Upvote Upvoted 0 Downvote Downvoted
Low cpu usage in game
posted in Q/A Help
1
#1
0 Frags +

Recently purchased a new laptop so I will not have to take a desktop on a train 6 times a year. It is an msi gs63vr (i7 7700HQ, gtx 1060) but I am not getting the framerate I was hoping for in tf2. On my i3 4330, R9 380 desktop I was getting around 120 fps and on the laptop I am getting close to 100fps. I would assume that the gpu is not the issue as it is straight out better than the 380 which led me to believe that the cpu must be the culprit. Looking at the cpu usage on the laptop each of the 4 cores has about 25% load which makes me think that the cpu may not be being utilized fully as when playing tf2 on my desktop the cpu there has approximately 60% load instead. Is there anything I can do based around this to increase my performance or am I stuck like this?

I understand that the laptop cpu does have a lower clock speed but I figured the much newer architecture and the higher core count would lead to better performance from it. Could it be that tf2 is much too focused on single core performance for the laptop to perform better?

Recently purchased a new laptop so I will not have to take a desktop on a train 6 times a year. It is an msi gs63vr (i7 7700HQ, gtx 1060) but I am not getting the framerate I was hoping for in tf2. On my i3 4330, R9 380 desktop I was getting around 120 fps and on the laptop I am getting close to 100fps. I would assume that the gpu is not the issue as it is straight out better than the 380 which led me to believe that the cpu must be the culprit. Looking at the cpu usage on the laptop each of the 4 cores has about 25% load which makes me think that the cpu may not be being utilized fully as when playing tf2 on my desktop the cpu there has approximately 60% load instead. Is there anything I can do based around this to increase my performance or am I stuck like this?

I understand that the laptop cpu does have a lower clock speed but I figured the much newer architecture and the higher core count would lead to better performance from it. Could it be that tf2 is much too focused on single core performance for the laptop to perform better?
2
#2
0 Frags +

You could try my config, it properly balances load for multicore CPUs and modern GPUs. http://www.teamfortress.tv/42867/mastercomfig-fps-customization-config

You could try my config, it properly balances load for multicore CPUs and modern GPUs. http://www.teamfortress.tv/42867/mastercomfig-fps-customization-config
3
#3
9 Frags +

x

x
4
#4
0 Frags +

its normal in TF2
oddly enough if you run TF2 Classic it uses more your CPU and GPU

its normal in TF2
oddly enough if you run TF2 Classic it uses more your CPU and GPU
5
#5
0 Frags +
sageoddly enough if you run TF2 Classic it uses more your CPU and GPU

tf2c uses source sdk 2013 i believe while tf2 uses 2009 so it's not really a fair comparison

windows core % isn't very reliable and is a simplification so i wouldnt trust it. i'd expect the desktop to outperform the laptop but make sure it's set to max performance and try giving tf2 slightly higher priority

[quote=sage]oddly enough if you run TF2 Classic it uses more your CPU and GPU[/quote]

tf2c uses source sdk 2013 i believe while tf2 uses 2009 so it's not really a fair comparison

windows core % isn't very reliable and is a simplification so i wouldnt trust it. i'd expect the desktop to outperform the laptop but make sure it's set to max performance and try giving tf2 slightly higher priority
6
#6
0 Frags +

Yeah the clock speed of each core on the desktop is higher than the laptop.

Feels bad that a £1.5k laptop gets outperformed by a 4-5 year old £700 desktop

Yeah the clock speed of each core on the desktop is higher than the laptop.

Feels bad that a £1.5k laptop gets outperformed by a 4-5 year old £700 desktop
7
#7
0 Frags +
gemm
tf2c uses source sdk 2013 i believe while tf2 uses 2009 so it's not really a fair comparison

windows core % isn't very reliable and is a simplification so i wouldnt trust it. i'd expect the desktop to outperform the laptop but make sure it's set to max performance and try giving tf2 slightly higher priority

No it uses Source 2013.

Firetf2 only uses 3 threads (more or less but ill let setsul explain the details), and i assume that a notebook i7 probably has lower single core performance than a desktop level i3.

you can always try to fiddle around with configs tho to increase your performance.

