• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

Fallout 4 CPU benchmark thread(need some Zen3 and Zen4 results!)

Caporegime
Joined
9 Nov 2009
Posts
25,012
Location
Planet Earth
Background

A few of us still play some of the old Creation Engine based games such as Fallout 4. Annoyingly a lot of reviewers don't seem to test older games anymore,and these generally benefit the most from better single threaded performance,faster memory speeds,etc.

Fallout 4 tends to be primarily limited by its graphics rendering thread,so higher drawcalls can collapse performance. Many reviews didn't bother testing the game properly in areas with higher drawcalls,which are usually the built up areas. Since Fallout 4 has user built settlements,it can cause very high drawcalls which can even cause performance dips on modern CPUs.

Another limitation is NPC AI which hogs another thread too,and built up areas and player settlements have another performance limitation.

However,a few years ago MajinCry on Anandtech forums did start a Fallout4 test thread:
https://forums.anandtech.com/threads/draw-call-performance-in-fallout-4.2501467/

This involved testing drawcall performance on different CPUs and dGPUs.Basically it involves using something called an ENB profiler,and it enables you to look at many stats including drawcalls and FPS.

The two tests,aim for 11700 and 8000 drawcalls respectively. This might sound like a lot,but I have seen something like 20000 in player built settlements,and it tanks CPU performance.

Since there is no real comparative reviews of the game on newer hardware,I would appreciate if anyone with the time,can run some of the benchmarks which were detailed in this thread. It will involve a bit of faff,but I will describe what you will need to do. If even a few data points can be provided,I would really appreciate it.

In my case it would certainly influence whether I upgrade to Zen3,or just hold off for a new platform in the future. So Intel benchmarks would be appreciated too!

Preparation

Open up the two directories which Fallout 4 uses
1.)C:\Users\Username\Documents\My Games\Fallout4
2.)steamapps\common\Fallout4

1.)Fresh Install of Fallout 4 with no mods

If you already have Fallout 4 installed,and don't want to muck it up,you can force a fresh install by renaming the two Fallout 4 directories to Fallout4Old.

Then you will find you cannot launch the game from Steam. If you click repair installation,the game will be re-downloaded into a new Fallout 4 directory.

2.)Launch Fallout 4

This will create the relevant ini files. Close the launcher down.

Now look in C:\Users\Username\Documents\My Games\Fallout4 for the Fallout4Prefs file. Open it and look for sD3DDevice. Note down what the name is. Mine was sD3DDevice="NVIDIA GeForce RTX 3060 Ti".

3.)Download the relevant test files from the thread over on Anandtech.

MajinCry in Anandtech thread said:
There's a couple things you'll need to do, before you can jump in:

1. Download these two save files, and place them in your "My Games\Fallout4\Saves" folder: https://mega.nz/#!bxljwZiQ!foEekBl-RWzM8XY5HZYztBOQrBeEpbloSsgZNVH8yLw

2. Backup & replace your .ini files in "My Games\Fallout4" with these: https://mega.nz/#!X88FDB6I!x9FCxGAB79AsFLXN16xyiI4mRjs_CQ_FPOBZ-zYIsJs

3. In FalloutPrefs.ini, replace: sD3DDevice="AMD Radeon HD 7800 Series" with the value from your original FalloutPrefs file.

In my case I replaced it with "NVIDIA GeForce RTX 3060 Ti".
4. Download and install the latest ENB mod, by placing the files inside the "WrapperVersion" folder, into the same folder as your Fallout4.exe file: http://enbdev.com/mod_fallout4_v0311.htm

5. Replace the ENB .ini files with these: https://mega.nz/#!3kUHTY7K!GnxkhGoqX1WjnqxUsVHMbPhfmO5zFmB5Ys8dcMva048

6. Disable VSync in your GPU driver settings program

As with any file check it with your antivirus and antispyware before opening.

The 4.) step links to an old ENB version which is not listed anymore,so get the latest one:
http://enbdev.com/mod_fallout4_v0468.htm

4.)Unlock the FPS.

In the Nvidia drivers,you can search for Fallout4 and make sure instead of VSync being on it is turned off.

Now in the C:\Users\Username\Documents\My Games\Fallout4 folder find the Fallout4.ini and Fallout4Pref.ini and change iPresentInterval=0 and save it. Now in the steamapps\common\Fallout 4 folder find Fallout4_Default.ini and do the same. In the steamapps\common\Fallout 4\Fallout4 folder the other copy of Fallout4Pref.ini and do the same.

The profiler window runs at a very low 640X480 resolution,so it maybe ideal to set the desktop resolution to something like 1280X800 just for the test tuns so its easier to see things.

Once you have done all that start up Fallout4. There will be two saves - the Corvega Plant and Diamond City.

MajinCry in Anandtech thread said:
And once you're in the game:

1. Load the save(s)

2. Press Shift+Enter to bring up the ENB overlay

3. Click on the "Profiler" category. Pan the camera around until you get the same number of draw calls, as the number in the save. So for "Corvega11700draws21fps", try and get as close to 11700 draw calls as possible. Same gig for the DiamondCity save.

4. Note down your framerate, and number of draw calls.

20220131183733-1.jpg


For the Corvega save pan the camera around until you get 11700 draw calls and note the FPS. In Diamond City pan the camera around until you get 8000 draw calls and note the FPS.

Current leaderboard(18/04/2023)

ITMZqIW.png
 
Last edited:
I did some tests. The CPU is a Ryzen 7 3700X and the dGPU is an RTX3060TI FE at stock.

OS was Windows 10 x64 21H2 and ENB version was v0.468 which is the latest one.

1.)497.09 and 32GB 2666MHZ DDR4

Diamond City had more fluctuations so did a 3 run average.

Corvega=53 FPS(Drawcalls=11710)
Diamond City= 63.43FPS(Drawcalls=8046)

2.)497.09 and 32GB 3600C16 DDR4

Corvega=60 FPS(Drawcalls=11700)
Diamond City= 73.93FPS(Drawcalls=8040)

3.)511.23 and 32GB 3600C16 DDR4

Did three runs of each.

Corvega=63 FPS(Drawcalls=11726.67)
Diamond City= 73.73FPS(Drawcalls=8013.33)
 
Last edited:
These tests are really more of a rendering tests although having lots of NPCs in the Diamond City save. If anyone can come up with an idea which would test NPC performance even more, and which could be reproduced without any mods (or maybe ENB like this one), it might make an interesting extra data point.

Thanks for adding your results. I would imagine one way to do it,would be to use a save with a large prebuilt settlement with lots of NPCs. I bought an extra copy of Fallout 4 for £3.50,and see if I could do something.

I would just like to applaud the purpose of this thread? I can't help, I don't have FO4 (yet), but I'm also interested in the results.

Its just useful to see the relative performance of the newer CPUs in the game. Too many reviews just concentrate on newer games! :(
 
v0.468? I had v0.461. Though that was the site offered me, but now I only see 468. Unsure if it makes a big difference, but I've now downloaded 468 too.

I would expect a Ryzen 5 3600 and Ryzen 7 3700X to be mostly the same,so I expect the difference with my system is down to the higher clockspeeds and faster RAM. My CPU was running at around 4.15GHZ~4.2GHZ during the tests.
 
Okay, I was going to put the clocks on a second monitor. Process Explorer might be able to tell me the threads but unsure if it can give a CPU time for each thread.

GPU also makes a difference even at 480P as the 470 was lot slower than the 3050, so your 3060Ti being twice as fast should make a difference too. CPU overhead is different for GPU drivers too.

5600X and 12600K CPU users are invited to contribute too!

From my own experiences of Fallout 4 on a Xeon E3 1230 V2/Core i7 3770,Ryzen 5 2600 and Ryzen 7 3700X with a RX470,GTX1080 and RTX3060TI at 1680X1050 and 2560X1400,the game had the same sorts of dips in FPS in my settlements at all the same places. The game gets insanely CPU limited in large settlements.

The difference you are seeing is 100% down to Nvidia DX11 drivers having slightly better DX11 overhead. However,the difference between your newer system and mine,is mostly likely down to the CPU. My RTX3060TI at the normal 60FPS capped framerates doesn't boost as high as when I running CB2077 for example.

I expect if you can pushed your Ryzen 5 3600 to around 4GHZ,and get the timings tightened down on your RAM,the difference will mostly vanish. Fallout 4 only really uses 4 threads a lot with limited usage of two extra ones.

I really would like to see some Zen3 and Alderlake results for this game!

I do have a Core i5 10400F in a SFF system,but it would require me to disassemble both my SFF systems which is a PITA.

The Core i5 10400F system does have a GTX960,but only has Linux on it. When I have some time I can try and put a copy of Windows on it to test.
 
Last edited:
12600k tuned with a stock 980ti and bdie tuned as well:

unknown.png

unknown.png

Ooh!! Some really nice scores there and thanks! What clockspeed was the Core i5 12600K at and what speed and timings was the system RAM at?

@Robert896r1 Slightly better than mine! :D
You might want to post that over on the original AT thread: https://forums.anandtech.com/thread...-call-performance-in-fallout-4.2548618/page-4
as you got better scores any of the posters there, including the one with a 12900K.
You might even be on the top of the FO4 leaderboard!


EDIT:
I did run it again (same score) but this time I had Process Explorer running and looked at threads:
3e0ZYzV.png
So the one thread (TID: 11840 in my run) does most of the work.
The second one does about half of that, the others are way behind.

No idea which thread does what but it looks like while Thread#1 limits things, the rest could scale up to about 9 threads.

My 3600 ran at a steady 3.95GHz.

EDIT: what was supposed to say "any" not "and".

Looking at some other comparisons for Fallout 4 performance,Skylake appears to be slightly ahead of Zen2 clock for clock(A Core i7 6700K was slightly ahead of a Ryzen 7 3700X in per clock performance). Looking at some reviews,performance seems to tail off after 6C.
 
It's a shame as I personally like comparing unique use cases than something vanilla like CB etc.

Same here,and testing an older game which is still played by a decent amount of people,and is not tested in modern reviews is actually a useful data-point. It's also a game which loves faster RAM with tighter memory timings.

It will be interesting to see how this fares with Raptor Lake and Zen4 once we get better DDR5 RAM kits!

I would test my other SFF rig which has a Core i5 10400F and RAM running at 2666MHZ,but it would involve taking apart two SFF PCs. Might try and put Windows on the other system and doing a test but only has a GTX960 4GB in it!

A final two from me.

First I want to bring in the rear by running it on my old machine:
OS Windows 10 x64 21H2
CPU i5-3570K @ 3.40 GHz (stock)
RAM 2 x 8GB DDR3 1600 10-10-10-27-41-2T
GPU Radeon 470 4GB (stock)
GPU driver 22.1.2
ENB Version V0.468
Save 1 Drawcalls 11,713
Save 1 FPS 33.1
Save 2 Drawcalls 8,006
Save 2 FPS 37.8
This was a bit less consistent than any of my previous runs to so I ran 3 times and averaged it.

And then since I had benched with an older ENB, I re-benched v0.461 vs v0.468 with my CPU locked to 4.1GHz:
OS Windows 10 x64 21
CPU Ryzen 5 3600 @ 4.1G
RAM 2 x 32 DDR4 3200 16-20-20-39-74
GPU GeForce 3050 8
GPU driver 511.
ENB Version V0.461 and V0.468
Save 1 Drawcalls 11,716 and 11,743
Save 1 FPS 55.6 and 52.0
Save 2 Drawcalls 8,040 and 8,071
Save 2 FPS 68.4 and 62.5
Looks like for me v0.468 is slower, but I didn't run many times to get an average.

Anyway, I've done a simple leader board for the results so far (for those who submitted multiple results I only took their best):

7h94c8S.png

The R's are in the lead.
The two 12600K's clocks aren't that far apart (5.6%) but the scores are way beyond that (+22% for Robert896r1).

Well, Fallout4 (and Skyrim etc.) is pretty heavy on memory speeds and while @RavenXXX2's DDR4-3800 CL16 is pretty good, @Robert896r1's DDR4-4166 CL15 is pretty unbeatable.

(Speaking of memory timing, I did play with setting DDR4-3600 on mine but wasn't willing to raise voltages so didn't get anywhere with that.)

Interesting the ENB seems to be making such a big difference! The RAM speed and tighter timings really show how much Fallout 4 loves decent RAM.

Even years ago it was one of the best games to test the effects of RAM on.
 
Last edited:
Yes, wasn't sure if that was just noise but did some averages. After 5 runs of both v0.461 and v0.468 I can say... Well the older is slightly faster.

b9818Xq.png
and as a rough idea of variance I used the stddev function too although I wouldn't put too much into that as it mainly shows that hitting 11,700 and 8,000 is not easy. Still 51.3/49.2 is 4.4% and 58.6/55.8 is 5.1% so there is something there.

I wonder why this is the case?
 
5800X stock

hhmmm... really not good, i don't know what extensions this game is using but the 5800X being clocked 15% higher than your Zen 2 core is only chucking out that many more FPS, at best, the IPC difference between Zen 2 and Zen 3 seems to account for nothing :cry:

Some sort of weird emulation going on because its missing the required extensions?

Not worth it for you mate :)

66.6
91.0

FA45.jpg


FA46.jpg

When I went from Zen+ to Zen2 the performance bump was quite noticeable. It does make me wonder whether something weird is going on.

My results were:

3.)511.23 and 32GB 3600C16 DDR4

Did three runs of each.

Corvega=63 FPS(Drawcalls=11726.67)
Diamond City= 73.73FPS(Drawcalls=8013.33)

The first test isn't almost the same but the second one for you is 25% more. So that definitely the extra IPC/8 core CCX is making some difference,but the first one isn't much higher for some weird reason!

What RAM speed are you using - maybe when you reset the CPU to stock it reset the RAM back too? I was running 3600C16 with a bit of tuning.

What performance are you get at the 5GHZ?? It could be the CPU isn't boosting properly because the load isn't enough?
 
Last edited:
3200MT/s
16-18-18-38-56

Its #### RAM, cost me £90 for 32GB, bargain at the time.

Did you reset the CPU back to stock,or is it running at 5GHZ? The second test was a decent improvement,still not certain what is happening with the first one! The 5GHZ Core i5 12600K was only 20% faster and is running at much higher RAM speeds,and Fallout 4 seems to benefit from it.
 
On a 5950x with 3800 CL16 at average of 4.85Ghz on what appears to be one thread in HWinfo ( its boost so it wavers) it I get ~86 and ~106 on the two levels, try as I might I could not catch the Intels....Boo :) , tried 1CCD and with SMT, no diff.

Thanks! It's not too far off the lower of the Core i5 12600K results.

With GPU do you have?

Stock, no idea....

The first result seems quite low considering the second test seems in line with the other Zen3 numbers so far.
 
Back
Top Bottom