Grand Theft Auto V Multi-GPU Performance Review Part 2 @ [H]

Quartz-1 said:
Can I pick you up on two points?

Firstly, please make the colours in the graphs consistent. Pick one colour for each card / set of cards and stick with it throughout.

Secondly, on the conclusion page you state that three Titan X cards will be required to max out the game. Yet I couldn't find any evidence in the article to back that up. If I haven't missed the absence, you should either remove that statement or provide evidence to back it up.
Click to expand...

Two TITAN X's cannot "max out" the game at 4K according to our testing shown. We are two settings away from maximum. Therefore, adding another card should allow us to raise that setting up two notches, maxing the game out at 4K.
 
Flip221 said:
Wait - Which AMD card was used? Was is R290 or R290x? Shouldn't the R290x have been used if comparing to the GTX980?
Click to expand...

AMD Radeon R9 290 CrossFire (two R9 290 cards)

AMD Radeon R9 295X2 (a single card that contains two R9 290X GPUs)

Those were the cards used.
 
Quartz-1 said:
Just to pick up on FXAA, does it really produce more of a quality improvement than other features with a similar performance hit? I'm particularly thinking of higher-dpi monitors.
Click to expand...

Technically it reduces aliasing on more things in the game than MSAA is capable of, this is just fact. It is the nature of how it works.

We will provide screenshot comparisons in Part 3.
 
Thanks Brent for spending time replying to us.

Me personally, i am going to have another look at FXAA instead of MSAA X2, going by what you have said in here, i may have dismissed it too early!

Thanks Again.
 
Bankie said:
It would be nice if they could include one older card (say a 660/70 or 7970) in the reviews just to see what an upgrade to the latest card will get you. I currently have a 670 and while I can ballpark it (it should be around 760/770/960 performance most of the time) it doesn't always work out that way.

I know this has been asked multiple times over the years and Kyle or Brent say that it just takes too much time that would be better spent on the latest hardware. But maybe they could include just one card from a couple of generations ago as a control.
Click to expand...

Yes - I often wonder how my 680s in SLI are holding up to the newer generation cards but it is hard to find comparisons. At some point i am going to have to bite the bullet and upgrade.
 
phorkz said:
Thanks Brent for spending time replying to us.

Me personally, i am going to have another look at FXAA instead of MSAA X2, going by what you have said in here, i may have dismissed it too early!

Thanks Again.
Click to expand...

I find MSAA x2 with NVidia MFAA (nv control panel) enabled along with TXAA = best image quality at decent performance. FXAA does fine for still images but as soon as you start moving = aliasing everywhere.
 
pandora's box said:
I find MSAA x2 with NVidia MFAA (nv control panel) enabled along with TXAA = best image quality at decent performance. FXAA does fine for still images but as soon as you start moving = aliasing everywhere.
Click to expand...

Yet to try TXAA, might give that a go too.
 
Brent_Justice said:
You are welcome to disagree.

Factually though, FXAA reduces aliasing on more things than MSAA does. FXAA reduces aliasing on polygon edges (objects) like MSAA, but also extends that to textures and alpha textures (trees, foliage, vegetation, fences, grass) any form of texture. FXAA is able to work inside polygons, where MSAA cannot. FXAA is a full-scene method. In a game like this, where view distance is important, FXAA does a better job "in the distance."

Technically therefore, FXAA reduces aliasing on more than MSAA does.

We will of course be looking at AA performance and IQ in Part 3.
Click to expand...

The question is whether you truly represented "highest playable settings" by eschewing MSAA in favour of a setting like "Grass Quality". My experience after multiple playthroughs is that you absolutely did not. MSAA has a much larger impact on the game's visuals than ultra Grass Quality, which only affects scenes outside the city in any appreciable fashion. (It would appear a lot of people agree that your choice of AA is insufficient.) Furthermore, the AA options are NOT exclusive, so there's no reason you couldn't have MSAA(4x TXAA) and FXAA enabled at the same time (as I do) to improve visuals even further. "Factually though" there is no denying that TXAA+FXAA would produce less aliasing than FXAA alone.

However, it appears the subjective answer is what we're discussing here, and the subjective answer is far more fuzzy (no pun intended).

