[gamegpu] Dragon Age Inquisition performance

[Deders]

My rig should be quite playable still not ideal. Is the game graphics really that good to justify bringing dual Titans to less than 40fps? I suspect poor optimization as usual, I hope either a patch or a driver improves NV's cards performance and AMD's too, although they fare very well comparatively. And I suspect there will be much more intense moments in the game then where they tested. Low 20 for sure at times, maybe I just lower settings, still seems a but ridiculous for me to have to lower settings to get good fps, maybe there's some graphics option which does almost nothing for IQ and lowers performance dramatically

You should be fine maxing it out if SLI scales well, I'm running it on a single 780 at 2560x1080 maxed except for MSAA and there are only a couple of small areas (so far, when you initially exit from where you were recovering in the beginning and Redcliffe) that go to 25fps but no lower, it's still very playable. Most of the time it's between 35 and 60fps.

Look at the benchmark from computerbase.de its very clear that the default inbuilt benchmark is NOT representative of gameplay results.

Again, another useless canned benchmark.

I would actually disagree here at least in my experience the benchmark results are very similar to what I get in game.

 

[Bateluer]

The in game benchmark . . . trying to see if this is just me here. When I start it, I stare at a black screen with the Inquisition loading icon in the lower right for a few minutes(with an audio loop/stutter), then a few blips of in game scenes and a frame counter that last only a few seconds, then another minute of black loading, then results. Anyone see this too?

 

[redzo]

720p for CPU testing ?? :thumbsdown:

I was wondering exactly the opposite. What's with all other high resolution cpu tests? Aren't cpu tests suppose to be run at lower resolution? At high resolution you usually get bottle-necked by the gpu and cpu performance becomes less obvious.

 

[Deders]

The in game benchmark . . . trying to see if this is just me here. When I start it, I stare at a black screen with the Inquisition loading icon in the lower right for a few minutes(with an audio loop/stutter), then a few blips of in game scenes and a frame counter that last only a few seconds, then another minute of black loading, then results. Anyone see this too?

Have you got it installed on a HDD? I've got mine hard linked to my SSD so loading times are soo much quicker. Don't think I could bear playing it on a HDD. That would at least explain the black screens.

As for the benchmark itself it seems to be synced with a fixed framerate so that the scene will move faster the higher the fps, unlike most benchmarks which sync with a more realistic speed. You can see what i mean if you enable fixed framerate in most 3Dmark benchmarks.

I was wondering exactly the opposite. What's with all other high resolution cpu tests? Aren't cpu tests suppose to be run at lower resolution? At high resolution you usually get bottle-necked by the gpu and cpu performance becomes less obvious.

It depends on what you are trying to test. If you are benchmarking the CPU itself then yes lower resolutions are the way forward but if you want to test how a game performs in real life situations and see how each CPU affects the game then naturally it's better to test with settings you are most likely to use.

 

[Bateluer]

Have you got it installed on a HDD? I've got mine hard linked to my SSD so loading times are soo much quicker. Don't think I could bear playing it on a HDD. That would at least explain the black screens.

As for the benchmark itself it seems to be synced with a fixed framerate so that the scene will move faster the higher the fps, unlike most benchmarks which sync with a more realistic speed. You can see what i mean if you enable fixed framerate in most 3Dmark benchmarks.

Yeah, its running off my mechanical drive. Given the concerns, though mostly baseless, about Denuvo DRM thrashing SSDs, I thought it best to install it to my mechanical drive initially.

 

[Deders]

Yeah, its running off my mechanical drive. Given the concerns, though mostly baseless, about Denuvo DRM thrashing SSDs, I thought it best to install it to my mechanical drive initially.

I did notice an increase in bytes written after playing the game for a while but it won't be too much and I can always take it off later.

 

[frozentundra123456]

I noticed a strange slowdown last night every time I transition to a new area. For about 2 minutes, CPU usage goes to 100 percent and the game stutters badly. After that, CPU usage drops back to 60 to 80 percent and the game runs normally. I have an i5 2320 and a HD7770. I had not noticed this before, although I just got out of the prologue, so was not really using fast travel I guess. Anyone else notice this?

 

[raghu78]

Another DAI review. unfortunately they use the integrated game benchmark tool instead of a fraps recording of actual gameplay. still for what its worth the R9 290 performs admirably matching GTX 780 Ti while the R9 290X is close behind GTX 980.

http://www.techspot.com/review/921-dragon-age-inquisition-benchmarks/page5.html

 

[Deders]

I noticed a strange slowdown last night every time I transition to a new area. For about 2 minutes, CPU usage goes to 100 percent and the game stutters badly. After that, CPU usage drops back to 60 to 80 percent and the game runs normally. I have an i5 2320 and a HD7770. I had not noticed this before, although I just got out of the prologue, so was not really using fast travel I guess. Anyone else notice this?

I can't say I have, usually my CPU is above 90% most of the time. Might be worth checking Vram usage in GPUz, there might be additional streaming going on initially in tight Vram conditions.