TF2 scales pretty well, https://www.youtube.com/watch?v=faNeipYH4cw

ElyassuYeah the clock speed of each core on the desktop is higher than the laptop.

Feels bad that a £1.5k laptop gets outperformed by a 4-5 year old £700 desktop

In addition to the config, you can try -dxlevel 98.

[quote=gemm]

tf2c uses source sdk 2013 i believe while tf2 uses 2009 so it's not really a fair comparison

windows core % isn't very reliable and is a simplification so i wouldnt trust it. i'd expect the desktop to outperform the laptop but make sure it's set to max performance and try giving tf2 slightly higher priority[/quote]
No it uses Source 2013.

[quote=Fire]tf2 only uses 3 threads (more or less but ill let setsul explain the details), and i assume that a notebook i7 probably has lower single core performance than a desktop level i3.

you can always try to fiddle around with configs tho to increase your performance.[/quote]
TF2 scales pretty well, https://www.youtube.com/watch?v=faNeipYH4cw

[quote=Elyassu]Yeah the clock speed of each core on the desktop is higher than the laptop.

Feels bad that a £1.5k laptop gets outperformed by a 4-5 year old £700 desktop[/quote]
In addition to the config, you can try -dxlevel 98.
8
#8
0 Frags +

You probably know this but when the laptop is unplugged the CPU and GPU will lower their clock speed because the battery can't deliver them enough power. Like with my laptop I run rocket league at 5fps unplugged and 150fps when plugged in.

You probably know this but when the laptop is unplugged the CPU and GPU will lower their clock speed because the battery can't deliver them enough power. Like with my laptop I run rocket league at 5fps unplugged and 150fps when plugged in.
9
#9
0 Frags +
mastercomsIn addition to the config, you can try -dxlevel 98.

Would -dxlevel 81 not give better performance? That is what I found was the case with my desktop anyway.

[quote=mastercoms]
In addition to the config, you can try -dxlevel 98.[/quote]

Would -dxlevel 81 not give better performance? That is what I found was the case with my desktop anyway.
10
#10
0 Frags +
ElyassuWould -dxlevel 81 not give better performance? That is what I found was the case with my desktop anyway.

No, most of the performance benefit from dxlevel 81 is by turning off effects that can be mostly turned off anyway in dx9, as well as giving an alternative rendering pathway to some old weak GPUs that can't handle dx9. The DX9 rendering pathway more effectively uses GPUs and it also enables a lot of other usages of modern PC hardware.

[quote=Elyassu]Would -dxlevel 81 not give better performance? That is what I found was the case with my desktop anyway.[/quote]
No, most of the performance benefit from dxlevel 81 is by turning off effects that can be mostly turned off anyway in dx9, as well as giving an alternative rendering pathway to some old weak GPUs that can't handle dx9. The DX9 rendering pathway more effectively uses GPUs and it also enables a lot of other usages of modern PC hardware.
11
#11
1 Frags +

2.8 GHz base vs 3.5 GHz, KBL vs HSW though, 100 vs 120 fps seems perfectly reasonable.
More than 3 cores or 2 core + SMT = 4 threads really doesn't do much for TF2.
~2 threads worth of load on a 4 thread CPU = 50% load, ~2 threads worth of load on an 8 thread CPU = 25% load, according to windows. Similar load, similar fps.
I don't see the problem here.

10% higher IPC of a newer architecture does not compensate for 20% lower clockspeed. As you've noticed (and I think it's been mentioned quite few times in other threads) the extra cores do nothing.

mastercomsTF2 scales pretty well, https://www.youtube.com/watch?v=faNeipYH4cw

How do ~30% CPU usage and 10-30% GPU usage imply good scaling? Scaling with what?

2.8 GHz base vs 3.5 GHz, KBL vs HSW though, 100 vs 120 fps seems perfectly reasonable.
More than 3 cores or 2 core + SMT = 4 threads really doesn't do much for TF2.
~2 threads worth of load on a 4 thread CPU = 50% load, ~2 threads worth of load on an 8 thread CPU = 25% load, according to windows. Similar load, similar fps.
I don't see the problem here.

