• Hey, guest user. Hope you're enjoying NeoGAF! Have you considered registering for an account? Come join us and add your take to the daily discourse.

Watch Dogs 2 PC performance thread

Megasoum

Banned
So Uplay never recognize the "Remember Me" setting on the login screen so everytime my pc comes out of sleep mode I have been logged out because of innactivity and I have to relog... I then have to re-type my whole email and password and then the 2 factor code EVERY. SINGLE. TIME.


So stupid
 

Iceternal

Member
Do you guys have any issues locking to 30 fps ?

It's been a pain for me.

The third Vsync option that"'s supposed to do it doesn't work.

And when I lock it with Nvidia inspector, all the FMVs randomly drop FPS.

It goes from 30 fps to like 10 every minute.
 

Tecnniqe

Banned
You don't say!



EDIT: Top of a new page, dammit.
Okay
So Uplay never recognize the "Remember Me" setting on the login screen so everytime my pc comes out of sleep mode I have been logged out because of innactivity and I have to relog... I then have to re-type my whole email and password and then the 2 factor code EVERY. SINGLE. TIME.


So stupid
Do you exit uPlay before you put it to sleep?
 

kaskade

Member
I have a 1070 and played the first mission on ultra and was staying pretty good in terms of fps. Usually around 80. Should I keep temporal filtering on or off? I never really used it in a game before so I'm not sure what it does much or how much of a positive or negative effect it has.
 

dr_rus

Member
I have a 1070 and played the first mission on ultra and was staying pretty good in terms of fps. Usually around 80. Should I keep temporal filtering on or off? I never really used it in a game before so I'm not sure what it does much or how much of a positive or negative effect it has.

Temporal filtering is basically the same thing as checkerboarding on PS4Pro: it renders 1/2 of the image and use the data from a previously rendered other 1/2 of the image to reconstruct the final full image.

This provides a great performance boost (as you essentially render only half of the final image in each frame now) but as any other temporal reconstruction solution lead to some artifacting, especially when moving the camera around, mostly visible on high contrast edges.

From my experience with WD2 it's better to leave it on and add supersampling (resolution scale) on top of it - this will take care of both temporal artifacting and general aliasing. With it being off it's really hard to find any AA option which will result in playable framerates while dealing with aliasing properly. (I have no idea why they didn't add TAA btw, it's basically a must with this type of temporal resolution reconstruction).
 
Temporal filtering is basically the same thing as checkerboarding on PS4Pro: it renders 1/2 of the image and use the data from a previously rendered other 1/2 of the image to reconstruct the final full image.

This provides a great performance boost (as you essentially render only half of the final image in each frame now) but as any other temporal reconstruction solution lead to some artifacting, especially when moving the camera around, mostly visible on high contrast edges.

From my experience with WD2 it's better to leave it on and add supersampling (resolution scale) on top of it - this will take care of both temporal artifacting and general aliasing. With it being off it's really hard to find any AA option which will result in playable framerates while dealing with aliasing properly. (I have no idea why they didn't add TAA btw, it's basically a must with this type of temporal resolution reconstruction).

temporal filtering is 1/4th with 2xmsaa not 1/2. from the media ive seen of watch dogs 2 it also looks nowhere near as good as checkerboarding. watch dogs 2 is an aliased, shimmery mess

Its similar to the method used in quantum break but the implementation here produces worse results
 

kaskade

Member
Temporal filtering is basically the same thing as checkerboarding on PS4Pro: it renders 1/2 of the image and use the data from a previously rendered other 1/2 of the image to reconstruct the final full image.

This provides a great performance boost (as you essentially render only half of the final image in each frame now) but as any other temporal reconstruction solution lead to some artifacting, especially when moving the camera around, mostly visible on high contrast edges.

From my experience with WD2 it's better to leave it on and add supersampling (resolution scale) on top of it - this will take care of both temporal artifacting and general aliasing. With it being off it's really hard to find any AA option which will result in playable framerates while dealing with aliasing properly. (I have no idea why they didn't add TAA btw, it's basically a must with this type of temporal resolution reconstruction).

Thank you, makes perfect sense now.

temporal filtering is 1/4th with 2xmsaa not 1/2. from the media ive seen of watch dogs 2 it also looks nowhere near as good as checkerboarding. watch dogs 2 is an aliased, shimmery mess

Its similar to the method used in quantum break but the implementation here produces worse results

Yeah that's what I noticed, it looks good overall but the aliasing is pretty bad.
 
I used TF for a little bit but then with something I changed the ghosting became ridiculous when moving that I just disabled it and running at 36FPS (1/4 refresh) due to my poor 2500K at 4.5GHz was getting pummeled at 60, it's not like I need it either since the game isn't doing a ton now to push my 1070.

