GTAV performance thread.

[Phixsator]

I have never seen a game other than gta 5 hit the vram usage so hard when enabling MSAA.. 2x MSAA made my vram go from 2800ish to 3900ish, which is way above the available vram on my 780, and everything turned into a slide show of 5 fps for 10 secs and then a total crash when switching to 2x msaa and TXAA .

 

[CaptainRAVE]

http://www.tweaktown.com/tweakipedia/83/grand-theft-auto-benchmarked-1080p-1440p-4k/index.html

Very interesting benchmark, the 295x2 not looking that good against AMDs single card solutions and Nvidias cards. The minimums suffering the most. Looks to me like GTAV needs better crossfire drivers, just a shame it always takes so long for this to happen with AMD.

 

[savagesam117]

I'm surprised about the Titan X in Sli @ 4K only being a little bit quicker.

 

[Ayahuasca]

The game doesn't seem to reap much benefit from SLI or CF looking at that review.

 

[LtMatt]

The game doesn't seem to reap much benefit from SLI or CF looking at that review.

Crossfire benefits greatly.

Settings as per the screenshots. Tess = Very High, advanced graphics all off.

Spoiler

290X

Spoiler

2x290X

Spoiler

3x290X

Spoiler

4x290X

Spoiler

 

[Ayahuasca]

We could all post screenshots of specific scenarios with unified memory usage Matt

Clearly doesn't scale like that in all parts of the game or for every crossfire user.

At the very least you could run the benchmark with the same settings the review used to see if it holds any truth.

 

[LtMatt]

We could all post screenshots of specific scenarios with unified memory usage Matt

Clearly doesn't scale like that in all parts of the game or for every crossfire user.

At the very least you could run the benchmark with the same settings the review used to see if it holds any truth.

Why? The benchmark is not representative of actual gameplay though. That screenshot is what the game plays like. Lowest gpu usage will drop to is around 80%-85% across four gpu's and that's worst case scenario.

EIDT

Minimum fps barely drops below 45, never mind down to single figures. It's not an accurate representation of performance, that's the problem.

 

[shankly1985]

Its nice seeing the Frame rate increase you got going from single to 4 way..

 

[Ayahuasca]

Why? The benchmark is not representative of actual gameplay though. That screenshot is what the game plays like. Lowest gpu usage will drop to is around 80%-85% across four gpu's and that's worst case scenario.

EIDT

Minimum fps barely drops below 45, never mind down to single figures. It's not an accurate representation of performance, that's the problem.

Then take a screenshot with everything maxed bar AA like the review in a field of grass with a lot of distance in the view as well.

Performance in the city is fine.

 

[TheRealDeal]

Then take a screenshot with everything maxed bar AA like the review in a field of grass with a lot of distance in the view as well.

Performance in the city is fine.

Which would not prove anything. A video would be much better as we could see how variable the frame rate is from low to high.

 

[Ayahuasca]

Which would not prove anything. A video would be much better as we could see how variable the frame rate is from low to high.

It would show true minimums actually.

I agree though that a video would be better in showing variation.

 

[LtMatt]

Then take a screenshot with everything maxed bar AA like the review in a field of grass with a lot of distance in the view as well.

Performance in the city is fine.

Okay but performance is still fine, gpu usage did drop to the lowest I've ever seen it though. Thankfully i don't spend much time on top of the vinewood sign looking across the whole city.

Spoiler

Spoiler

Spoiler

Spoiler

Spoiler

Spoiler

 

[CaptainRAVE]

Why? The benchmark is not representative of actual gameplay though. That screenshot is what the game plays like. Lowest gpu usage will drop to is around 80%-85% across four gpu's and that's worst case scenario.

EIDT

Minimum fps barely drops below 45, never mind down to single figures. It's not an accurate representation of performance, that's the problem.

I often see drops into the 60 and 70%s on my 295x2. There is definitely room for improvement with the drivers. I would agree that my minimums are rarely, if ever, below 40fps, but then I also don't have everything maxed out like they did in those benchmarks. For me it is the random stutter that is more annoying, but perhaps Freesync will help?

 

[Ayahuasca]

That's better.

Was that with every setting maxed bar AA?

Vinewood sign isn't where I see the lowest framerate though to be fair, it's the fields in the areas further north where it really plummets.

 

[LtMatt]

I often see drops into the 60 and 70%s on my 295x2. There is definitely room for improvement with the drivers. I would agree that my minimums are rarely, if ever, below 40fps, but then I also don't have everything maxed out like they did in those benchmarks.

Just because gpu usage drops, does not mean the drivers need improvement. Gpu usage is tied to more than just how well a driver performs. Think API limitations.

That's better.

Was that with every setting maxed bar AA?

Vinewood sign isn't where I see the lowest framerate though to be fair, it's the fields in the areas further north where it really plummets.

Yep everything maxed apart from AA.

 

[africanos23]

AMDMatt, those are worst case scenarios btw with the myst in the sky. It totally tanks performance vs any other weather or day night cycle.

 

[LtMatt]

AMDMatt, those are worst case scenarios btw with the myst in the sky. It totally tanks performance vs any other weather or day night cycle.

Really? Didn't know that just went up there and the storm rolled in.

 

[Ayahuasca]

Just to clarify I'm not saying this is an AMD or nVidia issue, it's more down to the Ultra grass setting tanking performance whilst barely looking any different to very high.

The game is pretty well optimised and it looks like that reviewer was running with crossfire not even working.

 

[LtMatt]

Just to clarify I'm not saying this is an AMD or nVidia issue, it's more down to the Ultra grass setting tanking performance whilst barely looking any different to very high.

The game is pretty well optimised and it looks like that reviewer was running with crossfire not even working.

That i can agree on. Grass heavy areas send Quad 290X's up to continuous 100% gpu usage.

 

[GAM8LE]

Im finally settled now with my settings, to the point I don't need to have anything open to monitor temps/FPS. I just open MSI to set my fan profile and its good to go.

Im upgrading my cards soon though so the process will probably reset and I will start tinkering again haha