10% higher IPC of a newer architecture does not compensate for 20% lower clockspeed. As you've noticed (and I think it's been mentioned quite few times in other threads) the extra cores do nothing.

[quote=mastercoms]TF2 scales pretty well, https://www.youtube.com/watch?v=faNeipYH4cw[/quote]
How do ~30% CPU usage and 10-30% GPU usage imply good scaling? Scaling with what?
12
#12
0 Frags +

What about the effect from intel turbo boost. The 7700hq can turbo up to 3.8 GHz but I see that it is not always at this speed. Is there a way to force this boost to always be on with safe temperatures or am I stuck with how it is set currently?

What about the effect from intel turbo boost. The 7700hq can turbo up to 3.8 GHz but I see that it is not always at this speed. Is there a way to force this boost to always be on with safe temperatures or am I stuck with how it is set currently?
13
#13
0 Frags +

too bad you didn't ask setsul before buying the laptop :/

But anyway, couldnt you get away with some kind of mini itx system + 1 monitor at each of your places (i'm assuming you're a college student?)

@setsul do you know any computer monitor made with a laptop screen but hookable to regular computers?(so that it can be folded and transported easily)?

too bad you didn't ask setsul before buying the laptop :/

But anyway, couldnt you get away with some kind of mini itx system + 1 monitor at each of your places (i'm assuming you're a college student?)

@setsul do you know any computer monitor made with a laptop screen but hookable to regular computers?(so that it can be folded and transported easily)?
14
#14
0 Frags +
SetsulHow do ~30% CPU usage and 10-30% GPU usage imply good scaling? Scaling with what?

I meant the cores. However, even though the Source Engine scales its usage of cores, it doesn't use them for very important things and when it does multithread, there's a lot of throttling and bottlenecking going on in the Source Engine for synchronization purposes. I was just trying to point out that TF2 does use more than 3 cores.

Also keep in mind that I was at a low resolution.

[quote=Setsul]
How do ~30% CPU usage and 10-30% GPU usage imply good scaling? Scaling with what?
[/quote]
I meant the cores. However, even though the Source Engine scales its usage of cores, it doesn't use them for very important things and when it does multithread, there's a lot of throttling and bottlenecking going on in the Source Engine for synchronization purposes. I was just trying to point out that TF2 does use more than 3 cores.

Also keep in mind that I was at a low resolution.
15
#15
2 Frags +

#12
That is already how Turbo Boost works. If both the temperature and TDP allow it it will boost and in the long term neither will on a laptop.
Also 3.8 GHz is the single core turbo, 3.4 all core, so thanks to windows properly spreading the threads out across cores it might actually be faster with half the cores disabled.

#13
You know that the bottom part of a laptop is the computer? Why would a monitor come with that?

#14
It uses more cores, but it does not scale.
Doesn't matter if they are important things or not, they simply don't take much time.
If you give TF2 100 cores to use then I'm pretty sure it will use all of them for a split second, but even with 8 threads the synchronization takes up more time than is saved. It's barely noticeable because those parts of the code take <1% of the time anyway, so even if they are slower it doesn't really matter. I've explained it in detail before but I really don't want to go over this in every single thread.
I can say "TF2 doesn't utilize more than 3 cores properly" and someone will still post "TF2 doesn't use more than 3 threads", if they haven't done so already. I can then correct them only to tell them that it doesn't make a difference anyway because TF2 still doesn't scale. It's just not worth making the distinction for most.
But you should now better. Using an infinity number of threads is not the same as performance scaling with the number of threads. Because TF2 really doesn't.

Resolution doesn't matter, either you're limited by the CPU or the GPU or weird shit. The only reason to hit a CPU limit at 30% usage is bad/lack of multithreading.

#12
That is already how Turbo Boost works. If both the temperature and TDP allow it it will boost and in the long term neither will on a laptop.
Also 3.8 GHz is the single core turbo, 3.4 all core, so thanks to windows properly spreading the threads out across cores it might actually be faster with half the cores disabled.

