Gears of War 5 Benchmark thread - First true NAVI optimised game?

[bigmike20vt]

Computer Base had done benchmarked Gears 5, if it is first true Navi optimised game then it is very poorly optimised for AMD cards as Nvidia Turing RTX 2070 SUPER is faster than Radeon VII and RX 5700 XT.

https://www.computerbase.de/2019-09/gears-5-benchmark-test/

Spoiler

Excuse my ignorance here however isn't that to be expected? I thought the 5700xt came out to best the 2070 which is exactly what it did. I thought that was kind of why NV bought out the super series as a response to that... Or are you saying the 5700xt is meant to be AMD answer to the 2080?

 

[VincentHanna]

Sort it out.

I might have a fiddle.

I don't often use automatic resolution scaling but I imagine the pixel drop will be so minuscule to get from 557-58 to 60 that I wouldn't even notice it.

Or somebody will release a video saying "The difference between 'High' and 'Ultra' setting is a 10% drop in frame rates but you won't notice the visual difference"

 

[HRL]

Probably get away with a single setting change if you pick the right one.

I’ve still got to play through GOW4 so I won’t be picking this up for a while yet.

 

[opethdisciple]

Excuse my ignorance here however isn't that to be expected? I thought the 5700xt came out to best the 2070 which is exactly what it did. Ok thought that was kind of why NV bought out the super series as a response to that... Or are you saying the 5700xt is meant to be AMD answer to the 2080?

No. I think it's just an example of a game that seems to heavily favour AMD's architecture over Nvidia's. In other words it's over performing in this one game.

 

[TheRealDeal]

No. I think it's just an example of a game that seems to heavily favour AMD's architecture over Nvidia's. In other words it's over performing in this one game.

No he is right. The 5700xt was brought out to beat the 2070 so Nvidia trumped it with the super. Overall the 5700xt beats a 2070.

 

[tyler_jrb]

You can also see the benchmarks on the oc3d website.

https://www.overclock3d.net/reviews/software/gears_5_pc_performance_review_optimisation_guide/1

Really looking forward to gears 5.

 

[shankly1985]

You can also see the benchmarks on the oc3d website.

https://www.overclock3d.net/reviews/software/gears_5_pc_performance_review_optimisation_guide/1

Really looking forward to gears 5.

Part of the concussion
"When it comes to the AMD VS Nvidia debate, Gears 5 offers no clear answers. Our ROG Strix GTX 1060 handily beats AMD's Radeon RX 580 while AMD's Radeon RX Vega 56 is on par with Nvidia's RTX 2060. Moving on to Navi and we can see AMD take a huge lead over its Nvidia rivals. Powercolor's RX 5700 XT Red Devil gave Nvidia's RTX 2070 Super a run for its money. AMD's Radeon RX 5700 (non-XT) Red Devil also bests Nvidia's standard RTX 2060 with ease."

I think the issue we have going on here with results is how the game is being tested.
Some will use the benchmark, some will use parts of the game.

Example the benchmark tool might not even show real world number, while the start of the game might show nvidia coming away strong while later in the game might have parts showing Amd doing really well.

So far from what I seeing Navi is doing well. So I will stand by what I said in the title that this game is Amd navi first true optimised game.

Next up is borderlands 3

 

[bru]

If a game has a built in benchmark, why on earth would you not use it?
If a reviewer wants to use his own section of the game as well then fine but for that sake of apple to apples, not using the built in benchmark just seems stupid.

 

[shankly1985]

If a game has a built in benchmark, why on earth would you not use it?
If a reviewer wants to use his own section of the game as well then fine but for that sake of apple to apples, not using the built in benchmark just seems stupid.

It does and to some part I agree with you. But at the same time they is countless games out there that have benchmarks that do not show the game at its real performance.
Metro Exodus is a prime example.

 

[greatash]

can never trust youtuber benchmarks, usually rushed and done in non optimal conditions/settings. basically just a guy in his bedroom

 

[imginy]

The benchmark is no where near real game numbers.

For me 60fps in bench is 85fps in game.

 

[Poneros]

If a game has a built in benchmark, why on earth would you not use it?
If a reviewer wants to use his own section of the game as well then fine but for that sake of apple to apples, not using the built in benchmark just seems stupid.