How much system Ram do you have?

 

[RussianSensation]

another review for DAI. GTX 980 is the clear leader. Both Nvidia and AMD cards perform well. A fact which people have to appreciate is Gaming Evolved titles run very well on Nvidia cards while the contrary cannot be said about TWIMTBP titles which use Gameworks features.
http://www.sweclockers.com/artikel/19653-snabbtest-grafikprestanda-i-dragon-age-inquisition

No surprises really as that's been the case for 2+ years now.

3+ months old themes continue:

- Kepler performance is falling off (280X/7970Ghz > 680/770, 290/290X > 780 and 780Ti is barely up there with a 290X). Considering how expensive 770/780/780Ti were vs. 280X/290/290X, this underlines how paying extra for those NV cards ended up providing no tangible performance in modern games coming out now. That's why I keep advocating buying a card with 80-85% of the top-end and just upgrading more often as the extra 15-20% performance at the top amounts to little in 1-2 years.

- 980 continues to perform at the top but given its $550 price vs. $300 for 290X and $250 for an after-market 290, the performance advantage is honestly very underwhelming (the only bright spot is the low power usage of the 980). 980 is one of the most overpriced NV cards in years on the price/performance metric, far worse than 480 was vs. 5870, 580 was vs. 6970 or 780Ti was vs. 290X. 780Ti was 7-15% faster for $150 more, but 980 is $250 more for 20% more and with some after-market 980's priced at $580-600, you are asked to pay nearly double for 20% more performance. 980 SLI performance in multi-monitor and 4K is also not convincingly better for nearly double the price. D: If 980 was an AMD card, it would be reaped apart to shreds. Reminds me of all those "7970 undelivered threads" 3 years ago but 7970 vs. 580 really puts 980 in a very bad light against the 290X, minus the same backlash of course....figures.

- 980's lead over 970 is very consistent in the latest games, suggesting 970 is clearly TMU and/or shader performance bottlenecked (both of those are cut by 20% from the 980).

It's really shocking to see that in the last 3 months Kepler performance took a major dive, probably because the driver team is now focused on optimizing for GM204 but given how few people out of the overall market own GM204 vs 2+ year old Kepler cards, this seems short-sighted on NV's part to stop focusing on Kepler this early. There is no way we should have a 780TI performing similar to 290X but that's exactly what's happening. Not only that but 780TI is having trouble outperforming the 970 and 780 is often barely keeping up or beating a 280X/7970Ghz.

Overall we have 280X/290/290X performing very well vs. NV's MUCH more expensive cards in 9 big games this quarter: DAI, CIV BE, FC4, The Evil Within, COD : AW, Vanishing of Ethan Carter, Alien Isolation, Ryse Son of Rome and Shadow of Mordor. The only notable stand-outs for NV are Lords of the Fallen and Unity.

It's possible the rumours of Fiji 380X competing with 980 and Bermuda XT R9 390/390X going against GM200 parts might be true because 980's lead over 290X in newer games is decreasing.

 

[Silverforce11]

It definitely looks to me like NV isn't working hard to optimize Kepler performance in all those games you've mentioned. The performance gap between the 780ti vs 980 on launch was ~5%, and the 970 was never so close to the 780ti. Now they are even and the 980 has a ~20% performance lead.

@RS you got a good point, I did notice in many games recently that R290/X is beating the 780/ti, I wouldn't say GCN is more future proof than Kepler, certainly most here wouldn't think that either. So it has to be lack of optimization for Kepler since GM204 launched.

 

[96Firebird]

Did anyone ever come to think that Kepler driver improvements are tapped out and Maxwell driver improvements are just beginning?

 

[Silverforce11]

Did anyone ever come to think that Kepler driver improvements are tapped out and Maxwell driver improvements are just beginning?

If that's true, then it means GCN isn't tapped out since its now beating Kepler across the range. I find that hard to believe after how well Kepler has kept its lead, to suddenly tank after GM204 launch. :/

 

[96Firebird]

Perhaps AMD decided to put their driver improvements into overdrive, since they were being beat by Maxwell and it doesn't look like they have any new GPUs before years end. They can't just keep lowering prices, they need to attack Maxwell on both fronts.

 

[caswow]

Perhaps AMD decided to put their driver improvements into overdrive, since they were being beat by Maxwell and it doesn't look like they have any new GPUs before years end. They can't just keep lowering prices, they need to attack Maxwell on both fronts.

excuse me? i still own a kepler card and i wont upgrade anytime soon. especially nvidia which is praised as the software overlord i cant accept this. i hope it wont continue like that because i change between amd and nvidia and depending on amds next step i will consider amd or nv...this is not the first time amds older cards hold their position better than their counterparts for its time.