#13
You know that the bottom part of a laptop is the computer? Why would a monitor come with that?

#14
It uses more cores, but it does not scale.
Doesn't matter if they are important things or not, they simply don't take much time.
If you give TF2 100 cores to use then I'm pretty sure it will use all of them for a split second, but even with 8 threads the synchronization takes up more time than is saved. It's barely noticeable because those parts of the code take <1% of the time anyway, so even if they are slower it doesn't really matter. I've explained it in detail before but I really don't want to go over this in every single thread.
I can say "TF2 doesn't utilize more than 3 cores properly" and someone will still post "TF2 doesn't use more than 3 threads", if they haven't done so already. I can then correct them only to tell them that it doesn't make a difference anyway because TF2 still doesn't scale. It's just not worth making the distinction for most.
But you should now better. Using an infinity number of threads is not the same as performance scaling with the number of threads. Because TF2 really doesn't.

Resolution doesn't matter, either you're limited by the CPU or the GPU or weird shit. The only reason to hit a CPU limit at 30% usage is bad/lack of multithreading.
16
#16
0 Frags +

Setsul, perhaps you misunderstood what I was trying to assert. I was trying to say that TF2 does scale in the sense that it will use more cores as available, since the person I was replying to was saying TF2 only uses 3 threads. I was trying to correct that error. In my previous post, I thought I made it clear I was already aware of the synchronization issues so not sure why you told me about that.

Setsul, perhaps you misunderstood what I was trying to assert. I was trying to say that TF2 does scale in the sense that it will use more cores as available, since the person I was replying to was saying TF2 [b]only[/b] uses 3 threads. I was trying to correct that error. In my previous post, I thought I made it clear I was already aware of the synchronization issues so not sure why you told me about that.
17
#17
0 Frags +

There's a program called Process Lasso that stabby has in his list of programs that increase TF2 performance.
https://www.reddit.com/r/tf2/comments/31c65p/stabbys_everything_pack_final_releaseall_the_tf2/

stabbyAutomatically manage Core Affinities to counter TF2's bias towards Core0
There's a program called Process Lasso that stabby has in his list of programs that increase TF2 performance.
https://www.reddit.com/r/tf2/comments/31c65p/stabbys_everything_pack_final_releaseall_the_tf2/
[quote=stabby]Automatically manage Core Affinities to counter TF2's bias towards Core0[/quote]
18
#18
0 Frags +

Like I said, using more threads != performance scaling with more threads.
The only thing that scales with the number of threads TF2 uses is the number of threads TF2 uses. That is not scaling and definitely not good scaling.
You should know that terminology. It doesn't help correcting a false statement with another false statement.

Firetf2 only uses 3 threads (more or less but ill let setsul explain the details)

He's aware that it's not quite the truth but there is nothing to be gained by saying

mastercomsTF2 scales pretty well

All it leads to is you getting quoted when someone is mad about their new 16 core CPU not getting any more fps.
Which will lead to another thread like this.
Stop perpetuating that cycle.

Like I said, using more threads != performance scaling with more threads.
The only thing that scales with the number of threads TF2 uses is the number of threads TF2 uses. That is not scaling and definitely not good scaling.
You should know that terminology. It doesn't help correcting a false statement with another false statement.

[quote=Fire]tf2 only uses 3 threads (more or less but ill let setsul explain the details)[/quote]
He's aware that it's not quite the truth but there is nothing to be gained by saying [quote=mastercoms]
TF2 scales pretty well[/quote]
All it leads to is you getting quoted when someone is mad about their new 16 core CPU not getting any more fps.
Which will lead to another thread like this.
Stop perpetuating that cycle.
19
#19
0 Frags +

Yeah admittedly I shouldn't have used the word scale. My bad.

Yeah admittedly I shouldn't have used the word scale. My bad.
20
#20
1 Frags +

yea probably should check ur ram, throtels a lot

yea probably should check ur ram, throtels a lot
Please sign in through STEAM to post a comment.