First rule of forum discussion: don't read further back than bottom of the last page, starting at the post you just made, duh :p

Wait, am I the only one that reads 99% of the posts in a thread before posting one myself?

My replies were pretty pissy but JaseC even added it to the initial post. ¯\_(ツ)_/¯

Besides it wasn't about my issue with it (haven't had any after the problems of the first couple days) it was just that EAC is a kernel level driver so it's entirely possible for it to cause BSOD from its crashes and possible conflicts.
 

Urthor

Member
Its similar to the method used in quantum break but the implementation here produces worse results

When Gaf says "worse results than Quantum break" I realise it's time to put my wallet away and write this game off until next Christmas AC: Unity style

How 2016 can give us BF1 on the one hand and ports like Forza/Quantum Break/Watch Dogs/Dishonored on the other amazes me. 2018 and my full CPU/GPU upgrade can't come soon enough
 
When Gaf says "worse results than Quantum break" I realise it's time to put my wallet away and write this game off until next Christmas AC: Unity style

How 2016 can give us BF1 on the one hand and so ports like Forza/Quantum Break/Watch Dogs/Dishonored on the other amazes me. 2018 and my full CPU/GPU upgrade can't come soon enough

The sad thing is, I just DID a full upgrade (of the GTX 1080 + i7 4600k variety) and I can't play those games without issues. It really pisses me off.
 
When Gaf says "worse results than Quantum break" I realise it's time to put my wallet away and write this game off until next Christmas AC: Unity style

How 2016 can give us BF1 on the one hand and so ports like Forza/Quantum Break/Watch Dogs/Dishonored on the other amazes me. 2018 and my full CPU/GPU upgrade can't come soon enough

Just to be clear im only talking about the IQ and artifacts of the 2 reconstruction techniques. Qb gives you a softer but stable image in motion with artifacts not being obvious on most edges. Wd2 is sharper but very fuzzy and in motion almost every single pixel of the image will shimmer like hell when in motion. Edge artifacts are also very visible

http://www.gamersyde.com/leech_38760_1_en.html

1080p ultra settings, no nvidia shadows with temporal filtering on, 20% extra details and 125% resolution scaling. Judge for yourself. A 1080 cant keep 60 fps during demenanding areas at this level
 

jorimt

Member
My replies were pretty pissy but JaseC even added it to the initial post. ¯\_(ツ)_/¯

Ha, don't misunderstand, it was more of a dig at his comment, not yours, and not even a dig at that; just a playful observation.

Too many of us are often guilty of asking a question that was just answered two posts above ours. And poor JaseC's OP; I read almost every comment, and even I forget to check the ever useful OP at times, though that goes for any thread.
 
Well, I'm perfectly happy with 4K Temporal Aliasing and everything at ultra minus shadows (very high). I do have pixel density at 84% mind you. I get 50fps 99% of the time and yes, there are some jaggies but it looks better than 4K with no MSAA and 84% pixel density and almost double the frame rate to boot. This is with Ultra Screen Space Reflections too.
But if I turn SSR off I can get 100% pixel density and a slightly sharper less aliased picture.
1080, 4790@4.4ghz, 16GB 2133mhz DDR3.

q9Be003.jpg
It's not perfect but it's fine for me.
 

dr_rus

Member
temporal filtering is 1/4th with 2xmsaa not 1/2. from the media ive seen of watch dogs 2 it also looks nowhere near as good as checkerboarding. watch dogs 2 is an aliased, shimmery mess

Its similar to the method used in quantum break but the implementation here produces worse results

1/4 of spatial resolution with MSAA 2x is equal to 1/2 of final frame buffer resolution. MSAA isn't used as MSAA in this case, it's there to make alternating checkerboard pattern, nothing more.

I've asked around and it's the exact same way as Pro's checkerboarding. The reason there are more visible artifacts is because for some mysterious reason they decided to not implement any kind of TAA on top of the temporal color accumulation - which is very strange since TAA should be pretty much "free" in this case.

QB is using a different method actually. They are using MSAA 4x on a frame which is only 1/2 the spatial size of the final image and combining color from four previous frames. It's obviously a much more performance intensive option which will produce better results.
 

kaskade

Member
Yeah, once I got into the open world my frames were tanking. I kinda wish I could just lock it at 30.

I turned down shadows and it seemed to put me at a much more stable rate though closer to 60.
 
1/4 of spatial resolution with MSAA 2x is equal to 1/2 of final frame buffer resolution. MSAA isn't used as MSAA in this case, it's there to make alternating checkerboard pattern, nothing more.

