Jump to content
Romby

Getting 50-70 FPS with a 1070 1440p

Recommended Posts

Hey

 

im getting around 50-70 fps in squad on a server with 76-80 people. Im not sure if this is low or to be expected with a GTX 1070 (without OC), I7-4770K OC'ed to 4,5Ghz and 8 gb Ram. Game is installed on a SSD. My settings are 1440p, no Super Sampling, Screen Sharpening 0.50 AA SMAA rest Cinematic/Epic. What do you guys think? Reading the Stat Unit number im guessing that my GPU is the bottleneck (GPU number is the highest). Is this normal?

 

Here you have few screenshots with Stat fps and Stat unit (from Al Basrah and Chora).

9xDvsZ4.jpg

KfaGbYx.jpg

Yf4hsQN.jpg

 

Picture of my settings

bAcRAOr.jpg

3mpuGZu.jpg

 

Thanks for the help!!

Share this post


Link to post
Share on other sites

I'm guessing these numbers are about right. I run a 6700k stock (4.2ghz) with a 1070 at 1080p. I run super sampling 1.5x and my game runs same numbers as yours. I'm looking to upgrade my cpu cooling and then OC to 4.6 - 4.7. I'll then OC my 1070 as well. I'm not going to do this upgrade until I upgrade to 1440p or 4k though.

Share this post


Link to post
Share on other sites

Not really on topic, i am planning to upgrade from 1080 to 1440. Is it worth the difference? Also have the 1070 and i54670k 4,5ghz and 16gb ram.

Share this post


Link to post
Share on other sites
2 hours ago, Rob86 said:

Not really on topic, i am planning to upgrade from 1080 to 1440. Is it worth the difference? Also have the 1070 and i54670k 4,5ghz and 16gb ram.

yes playing in 1440p is great compared to 1080p IMO.

Share this post


Link to post
Share on other sites

Whats limiting your FPS is not the GPu ( draw around 12ms ) but game thread ( 19 ms ) so in conclution its a CPU bound  Limit

Share this post


Link to post
Share on other sites

No GSync monitor I assume? I over estimated how many frames my new PC would get in Squad, but even when the frames dip I think GSync is keeping things silky.

I'd say your frames are pretty damn good, but if you are like me, you had higher expectations of your upgrades.

Share this post


Link to post
Share on other sites
16 hours ago, moppel said:

Whats limiting your FPS is not the GPu ( draw around 12ms ) but game thread ( 19 ms ) so in conclution its a CPU bound  Limit

 

OK. Didn't know how to read those numbers. Tried running game with external measuring software. According to that my gpu was running at 100% and CPU core 1 at 95%. So thought the gpu was the limiting factor. 

Share this post


Link to post
Share on other sites
2 hours ago, Romby said:

 

OK. Didn't know how to read those numbers. Tried running game with external measuring software. According to that my gpu was running at 100% and CPU core 1 at 95%. So thought the gpu was the limiting factor. 

 

You are limited by your GPU according to your screenshots, even though only by little. GPU has highest time on all screenshots = GPU bound. https://docs.unrealengine.com/latest/INT/Engine/Performance/Overview/ "Identify What You Are Bound By"

Share this post


Link to post
Share on other sites

Actually, those numbers on stat unit can be a little bit deceiving. Although the game thread and the gpu thread are seperate, they still have to wait for each other to make a frame at the end. So it can be that the gpu is waiting on the cpu, even tho it is showing times only a few ms more. One way to tell, is to look at the game thread, if its pretty high, up close to the gpu time, but the draw thread is quite low, chances are the gpu is waiting on the game thread.

 

Another better way is to join an empty server, run around a bit and printscreen your stat unit and fps

then join the same map on a full server, and look at the difference. If you are getting much better fps on the empty server, your gpu is waiting on the cpu on a full server.

 

On an empty server, your gpu will be the bottleneck, and on a full server your cpu will be currently. You should also get much better fps on an empty server.

 

 

If none of the above is true, then maybe a 1070 is too slow for 4k, but i doubt it.

