• 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

thalixte

Member
Or you can keep temporal filtering on with the 1080p resolution, and increase pixel density (at least 1.25, if not 1.5 with the latest patch).

Another solution is to disable temporal filtering, and enable 2xTXAA, which produces a really good result, at the cost of sharpeness and some fps. You can adjust the sharpening setting to 20 % to unblur the picture.
 
its really not a very good port

wd_2_bay_bridge.png


this isnt even using the extra lod slider
 

JaseC

gave away the keys to the kingdom.
Technical requirements for TWIMTBP games? Either they just came up with that or they're hoping everyone forgets about Arkham Knight and Dishonored 2, which were arguably the worst PC ports of the years they were released.

That appears to be the case:

We take a look at what has become the TWIMTBP Technical Requirements Checklist, an NVIDIA view on what technical readiness means on the PC.
 

Agent_4Seven

Tears of Nintendo
Or you can keep temporal filtering on with the 1080p resolution, and increase pixel density (at least 1.25, if not 1.5 with the latest patch).

Another solution is to disable temporal filtering, and enable 2xTXAA, which produces a really good result, at the cost of sharpeness and some fps. You can adjust the sharpening setting to 20 % to unblur the picture.
Nah, I'll be playing in 1440p and with current settings, thanks.

its really not a very good port

wd_2_bay_bridge.png


this isnt even using the extra lod slider
But it sure as hell much better than Dishonored 2.
 
I REALLY need to OC my 6600K. Don't think I've ever seen a game this CPU-hungry, and had I known this would be the case I would've gone with a 6700K instead. I really thought the 6600K would be a safe bet for high-end PC gaming, especially with a 1080 :(
 

thalixte

Member
Nah, I'll be playing in 1440p and with current settings, thanks.

Ok, but you know that with temporal filtering on and pixel density set to 1.00, there is a downscale, so you are not playing at 1440p, but nearly at 1080p with those settings.
That's the reason why, with resolution set to 1080p, temporal filtering on, and pixel density set to 1.00, the image is so ugly, because you play in this case in nearly 720p.
 

J_Viper

Member
I'm getting hit with some bad hard locks when starting the game.

I've had two instances in a row where I'd start, fast travel, then the game will crash a few seconds later.

Is this an issue caused by a recent patch?

EDIT: Make that three crashes.

I'm glad I got this for free. As much as I'm enjoying the game, this is fucking ridiculous.
 

Durante

Member
its really not a very good port
How does the image you posted indicate whether it's a good or bad port?

As far as I know, it's a game which runs at 30 FPS (not 100% stably either) on consoles, with much lower settings (in particular those affecting CPU load) than the "Ultra" represented there. PC CPUs are faster, but not infinitely so.
 

Pjsprojects

Member
After completing the PS4 version I noticed the PC deluxe version had 50% off, so wanting to see how improved over the console version it is I jumped in.

It's now looking like my 4gb vram card will need replacing this year....
With everything on ultra/max with the texture pack installed,car headlight shadows disabled and only set at 1080p I'm just under my vram limit.
Running without vsync I get between 35 and 50fps so it's best run at the 30fps lock to keep things smooth.

An 8gb vram card is looking like my best bet but without needing to remortgage the house!

Current spec 8350@4.2,R9-290X@stock and 16gb@1833
 

M_A_C

Member
I'm getting hit with some bad hard locks when starting the game.

I've had two instances in a row where I'd start, fast travel, then the game will crash a few seconds later.

Is this an issue caused by a recent patch?

EDIT: Make that three crashes.

I'm glad I got this for free. As much as I'm enjoying the game, this is fucking ridiculous.

Same here. It locks and completely reboots my computer.
 
How does the image you posted indicate whether it's a good or bad port?