I've asked around and it's the exact same way as Pro's checkerboarding. The reason there are more visible artifacts is because for some mysterious reason they decided to not implement any kind of TAA on top of the temporal color accumulation - which is very strange since TAA should be pretty much "free" in this case.

QB is using a different method actually. They are using MSAA 4x on a frame which is only 1/2 the spatial size of the final image and combining color from four previous frames. It's obviously a much more performance intensive option which will produce better results.

Any info you can provide that checkerboarding in ps4 pro uses msaa? I thought they were using the new custom buffers
 

dr_rus

Member
Any info you can provide that checkerboarding in ps4 pro uses msaa? I thought they were using the new custom buffers

What's "custom buffers"? It's what I've been told so you either take it or not. Note that MSAA in this case is not really MSAA as it forces full subpixel shading and rasterization, the reason MSAA is used is because it's allowing you to program the subsamples positions which in turn allows the creation of said "checkerboard" when rendering a frame and another "checkerboard" with different subsamples positions for another frame.

Pro likely have something which helps with temporal reconstruction though - the primitive ID buffer or what's it's called - which may be a reason why the artifacting from checkerboarding is less pronounced in the Pro titles than in WD2.
 
Friendly reminder that you don't have to use temporal filtering to get 60fps, but can instead gain the extra frames needed by not using screen space reflections. It made the difference between 35fps and 55-70fps for me. (Also dropped shadows to Very High for a touch more stability).

Do we know why all CPU's are being maxed out though? Is it regardless of settings used? My i7 6700 is constantly between 90-100% usage which is insane. How did they get the game to run at all on consoles?!
 
What's "custom buffers"? It's what I've been told so you either take it or not. Note that MSAA in this case is not really MSAA as it forces full subpixel shading and rasterization, the reason MSAA is used is because it's allowing you to program the subsamples positions which in turn allows the creation of said "checkerboard" when rendering a frame and another "checkerboard" with different subsamples positions for another frame.

Pro likely have something which helps with temporal reconstruction though - the primitive ID buffer or what's it's called - which may be a reason why the artifacting from checkerboarding is less pronounced in the Pro titles than in WD2.

http://www.eurogamer.net/articles/d...tation-4-pro-how-sony-made-a-4k-games-machine

it really seems like something entirely different
 

NoPiece

Member
So Uplay never recognize the "Remember Me" setting on the login screen so everytime my pc comes out of sleep mode I have been logged out because of innactivity and I have to relog... I then have to re-type my whole email and password and then the 2 factor code EVERY. SINGLE. TIME.


So stupid

I hate this so much...
 
Friendly reminder that you don't have to use temporal filtering to get 60fps, but can instead gain the extra frames needed by not using screen space reflections. It made the difference between 35fps and 55-70fps for me. (Also dropped shadows to Very High for a touch more stability).

Do we know why all CPU's are being maxed out though? Is it regardless of settings used? My i7 6700 is constantly between 90-100% usage which is insane. How did they get the game to run at all on consoles?!

SSR adds so much depth to the world that it's not worth the tradeoff imo. I'll take the shimmering over the lack of reflection any day.
 

dr_rus

Member

Not at all. The ID buffer is "super stencil" while stencil is a part of the Z buffer which in turn is what is being filled by MSAA h/w. Seems perfectly in line with what Ubi is doing with their "MSAA trick" temporal reconstruction. ID buffer likely helps the Pro's GPU to better identify the primitive edges between frames which in turn should help in better temporal reconstruction and lessen the artifacts typical in such approaches.

I'm actually quite a bit interested in how "h/w" this ID buffer really is as this may all be just s/w really re-using the same old MSAA h/w in a different way.

Clearly if the camera isn't moving we can insert the previous frame's colours and essentially get perfect 4K imagery. But even if the camera is moving or parts of the scene are moving, we can use the IDs - both object ID and triangle ID to hunt for an appropriate part of the previous frame and use that. So the IDs give us some certainty about how to use the previous frame.
This pretty much sounds like inserting an alternated MSAA pattern of a previous frame into the new frame. Plus the temporal accumulation assisted by the indexed Z buffer data (super-stencil / ID buffer).

But we're doubling pixel shader workload, there are other overheads as well and it may not be possible to from 1080p native all the way up to 2160p checkerboard.
The easiest way of doubling the pixel shader workload from 1080p is to use MSAA 2x on it and force subsample shading (use SGSSAA 2x essentially).

The big question is how custom this h/w is really - it may be a slight modification of the standard depth / MSAA h/w adding some bits to depth writes to fill the ID buffer - or it may not even be h/w based at all.
 

JaseC

gave away the keys to the kingdom.