 

 

And yes, cpu optimisations are on our immediate hitlist :D

Share this post


Link to post
Share on other sites
5 hours ago, Drav said:

Actually, those numbers on stat unit can be a little bit deceiving. Although the game thread and the gpu thread are seperate, they still have to wait for each other to make a frame at the end. So it can be that the gpu is waiting on the cpu, even tho it is showing times only a few ms more. One way to tell, is to look at the game thread, if its pretty high, up close to the gpu time, but the draw thread is quite low, chances are the gpu is waiting on the game thread.

 

Another better way is to join an empty server, run around a bit and printscreen your stat unit and fps

then join the same map on a full server, and look at the difference. If you are getting much better fps on the empty server, your gpu is waiting on the cpu on a full server.

 

On an empty server, your gpu will be the bottleneck, and on a full server your cpu will be currently. You should also get much better fps on an empty server.

 

 

If none of the above is true, then maybe a 1070 is too slow for 4k, but i doubt it.

 

 

And yes, cpu optimisations are on our immediate hitlist :D

Thx a lot for the this. Kinda hard to understand these Stat Unit readouts. Can i achive the same if i just load a map with training field and console or does it need to be a server?

 

Btw im not playing 4k. Only 1440P :D.

Share this post


Link to post
Share on other sites
6 hours ago, Drav said:

Training field and console should work fine. Post some results, im always interested

I tested it. So i pretty much have the same FPS in Training Field changed to Gorodok as i do on a 80 person server. Perhaps i need to do some more test on different maps. But this position on Gorodok Fps is the same.

 

On 78 player Server:

Nqb1XWf.jpg

 

Singleplayer

bGwUZhd.jpg

Share this post


Link to post
Share on other sites

EDIT : I followed @Skul guide here : 

 

And I have to admit, it did help regarding frames!! ( speed of pc also improved a bit,)

I also overclocked the ram to around 2900Mhz from 2133Mhz.

Could get it working on 2666mhz with looser timings.. and 1 Nanosecond faster than 2900mhz, stil.. 2900Mhz is too low to my taste's.

In some game's 3600Mhz ram can be up to 20% performance.

It seems its stable at 2900Mhz, with cheap cheap Crucial Micron sticks, ( without heatsinks )

 

This is done on a pretty cheap motherboard.

 

 

 

Got pretty much the same with a TI. seems like everyone is hitting those frames in full server, intense teamfights.

does not matter the setting i use.. i use 1.75x or 2.00x super sampling to get my gpu at 99% usage. 

dropping some settings yield maybe 6-7fps tops... but when your in the firefight/bottleneckfest it does not that much matter.

Edited by ChiefSOSA

Share this post


Link to post
Share on other sites
On 2017-07-13 at 10:41 AM, Drav said:

Actually, those numbers on stat unit can be a little bit deceiving. Although the game thread and the gpu thread are seperate, they still have to wait for each other to make a frame at the end. So it can be that the gpu is waiting on the cpu, even tho it is showing times only a few ms more. One way to tell, is to look at the game thread, if its pretty high, up close to the gpu time, but the draw thread is quite low, chances are the gpu is waiting on the game thread.

 

Another better way is to join an empty server, run around a bit and printscreen your stat unit and fps

then join the same map on a full server, and look at the difference. If you are getting much better fps on the empty server, your gpu is waiting on the cpu on a full server.

 

On an empty server, your gpu will be the bottleneck, and on a full server your cpu will be currently. You should also get much better fps on an empty server.

 

 

If none of the above is true, then maybe a 1070 is too slow for 4k, but i doubt it.

 

 

And yes, cpu optimisations are on our immediate hitlist :D

I have an i5 2500k. I suspect I am bottlenecked by cpu. I get descent fps at 1920x1080. If I increase resolution to 2560x1440 then I get fps drop. My question is now: If I buy a new much more powerful gpu and increase the resolution to 2560x1440, can I still play at the same fps as before, or will the increase in resolution also load down the cpu a bit extra, bottle-necking me again by cpu, but this time even lower than before?

 

In short: Does an increased resolution load the cpu more? It seems it shouldn't but I don't know for sure.

Edited by SpecialAgentJohnson

Share this post


Link to post
Share on other sites
On 2017-07-13 at 10:41 AM, Drav said:

Actually, those numbers on stat unit can be a little bit deceiving. Although the game thread and the gpu thread are seperate, they still have to wait for each other to make a frame at the end. So it can be that the gpu is waiting on the cpu, even tho it is showing times only a few ms more. One way to tell, is to look at the game thread, if its pretty high, up close to the gpu time, but the draw thread is quite low, chances are the gpu is waiting on the game thread.

