The actual game is really good and better than it’s predecessor. The PC port however, like it’s predecessor is straight up garbage.
I am only reporting primarily what is happening on my rig, which is the following.
I7 8770K @4.7GHz on all cores at 1.2vcore
Asus TUF Z370-PRO GAMING Mobo
16GB of Low Latency DDR4 Ram @ 3600MHz
ASUS RTX 3080 TUF OC
The game is installed on a Samsung 970 PLUS 1TB PCIe 3 NVME SSD
I also want to state I see what I consider a lot of misunderstandings, or at least the incorrect viewpoint in relation to performance. Yes high GPU usage is a good thing generally speaking, However when there is nothing demanding going on and it is using all the GPU and only pulling 120 FPS in a PS4 game at heart, that is unacceptable. Yes I can pull 120fps on this beast of a 3080 machine (with a lot of caveats which I will get to later), the problem is how much GPU it is having to use to achieve that.
This is a 3080 this is no slouch of a GPU and can even pull 61-80+ fps in Minecraft with full path ray tracing, which is incredibly demanding, all without enabling DLSS. It can play cold war at 110-140 fps without RT (I have it capped at 140), using less than 90% of the GPU most of the time. This is what people should be looking at, how much of the GPU it is using to pull the FPS it is, and what is actually going on with the graphics aka types of shadows, texture filtering, aa solutions, type of lighting etc.
The game is 76.35GB installed which is absurd given the amount of repeated assets in the game.
With that out of the way, here is my PC report.
Graphics settings
When booting the game and going to the options, the alarm bells were already ringing, they consist of the following.
Display Mode, screen Resolution, rendering resolution percentage, which only goes up to 100, frame rate cap 30 – 120, shadows off to high, motion blur on or off, dynamic reflections on and off, texture quality high or low, and effects high or low. No AF and no AA options… So you want to add in AF, which can be done by the Nvidia Control Panel if you have an Nvidia GPU. Also in an age of 144-265hz+ monitors, a 120fps cap for a PC port is undeniably POOR.
Graphics
NIOH 2 is a very ugly extremely basic game that is a PS4 game at heart, it does not look good, it is serviceable but not good. It looks pretty much like NIOH 1, bar slightly better lighting and improved models. Without AF it looks flat out ugly and incredibly muddy, with AA added at a driver level it is ok at best. There are also egregious and indefensible amounts of pop in, you can clearly see a render line, which is closer to the player than it should be.
Performance
Owners of Nvidia GPUs must be aware of a HUGE ISSUE with the game, it HATES Nvidia GPUs. I have seen multiple reports now and a chart from the website gamegpu, showing that AMD GPUs perform significantly better. To put this into a bit more detail, a 5700XT outperforms a 2080TI in NIOH 2, and is only 25% away from a 3080. Any other title you would expect about a 35%+ leads in the 2080 ti’s favour, so something is MAJORLY WRONG here.
Putting effects to low resolves some extreme drops, I have gone as low as 42 fps with effects on high, admittedly this was during a massive attack by a Yokai that kicked up an absolute ton of dust close to the camera, but it is unacceptable. Putting effects to low, has seemed to fix this.
I also want to add I have read reports that the ultra-wide support they bragged about in the trailer is also reportedly half baked, many users are reporting it does not go up to their full resolution, some reported reporting it does not even go up to the common 1440 Ultra-Wide resolution. So already you are getting a picture here, and it is not a particularly good one.
Frame timings are surprisingly another issue, but this is common in a lot of games these days, the in game fps cap is garbage, as pretty much all software ones are. You should use rivatuner statistics server, which implements a driver level cap on the framerate and smooths out your frame timings, if you are constantly hitting that FPS. Even with VRR you still get this issue, which is a rant about games for another time and not specific to NIOH 2.
Load times are fast on an SSD as they should be. Although When amission has a cutscene at the beginning it takes far longer than it should to load.
At 4k which is the strong part for the new RTX 3000 architecture, GPU usage is around 58-75% Usage in game for 60fps. So 4K 60 is possible, but for a game like this 4K120 should be possible easily. A game designed to run on a incredibily inefficient 1.84 TF Tablet APU console should run a lot better than this on a modern cutting edge architecture lightyears ahead of last gen, on a GPU that has a total of 29.7 TeraFlops of compute power.
GPU Usage
Now here is where the major issues with the game exist, and I have had people say “you are near a 120 luck be happy that is fine”. While that statement is not technically wrong the issues is, one I am dropping two the amount of GPU being used to get this frame rate and three this is at it’s very heart a PS4 GAME. There is nothing going on graphically that should be maxing out my GPU and struggling to pull 120 frames, as mentioned previously High GPU usage is good, generally speaking you want a GPU to be utilized as much as possible.
However, a game from the early 2000s or for a modern example Minecraft without RT will not use 100% of the GPU or anywhere near 90% (more like 30% for old games at best) while pulling 120fps, because the GPU does not need to work that hard to render the workload. On the bright total Vram usage for windows and the game is around 5GB, which is a bit too high for the visuals but not the worst offender I have seen. There are also non frame time related stutters that happen on the missions.