As far as I know, it's a game which runs at 30 FPS (not 100% stably either) on consoles, with much lower settings (in particular those affecting CPU load) than the "Ultra" represented there. PC CPUs are faster, but not infinitely so.

without using the extra lod slider theres not much improvement at all over the console versions when it comes to lod. What other cpu affecting settings are there? I struggle to see why a 8 core 16 thread top of the line intel cpu at 4.5 ghz is a bit more than twice as fast a 6 core 6 thread jaguar cpu at 1.6 ghz. Not to mention the lower performance of various other cpus there

The performance of the fx6350 for example
 

Agent_4Seven

Tears of Nintendo
Ok, but you know that with temporal filtering on and pixel density set to 1.00, there is a downscale, so you are not playing at 1440p, but nearly at 1080p with those settings.
That's the reason why, with resolution set to 1080p, temporal filtering on, and pixel density set to 1.00, the image is so ugly, because you play in this case in nearly 720p.
Downscale? There is no indication in game options that temporal filtering is indeed downscale.

Anyway, I've disabled temporal filtering and.... the game does not run very good at all even in 1080p with current settings (which is nowhere near max settings btw). I've seen drops to like 40s and the game cannot maintain 60 FPS most of the time and thats not while I'm driving a car very fast.

I'm afraid that I can't say WD2 is a good PC port at all in terms of performance in native resolutions (1080p / 1440p) and temporal filtering should not be a fix to this problem (if it's indeed a downscale), the game should run in native 1080p 60 FPS on hi-end hardware and it's not most of the time even on not max settings.

What I can say is that the game looks good and performs okay in 1440p with temporal filtering on and it looks better that in native 1080p without temporal filtering. So there's that.
 
Downscale? There is no indication in game options that temporal filtering is indeed downscale.

Anyway, I've disabled temporal filtering and.... the game does not run very good at all even in 1080p with current options which is nowhere near max settings btw. I've seen drops to like 40s and the game cannot maintain 60 FPS most of the time and thats not while I'm driving a car very fast.

I'm afraid that I can't say WD2 is a good PC port at all in terms of performance in native resolutions (1080p / 1440p) and temporal filtering should not be a fix to this problem (if it's indeed a downscale), the game should run in native 1080p 60 FPS on hi-end hardware and it's not most of the time even on not max settings.

What I can say is that the game looks good and performs okay in 1440p with temporal filtering on and it looks better that in native 1080p without temporal filtering. So there's that.

hes just using the wrong terminology. temporal filtering cuts the resolution of each axis in half(1/4th the total number of pixels) and reconstructs edges using 2x msaa.

so 1080p w/ temporal filtering is 960x540 w/ 2xmsaa
 

thalixte

Member
You perfectly right, that was a shortcut in order not to explain the whole process of temporal filtering, sorry if it leads to a misunderstood, but the visual result is a lost in the overall visual quality, which looks like a reduction of the resolution, if pixel density is set to 1.00.
However, i found some settings that produce quite good results for my rig (and perhaps for Agent_4Seven's one):
i put the resolution to 4k, and lower the pixel density to 0.75, applying the temporal filtering w/SMAA.
These settings provide better performances than no temporal filtering + 2xTXAA + pixel density to 1.00, with almost the same IQ, if not better.
I suggest Agent_4Seven to try it.
 

dr_rus

Member
without using the extra lod slider theres not much improvement at all over the console versions when it comes to lod. What other cpu affecting settings are there? I struggle to see why a 8 core 16 thread top of the line intel cpu at 4.5 ghz is a bit more than twice as fast a 6 core 6 thread jaguar cpu at 1.6 ghz. Not to mention the lower performance of various other cpus there

The performance of the fx6350 for example

Why do you think that this graph is in any way comparable to the game's console performance?