 

Another better way is to join an empty server, run around a bit and printscreen your stat unit and fps

then join the same map on a full server, and look at the difference. If you are getting much better fps on the empty server, your gpu is waiting on the cpu on a full server.

 

On an empty server, your gpu will be the bottleneck, and on a full server your cpu will be currently. You should also get much better fps on an empty server.

 

 

If none of the above is true, then maybe a 1070 is too slow for 4k, but i doubt it.

 

 

And yes, cpu optimisations are on our immediate hitlist :D

If cpu optimisations are prioritised should I stick with my old i5 2500k a while longer? What kind of optimisation can be expected if multi-threading or whatever is improved? 2%, 20%, 50% ? What is the potential for max cpu optimisation?

Share this post


Link to post
Share on other sites
10 minutes ago, SpecialAgentJohnson said:

If cpu optimisations are prioritised should I stick with my old i5 2500k a while longer? What kind of optimisation can be expected if multi-threading or whatever is improved? 2%, 20%, 50% ? What is the potential for max cpu optimisation?

All unknown at this stage, we need all the major game features in place. Then the real optimisation can take place during beta.

Personally I dream of a absolute min 90fps (on some maps I can drop to mid 50's) in 1440p with Squad, I have a high end system and feel it may be possible.

Granted most of the time I am running around at 85fps, so it's not all bad news!

Share this post


Link to post
Share on other sites
On 8/17/2017 at 4:46 PM, SpecialAgentJohnson said:

I have an i5 2500k. I suspect I am bottlenecked by cpu. I get descent fps at 1920x1080. If I increase resolution to 2560x1440 then I get fps drop. My question is now: If I buy a new much more powerful gpu and increase the resolution to 2560x1440, can I still play at the same fps as before, or will the increase in resolution also load down the cpu a bit extra, bottle-necking me again by cpu, but this time even lower than before?

 

In short: Does an increased resolution load the cpu more? It seems it shouldn't but I don't know for sure.

Just for your info, i have a I5 2500K still and i've only got it running at 4.2 ghz.

I've recently added a 1440p monitor and upgraded the GPU from a 760 to a 980ti.

 

This has given me considerable improvements in graphics and FPS and i'm happy running the game at about 50-60 average FPS with high/ultra settings and super sampling at 1.25 (considering it's a old CPU).

 

I was considering a new build but the game is now perfectly playable so i'll just see what happens with optimisation. 

Share this post


Link to post
Share on other sites
8 hours ago, JD85 said:

Just for your info, i have a I5 2500K still and i've only got it running at 4.2 ghz.

I've recently added a 1440p monitor and upgraded the GPU from a 760 to a 980ti.

 

This has given me considerable improvements in graphics and FPS and i'm happy running the game at about 50-60 average FPS with high/ultra settings and super sampling at 1.25 (considering it's a old CPU).

 

I was considering a new build but the game is now perfectly playable so i'll just see what happens with optimisation. 

Thanks Bro! Thinking about going for a  msi 1070. Only problem is msi 1080 is only slightly more expensive, though I think 1080 might be a waste (?) if I stick with my i5 2500k. If I buy a new processor I have to update motherboard and my watercooler and ram and all probably. Don't feel like going for that kind of investment just to play one game. Monitor can only do 1440p by the way.

Share this post


Link to post
Share on other sites
8 hours ago, SpecialAgentJohnson said:

Thanks Bro! Thinking about going for a  msi 1070. Only problem is msi 1080 is only slightly more expensive, though I think 1080 might be a waste (?) if I stick with my i5 2500k.

 

Just use the console command stat unit (and stat fps if you want) and check which of the figures is highest..

 

You can use this article to figure out which is your issue

https://www.unrealengine.com/en-US/blog/how-to-improve-game-thread-cpu-performance

 

So try at 1080, check your figures, then try at 1440.. if the game thread time doesn't change but the draw time goes up to higher than the game thread at 1440p, then having a more powerful GPU will allow you to get roughly the same fps @1440p as at 1080p

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

×