Thanks for the response.
 
pandora's box said:
I find MSAA x2 with NVidia MFAA (nv control panel) enabled along with TXAA = best image quality at decent performance. FXAA does fine for still images but as soon as you start moving = aliasing everywhere.
Click to expand...

Hopefully they'll get MFAA working with SLI, and DSR working with SLI+GSync so I can try some of those other options.

I completely agree that FXAA is insufficient in this game as aliasing in motion is extremely pronounced. FXAA should be used in addition to other AA methods, not as the ONLY AA method.
 
I am uploading a video of my 2 290x and 1 290 crossfire running 4k max settings with aa turned off & over 12gbs of ram usage.. No stutter and almost consistent 60fps.. for 800$..
 
Prettiest well defined settings in my view- resolution agnostic.
turn max everything- 8x MSAA and 8x reflections. shadows- softest (not NVidia/AMD)
Disable depth of field and motion blur at minimum.

I recently moved to the Titan X. I play 1280x720 @ 4.4GB VRAM and 25FPS rain + vegetation.

Use the centered timing within NVIDIA control panel and it is like playing (visuals and fps) on a super futuristic tablet sized screen from 5yrs in the future and a ppi density from 5 years in the past (given the crazy flagship smartphones ppi densities today). I say- simply brilliant;) the diagonal is just over 15" on my FHD screen and it looks like nothing Rockstar have done in the past!
 
Brent_Justice said:
We are two settings away from maximum. Therefore, adding another card should allow us to raise that setting up two notches, maxing the game out at 4K.
Click to expand...

Well yes, adding a third TX should allow you to run at max settings, but does it? If you can't demonstrate that then I think you shouldn't make the claim.
 
ApeXiTT said:
I am uploading a video of my 2 290x and 1 290 crossfire running 4k max settings with aa turned off & over 12gbs of ram usage.. No stutter and almost consistent 60fps.. for 800$..
Click to expand...

No doubt this can be done in certain sections of the map. If you read Part 1 of our reviews on GTA V, you will see that we have found a worst case scenario in terms of framerate. You might very well be comparing apples to oranges.
 
Quartz-1 said:
Well yes, adding a third TX should allow you to run at max settings, but does it? If you can't demonstrate that then I think you shouldn't make the claim.
Click to expand...


This is not a "claim." He was asked his thoughts and opinions and he shared those....just like you are doing.
 
I may be missing it, but was there a description of the way the game was tested? I can't find a description in either the preview part 1 or this part 2.

Well anyway, using the settings described in this review for the TITAN X at 1440p I was seeing framerates in the 80-110 FPS range during a normal playthrough with a stock i7-4770. I did not get to where the grass gets heavier in the outer and mountainous regions of the game, this was doing story missions in the city proper. I don't have the exact numbers on me right now, but I remember the benchmark showed a minimum framerate of 47, maximum of 96, and average of 61 on my system.

I think there is enough of a budget there left to add full-scene antialiasing. 4xMSAA+MFAA+FXAA brought the framerate down to around 70'ish during the daytime. That may change later in the game, but I think frame times in this game are good enough to play in the 60 FPS range, while I prefer to keep it around 80 (100, optimally).

One interesting thing this game has exposed: I need to repaste my CPU and HSF! This game makes your CPU run hot as it pegs usage of every core even after the patch. In my case it went over 90C, with Core 0 and 2 hitting 94C max. I may have to delid it if a HSF reseat doesn't help...
 
"We started with a preview of the game's performance and image quality settings when the game was released. We recommend starting here if you haven't already, to learn what makes this game unique in terms of features on the PC."

Full explanation as linked in Part 2.
 
Great review as always, [H]. Thanks for spending the time to do this for the enthusiast community.

I'm a bit put off by the brand agnostic crappy multi-GPU scaling. Hope that Rockstar and the GPU driver teams get their shit together and quickly addresses it. Would be nice to see 80%+ scaling for all setups. I feel bad for Titan X owners that dropped $2000+ on their multi-GPU setups and get the worst scaling out of all configurations tested.
 
Brent_Justice said:
Will I test MSAA performance and image quality? Heck yeah! Look for it in part 3.
Click to expand...

