Without texture streaming (Al Basrah - 74 players. Insurgents side, joined mid-game, spawned in the town, random encounter moving into the buildings, long range fight trying to get close to apartments, had to fall back on refinery, heavy fighting - 20+ soldiers.)

3d2594d3f2b85cee5e0fe8c70004dc9ab2cf87bf


Overall, there was more action in the second recording and no driving time altering the average fps, but i think there is more spikes without texture streaming.

 

Standing in the fire range, not moving at all, texture streaming on...

 

7cd1f797d1188dcfa9d9402462239bcae4067959

 

And off.

 

a50650d36f6b7cb1731ce301d503b0c1015be6d4


I still got spikes! There not as high / frequent as regular gameplay, but there is definitely something.

 

Same firing range position, 200%

190365ff498aada53dbba5c3c322d8003c372fec

No lag spike anymore. The fps are obviously lower, but they're also way more stable.

I'm a bit confused. I was thinking the spikes could be caused by the gpu switching frequency / vcore due to the low use, but it do not seems to change at all under gpuZ (maybe it's so fast it's not recorded?)

Edited by Ouiche

Share this post


Link to post
Share on other sites
moppel   

Is your 1070 Oced , maybe it runs into Power/Templimit and causes these spikes ?

do you get these spikes to when you limit your FPS to like 60 at FHD?

Share this post


Link to post
Share on other sites
Ouiche   

No oc (factory frequency), temperature is far from the limit (around~65-68c° in gaming, the card can go up to 82 without throttle) and the power limit... i don't think so, in short the Pascal GPU is designed to throttle constantly. (it automatically adjust / reduce the maximum frequency according to the maximum TDP, temperature, voltage and load, it's always hitting a limit). It's possible to increase that power limit but there is little to no benefit unless you're overclocking.


I would be interested to see how other computers are doing on that matter.

Edited by Ouiche

Share this post


Link to post
Share on other sites
moppel   

Ill check this evening when i have time to Fire up Gaming rig , ill suggest to test Firingrange at first because its a bit more reproucable situation.

 

Sadly i have no ryzen CHIP:

 

CPU wise i got : I7 5820k,I7 4790,2xXEON E5-2687W

GPU wise:GTX1080,GTX1060 and a 290X ( but its WC a swapp would mean to much efford)

 

 

 

Share this post


Link to post
Share on other sites
moppel   

Had only limited time so made a short AB log and transfered to Calc. Basicly i see also a big fluctuation in Frametime even in Firingrange. 1440P starts to be GPU bound. Resolution is a little low ( 1 s ) but i think ist good enough , maybee ill make a better one with more time.

 

 

 

 

Squadfps.png

 

Share this post


Link to post
Share on other sites

I can also confirm that HW Monitor (CpuZ) reports average utilizations for CPU 52% and GPU 66% while playing Squad at 35-42 fps (my setup below)

 

Edited by Polo Besscobar

Share this post


Link to post
Share on other sites
moppel   

Jepp thats the avarage observation in Squad, GPU is limited by CPU even with high IPC + clock chips like 6700K/7700K. Squad chews hard on some serial code it can not distribute over the cores therefor it creates a Bottleneck. looking forward to animations rework an be curious what it does.

 

 

Share this post


Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now

×