Fallout 4 CPU scaling

Page 4 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

Headfoot

Diamond Member
Feb 28, 2008
4,444
641
126
Notes from playing through with the new driver on my system [2500k @ 4.5, 16 GB DDR3 1600 CL9, r9 290 @ 1100 Core, 1080p] at Ultra w/ Godrays @ high, Shadow Distance = 4000 in .ini, Tesselation capped at x16 in drivers, and SweetFX ReShade LumaSharpen effect on per the Fallout4 Tweak Guide guide.

I am not noticing any difference in gameplay walking around in the city. I'm still getting huge FPS drops while maintaining 60 fps most of the time. Whatever the driver improves doesn't seem to be affecting me much if at all, from how the gameplay feels and monitoring my CPU and GPU usage real-time. I did install the ReShade LumaSharpen mod since last patch but that isn't very resource intensive so I doubt its impacting my FPS drops here.

I'm reading on the internet the FPS drops may be related to the Vsync implementation in the game. I tried turning it off but I'm very sensitive to screen tearing and I had to turn it back on. For the small time I played vsync off (bPresentInterval = 0 in the .ini) the FPS drops seemed less drastic but that could just be due to where I was in the city.

My problem is definitely not average FPS, but rather the wild swings from minimum to maximum. I'm certain it's CPU/Memory related as both my CPU and Memory are getting dated. My average is probably pretty high since im capped at 60 75% of the time with the other 25% wildly swinging
 
Last edited:

MajinCry

Platinum Member
Jul 28, 2015
2,495
571
136
If you have an AMD card, and use VSync, get RadeonPro and enable triple buffering.

The new drivers do nothing for the insane number of draw calls being made. Nothing can be done about that, unless Bethesda uses the proper API for open world games (Mantle, Vulkan, D3D12).

I'm curious; how do NVidia cards fare on this title, with Phenom IIs, making the same number of draw calls? Would be interested in seeing the fps @ Diamond City's entrance, which makes around 6.3k draw calls with shadow distance @ 7000.
 

SPBHM

Diamond Member
Sep 12, 2012
5,066
418
126
DX11 was the only choice, Mantle is dead and GCN only, Vulkan is not a factor yet and it's to soon for DX12 (tons of people on Windows 7/8.1, immature drivers and OS), if it was 11/2016 complaining about the use of DX11 could make sense,

Skyrim 4 years ago was still DX9, 2 years after DX11 became available (let' not even mention DX10)
 
Last edited:

TheELF

Diamond Member
Dec 22, 2012
4,027
753
126
The new drivers do nothing for the insane number of draw calls being made. Nothing can be done about that, unless Bethesda uses the proper API for open world games (Mantle, Vulkan, D3D12).
The draw calls only account for the "normal" FPS drop,look left -huge amount of stuff- low fps,look right -less stuff- high fps.

The walking around in a low demanding setting and getting huge frame drops comes from disk access,same as any other console port, texture streaming doesn't work in Dx11, yet every game is coded to use it.
 

Head1985

Golden Member
Jul 8, 2014
1,867
699
136
I dont have any FPS drops.I think most fps drops are because slow cpu.Only skylake can provide 60fps in this game.
6700k in this game at 3Ghz=almost match 5960x at 4.2Ghz....
 
Last edited:

Face2Face

Diamond Member
Jun 6, 2001
4,100
215
106
I dont have any FPS drops.I think most fps drops are because slow cpu.Only skylake can provide 60fps in this game.
6700k in this game at 3Ghz=almost match 5960x at 4.2Ghz....

Since I'm playing FO4 exclusively now, this test almost made me pick up a 6700K over the 5820K, but I decided I wanted the extra cores and threads for the lot of video encoding I'm doing now. I grabbed some cheap 3000MHz CL15 Corsair LPX...I'm also going to get the MSI X99S MB and I'll be set.
 

ctk1981

Golden Member
Aug 17, 2001
1,464
1
81
Has there been any tests conducted with FO4 and the i7-5775c to see if its edram helps with the amount of draw calls, or in general the performance?
 
Last edited:
Aug 11, 2008
10,451
642
126
Skylake is only about 10 to 15 % faster than haswell at best. Not sure where the poster got that information, but I find it hard to believe Skylake at 3ghz is equal to 5960x at 4.2. Even 5960x at stock is at the top of most game.gpu charts, probably because of the huge cache. Not saying it is impossible, but maybe an artifact of some sort or skylake had faster ram? The game is insanely sensitive to ram speed.
 

ShintaiDK

Lifer
Apr 22, 2012
20,378
145
106
Skylake is only about 10 to 15 % faster than haswell at best. Not sure where the poster got that information, but I find it hard to believe Skylake at 3ghz is equal to 5960x at 4.2. Even 5960x at stock is at the top of most game.gpu charts, probably because of the huge cache. Not saying it is impossible, but maybe an artifact of some sort or skylake had faster ram? The game is insanely sensitive to ram speed.

