[gamegpu] Dragon Age Inquisition performance

[bystander36]

The game runs well on an i7 @ 3.8Ghz. It's almost always well over 60 FPS, with a few drops below. I was playing on all high with ultra textures, then zoned to the Hinterlands again, and noticed Tessellation killed performance. I'm playing on medium tessellation now.

 

[Red Hawk]

The game runs well on an i7 @ 3.8Ghz. It's almost always well over 60 FPS, with a few drops below. I was playing on all high with ultra textures, then zoned to the Hinterlands again, and noticed Tessellation killed performance. I'm playing on medium tessellation now.

What's your graphics card bystander?

 

[bystander36]

680 SLI. I wonder if the high tessellation FPS drop was VRAM related.

 

[Red Hawk]

680 SLI. I wonder if the high tessellation FPS drop was VRAM related.

Possibly. With 2 GB 680s in SLI, the bottleneck you're most likely to run into is VRAM. Two 680s should have plenty of raw tessellation power.

I got my new 1080p monitor, and ran the benchmark test on it. 38.2 frames per second with max settings and no MSAA, it reported. To compare, I got 44.4 frames per second at 1440x900 with no MSAA, 39.5 FPS with 2x MSAA at 1440x900, and 35.9 FPS with 4x MSAA. So in summary, just 2x MSAA at 1440x900 is comparable to a jump from 1440x900 to 1080p, and 4x MSAA exceeds that jump.

 

[bystander36]

Possibly. With 2 GB 680s in SLI, the bottleneck you're most likely to run into is VRAM. Two 680s should have plenty of raw tessellation power.

I got my new 1080p monitor, and ran the benchmark test on it. 38.2 frames per second with max settings and no MSAA, it reported. To compare, I got 44.4 frames per second at 1440x900 with no MSAA, 39.5 FPS with 2x MSAA at 1440x900, and 35.9 FPS with 4x MSAA. So in summary, just 2x MSAA at 1440x900 is comparable to a jump from 1440x900 to 1080p, and 4x MSAA exceeds that jump.

I found I could run at high tessellation in some zones, but others had immediate problems. But keep in mind, I'm downsampling from 1440p as well, which might run me into that VRAM limit sooner. The difference between medium and high appears to be in how many rocks had shape. I can live with that.

 

[exar333]

I found I could run at high tessellation in some zones, but others had immediate problems. But keep in mind, I'm downsampling from 1440p as well, which might run me into that VRAM limit sooner. The difference between medium and high appears to be in how many rocks had shape. I can live with that.

Yes, my understanding is that the tessellation setting affects distance rather than level of. I run medium instead of high.

 

[Red Hawk]

I've played a bunch with my 270X and 1080p monitor and I think that, subjectively, the 270X is fine. Not 60 FPS, of course, but it's acceptable. The most trouble I get is in an area with a lot of NPCs, like Redcliffe. There's some stutters here and there, but it's not in combat, so I find that acceptable.

 

[tential]

I've played a bunch with my 270X and 1080p monitor and I think that, subjectively, the 270X is fine. Not 60 FPS, of course, but it's acceptable. The most trouble I get is in an area with a lot of NPCs, like Redcliffe. There's some stutters here and there, but it's not in combat, so I find that acceptable.

Wow what a horrible engine. AC Unity's engine allows it to have 0 effect on FPS when there are a lot of NPCs on the screen...

=D

 

[exar333]

Wow what a horrible engine. AC Unity's engine allows it to have 0 effect on FPS when there are a lot of NPCs on the screen...

=D

I am assuming this was sarcasm? AC Unity engine is about the most buggy engine released in years. Not something to aspire to. I will take slower and reliable vs. 'fast' and terrible. LOL

Easy to handle NPCs when they look like this:

 

[Red Hawk]

Yeah, I'm pretty sure that was sarcasm.

Just to revisit the point about what's needed for console parity: Sony never officially stated the clock speed for the PS4's graphics component, but they did state the theoretical performance in teraFLOPS, which is derived from clock speed and architecture. The PS4 has a theoretical graphics performance of 1.84 teraFLOPS. We can pretty easily compare theoretical performance of the PS4 and Xbox One to AMD cards, since PS4 and Xbox One both use the GCN architecture (such a comparison isn't practical with Nvidia). To compare to my 270X, I have it set at 1120 MHz, which was the advertised boost clock of the card I bought. (side note: the price has gone up a bit since I bought it ) Using GPUreview.com's page for the 7870 and adjusting for clock speed since the 270X is the exact same Pitcairn chip as the 7870, I get a total of 2.8672 teraFLOPs in theoretical performance. So with that clock speed, I'm getting a whole extra teraFLOP, more that 150% of the PS4's theoretical performance. No wonder I can add in higher quality depth of field, "HBAO Full", higher tessellation, etc., and still be playable at 1080p.