"A good port" is mostly about issues a game may have like a fixed rendering resolution or the need to go through a m+k launcher on a game's start. There's no requirement for "a good port" to perform to your liking, never was, never will be.

hes just using the wrong terminology. temporal filtering cuts the resolution of each axis in half(1/4th the total number of pixels) and reconstructs edges using 2x msaa.

so 1080p w/ temporal filtering is 960x540 w/ 2xmsaa

This is completely misleading as well since it's not in fact "MSAA" as it forces the shading of all rendered subsamples - which MSAA doesn't being MSAA. So you're looking at half the spatial resolution being rendered each frame, not 1/4th - it would be inaccurate to claim any sort of number since we're dealing with a checkerboard grid here but you could say "1920x540" which would at least be closer to the actual number of pixels processed each frame.
 
Why do you think that this graph is in any way comparable to the game's console performance?

"A good port" is mostly about issues a game may have like a fixed rendering resolution or the need to go through a m+k launcher on a game's start. There's no requirement for "a good port" to perform to your liking, never was, never will be.



This is completely misleading as well since it's not in fact "MSAA" as it forces the shading of all rendered subsamples - which MSAA doesn't being MSAA. So you're looking at half the spatial resolution being rendered each frame, not 1/4th - it would be inaccurate to claim any sort of number since we're dealing with a checkerboard grid here but you could say "1920x540" which would at least be closer to the actual number of pixels processed each frame.

So why was quantum break on dx12 a bad port?
 

dr_rus

Member
So why was quantum break on dx12 a bad port?

Because its performance profile between different hardware vendors clearly pointed to it being badly optimized - the fact which was proven by the DX11 version half a year later. There are no such facts in case of WD2.

QB also had its share of pure porting issues like the lack of exit option, inability to turn off temporal reconstruction, etc. - which all are what is considered signs of "a bad port" more than performance.
 
Because its performance profile between different hardware vendors clearly pointed to it being badly optimized - the fact which was proven by the DX11 version half a year later. There are no such facts in case of WD2.

QB also had its share of pure porting issues like the lack of exit option, inability to turn off temporal reconstruction, etc. - which all are what is considered signs of "a bad port" more than performance.

dude it was considered a bad port primarily because of its performance. and in the case of watch dogs 2, you have a fx6350, which is more than 2x faster than 6 jaguar cores at 1.6ghz performing worse than the console version
 

Agent_4Seven

Tears of Nintendo
QB also had its share of pure porting issues like the lack of exit option, inability to turn off temporal reconstruction, etc. - which all are what is considered signs of "a bad port" more than performance.
But QB is still not a good PC port in terms of performance too in native resolutions (1080p / 1440p) and without temporal reconstruction. It is better than DX12 version for sure, but not by a whole lot cuz the game is not performing very well without temporal reconstruction in native 1080p / 1440p on any hardware. There's simply no hardware available that could run this game with constant 60+ FPS all the time without temporal reconstruction.
 

dr_rus

Member
dude it was considered a bad port primarily because of its performance. and in the case of watch dogs 2, you have a fx6350, which is more than 2x faster than 6 jaguar cores at 1.6ghz performing worse than the console version
Nope, the main reasons where lack of options, including even vsync off b/c of UWP 1.0. People seems to forget these things awfully quickly. Performance was bad in comparison, sure - but this has no relation to WD2 as there are no reason to suspect that WD2 is somehow badly optimized for any modern PC h/w. I have no idea what this test of FX6350 is doing.

But QB is still not a good PC port in terms of performance too in native resolutions (1080p / 1440p) and without temporal reconstruction. It is better than DX12 version for sure, but not by a whole lot cuz the game is not performing very well without temporal reconstruction in native 1080p / 1440p on any hardware. There's simply no hardware available that could run this game with constant 60+ FPS all the time without temporal reconstruction.

I played it in 1080p native with almost stable 60 fps and I've had most options on Ultra while doing this. No idea what's wrong with its performance now - seems like a very solid engine for its output in DX11 version. Its biggest issue is probably the lack of good scaling with lower settings meaning that if you're running something less than mid range you're totally out of luck. But then you must consider that the game is running on a mix of low and medium settings with "upscaling" and on 30 fps on XBO which can be considered low end by modern PC standards.
 
