Just got a GTX 1070 but BF1 runs only a little better why?

N

ng4ever

2[H]4U
Joined
Feb 18, 2016
Messages
3,432
Here is my system specs.

i5 2500k @ 4Ghz (for right now)
8 GB of ram
GTX 1070

I know my i5 2500k is old but still I don't feel like I should be getting only Avg: 63.517 - Min: 39 - Max: 111 recorded by fraps.

This is on a multiplayer Conquest server with 64 people at 1920x1080 resolution all details on Ultra.

Should I wait for the patches for BF1 and the newest and greatest nvidia drivers to come out before I make a decision ?
 
While I wouldn't say a minimum of 40 FPS is bad per se, I would say that a 2500K at 4GHz is only equivalent to a modern, stock i5 6500 in speed. So yeah, slight CPU bottleneck on the 64-player servers, which are inherently CPU intensive. You will probably see your minimums hop up with a newer CPU.
 
btw this was my fps with my older GTX 770 4GB

Avg: 57.193 - Min: 33 - Max: 79
 
ng4ever said:
btw this was my fps with my older GTX 770 4GB

Avg: 57.193 - Min: 33 - Max: 79
Click to expand...

That just confirms the CPU bottleneck: Your GPU power nearly tripled, but your minimums stayed roughly the same. I bet if you turned on VSR, your minimums would stay around the 30-40FPS mark.
 
You can find why by monitoring CPU and GPU use.
I use MSI Afterburners graphing tool.
See if any part is near maxing out.
 
KazeoHin said:
While I wouldn't say a minimum of 40 FPS is bad per se, I would say that a 2500K at 4GHz is only equivalent to a modern, stock i5 6500 in speed. So yeah, slight CPU bottleneck on the 64-player servers, which are inherently CPU intensive. You will probably see your minimums hop up with a newer CPU.
Click to expand...

Did you see what my GTX 770 4GB was able to do at the same settings above ? I feel like upgrading to the GTX 1070 was a waste but maybe I am overreacting.

What do you get on a BF1 multiplayer server ? If you played.

I thought I would have at least 60 minimum with a GTX 1070.
 
Bump up your OC to 4.5ghz or higher and see if that improves any reasonable percentage.

Just to be clear, did the settings for the game change as well? (If you left it automatic?? maybe it changed itself to suit the card?)
 
KazeoHin said:
That just confirms the CPU bottleneck: Your GPU power nearly tripled, but your minimums stayed roughly the same. I bet if you turned on VSR, your minimums would stay around the 30-40FPS mark.
Click to expand...

What is VSR please ?
 
ReaperX22 said:
Bump up your OC to 4.5ghz or higher and see if that improves any reasonable percentage.

Just to be clear, did the settings for the game change as well? (If you left it automatic?? maybe it changed itself to suit the card?)
Click to expand...

Ok. No.
 
ng4ever said:
Another question could a new nvidia driver help or BF1 patch ?
Click to expand...

Drivers are always being updated, so long as you have the most recent Nvidia driver, you should be getting the optimal perf available.

This is sounding more and more like a CPU bottleneck.
 
KazeoHin said:
Drivers are always being updated, so long as you have the most recent Nvidia driver, you should be getting the optimal perf available.

This is sounding more and more like a CPU bottleneck.
Click to expand...

Ok thank you.

Will DX12 help any ?
 
If you have not uninstalled the previous drivers prior to installing the 1070 then that would be a route I'd take, then install newer drivers. CPU limitation could be a factor.
 
I would be overclocking that 2500k more myself. At the very least it would answer the question whether you are cpu limited. 4.5ghz for sandy really isn't that hard an overclock honestly. Some people spent years bouncing their sandys off 5ghz.
 
DX12 is showing a decrease in performance on BF1. The only BF which uses dx12. Honestly, the frostbite engine makes very good use of the cpu, I doubt you will ever see dx12 make a noticable impact on frostbite vis a vis the cpu. GPU is yet to be determined.
 
have you looked in your Driver control panel ?

tried other drivers ?..oh wait ya cant on a 1070 ..ask Nvidia
or lower your settings for now
 
ng4ever said:
So what is VSR ?
Click to expand...