I look forward to it. I tried running 2x MFAA (forcing it in Nvidia control panel) with FXAA on and was fine in the city, running 45-60+fps in 1440p on a GTX 980 with all other settings maxed except grass very high and only high resolution shadows.

But I just got to the outer limits with Trevor and am struggling to keep fps over 40, now average 35fps, which sucks. I'm scaling back AA to just FXAA for now
 
"To find the lowest framerates we actually had to drive out of the city, around the coast, and through the mountains filled with vegetation, grass, and other towns along the way. We experienced framerate drops down to 40 FPS in these areas at the same settings that we were getting 80 FPS in the city downtown and 100+ FPS indoors."

Hence my previous post. Doesn't this skew the testing too much towards the various architectures/cards ability to render grass in GTA V? Not only is ultra grass setting a performance killer will relatively small visual gains, you've gone and found some extra grassy areas to test in?

I've been reading you guys for way long enough to know you're not biased or stupid so these are just thoughts and nothing more. I understand time is very finite so you have to make a stand somewhere so to speak. Thanks.
 
They're trying to create a test playthrough of the game that will represent the entire experience. If they just ran around the city it's my opinion that the [H] wouldn't have been doing their job. The fact is grass is the biggest performance hit in this game. I'm sure in the part 3 features breakdown that they will show how you can turn the grass setting down without losing much, if anything, in the visual experience.
 
Last edited:
AllBlackFan said:
"To find the lowest framerates we actually had to drive out of the city, around the coast, and through the mountains filled with vegetation, grass, and other towns along the way. We experienced framerate drops down to 40 FPS in these areas at the same settings that we were getting 80 FPS in the city downtown and 100+ FPS indoors."

Hence my previous post. Doesn't this skew the testing too much towards the various architectures/cards ability to render grass in GTA V? Not only is ultra grass setting a performance killer will relatively small visual gains, you've gone and found some extra grassy areas to test in?

I've been reading you guys for way long enough to know you're not biased or stupid so these are just thoughts and nothing more. I understand time is very finite so you have to make a stand somewhere so to speak. Thanks.
Click to expand...

If you want to have to stop your game every time you get out into one of these areas to reset your graphics settings to get playable framerates, by all means do that. However we are not going to do it that way.

That said, I think we were VERY CLEAR on how we were doing things...if you read. You may not agree with that. As always we are transparent on all of this and if you disagree with it, we accept that. We give you a forum here to tell us we are full of shit.

Grand Theft Auto V is a game of epic scale. This game has a massive open world, and also includes indoor as well. This game can be played in first person mode or third person mode. Considering the scope of this game, we have to test it in a way that takes into account the worst performance you'll experience in the game. By testing in scenarios that have the lowest performance, and determining what is playable from those parts of the game then everything else in the game will most definitely be playable, as the rest of the game renders with much higher performance.

GTA V has a diverse world scope that makes performance vary by drastic amounts on the same video card. Let's take a GeForce GTX 980 as an example. With a GeForce GTX 980 installed performance can vary in-game (at the same settings) from 40 FPS to 140 FPS. Yep, there is that much variance in framerate in this game depending on where you are and what you are doing. If we were to test only in the high-FPS areas, then when you hit the low-FPS areas the game would be unplayable.

Therefore, we had to find out first where the low-FPS areas are compared to the high-FPS areas and test in those areas for our actual run-through of data and base the highest playable settings off of that. In that way, you are assured the game is playable from the lowest framerates, to the highest framerates. Hopefully a "set it and forget it" approach, as no one wants to stop their gaming to figure out what is going on with frame rates that all of a sudden break the immersion.
Click to expand...
 
The 8gb 290x at $379
http://www.newegg.com/Product/Product.aspx?Item=N82E16814127866

Not much more than a 970 and definitely less than a 980. Go tri with 8gb 290x would be very interesting at 4K, much cheaper than 2 Titan X and even cheaper than two 980's.

That would be an interesting test plus could be the ticket at highest settings at 4K.

Still the bottom line is actually how it feels and play and not so much the numbers in the end.
 
Now I hope HardOCP gets two 390x (what ever the high next generation Radeons will be called) to do this over again with this game in the near future (like June). This game really does push gaming forward.
 