Just to throw out some more comparisons:

• The R9-270, the slowest full Pitcairn chip at a stock clock speed of "Up to 925 MHz", according to AMD's website, has a theoretical performance of 2.368 teraFLOPS.
• The cut down Pitcairn chip has 1.76128 teraFLOPS on the 7850, 1.8944 on the R7-265.
• GPUReview stopped being updated before any Bonaire card was released, so there's no easy way to calculate its theoretical performance (that I know of). AMD has provided the theoretical FLOPS performance for the 7790 on its website, though: 1.79 teraFLOPS. The 260X is slightly higher clocked so will have slightly higher performance, while the 260 is a cut down chip so will have considerably lower performance.
• The Xbox One's graphics component, which has the same amount of shaders as an R7-260, has a theoretical performance of 1.31 teraFLOPS.
• The R7-250x has the same clock speed and thus the same performance as the Radeon HD 7770, 1.280 teraFLOPS.

Note that Bonaire seems to punch close to the 7850/265. Keep in mind though that Bonaire is held back by a 128 bit wide memory bus, while all versions of Pitcairn have a full 256 bit wide bus, as does the PS4. So while Bonaire is close in theoretical performance to the PS4, you're likely to run into a memory bottleneck with Bonaire in scenarios that the PS4 would have the memory bandwidth to power through. For parity with the PS4, I would recommend equivalent memory bandwidth and a comfortable advantage in FLOPS, the R9-270 at minimum (a 7870 actually has significantly less memory bandwidth than the 270 and the PS4)

The Xbox One is a bit harder to compare because of its memory setup. Theoretically, its raw computational performance is comfortably behind Bonaire and just a bit above the 7770/250X's Cape Verde. However, it uses a combination of 8 GB of slow DDR3 RAM on a 256 bit wide bus combined with a 32 MB pool of super-fast ESRAM (by comparison, the PS4 uses a straightforward 8 GB pool of GDDR5 RAM). It's hard to say if it would have better or worse memory performance than a 2 GB Bonaire chip. Either way, you will want at least a 2 GB Bonaire card for parity with the Xbox One, as Cape Verde falls short in theoretical performance to begin with. Cape Verde also has half the geometry engines and rasterizers as the rest of the chips talked about here, which would lead to other practical performance deficits.

 

[Majcric]

I've played a bunch with my 270X and 1080p monitor and I think that, subjectively, the 270X is fine. Not 60 FPS, of course, but it's acceptable. The most trouble I get is in an area with a lot of NPCs, like Redcliffe. There's some stutters here and there, but it's not in combat, so I find that acceptable.

Same here. In the benchmark at 1080p everything on Ultra, no MSAA, I get 40min and 52 average. But as soon as I head to redcliffe I have seen dips in the 20's.

 

[Red Hawk]

Ha! The game just crashed and gave me a "Direct3D has no memory available" error message, or some such. When I closed the game out, Windows was giving me a "low on memory" message, too. Maybe I should have gotten a 3 GB card. Was in Crestwood, if that makes any difference.

 

[bystander36]

Ha! The game just crashed and gave me a "Direct3D has no memory available" error message, or some such. When I closed the game out, Windows was giving me a "low on memory" message, too. Maybe I should have gotten a 3 GB card. Was in Crestwood, if that makes any difference.

I think it is more likely due to a memory leak. I'm playing on high, with Ultra textures and medium tessellation. I have run for hours without a crash, but ever once in a while, in places I've spent plenty of time in, I'll just crash. It's just part of the game we have to live with for now. I'm using 2Gb cards.

 

[Red Hawk]

If it was a memory leak, then it doesn't seem to be a particularly bad one. I've been playing the heck out of this game, and that's the first time I've seen it crash. Overall the game seems very polished to me -- I haven't had blatant issues with bugs.