VSR is driver-force supersampling. It essentially makes the GPU render at higher resolutions and then samples the image down to run at native. Its a visual-quality setting that can help make CPU bottlenecks more apparent.

Enable 2x VSR and minimum FPS stays the same: major CPU bottleneck.
 
Now, what would you be making a decison on? Your opening post is not clear. Buying a new CPU?

Now, Perhaps I'm being a little too cautious but I would say wait till general release (Oct. 21) and for drivers to be released before confirming your current performance is the best possible. As far as I know the latest nvidia drivers do not mention BF1, and I'm sure when Nvidia releases their Game Ready drivers, they will make sure everyone else knows it.
 
Yes buying a new cpu but I will wait.

I just tried BF1 single player campaign and getting easily 90 fps max with only 63 fps minimum ! So I am happy about that.

Going to try The Witcher 3 next.
 
KazeoHin said:
VSR is driver-force supersampling. It essentially makes the GPU render at higher resolutions and then samples the image down to run at native. Its a visual-quality setting that can help make CPU bottlenecks more apparent.

Enable 2x VSR and minimum FPS stays the same: major CPU bottleneck.
Click to expand...
NVidias version is DSR.
 
Trying Witcher 3 is a good idea, its completely patched, so you can test your CPU without wondering if perhaps DICE will release more optimization patches following release day. (Which they will.)
 
I agree with the others - the CPU is somewhat limiting you. I have two different systems, one for me and one for my mini-me. I can see the performance jump from a 970 to a 1070 so there is no reason why you shouldn't coming from a 770.
 
huge CPU bottleneck, that simple. my 4770K @ 4.5 GHz was at constant 80% usage or higher at 100-130 fps. stock v stock a 2500K is almost 40% slower than a 4770K in properly multithreaded apps which Frostbite engine games are. your performance makes perfect sense. Frostbite games, specifically BF3 and BF4, were a big reason why i upgraded from a 2500K. even in BF3 a 2500K @ 5 GHz would cap you out around 90-100 fps most of the time or even lower on certain maps. BF4 is even harder on the CPU, and BF1 is even harder still.
 
Thank you I will look into getting another processor but I will have to get a new motherboard too.
 
I'm sorry but your CPU won't cut it @ 1080p. Either increase the resolution to 2560x1440 or get a better CPU - 2600k/3770k. The game runs like a boss on my 3770k @ 4.6Ghz/ EVGA GTX 1060 SC and it RUNS exceptionally well on my hexa core sig rig. 100 FPS and up depending on the settings.
 
What speed is your RAM running at?

If anything below 1600, then that could be part of your problem as well. For an older dual channel setup, (you have more than 1 stick of RAM, and in the correct slots to run dual channel, correct?), the higher the speed the better.
 
Nenu said:
You can find why by monitoring CPU and GPU use.
I use MSI Afterburners graphing tool.
See if any part is near maxing out.
Click to expand...

Even though this is helpful its not an entirelly bulletproof way to find CPU bottleneck.a software running 25% on each core of an i5 2500k can still be CPU bottle necked.
for proper detection of CPU bottleneck use process explore.
Check the thread utilization on the process you want to diagnose. if any of those hits (100%/number of logical cores), you have a CPU bottleneck.

Here is an example with 7-zip that is 100% CPU bottlenecked (keeping it at one thread for illustration purpose) in this case since there is few thread than logical core hitting the full core utilization, it is bottleneck in height aka in mhz ( among others) so a faster CPU in height would give faster performance but a faster CPU in width ( Cores) would not help

https://s16.postimg.org/xrvalnirp/CPUbottleneck.png
CPUbottleneck.png


It is easy to misinterpret this as a situation with no CPU bottleneck, when indeed it is CPU bottlenecked, due to how multicore and threads works.
 