I saw a snippet of the Steam Community announcement in Steam's Detail View and launched the game thinking I'd be able to "cheese" the dog fighting mission to quickly gain cash, only to realise that the patch hadn't yet released. Ubi's testing branch was last updated just a few hours ago for the first time since the weekend and the time frame is a rather vague "later this week", so I'm assuming the patch is actually ready but Ubi is waiting for the console variants to pass certification -- something I observed with AssCreed Unity. I can understand desiring patch release parity for games with cross-platform multiplayer, but otherwise it's rather pointless and introduces an unnecessary delay for PC players.

In short: Bah. Let me finish buying up the car dealerships already.

Edit: Oh, hey, the patch was pushed out shortly after I made this post.
 
SSR adds so much depth to the world that it's not worth the tradeoff imo. I'll take the shimmering over the lack of reflection any day.

You still get SSR in big bodies of water, but also the reflections the game without SSR do a decent enough job of showing the different materials, like vents, polished cars, etc...

The main reason I turned it off was actually because I was driving a very reflective car which had hundreds of aliased lines drawn across it because of the SSR. It just looked horrendous in that instance.

But it does add a lot to the environment and indoor locations.
 
This latest patch finally fixed the framerate issues that the first patch introduced. I was playing with an FPS of around 60 in 1440p for the first few days, and then the patch dropped and halved my FPS no matter what I did. Today's patch brought my framerate back up to where it used to be again, which is going to make the game much more pleasant to play.
 

Iceternal

Member
They didn't fix the shadows setting problem.

You have to restart the game to see the change.

And it's kinda crazy that I can't get steady 30 fps on max settings with a 980 Ti, at 1080P.
 
not every scene takes place in the open world. theres several like for like scenes with no place for variability

That's not necessarily how things work on PC, my own results with the patch serving as an example of why. It would be more correct to say "performance is worse for this youtuber". Doesn't make it true for other people if their hardware is different, though it could be indicative of that.

im going to take an actual video showing performance as more likely to represent reality. these threads are always filled with comments claiming every new patch transformed performance without a shred of evidence and never being shown by actual benchmarks.
 
im going to take an actual video showing performance as more likely to represent reality.

You missed my point. I'm saying one person's benchmarks are not representative of real world performance changes on PC since so much depends upon what combinations of hardware someone has. I'm not denying that the youtuber linked to had worse performance after the patch, but that might not mean anything for anyone else. I myself pretty much doubled my performance with this patch (admittedly because the previous patch halved it, but hey), despite your underhanded attempt to tell me that you don't trust me. That very few other people reported the same framerate problems as I did last patch should be a pretty good indication that performance changes are sometimes restricted to certain groups of people with certain system configurations.
 

JaseC

gave away the keys to the kingdom.
not every scene takes place in the open world. theres several like for like scenes with no place for variability

Well, I'd say there are only a couple (the very beginning of the game and the subsequent cutscene), with the rest offering little room for variation, but that's immaterial as the sole scene that performs consistently outside of what could be considered a margin of error is the one with Marcus overlooking the bay... only the gap is understandable as the same scene in the v1.07 benchmark is rendering noticeably -- not a lot, but noticeably -- more fog.
 
Well, I'd say there are only a couple (the very beginning of the game and the subsequent cutscene), with the rest offering little room for variation, but that's immaterial as the sole scene that performs consistently outside of what could be considered a margin of error is the one with Marcus overlooking the bay... only the gap is understandable as the same scene in the v1.07 benchmark is rendering noticeably -- not a lot, but noticeably -- more fog.

small degradation but big and consistent enough to not be classified as margin of error
 

JayB1920

Member
So is HMSSAO or SSBC supposed to be better? The game says HMSSAO is superior while the geforce guide says SSBC is superior. On another note TXAA works really well in this game. Too bad I can't get good performance with it on my GTX 980. Game pretty much needs either 4K or TXAA
 

kaskade

Member
Could my CPU be the issue? I have a 3570k OC to 4.4. My performance does seem a little worse post patch though. I'm at the mission where you race that car from the movie around the city and frame drops are insane. I even dropped my settings substantially and I still get the drops. I don't know if I'd even call it a drop it's more like a brief freeze. It's almost unplayable at this point.
 
I know when I first started the game early this morning there was a huge spike in CPU usage initially but leveled off after a few minutes but beyond that I didn't pay super close attention to utilization.
 

sertopico

Member
Looks like they finally fixed the nvidia cloud feature that I complained weeks ago, which no one seemed to bother much about it.

Yeah, finally I've been able to see this fog in action, it's pretty impressive I must say, a step forward if compared to fog techniques used in other games. It's really demanding though.

Regarding perfs, after the patches I am getting a more stable framerate and that annoying stuttering occurring when you drive through the city with Ultra textures has disappeared. Much more enjoyable now.
 
Top Bottom