Upcoming graphical powerhouses

[Carfax83]

Nvidia's parternship with Ubisoft did nothing to improve experience in the game, and did nothing to show Ubisoft gives a crap about PC.

On the contrary, AC IV has significant PC enhancements including 4K support, TXAA, PhysX, HBAO+, PCSS etcetera..

Watch Dogs has 4K support, and ultra level detail which is much greater than what you find on the console versions and looks a lot better.. It's a shame that the engine has so many issues with resource management on the PC, as the game can look beautiful at times..

They purposely gimp the game to make consoles look good. They release patches when they feel like it. DX11 multithreading did pretty much nothing to benefit the game, as Nvidia and AMD are neck and neck in performance. TXAA looks like the same old smeared crap that it always has.

I don't think the game was purposely gimped. I think the engine has a lot of performance issues and they cut some features or didn't fully implement them due to lack of time. Also DX11 multithreading does help, in CPU limited situations as shown by pcgameshardware.de's CPU and driver overhead benchmark.

http://www.pcgameshardware.de/scree..._Dogs-CPU-Driver-Overhead-Benchmarks-pcgh.png

The only benefit TW3 has going for it is PhysX, because I think CDPR will make good use of it, and not waste it on the stupidity of billions unrealistic particles.

I'm surprised a rabid anti NVidia guy like you would think PhysX a benefit in the Witcher 3

 

[cmdrdredd]

Watch Dogs looks better on PC than console by far. Anyone who says different is blind

 

[chimaxi83]

On the contrary, AC IV has significant PC enhancements including 4K support, TXAA, PhysX, HBAO+, PCSS etcetera..

Watch Dogs has 4K support, and ultra level detail which is much greater than what you find on the console versions and looks a lot better.. It's a shame that the engine has so many issues with resource management on the PC, as the game can look beautiful at times..



I don't think the game was purposely gimped. I think the engine has a lot of performance issues and they cut some features or didn't fully implement them due to lack of time. Also DX11 multithreading does help, in CPU limited situations as shown by pcgameshardware.de's CPU and driver overhead benchmark.

http://www.pcgameshardware.de/scree..._Dogs-CPU-Driver-Overhead-Benchmarks-pcgh.png



I'm surprised a rabid anti NVidia guy like you would think PhysX a benefit in the Witcher 3

Rabid? Wtf are you talking about? I'm not anti Nvidia, I'm anti dumbass. I'm not a rabid anything, unlike the retard loyalist crews in VC&G. I call out idiots when I see them, doesn't matter who they wave pom poms for. I buy whatever looks best when I'm in the market, plain and simple.

So DX11 multithreading helps in the same fashion that Mantle helps supported games? Lower end CPUs get the most benefit?

Watch Dogs looks better on PC than console by far. Anyone who says different is blind

It doesn't look bad by any stretch, it just didn't live up to the hype until you use a mod to enable the "missing" features.

 

[Carfax83]

Rabid? Wtf are you talking about? I'm not anti Nvidia, I'm anti dumbass. I'm not a rabid anything, unlike the retard loyalist crews in VC&G. I call out idiots when I see them, doesn't matter who they wave pom poms for. I buy whatever looks best when I'm in the market, plain and simple.

OK fine, perhaps rabid was exaggerating things a bit. I've had some many dust ups with hardcore AMD fans in the VC&G that I've lost count of who are the real fanatics, and who are just AMD supporters..

So DX11 multithreading helps in the same fashion that Mantle helps supported games? Lower end CPUs get the most benefit?

DX11 multithreading isn't nearly as efficient as Mantle, so I'm not certain that I would say that lower end CPUs get the most benefit. Mantle has FAR less lower overhead and so the performance boost is bigger than with DX11 multithreading for low end CPUs. Still, for a feature that is part of the DX11 spec, it's definitely better than manual threading..

The main purpose of DX11 multithreading is that it allows much easier scaling of rendering to multiple threads than manual threading.. But it has a much higher operating cost than Mantle, and it relies on the quality of the drivers for effective management..