Last edited:
cyclone3d said:
What speed is your RAM running at?

If anything below 1600, then that could be part of your problem as well. For an older dual channel setup, (you have more than 1 stick of RAM, and in the correct slots to run dual channel, correct?), the higher the speed the better.
Click to expand...

1600 and yes I have more than 1 stick of RAM plus they are in the correct slots to run dual channel.
 
I'm actually trying to make an easy tool that will run in the background of you game and report back if it reported CPU bottleneck.s how many time and what the longest period is.
The only problem is I'm having a hard time finding out how to get the TID of a PID and get the CPU usage info.

I'm hoping that it will help people a lot by gettinh actually measurements more than guesses and estimate from "strangers" that might run a similar setup but not entirely etc etc
 
SvenBent said:
I'm actually trying to make an easy tool that will run in the background of you game and report back if it reported CPU bottleneck.s how many time and what the longest period is.
The only problem is I'm having a hard time finding out how to get the TID of a PID and get the CPU usage info.

I'm hoping that it will help people a lot by gettinh actually measurements more than guesses and estimate from "strangers" that might run a similar setup but not entirely etc etc
Click to expand...

Cool please let us know when it is out! Thank you.
 
SvenBent said:
Even though this is helpful its not an entirelly bulletproof way to find CPU bottleneck.a software running 25% on each core of an i5 2500k can still be CPU bottle necked.
for proper detection of CPU bottleneck use process explore.
Check the thread utilization on the process you want to diagnose. if any of those hits (100%/number of logical cores), you have a CPU bottleneck.

Here is an example with 7-zip that is 100% CPU bottlenecked (keeping it at one thread for illustration purpose) in this case since there is few thread than logical core hitting the full core utilization, it is bottleneck in height aka in mhz ( among others) so a faster CPU in height would give faster performance but a faster CPU in width ( Cores) would not help

https://s16.postimg.org/xrvalnirp/CPUbottleneck.png
CPUbottleneck.png


It is easy to misinterpret this as a situation with no CPU bottleneck, when indeed it is CPU bottlenecked, due to how multicore and threads works.
Click to expand...

Ah yes, I forgot that a single thread can now be moved between cores.
Although when I had a 2500k with Gigabyte mobo, it didnt exhibit that behaviour, I only noticed it when moving to my 6700K with an Asus mobo.
I saw mention in the motherboard blurb that this is a feature of the motherboard (Maximus Hero VIII).
Not sure if other set ups can do it.

Looking forward to your tool.
 
Nenu said:
Ah yes, I forgot that a single thread can now be moved between cores.
Although when I had a 2500k with Gigabyte mobo, it didnt exhibit that behaviour, I only noticed it when moving to my 6700K with an Asus mobo.
I saw mention in the motherboard blurb that this is a feature of the motherboard (Maximus Hero VIII).
Not sure if other set ups can do it.

Looking forward to your tool.
Click to expand...

I have to disagree wit you.
Threads have always been able to move between cores. They get assigned one quantua ( a time slice of the CPU/cores attention) at a time but can only get from one core in one "instance".
threads moving between cores has been the way it works since i ran dual celeron 366 on abit BP6... If i recall correctly
and defiantly always on an i5 2500k. If it didn't its probably because the software set some kind of affinity itself.

-- edit --
oh I forgot the above screen shoots is from an i7 930 which is my work computer. so threads moving around between cores is definatly not something that is newer than the i5 2500k
-- edit --

The man in charge of how to share out the CPU ressources among threads is in this case Microsoft Windows and it does it differently compared to *nix Os which have an abundent of different CPU sharing methods.
Microsoft uses a heavily modified round robing method ( modified for priority and exception to avoid threading deadlocks etc. but funny enough not for thread collisions on a CMT/SMT CPU)

So basically the harware is in charge of how the ressources are represented. But windows decides how and where to put the software threads.... roughly.


I'll post my tool as soon as it figured out how to get those darn TID's from a process
 
Last edited:
You must log in or register to reply here.
Back
Top