 

[frozentundra123456]

Another DAI review. unfortunately they use the integrated game benchmark tool instead of a fraps recording of actual gameplay. still for what its worth the R9 290 performs admirably matching GTX 780 Ti while the R9 290X is close behind GTX 980.

http://www.techspot.com/review/921-dragon-age-inquisition-benchmarks/page5.html

So looking at the cpu tests, how did they manage to get results for an Athlon X2? Thought the game did not run on dual cores without hyperthreading? Maybe the benchmark does but not the game? If true, seems like the game should at least run on a dual core.

 

[96Firebird]

excuse me? i still own a kepler card and i wont upgrade anytime soon. especially nvidia which is praised as the software overlord i cant accept this. i hope it wont continue like that because i change between amd and nvidia and depending on amds next step i will consider amd or nv...this is not the first time amds older cards hold their position better than their counterparts for its time.

Ok?

Don't accept it, I don't care. I'm just sharing my opinions on the matter.

 

[Spjut]

I don't doubt that Nvidia is prioritizing Maxwell driver optimisations, but we also know that Maxwell has some certain advantages over Kepler and Fermi.

For example, Nvidia could perhaps simply be encouraging the use of compute, which both GCN and Maxwell typically are better at than Kepler.

Perhaps Maxwell and GCN also get an advantage due to having full support for DX11.1/11.2? Frostbite 3 supported DX11.1 already for BF4, perhaps it's been improved further since last year.

 

[bystander36]

I noticed a strange slowdown last night every time I transition to a new area. For about 2 minutes, CPU usage goes to 100 percent and the game stutters badly. After that, CPU usage drops back to 60 to 80 percent and the game runs normally. I have an i5 2320 and a HD7770. I had not noticed this before, although I just got out of the prologue, so was not really using fast travel I guess. Anyone else notice this?

I've had that happen, though it usually takes many transitions before it does. If you don't reload right away, you will speed up over a few minutes of play, but other bugs start to show itself.

 

[exar333]

I don't doubt that Nvidia is prioritizing Maxwell driver optimisations, but we also know that Maxwell has some certain advantages over Kepler and Fermi.

For example, Nvidia could perhaps simply be encouraging the use of compute, which both GCN and Maxwell typically are better at than Kepler.

Perhaps Maxwell and GCN also get an advantage due to having full support for DX11.1/11.2? Frostbite 3 supported DX11.1 already for BF4, perhaps it's been improved further since last year.

Keep in mind that all new Maxwell cards are 4GB+. I think the console port VRAM usage will continue to hamper older 2GB and 3GB cards.

Case in point, DAI is using almost 4GB of VRAM on my rig, with NO MSAA enabled...crazy.

 

[cytoSiN]

Anyone running at 5760x1200 (or even 1080)? If so, what settings and hardware? I'm hoping to try it with 2x970 on first sig rig (if they ever come back in stock) at highest possible settings.

 

[frozentundra123456]

I've had that happen, though it usually takes many transitions before it does. If you don't reload right away, you will speed up over a few minutes of play, but other bugs start to show itself.

Yea, like I said, it does eventually go away. Edit: I was seeing it pretty much every time I transition to a new area (fast travel). I first noticed it yesterday when they were having server problems as well, but I dont see how that would cause this effect. In response to the other poster who responded, I have only a HD7770, so 1gb of vram. I have 8 gb of system ram, and the game uses more than most I have seen, but only up to maybe 60% ram usage, so I dont think I am running out of system ram, vram could be a problem though for sure. All in all, I expected Frostbyte to be more optimized, but I guess it is better than most of the ubisoft recent releases. I also cant get mantle to work with my card. The options screen says only DX11 is available. Dont think I downloaded the latest beta drivers though.

 

[TheUnk]

Anyone running at 5760x1200 (or even 1080)? If so, what settings and hardware? I'm hoping to try it with 2x970 on first sig rig (if they ever come back in stock) at highest possible settings.

I tried 5760x1080 on 2x980s and gave up. Could not find a balance I liked in quality and smooth fps. Some areas would run smooth well over 60fps and then others would drop the 30 and below. There is something about this game that when your fps gets low, rotating the camera has a severe stuttering effect.

The most frustrating thing about it is that the majority of graphic options all tell you they require a restart, makes it very hard to determine which settings I should adjust.

 

[escrow4]

20hrs in, 780 Ti GHz sustains 60FPS in smaller areas, drops to 50 ish in Redcliffe or higher, some areas around 55FPS maxed out @ 1200p with AA disabled (post process enabled). VRAM pegged closed to 3GB, RAM usage a bit over 3GB just for the game, and all six cores of my 5930K are evenly loaded to 3.7GHz consistently barring cutscene drops. On the contrary this game is optimised . . . . . if you have the hardware.

 

[cytoSiN]

I tried 5760x1080 on 2x980s and gave up. Could not find a balance I liked in quality and smooth fps. Some areas would run smooth well over 60fps and then others would drop the 30 and below. There is something about this game that when your fps gets low, rotating the camera has a severe stuttering effect.

The most frustrating thing about it is that the majority of graphic options all tell you they require a restart, makes it very hard to determine which settings I should adjust.

Thanks for the info. That's depressing though.