Still, the benefit is similar in that CPU bottlenecks are drastically diminished. If you look at that graph, just enabling HT results in a 18% performance increase on NVidia hardware, but on AMD hardware, enabling HT is only 1% gain.. So the NVidia driver is able to utilize more rendering threads courtesy of DX11 multithreading, compared to the AMD driver which is likely limited to only two rendering threads..

This doesn't mean that NVidia is going to be faster than AMD in every situation of course, but in CPU limited circumstances, NVidia will have the advantage.. You would see a similar occurrence in SLI. At 1600p and below, NVidia should have much better SLI performance than AMD in xFire.

Radeons in xFire in Watch Dogs:

NVidia in SLI in Watch Dogs:

Actually to be honest, NVidia's advantage in Watch Dogs is seriously blunted by the engine's inefficiency when it comes to resources. The game says you need 3GB of VRAM for ultra textures, but you really need 4GB; especially at resolutions beyond 1080p..

 

[RussianSensation]

Actually to be honest, NVidia's advantage in Watch Dogs is seriously blunted by the engine's inefficiency when it comes to resources. The game says you need 3GB of VRAM for ultra textures, but you really need 4GB; especially at resolutions beyond 1080p..

Problem is HardOCP found the opposite - NV cards perform worse in Watch Dogs. That's not related to textures either:

1) AO - Believe it or not the GeForce GTX 780 Ti is taking the greater performance hit enabling HBAO+ High versus no AO. GeForce GTX 780 Ti takes a 20% performance hit, while XFX Radeon R9 290X DD takes a lesser 15% performance hit.
http://www.hardocp.com/article/2014...formance_image_quality_review/10#.U_zuhPmSzsc

2) Shader based AA in this game also takes a performance hit on the GeForce GTX 780 Ti. The impact on performance is greater than that on AMD Radeon R9 290X. Temporal SMAA for example has a 12% hit here, whereas it is 9% on AMD Radeon R9 290X.
http://www.hardocp.com/article/2014...formance_image_quality_review/11#.U_zuuvmSzsc

While in theory the multi-threading should help NV's cards in SLI, in reality R9 290X CF crushes 780TI SLI in WD by 33-40%.

http://www.hardocp.com/article/2014/08/18/watch_dogs_performance_image_quality_review/7#.U_zvLPmSzsc

Ubisoft's inability to make NV-centric features run faster on NV and not being able to fit such mediocre textures in a 3GB VRAM is seriously disappointing. The High Textures in the game look like they came out from a 2007 PC game. This is why I am going in far more pessimistic for AC Unity vs. Witcher 3.

 

[tidy]

.

 

[Carfax83]

Problem is HardOCP found the opposite - NV cards perform worse in Watch Dogs. That's not related to textures either:

1) AO - Believe it or not the GeForce GTX 780 Ti is taking the greater performance hit enabling HBAO+ High versus no AO. GeForce GTX 780 Ti takes a 20% performance hit, while XFX Radeon R9 290X DD takes a lesser 15% performance hit.
http://www.hardocp.com/article/2014...formance_image_quality_review/10#.U_zuhPmSzsc

You think a 5% difference, especially in an unoptimized game like Watch Dogs is significant? I bet if they reran the benchmarks seven times they'd get seven different results, that's how unpredictable this game is..

And that extra 1GB of VRAM makes a huge difference in preventing stuttering with ultra textures believe me..

While in theory the multi-threading should help NV's cards in SLI, in reality R9 290X CF crushes 780TI SLI in WD by 33-40%.

You know full well that AMD has better performance at 4K rez than NVidia.. Thats in practically every title, and Watch Dogs is no different..

Also, DX11 multithreading would be a non factor at GPU bound resolutions anyway..

Ubisoft's inability to make NV-centric features run faster on NV and not being able to fit such mediocre textures in a 3GB VRAM is seriously disappointing. The High Textures in the game look like they came out from a 2007 PC game. This is why I am going in far more pessimistic for AC Unity vs. Witcher 3.

I don't think Watch Dogs is indicative of Ubisoft games in general. I just think that it's indicative of that particular team. Ubisoft is a huge developer and they have multiple teams working on a game in unison.

The Montreal studio did the PC version for Watch Dogs, and they screwed up in a catastrophic manner. Or perhaps the engine just wasn't ready for prime time..

