Video card tests at less than maximum graphics settings

[guachi]

Do these tests exist in wide numbers?

We see test at multiple resolutions. Why don't we see tests at different graphics settings widely done?

Say, 1080 gets max settings, 1440 gets one step below that, and 2160 one step further down.

As a buyer of a card I'd be really interested in seeing what setting a game needs to hit something playable.

I've found that *sync makes games vastly more playable at low frame rates, so I could accept playing a game if 99.5% of the frames were at or above, say, 35 FPS.

For example, the GTX 1080 (ironically named because why would you use it at 1080 gaming?) is the only card I see that can handle 3840x2160 gaming at max settings. It rarely tops 60 fps in a game (which is fine as the monitors don't go above 60 Hz) but stays at 40+ in basically everything.

What settings would other cards need to achieve that performance? Can they even do it?

 

[Bacon1]

Some sites do it. HardOCP does it a lot, but it seems random. They used to go for solid 60 avg FPS, but now their "target fps" is completely different per game, sometimes 30, sometimes 40, 50, etc.

Some sites run "Max" with AA and everything, while others will run "Mostly max" settings with maybe a few here or there that kill FPS turned off. Makes cross-comparison harder too.

 

[bystander36]

I'd definitely like it if settings were used to get 60+ FPS for the GPU being reviewed. Things can change a lot when playable settings are used.

 

[tential]

I'd definitely like it if settings were used to get 60+ FPS for the GPU being reviewed. Things can change a lot when playable settings are used.

I'd like sites to look for optimal settings and test that. Show image differences between maxed out settings and optimal settings and benches between those.

I think hardocp is a step in the right direction. Reviewers need to differentiate their testing more and have something that makes them unique.

The reason I look at hardocp is because it's very different than other sites. Also I think if you only have to drop one setting that's minimal between gpus with a 15-30% price difference why not? Hardocp can help with that decision.

More variety of testing would be amazing.

 

[3DVagabond]

I'd like sites to look for optimal settings and test that. Show image differences between maxed out settings and optimal settings and benches between those.

I think hardocp is a step in the right direction. Reviewers need to differentiate their testing more and have something that makes them unique.

The reason I look at hardocp is because it's very different than other sites. Also I think if you only have to drop one setting that's minimal between gpus with a 15-30% price difference why not? Hardocp can help with that decision.

More variety of testing would be amazing.

The only problem with that is it gets very subjective.

All you can really do is read a lot of reviews.

 

[bystander36]

I'd like sites to look for optimal settings and test that. Show image differences between maxed out settings and optimal settings and benches between those.

I think hardocp is a step in the right direction. Reviewers need to differentiate their testing more and have something that makes them unique.

The reason I look at hardocp is because it's very different than other sites. Also I think if you only have to drop one setting that's minimal between gpus with a 15-30% price difference why not? Hardocp can help with that decision.

More variety of testing would be amazing.

Of course, "optimal" settings can be subjective, but it would be nice to possibly see a range, like 40 FPS, 60 FPS and 80 FPS, or if it's too much work 60 FPS.

 

[bystander36]

The only problem with that is it gets very subjective.

All you can really do is read a lot of reviews.

The desired FPS may be subjective, but it's still better than "maxed" if you are looking at results that are unplayable.

 

[brandonmatic]

I'd like sites to look for optimal settings and test that. Show image differences between maxed out settings and optimal settings and benches between those.

I think hardocp is a step in the right direction. Reviewers need to differentiate their testing more and have something that makes them unique.

The reason I look at hardocp is because it's very different than other sites. Also I think if you only have to drop one setting that's minimal between gpus with a 15-30% price difference why not? Hardocp can help with that decision.

More variety of testing would be amazing.

This is a great idea. It would be a review site that applies its own subjective view of what graphics options are optimal from an IQ vs. performance basis and then tests using those settings. The most important part of this would be to explain why a certain setting was chosen and use screenshots to show the difference. I find this invaluable in the Nvidia tweak guides, which are awesome.

http://www.geforce.com/whats-new/gu...-hunt-graphics-performance-and-tweaking-guide

 

[nerp]

Why not just some different quality settings in addition to resolutions and provide screenshots so people can decide if the quality tradeoff is worth the FPS increase. It might also give insight into which games benefit from dialing back settings and which don't. But this could add days of time for someone to produce a review.

 

[guachi]

I stumbled across this at the pcworld review of the Nitro 480 4gb

http://www.pcworld.com/article/3098...iew-polaris-rethought-and-refined.html?page=4

Some games are tested at different graphics settings. Some aren't. RotTR is the first game that has that distinction so I've linked directly there.

The weak cards gain A LOT from lower settings.

In RotTR:

The 960 gains 43.2% or 11.9 FPS switching from very high to high at 1440.
The 380 gains 36.5% or 10.3 FPS (and is faster at very high vs. the 960 but slower at high)

At very high, neither are particularly playable as both are at ~28 FPS. At high, both could be played on a *sync monitor fairly easily as both average 39 FPS as both would be smooth(ish).

On the other hand:
The Nitro 480 gains 9.4% or 5.0 FPS
The 1060 gains 9.8% or 5.6 FPS

Gains are similar at 1080.

Far Cry: Primal @ 1080 Gains from switching to high from ultra.
960: +43%/17 FPS
380: +41%/18 FPS

1060: +31%/20 FPS
480: +44%/26 FPS

In this game, lowering graphics benefits EVERY card, not just the weakest.

Wouldn't you, the reader, find this useful?

Wouldn't you want to know the 480 can hit 57 FPS in Far Cry: Primal at high settings in 1440 and 60 FPS at ultra in 1080?