A heavily improved uncore at 4.1Ghz plus caches.

If the HEDT cant perform, I doubt the Broadwell-C can. These share the same benefits so to say, tho different implementation.
 

dark zero

Platinum Member
Jun 2, 2015
2,655
140
106
I dont have any FPS drops.I think most fps drops are because slow cpu.Only skylake can provide 60fps in this game.
6700k in this game at 3Ghz=almost match 5960x at 4.2Ghz....
Only Skylake? Only 6700K.. And that is definately a really bad optimized game if it only works on one chip that is crazy expensive...

It's like releasing a game for the Neo Geo.... Without the charm of the console
 

Head1985

Golden Member
Jul 8, 2014
1,867
699
136
Last edited:

Udgnim

Diamond Member
Apr 16, 2008
3,680
124
106
FO4 has been shown to be responsive to RAM speed

sure FO4 is not responding to DDR4 RAM speed instead of uncore?
 

Face2Face

Diamond Member
Jun 6, 2001
4,100
215
106
I finally have Haswell-E results
5960x 4.2ghz 4x4 quad channel DDR4 2800mhz GTX980TI
58.3Fps

I also have results with new AMD drivers
4930K 4.4Ghz quad channel 2400DDR3 cl11 + r9 290 46fps(old was 33fps)

All results:
6700k 4.5Ghz 3000DDR4 cl14 70fps
5960x 4.2ghz 4x4 quad channel DDR4 2800mhz 58.3fps
6700k 3Ghz 3000DDR4 cl14 54fps
3570K 4.8Ghz 1866MHz DDR3 9-9-9-24 52.8fps
Xeon 6core sandy bridge 4.3ghz quad channel DDR3 1600cl9 52fps
4670k 4,7Ghz 2400DDR3 cl11 52fps
2600K 4.5Ghz 1600DDR3 cl9 49fps
3570k 4.2Ghz 1600DDR3 cl9 43fps

AMD cards
4930K 4.4Ghz quad channel 2400DDR3 cl11 + r9 290 + new driver 46fps
4930K 4.4Ghz quad channel 2400DDR3 cl11 + r9 290 33.3Fps
I7 4820K 4,2Ghz quad channel DDR3 1866 + r9 290 28fps
AMD FX 8370E 4,5GHz 1600DDR3 cl11 + r9 290 22fps

Alright, time to add another Haswell-E result.

5820K @ 4.5GHz - Uncore @ 4.4Ghz -
DDR4 2666MHz - CL15

 

wilds

Platinum Member
Oct 26, 2012
2,059
674
136
Would love to see an unlocked Broadwell get benched for this game... Anyone want to send me one? :)
 

cbn

Lifer
Mar 27, 2009
12,968
221
106
Would love to see an unlocked Broadwell get benched for this game... Anyone want to send me one? :)

I would love to see a E5-2670 get benched for this game. (re: its been dropping in price pretty fast over the last two months and now starts @ 104.99 in ebay buy it now listings)

EDIT: Here was an earlier guess I made based on extrapolation:

http://forums.anandtech.com/showpost.php?p=37847934&postcount=111

It might even work well for gaming on a budget level (particularly if we see the price drop even more....and I expect it will).

Looking at these results for Fallout 4, it will be slower than the i7 3970X due to having 900 Mhz lower base clock and 700 Mhz lower turbo:

http--www.gamegpu.ru-images-stories-Test_GPU-RPG-Fallout_4-test-proz.jpg


2.6 Ghz base (for E5 2670) works out to be 74% of the 3.5 Ghz base clock of the i7 3970X.

If we extrapolate that % to frame rates (and assume it scales linearly) it would work out to 78 minimum FPS and 94 avg. FPS (compared to i7 3970X).

Of course, that simple estimate doesn't take into the fact the turbo delta (4.0 Ghz vs. 3.3 Ghz) is smaller than the base clock delta (3.5 Ghz vs. 2.6 Ghz). So that might make the gap a bit smaller (in favor of the E5 2670). Also the E5 2670 has two more cores and 5MB more cache than the i7 3970X (which is another variable I didn't factor into my estimate).

With that mentioned, I see this type of CPU more as a budget work horse for virtualization set-ups. (Example: NAS and Desktop together)

Taking that a step further....I wonder how should the memory be configured for max performance?

Quad channel DDR3 with low clocks and tight timings or higher clocks and looser timings?
 
Last edited:

toyota

Lifer
Apr 15, 2001
12,957
1
0
Alright, time to add another Haswell-E result.

5820K @ 4.5GHz - Uncore @ 4.4Ghz -
DDR4 2666MHz - CL15


Fallout 4 was patched since he ran his tests and is less cpu intensive now. places I dropped to mid 40 fps I can usually maintain 60 fps now. worst case before patch was 40 fps on the nose and that same spot is now mid 50s. when I actually play the game though I go ahead and put shadow distance to medium so I never have to worry about dropping below 60 fps.