That is where there seems to be fundamental misunderstandings. Here is a screenshot of the game maxed, 96% GPU usage, and 116FPS for what is going on screen is incredibly poor.
Crysis 3 without AA (even with AA) uses less GPU to render significantly more GPU workload. Crysis 3 has far higher quality lighting, shadows and everything across the board. In addition to light refractions off water, tessellation including on the water, and naturally being a Crysis game, pushes the graphical envelope to insane degrees for it’s time across the board. So NIOH 2 an extremely basic graphically demanding game and a PS4 game at its heart, with very basic lighting, shadows and so on should not be as demanding as it is, especially on a 3080.
There are times where GPU usage is anywhere between 65-80% while pulling 120fps, but again for what is happening on screen, this is absurdly high. I can only imagine how poor performance is on lower tier Nvidia GPUs, especially people with 970s (which is still a solid low end card). Although a friend is reporting with an RX 580 and Ryzen 1600, he is pulling around 80fps at 1080p with shadows and AO off (which is still not a pretty picture), I have asked him to get back to me in detail and will update accordingly.
If I was to take a guess at some of the things going on the game to cause the poor performance it would be the following. Fundamentally team ninja do not know how to make modern games using modern techniques, but beyond that they seem to fundamentally lack basic skills any game dev should had. There is clearly 0 culling going on in the game, which results in drops in fps and increase in GPU usage just by moving the camera to a different angle.
The game is clearly still putting the entirety of what is not on screen in the GPU workload, normally you would have a render cone if you will, where you will render stuff outside the player view, for obvious reasons. The shadow rendering technique also seems bugged with my character not emitting a full shadow, AO also seems to disable the character shadow. Suggesting some inefficient and straight up poor rendering techniques being used.
So again, there just seems to be a fundamental lack of skills at the dev team inside Team Ninja who do all the behind the scenes work, which Is resulting in incredibly poor performance. The PS5 reportedly drops a slow as 70fps at 1080p in its performance mode, which again given the GPU in that machine is pretty much 2080 level is abysmal. It also shows that this is not simply an issue with the PC version of the game, it is a fundamental problem with how they are using UE4 and the tech guys at TN.
Conclusion
The game leaves a lot to be desired on the performance front, which is a massive shame, given how good the actual game is. You could argue the port is fine, as it’s huge performance flaws are due to how incompetant TN are at making engines. I however disagree with that statement entirely, the game is £50 and labelled as a PC game, none of these issues should exist with all that considered. Defending it is just making excuses for KT and TN at this point, for the price tag THIS SHOULD BE BETTER.
I must issue a BUYER BEWARE warning on NIOH 2 for PC. You should probably check the steam forums for people with similar rigs (including the brand of GPU) and their performance before buying. There is no guarantee this game will run fine for you, even though on paper and it’s recommended requirements say it should. Just remember to force AF if you can and cap the framerate with RTSS. This isn’t TN’s first rodeo, yet they have learned ABSOLUTELY NOTHING.
Update
Further snooping around forums and trying tweaks has provided a significant update, which tackles exactly the point I was making with my preface to this report and makes videos like fextra life’s calling this port “ok” as complete shams. It does also show the port actually is not as bad as I initially thought, is still mediocre at best.
As pointed out by Useless Foetus, who has 2 rigs (see thread for full details https://steamcommunity.com/app/1325200/discussions/0/3113645835775825746/), one rocking a R5 3600 and 1080ti the other R5 3600 and 5600XT. The 5600xt is getting 105fps at 1080p and the 1080ti is getting 65fps, just looking at an item at the very start of the game, which the effects setting will have no major effect on . The 1080TI on average is faster than the 5600XT by 10-15% yet the 1080ti is getting which is just over 60% of the 5600XTs frame rate. This means going by law of average performance gain of at least 10% in the 1080Tis favour it should be pulling just shy of 120 at 1080p, if not 120, which means it is around 50% slower than it should be, as it is only pulling 65. To add to this further, although not exact math if the 2080ti is short around the same percentage that means the 2080ti should be pulling around 135-180fps (again not exact math just rough law of averages) at 1440p.
Making the 90-120 that Fextralife reports completely unacceptable and far from ok. This clearly shows and proves my point that what is going on screen and the GPU usage and performance OBJECTIVELY does not match at all, and while a 50% performance increase would make performance a lot closer to where it should be, it is still far off, but that there is the limits of the game’s engine, which is just horrible.
This is also confirms what I reported in my original article, where I stated that a 5700XT was beating a 2080ti by 45% when the 2080ti should be around 15% ahead. It also confirms that lowering effects to low, does not resolve the poor performance issues on Nvidia GPUs, and something is fundamentally wrong here. Hopefully Nvidia can do some hardwork and release a driver that can resolve a lot of TN’s incompetencies.
My friend on his RX580 is also reporting that despite a solid 60, and solid frametimes and using RTSS to cap the fps at 60, they are still getting hitching and stuttering, indicating even on at least some AMD cards there are still issues with the game.
As of 2019 75% of the GPUs steam users had were NVIDIA GPUs, so there is really no excuse to mess up performance with NVIDIA GPUs, and how this got missed is BEYOND ME.