AllBlackFan said:
Why would I need to stop and reset my settings? The areas that performance isn't slowed by the ultra grass setting don't have much, if any, grass in them so why would I want to turn it back up to ultra then anyway?
Click to expand...

Roger that. I am sorry you do not like the way we test. There are plenty of other sites that will likely have the GTA V testing you are looking for.
 
Hi I'm rocking 4770k at 4.7, 16gb and 295x2 in crossfire (4GPU) at 4k and it stays at 60 fps almost always. It's honestly the most sublime and believable experience I've had on PC. I get aliasing and slow down when racing fast probs cos it's on WD 1TB. It's really really beautiful and all sliders are max (view etc ) but no MSAA. That is max grass etc all v high or ultra but I suspect it would do better on 850pro in raid. I'm using 14500 of vram of a possible 16gb. Today I saw a highly polished Audi in the street and it reminded me of GTA5 . I also know I'm CPU bound and VRAM bound. I can't run MSAA because memory limitations.
 
Excellent reviews as always.

One question on the scaling part with the 295x2 and the 290x for the single gpu comparison.

Wouldn't the single gpu 290x perform a bit better than each of the 2 gpus' crammed on the 295x2? Is there some way to temporarily disable 1 gpu on a 295x, to see how that scales?

It might bear out that the scaling is actually a bit better (technology wise).

Of course the 295x2 is what it is, and the performance being slightly hindered by being 2 gpu's crammed on to one card is what you get when you use one.

I suppose the real question I am asking, would be to compare scaling using 2 290x in crossfire compared to a single 290x.

(Just curious)
 
Daroller said:
I disagree with the [H] conclusion that FXAA is sufficient in this game. There is far too much aliasing in this game for me to consider it playable at less than 4xMSAA. I've been bouncing back and forth between 8xMSAA and 4xTXAA. They both have their uses, and the performance hit is roughly identical. MSAA is slightly slower, but not noticeable in gameplay.
Click to expand...

I agree. MSAA makes a huge difference in especially hair and beards. They look awful with FXAA or SMAA (via ReShade+SweetFx). Personally I find that 4xTXAA + SweetFx Lumasharpen filter gives a great combination of good antialiasing and still a sharp image.

This is probably the first game where I've been significantly bothered by aliasing. Usually SMAA is good enough for me. I'd love to try DSR as an alternative but unfortunately Nvidia just doesn't allow DSR on SLI if a G-Sync display is connected (no matter if G-Sync is enabled or not).

I run the game with an overclocked 970 SLI setup at near max settings (advanced distance scaling is slightly under half and of course not using max AA) at 1440p quite happily. I don't know how much fps I'm pulling and it definitely drops in the more forested areas thanks to ultra grass but in general it runs well enough to not bother me.

It would be nice to see how the game performs with overclocked cards. For example 970 SLI and 980 SLI seemed to be neck and neck at stock settings but it would be cool to see how the 980 SLI pulls away when overclocked and how AA settings affect it.
 
kasakka said:
I agree. MSAA makes a huge difference in especially hair and beards. They look awful with FXAA or SMAA (via ReShade+SweetFx). Personally I find that 4xTXAA + SweetFx Lumasharpen filter gives a great combination of good antialiasing and still a sharp image.

This is probably the first game where I've been significantly bothered by aliasing. Usually SMAA is good enough for me. I'd love to try DSR as an alternative but unfortunately Nvidia just doesn't allow DSR on SLI if a G-Sync display is connected (no matter if G-Sync is enabled or not).

I run the game with an overclocked 970 SLI setup at near max settings (advanced distance scaling is slightly under half and of course not using max AA) at 1440p quite happily. I don't know how much fps I'm pulling and it definitely drops in the more forested areas thanks to ultra grass but in general it runs well enough to not bother me.

It would be nice to see how the game performs with overclocked cards. For example 970 SLI and 980 SLI seemed to be neck and neck at stock settings but it would be cool to see how the 980 SLI pulls away when overclocked and how AA settings affect it.
Click to expand...

Seems some others users have been able to get that combination working

Http://forums.guru3d.com/showthread.php?t=397491


http://forums.guru3d.com/showthread.php?t=398128
 
You must log in or register to reply here.
Back
Top