If you owned a 480 (or 1060. They perform basically the same in Far Cry) and a 1440 monitor, you might say "no, can't play Far Cry" if you only saw the 43 FPS at 1440/ultra. But you'd probably say "yes, can play" if you also say 57 FPS at 1440/high.

They test different resolutions, but it's far easier for me to change a graphic setting than it is to have different monitors lying about and switch based on the game I'm playing.

[EDIT: Anyone know if AMD/nVidia in their "reviewers guides" ask reviewers to review at max settings to make it look like buyers need to spend more money on a card to play a game?]

 

[tential]

Of course, "optimal" settings can be subjective, but it would be nice to possibly see a range, like 40 FPS, 60 FPS and 80 FPS, or if it's too much work 60 FPS.

That's true it will be subjective and there is nothing wrong with that. Another reviewer may conclude something different and through more interest in optimizing settings for game performance we will come to a group of subjective conclusions that people can try out and then tweak to their own liking.

User below also has some good points about the screenshots and the Nvidia tweak guide which I actually still use when I can't find anything else.

Neogaf is a good resource for this actually guys now that I think of it.
Check out their PC performance threads sometimes people do post pics of all the settings or a little guide if you don't want to look yourself.

This is a great idea. It would be a review site that applies its own subjective view of what graphics options are optimal from an IQ vs. performance basis and then tests using those settings. The most important part of this would be to explain why a certain setting was chosen and use screenshots to show the difference. I find this invaluable in the Nvidia tweak guides, which are awesome.

http://www.geforce.com/whats-new/gu...-hunt-graphics-performance-and-tweaking-guide

I would enjoy this a lot. Especially because it you then have extra gpu horsepower leftover you can know which settings give the best image to performance ratio or whether even to render at a higher resolution. Yes it's subjective but I definitely would like people looking into this and trying to optimize performance to image quality ratio rather than just ultra everything I don't care if this only changes 1 pixel on the screen but I lose 50 fps.

 

[JoeRambo]

User below also has some good points about the screenshots and the Nvidia tweak guide which I actually still use when I can't find anything else.

Nvidia tweak guides are GOLD mine. Just look at their Fallout 4 stuff, epic quality with amazing amount if screenshots. Heck they even have settings (like uGrids) listed for GTX 1380 TI cards

 

[mohit9206]

I remember when review sites used to actually follow this methodology of testing lower end cards at low or medium settings not that long ago.
For some reason they stopped doing it and went the highest details route.
Also how about using entry level gpu to test entry level cards. Everyone knows that AMD cards have cpu driver overhead in dx11 so testing cards like rx460 with $1000 cpu does not give the readers a clear picture of what the card will actually perform like in real life as nobody buying a card like rx460/470 will be running i7 cpu or higher. How about testing these lower end cards with i3 and i5?
But of course they won't do that citing "we make sure there are no bottlenecks present" or "we do this for the sake of consistency".
Yeah sure you'd rather lie with these performance figures that nobody will actually get in real life for the sake of consistency.

 

[SPBHM]

testing with lowered settings is dangerous, people might notice there is not a huge visual quality difference on most cases and that they don't really have to buy expensive graphics cards or upgrade all the time.

also it's easier to just select "ultra" and run a quick benchmark compared to testing with multiple settings and trying to customize settings for the best compromise.

but I can remember a few quickly, Digital Foundry at times uses custom settings on tests trying to match consoles or something else (like a specific FPS target) but they use "ultra" settings for typical reviews

pclab.pl tests games with more settings (but only use "ultra" for typical reviews)
like low and medium here:
http://pclab.pl/art70150-4.html
http://pclab.pl/art70140-5.html

 

[SolMiester]

I stumbled across this at the pcworld review of the Nitro 480 4gb

http://www.pcworld.com/article/3098...iew-polaris-rethought-and-refined.html?page=4

Some games are tested at different graphics settings. Some aren't. RotTR is the first game that has that distinction so I've linked directly there.

The weak cards gain A LOT from lower settings.

In RotTR:

The 960 gains 43.2% or 11.9 FPS switching from very high to high at 1440.
The 380 gains 36.5% or 10.3 FPS (and is faster at very high vs. the 960 but slower at high)

At very high, neither are particularly playable as both are at ~28 FPS. At high, both could be played on a *sync monitor fairly easily as both average 39 FPS as both would be smooth(ish).

On the other hand:
The Nitro 480 gains 9.4% or 5.0 FPS
The 1060 gains 9.8% or 5.6 FPS

Gains are similar at 1080.

Far Cry: Primal @ 1080 Gains from switching to high from ultra.
960: +43%/17 FPS
380: +41%/18 FPS

1060: +31%/20 FPS
480: +44%/26 FPS

In this game, lowering graphics benefits EVERY card, not just the weakest.

Wouldn't you, the reader, find this useful?

Wouldn't you want to know the 480 can hit 57 FPS in Far Cry: Primal at high settings in 1440 and 60 FPS at ultra in 1080?

If you owned a 480 (or 1060. They perform basically the same in Far Cry) and a 1440 monitor, you might say "no, can't play Far Cry" if you only saw the 43 FPS at 1440/ultra. But you'd probably say "yes, can play" if you also say 57 FPS at 1440/high.

They test different resolutions, but it's far easier for me to change a graphic setting than it is to have different monitors lying about and switch based on the game I'm playing.

[EDIT: Anyone know if AMD/nVidia in their "reviewers guides" ask reviewers to review at max settings to make it look like buyers need to spend more money on a card to play a game?]

Great post and on the money!:thumbsup: