Benchmarking Fallout: New Vegas with and without various mods, programs, Win7 vs Win10

[MildlyShoey]

You can view the raw benchmarks here: https://docs.google.com/document/d/1ivERHUX35O0shJlyV377p3xaNUk1yRhUhoG_bcmhxNo/edit?usp=sharing

 

This isn't really a guide, but I'll go over my notable results and make a lists of the no/low/high impact mods I tested.

 

Important Notes:

 

1) Messing with priorities or limiting the game to X threads does nothing but hinder performance.

2) Texture packs shouldn't impact anything other than load times unless you break your VRAM/RAM limit.

3) Windows 7 by itself was a 4FPS gain for me, with NVSR being another 2FPS or so (and substantially reducing lows). This isn't very surprising to those who know of those spectre/meltdown exploits, but if you absolutely must have the best performance the only thing I'll suggest is make sure your BIOS is set to legacy or UEFI with CSM if you have a modern computer/motherboard, otherwise you won't be able to install Windows 7.

4) Shadowplay, Firefox, etc don't impact framerate substantially. YouTube, however, even with v-sync forced off would result in the game attempting to sync to 30/60. I'm unsure if this is a driver bug or a YouTube bug, but it is likely one of the two.

5) Even with the appropriate INI tweaks, New Vegas will only slam one core, though you can expect up to 3 others reaching 50% utilization (though not all at once). If you want "max performance", disabling all but 4 cores then overclocking those as far as you can isn't the worst idea, though if you don't already know how to overclock I can't say I'd recommend it. This isn't shown in the benchmarks (for the most part) but was shown by MSI Afterburner while I was doing them.

 

Now, for the mods;

 

Notes:

1) Any quest mods listed were sitting at the "haven't started" stage. Quest mods will occasionally impact performance differently depending on what stage you are on.

2) Some of these will be redundant if you are familiar with modding, but I'm choosing to include them anyways.

3) Pretty much all of these tests were performed with New Vegas Uncut - Freeside Open, as it is the only natural, reliable way to get the game to drop below 70FPS on my computer other than using The Strip Open.

 

No Measurable Impact during non-combat gameplay:

 

NMC Large

Realistic Wasteland Lighting

Yukichigai's Unofficial Patch

JSawyer Ultimate

EVE/EXE/Impact - CE

B42 Quickthrow

Enhanced Item Info

Spice of Life

Inventory Search

Universal Item Sorter

Sticky Ragdoll Camera

Mojave Arsenal

Mojave Raiders

Mojave Wildlife

JIP Companion Command and Control

JIP Improved Recipe Menu

All Weapon Sounds Overhaul + Scripted Patch

One HUD

Immersive Minigames

The Weapon Mod Menu

Roberts NV

exeter's Type3 Base Game + DLCs

Autumn Leaves

DEIMOS

New Vegas Bounties I

Russell

FOV Slider

Securitrons in CRT

 

Low Impact - these mods can be expected to shave 1-2 FPS off each if you aren't consistently hitting 60.

 

Wasteland Flora Overhaul

CASM - will shave 2FPS off whenever the player is moving.

The Living Desert

The North Road Remastered

The Initiation + Eliza

New Vegas Bounties II

FNVLODGen + Resources + LOD additions and improvements + Much Needed LOD - shaves 2 FPS off a hundred, but that's a lost frame if you are under 60.

 

High Impact - these mods can be expected to shave 2-6 FPS off each if you aren't consistently hitting 60.

 

Qwinn's New Vegas Redesigned 3 - shaves 3-4FPS off of 65.

NV Interiors - shaves 6FPS off the Freeside route - see document for more details.

[madmongo]

2) Texture packs shouldn't impact anything other than load times unless you break your VRAM/RAM limit.

 

While loading high resolution texture packs along with mods that add NPCs or other things to an area can crash your game due to the game running out of memory, there is another issue here. FNV's texture caching system leaks memory. The common symptom is that the game gets slower and slower and eventually crashes due to the texture cache leaks. Using high resolution texture packs will increase memory leaks and will make this game slowdown and crashing happen faster.

 

A texture pack that changes textures without making them bigger won't have an impact, but any texture pack that uses higher resolution textures will definitely cause slowdown and crashing issues. I wouldn't say that it doesn't impact anything other than load times.

[MildlyShoey]

"Unless you hit your VRAM/RAM limit" is the important thing.

Between ENBoost and 4GB patch I don't think anyone will be hitting that during normal sessions.

[JimboUK]

ENBoost has made running out of memory a thing of the past but the more work it has to do the more it impacts performance, also stuttering/pop in becomes a thing even on a fast SSD, the old game engine just can't process all that.

[MildlyShoey]

While that's logical, it's also not strictly true. I've added 4 new benchmark situations reflecting it.

 

Excerpt:

Baseline:

08-05-2019, 19:09:51 FalloutNV.exe benchmark completed, 3681 frames rendered in 57.016 s
                     Average framerate  :   64.5 FPS
                     Minimum framerate  :   23.6 FPS
                     Maximum framerate  :   82.9 FPS
                     1% low framerate   :   17.5 FPS
                     0.1% low framerate :    9.7 FPS
08-05-2019, 19:12:29 FalloutNV.exe benchmark completed, 3756 frames rendered in 56.984 s
                     Average framerate  :   65.9 FPS
                     Minimum framerate  :   27.7 FPS
                     Maximum framerate  :   83.8 FPS
                     1% low framerate   :   21.0 FPS
                     0.1% low framerate :   11.5 FPS

Baseline w/o NMC Large:

08-05-2019, 19:14:45 FalloutNV.exe benchmark completed, 3754 frames rendered in 57.750 s
                     Average framerate  :   65.0 FPS
                     Minimum framerate  :   29.6 FPS
                     Maximum framerate  :   83.5 FPS
                     1% low framerate   :   28.7 FPS
                     0.1% low framerate :   16.0 FPS
08-05-2019, 19:16:48 FalloutNV.exe benchmark completed, 3810 frames rendered in 57.187 s
                     Average framerate  :   66.6 FPS
                     Minimum framerate  :   28.8 FPS
                     Maximum framerate  :   85.6 FPS
                     1% low framerate   :   25.7 FPS
                     0.1% low framerate :   17.4 FPS


Baseline w/ ENBoost:

08-05-2019, 19:29:04 FalloutNV.exe benchmark completed, 3768 frames rendered in 58.531 s
                     Average framerate  :   64.3 FPS
                     Minimum framerate  :   27.9 FPS
                     Maximum framerate  :   81.9 FPS
                     1% low framerate   :   27.0 FPS
                     0.1% low framerate :   17.0 FPS
08-05-2019, 19:33:21 FalloutNV.exe benchmark completed, 3662 frames rendered in 56.750 s
                     Average framerate  :   64.5 FPS
                     Minimum framerate  :   30.4 FPS
                     Maximum framerate  :   81.9 FPS
                     1% low framerate   :   26.8 FPS
                     0.1% low framerate :   14.3 FPS

Baseline w/ ENBoost w/o NMC Large:

08-05-2019, 19:36:01 FalloutNV.exe benchmark completed, 3692 frames rendered in 57.250 s
                     Average framerate  :   64.4 FPS
                     Minimum framerate  :   30.4 FPS
                     Maximum framerate  :   82.9 FPS
                     1% low framerate   :   29.2 FPS
                     0.1% low framerate :   14.5 FPS
08-05-2019, 19:37:55 FalloutNV.exe benchmark completed, 3725 frames rendered in 57.516 s
                     Average framerate  :   64.7 FPS
                     Minimum framerate  :   27.6 FPS
                     Maximum framerate  :   83.8 FPS
                     1% low framerate   :   27.4 FPS
                     0.1% low framerate :   13.3 FPS

I am curious as to the reasons why, though. I will admit it's not something I thought about. I'm going to hazard a guess it is related to New Vegas only slamming 1 thread regardless of configuration (atleast on a modern processor), like ENBoost not relying on that "main thread".

[JimboUK]

I wasn't talking about frame rate, I'm talking about stutter and pop in, something that happens because the engine simply cannot process a huge number of textures, it starts to choke.

 

The things that cause the biggest frame rate hits are large numbers of NPCs, scripts and multiple light sources. This was a test I did a while back, memory and textures here aren't an issue, they're clones and their assets are only memory once no matter how many there are. See the video at the bottom for what happens, the engine is so bogged down that the frame rate has gone to hell and it can't even load the next cell without a long pause. I've since redone the test with a 1080ti and the result is the same.

 

And yeah the game using 1 thread fully is a serious problem, I struggle to get 60 fps on the strip while my computer is mostly asleep because the game just doesn't use it.

 

[MildlyShoey]

"The more it impacts performance" means average framerates or lows that lasted more than a second in my mind (stutters too, but that's difficult to properly test considering the 1/0.1% lows being locked to just those percentages and there not being much of a way to prevent load screens from factoring into benchmarks).
With the lows that shows still being the same, I think that under typical circumstances it won't influence it heavily.
If I knew of a better place to benchmark it I would. I suppose The Strip Open would be a good place to start, even though it's... not very representative of "average gameplay being taken down to 60FPS by load".

[MildlyShoey]

Chiming in with a couple extra benchmarks (largely done because of convenience):

Baseline with NMC Large after 3 hours of gameplay:

11-05-2019, 13:49:51 FalloutNV.exe benchmark completed, 2665 frames rendered in 60.407 s
                     Average framerate  :   44.1 FPS
                     Minimum framerate  :   23.1 FPS
                     Maximum framerate  :   53.7 FPS
                     1% low framerate   :   17.6 FPS
                     0.1% low framerate :   11.1 FPS

Baseline with NMC Large after saving and restarting the game (after previous benchmark):

11-05-2019, 14:01:15 FalloutNV.exe benchmark completed, 3056 frames rendered in 61.188 s
                     Average framerate  :   49.9 FPS
                     Minimum framerate  :   17.1 FPS
                     Maximum framerate  :   65.8 FPS
                     1% low framerate   :   14.2 FPS
                     0.1% low framerate :    5.8 FPS

Baseline with NMC Small after 3 hours of gameplay:

11-05-2019, 17:09:36 FalloutNV.exe benchmark completed, 2128 frames rendered in 57.859 s
                     Average framerate  :   36.7 FPS
                     Minimum framerate  :   14.4 FPS
                     Maximum framerate  :   53.4 FPS
                     1% low framerate   :   10.9 FPS
                     0.1% low framerate :    3.5 FPS

Baseline with NMC Small after saving and restarting the game (after previous benchmark):

11-05-2019, 17:41:24 FalloutNV.exe benchmark completed, 2713 frames rendered in 58.797 s
                     Average framerate  :   46.1 FPS
                     Minimum framerate  :   16.3 FPS
                     Maximum framerate  :   58.1 FPS
                     1% low framerate   :   14.9 FPS
                     0.1% low framerate :   10.2 FPS

It's no where near controlled enough to say for certain, but with the lack of comparable benchmarks that I can find, I'm inclined to believe that the discrepancy others have seen is more due to the game not properly purging irrelevant data than the fault of any texture pack (or RAM limits being "just right" to cause it).

[JimboUK]

I think benchmarking New Vegas is very hard, the game is a mess with random frame rate drops that make any kind of test an act of futility. Too many light sources is something to avoid, we put a ton of lights in this on a timer, the lights come on around 9.00 or 9.30 and you can see the frame rate drop, there's no need to benchmark it. It gets worse depending on how many NPCs are wondering around, I guess because each one has to be lit according to the light at their location.

[MildlyShoey]

I didn't note that those benchmarks all used the same route (Freeside Open East Gate to Strip Gate).

I didn't control for time of day, but that's partly why I did the "after restart" benchmarks too.

Something important to note that I forgot to, was that in both cases the game hanged for approximately 7 seconds when opening the pause menu (after playing for 3 hours). If there was a large impact solely from texture packs I imagine that would've varied more, but that is just speculation.