AC Unity PC is being done by Ubisoft Kiev, and their track record is pretty decent..

 

[finbarqs]

do we still need mods for Watch dogs to look awesome? Or has it finally been patched where it looks pretty decent now? (at least as good as the mod)?

 

[Cerb]

The Witcher 3 is the only one I've bought, so I guess that makes it the one. But frankly, I don't really care about that aspect of it. I mean I care that they do a good job, but more by not having vast texture density variations, obnoxious clipping, bad animations, etc.. I could go back and play an old game like BG&E (original PC version), or Giants, or the pre-HD Serious Sams, and be fine with the graphics, because they 'got' making an aesthetic, rather than overworking artists for moar content.

 

[chimaxi83]

do we still need mods for Watch dogs to look awesome? Or has it finally been patched where it looks pretty decent now? (at least as good as the mod)?

You still need mods.

 

[SLU Aequitas]

I will say though, with the mods Watch Dogs is absolutely amazing. It also runs very smooth in ultra for me, much better than launch.

 

[Carfax83]

I must be the only person who thinks the game looks better on stock ultra. I tried many of the mods, but inevitably ended up going back..

The mods either made things too bright or too windy to the point where it looked unrealistic.. Also the headlight shadows look terrible, so it's better just to keep it off..

 

[boxleitnerb]

Problem is HardOCP found the opposite - NV cards perform worse in Watch Dogs. That's not related to textures either:

1) AO - Believe it or not the GeForce GTX 780 Ti is taking the greater performance hit enabling HBAO+ High versus no AO. GeForce GTX 780 Ti takes a 20% performance hit, while XFX Radeon R9 290X DD takes a lesser 15% performance hit.
http://www.hardocp.com/article/2014...formance_image_quality_review/10#.U_zuhPmSzsc

2) Shader based AA in this game also takes a performance hit on the GeForce GTX 780 Ti. The impact on performance is greater than that on AMD Radeon R9 290X. Temporal SMAA for example has a 12% hit here, whereas it is 9% on AMD Radeon R9 290X.
http://www.hardocp.com/article/2014...formance_image_quality_review/11#.U_zuuvmSzsc

While in theory the multi-threading should help NV's cards in SLI, in reality R9 290X CF crushes 780TI SLI in WD by 33-40%.

http://www.hardocp.com/article/2014/08/18/watch_dogs_performance_image_quality_review/7#.U_zvLPmSzsc

Ubisoft's inability to make NV-centric features run faster on NV and not being able to fit such mediocre textures in a 3GB VRAM is seriously disappointing. The High Textures in the game look like they came out from a 2007 PC game. This is why I am going in far more pessimistic for AC Unity vs. Witcher 3.

HardOCP has a major flaw in their review:
They don't list clocks. I bet the 780 Tis are not reaching their full boost potential, the 290X DDs much more likely always clock with 1 GHz. It has been proven that Nvidia and AMD cards with reference coolers can throttle depending on the case/airflow. This goes especially for MGPU. I find it unprofessional to compare a card with a 3rd party cooler with a reference model from the competition.

I don't think that [H] is a realiable source as long as they ignore the temperature dependency of clockspeeds. They should either benchmark default vs default (quiet mode on AMD) or maxed out power/temp target vs. uber mode or pitch models with similar cooling performance against each other. Everything else is nonsense.

 

[Carfax83]

HardOCP has a major flaw in their review:
They don't list clocks. I bet the 780 Tis are not reaching their full boost potential, the 290X DDs much more likely always clock with 1 GHz. It has been proven that Nvidia and AMD cards with reference coolers can throttle depending on the case/airflow. This goes especially for MGPU. I find it unprofessional to compare a card with a 3rd party cooler with a reference model from the competition.

I don't think that [H] is a realiable source as long as they ignore the temperature dependency of clockspeeds. They should either benchmark default vs default (quiet mode on AMD) or maxed out power/temp target vs. uber mode or pitch models with similar cooling performance against each other. Everything else is nonsense.

Good catch. I never really look at [H] as a reliable source anyway, because I've seen too many reviews where they list "highest playable" settings with frame rate drops below 30 or the average right at 30..

Playable settings is too subjective for my tastes..