Nope, the main reasons where lack of options, including even vsync off b/c of UWP 1.0. People seems to forget these things awfully quickly. Performance was bad in comparison, sure - but this has no relation to WD2 as there are no reason to suspect that WD2 is somehow badly optimized for any modern PC h/w. I have no idea what this test of FX6350 is doing.



I played it in 1080p native with almost stable 60 fps and I've had most options on Ultra while doing this. No idea what's wrong with its performance now - seems like a very solid engine for its output in DX11 version. Its biggest issue is probably the lack of good scaling with lower settings meaning that if you're running something less than mid range you're totally out of luck. But then you must consider that the game is running on a mix of low and medium settings with "upscaling" and on 30 fps on XBO which can be considered low end by modern PC standards.

its just traversing some part of the map with ultra preset, which leaves extra lod slider at 0.

edit - heres the scene for those results "As a place of test used to measure the performance of processors used the exit of the bridge Bay Bridge towards Haum Data Center. The bridge is crowded during the day by a large number of cars. Maneuvering between them in tact to escape from the police is not too difficult. The condition is, of course, adequate performance CPU, which at this point depends very much. About how many you will see on the fifth page."



heres another, again with extra lod slider at 0.
w3_proz.png


heres the benchmark scene

https://www.youtube.com/watch?v=IRVzZL5RvFk
 

dr_rus

Member
its just traversing some part of the map with ultra preset, which leaves extra lod slider at 0.

edit - heres the scene for those results "As a place of test used to measure the performance of processors used the exit of the bridge Bay Bridge towards Haum Data Center. The bridge is crowded during the day by a large number of cars. Maneuvering between them in tact to escape from the police is not too difficult. The condition is, of course, adequate performance CPU, which at this point depends very much. About how many you will see on the fifth page."



heres another, again with extra lod slider at 0.
w3_proz.png


heres the benchmark scene

https://www.youtube.com/watch?v=IRVzZL5RvFk

So the immediate question is - why are you comparing Ultra preset of PC version to what's running on consoles? Is there any evidence to assume that this is a valid comparison? Detail LODs are the easiest to crank up, the hardest to hit CPU and usually the most difficult to spot in-game as a notable quality increase.
 
So the immediate question is - why are you comparing Ultra preset of PC version to what's running on consoles? Is there any evidence to assume that this is a valid comparison? Detail LODs are the easiest to crank up, the hardest to hit CPU and usually the most difficult to spot in-game as a notable quality increase.

the only setting that really impacts cpu performance is the extra lod slider, which again was disabled for both benches. so basically youre SOL if you dont have a high end intel cpu

and the console versions match up petty close to the pc version when not using extra detail in terms of lod. DF even states geometry is the same
 

MaLDo

Member
the only setting that really impacts cpu performance is the extra lod slider, which again was disabled for both benches. so basically youre SOL if you dont have a high end intel cpu


WTF

This is bullshit.

Shadows, vegetation, geometry, terrain and texture streaming (resolution) affect CPU.
 

dr_rus

Member
the only setting that really impacts cpu performance is the extra lod slider, which again was disabled for both benches. so basically youre SOL if you dont have a high end intel cpu

and the console versions match up petty close to the pc version when not using extra detail in terms of lod. DF even states geometry is the same

Highly doubtful that extra LOD is the only thing affecting CPU, even by pure logic you should assume that LODs without "extra" is able to affect the CPU just as heavily - and that's even disregarding the rest of stuff going on in the game and graphics which can be quite different between PC and consoles.

There is some merit in saying that if these LODs are increased to no noticeable quality gain then this is just a bad decision on how to use the CPU power available on PC but then it hardly makes the game into a bad port - a demanding one sure, but not bad.
 
Highly doubtful that extra LOD is the only thing affecting CPU, even by pure logic you should assume that LODs without "extra" is able to affect the CPU just as heavily - and that's even disregarding the rest of stuff going on in the game and graphics which can be quite different between PC and consoles.