 

[ShintaiDK]

Ha! The game just crashed and gave me a "Direct3D has no memory available" error message, or some such. When I closed the game out, Windows was giving me a "low on memory" message, too. Maybe I should have gotten a 3 GB card. Was in Crestwood, if that makes any difference.

You ran out of system memory, not VRAM.

 

[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.

After 150 recorded hours on this game I can confirm that there was at least 300GB written to my SSD that hadn't been before. It's in perfect health still thankfully.

I've ruled out savegame writes as they are stored on a different disk in my documents. I experimented with having the game files on my HDD and deep linking the Data folder which contains 99% of the files onto my SSD and so far, 5 hours in and the data written to my SSD hasn't budged. The levels still load quickly and I don't have to worry about my SSD.

I Looked on the Denuvo website where it states:

"Does Denuvo Anti-Tamper affect my SSD or any other type of hard drives in any way?

No. As mentioned before, Denuvo Anti-Tamper does not constantly read or write any data to storage media."

But I can't see what else it could possibly be.
​

 

[Flapdrol1337]

But I can't see what else it could possibly be.
​

​

Probably the pagefile, especially with the mention of a memoryleak a few posts up.​

​

Having the game files installed on the ssd shouldn't matter, game files are only read during play, no writes.​

 

[Deders]

[/LEFT]

Probably the pagefile, especially with the mention of a memoryleak a few posts up.​

Having the game files installed on the ssd shouldn't matter, game files are only read during play, no writes.​

[/CENTER]

Except there is the controversy about the Denuvo anti tamper software which supposedly encrypts and decrypts the games .exe in the background and causes ssd wear.

I'll keep monitoring my ram usage but I've only seen it get close to 8GB on one occasion in this game. I have it monitored on my keyboard's screen so I can see it at all times.

If the writes don't increase on this next play-through then it can't be the swapfile.

 

[Worlocked]

So should I "upgrade" from my 2gb GTX 670 to a 4gb R9 270x? I seem to be terribly vram limited. Mantle is pretty appealing as I have only a stock clocked 2550k(case airflow is lacking). I'm also not sure if my 8gb system ram is cutting for the game either.

 

[Red Hawk]

So should I "upgrade" from my 2gb GTX 670 to a 4gb R9 270x? I seem to be terribly vram limited. Mantle is pretty appealing as I have only a stock clocked 2550k(case airflow is lacking). I'm also not sure if my 8gb system ram is cutting for the game either.

Going from a GTX 670 to a R9-270X isn't much of an upgrade. In a lot of games, you'll get worse performance. I wouldn't switch to AMD unless you're getting at least a 280/280X.

 

[toyota]

a 270x is a DOWNGRADE from a 670.

 

[Worlocked]

I know it'd be a slight downgrade in performance that's why I said "upgrade". I'm thinking more in terms of eliminating that vram bottleneck and being able to use mantle to improve DA:I cpu performance(I'm on win7 so the NVIDIA DX11.2 draw call improvements don't help me). I don't play many games anymore and the ones I've actually played recently seem to have favored AMD(DA:I and Alien Isolation).

The R9 280x is only 3gb, do you think that is enough for DA:I at ultra 1080P@30fps?

EDIT: If only they would push out some 4gb 285's it would make life so much easier... Where the heck are they??

 

[Deders]

I know it'd be a slight downgrade in performance that's why I said "upgrade". I'm thinking more in terms of eliminating that vram bottleneck and being able to use mantle to improve DA:I cpu performance(I'm on win7 so the NVIDIA DX11.2 draw call improvements don't help me). I don't play many games anymore and the ones I've actually played recently seem to have favored AMD(DA:I and Alien Isolation).

The R9 280x is only 3gb, do you think that is enough for DA:I at ultra 1080P@30fps?

I think the draw call improvments are universal, can you link to anything specific for 11.2?

The 280x should be fine and you'll have the advantage of Mantle. You may have to turn MSAA off but the shader based AA is pretty good and doesn't blur things. Some people turn the tessellation distance setting down for a bit of extra performance but I like it maxed.

 

[Red Hawk]

At 1080p I think you're going to run into bottlenecks besides memory usage with a 670 or 270X. I've been pretty satisfied with my 270X so far, with MSAA turned off of course. Have you been playing the game with your 670?

 

[bystander36]

You ran out of system memory, not VRAM.

Well, I have 2Gb less RAM than him, and in 130 hours, I've had about 5 crashes. I doubt it is System RAM, unless it is due to a memory leak/bug.