There is some merit in saying that if these LODs are increased to no noticeable quality gain then this is just a bad decision on how to use the CPU power available on PC but then it hardly makes the game into a bad port - a demanding one sure, but not bad.

Lowering geometry, terrain, vegetation and shadows from ultra to lowest nets you <20% perf improvement when cpu limited. Geometry acounts for about 15 of that <20
 

DjRalford

Member
Just downloaded a new patch, but the crashing upon start-up still hasn't stopped

Have they addressed this at all?

The crashing for me was always random, i could play anywhere between 10 minutes and 2 hours before getting a crash, very odd and a bit pisstaking too, i tried carrying on where i left off a few days ago and after 30 minutes had a crash, then played another hour or so without issue.
 

dcx4610

Member
I put in about 30 hours into the game. Zero crashes or problems. The lastest patch definitely improved performance. I'm on a 1080 and seem to get more consistent frame rates. I average in the 50s and 60s on Ultra at 1440p. It's a beautiful game but definitely challenging to run.

I started the first Watch Dogs after and it looks great for being a few years old now. At Ultra, I'm getting 80 to 100fps. Pretty disappointing to not hit 144hz but it looks like Ubisoft isn't the best at optimization.
 
Gettinng mostly 60fps on 970 i5 6600k with everything on high and temporal filtering off. Is that normal? I thought with that off it was really bad performance wise. Get a few hitches like when driving, and blowing up man holes but other than that it looks and plays well.
 
Gettinng mostly 60fps on 970 i5 6600k with everything on high and temporal filtering off. Is that normal? I thought with that off it was really bad performance wise. Get a few hitches like when driving, and blowing up man holes but other than that it looks and plays well.

New patch? Is your CPU OC'd?
 

J_Viper

Member
The crashing for me was always random, i could play anywhere between 10 minutes and 2 hours before getting a crash, very odd and a bit pisstaking too, i tried carrying on where i left off a few days ago and after 30 minutes had a crash, then played another hour or so without issue.

Yeah, it seems like after three failed launches, the game works fine, but having my computer lock up like that can't possibly be good for its health.

The last crash gave me a Blue Screen of Death, so I'm out until there's an actual fix.
 

DjRalford

Member
Yeah, it seems like after three failed launches, the game works fine, but having my computer lock up like that can't possibly be good for its health.

The last crash gave me a Blue Screen of Death, so I'm out until there's an actual fix.

Nasty, luckily i have only ever had a standard CTD without any error messages.
 
Oh god. This thread now makes me think I have this temporal filtering thing turned. is it on by default? I'm getting great performance with most things up full. I'm gonna cry when I get home, I know it.
 
Haven't seen any performance increases on my rig (GTX 1070 / i 7-6700K), unfortunately.

Still not terrible, but GTA V gives me at least DOUBLE the FPS with nearly everything maxed out, compared to Watch Dogs 1440p w/ Temporal Filtering and a lot of the settings reduced. I know GTA is on a different level - but it was also released 3 years ago... argh.

Question:

Is there ANY way to turn off cars, objects, people from FLASHING white? It's extremely annoying and totally unnecessary. It looks stupid and takes away from the visuals.

WHYYYYY.
 

Piggus

Member
Not really impressed with the temporal filtering. Even at close to 4k, there's a ton of pixel crawl on my 1080p TV. I get much better results (though a bit softer image) at native 1080p with TAA2x. Unfortunately the game won't even maintain a consistent 30 fps with that setting. It's somehow much more demanding than 1800p with temporal filtering.
 
Between temporal filtering & frame scaling - it's a bit difficult to tell what resolution we're actually getting.

What's the true equivalent of 1440p / 1.0 Pixel Density / Temporal Filtering on?
 
Yesterday I tried WD2 for the first time. I used geforce exp to optimize the game and it was terrible. The framerate drops below 10 fps and in one moment the game crashed. I have an i5 6600K, 980 Ti and 8GB DDR4. What can I do to get stable 1080p60 gameplay?
 
Top Bottom