Because they're not representative. Learned this myself the hard way meticulously benchmarking every setting in DXMD through the in-game benchmark and then later testing in real gameplay and finding out I wasted a whole day because **** didn't work the same. The simple fact is, there's never gonna be apples to apples. Games are too complex and scene to scene will give different results between GPU families. Ideally you'll have benchmark runs for each zone, but who's gonna bother? Computerbase comes closest but the rest, meh.

 

[Besty]

I ran the 3 cards (1080ti HOF, 5700XT AE, 2070S FE) together in gears just now and used OCAT in-game to do the capture.

https://forums.overclockers.co.uk/posts/32903367/

I will add the built-in bench results later.

 

[Gregster]

If a game has a built in benchmark, why on earth would you not use it?
If a reviewer wants to use his own section of the game as well then fine but for that sake of apple to apples, not using the built in benchmark just seems stupid.

Agreed. Whilst I don't give a hoot who is faster, the standardised way of testing would be to use the inbuilt bench, regardless of 'actual game performance'. If part of the game is tough on NVidia, then it will be tough on AMD also.

 

[bru]

Because they're not representative. Learned this myself the hard way meticulously benchmarking every setting in DXMD through the in-game benchmark and then later testing in real gameplay and finding out I wasted a whole day because **** didn't work the same. The simple fact is, there's never gonna be apples to apples. Games are too complex and scene to scene will give different results between GPU families. Ideally you'll have benchmark runs for each zone, but who's gonna bother? Computerbase comes closest but the rest, meh.

Fine if you think they are not representative then include a section of the game that is tougher 'AS WELL' if you don't include the built in benchmark than people will question the validity of you work, just as has happen to Jokers video, because he didn't include the built in benchmark.

 

[shankly1985]

Agreed. Whilst I don't give a hoot who is faster, the standardised way of testing would be to use the inbuilt bench, regardless of 'actual game performance'. If part of the game is tough on NVidia, then it will be tough on AMD also.

Again I not saying this is wrong built in benchmarks are good but they definitely shouldn't be used as fact for a games performance.
At the end of the day you don't play the benchmark.

Think I can count on a single hand what games have decent benchmarks that show close to the games actual performance.
Forza horizons 4 and far cry 5 both spring to mind.

 

[Gregster]

Again I not saying this is wrong built in benchmarks are good but they definitely shouldn't be used as fact for a games performance.
At the end of the day you don't play the benchmark.

Think I can count on a single hand what games have decent benchmarks that show close to the games actual performance.
Forza horizons 4 and far cry 5 both spring to mind.

I don't think anyone is using it as an indication of game performance? I see an inbuilt bench as a way to determine how card x, y, z will roughly cope in said game and determines which card is the fster overall.

 

[Besty]

I don't think anyone is using it as an indication of game performance? I see an inbuilt bench as a way to determine how card x, y, z will roughly cope in said game and determines which card is the fster overall.

@shankly1985 hey guys, yeah generally the problem with built in benchmarks is that manufacturers can and do optimise the benchmark in their drivers and the in-game perf is different.

in this particular game the in game perf and the benchmark are roughly similar at the moment with a 9fps difference at 1440p between the 2070S fe and the 5700XT AE with is roughly 10%. i think from memory, that the 5700XT driver does not currently support the tile based rendering option in gears 5, i will retest once AMD do a driver revision on the game.

 

[Grim5]

can never trust youtuber benchmarks, usually rushed and done in non optimal conditions/settings. basically just a guy in his bedroom

Most youtube benchmarks are fake anyway (the small nobody channels) - if they don't show you a full run with rivatuner on and just show you graphs, it's usually fake or just made up,

 

[Grim5]

You can also see the benchmarks on the oc3d website.

https://www.overclock3d.net/reviews/software/gears_5_pc_performance_review_optimisation_guide/1

Really looking forward to gears 5.

Cheers. Looking at the OC3D graphs and also what users here have posted I can only come to the conclusion that this game loves overclocks.
I'm getting around 20% more frames than OC3D's stock 2080ti fe and the same compared to other users here who have posted stock 2080ti numbers - my card isn't on water it's just manually overclocked with it's stock cooler.

So yeah if you need more frames, overclock your GPU or at least just raise the powerlimit, this game loves overclocks.

And despite the xbox one x claiming 4k 60fps, it can do that but it also can drop resolution to 1600p to maintain that 60fps.

And while sort of similar, the xbox one x is not running the same settings.

Here is a GTX1070 doing 1440p Ultra on the left vs Xbox One X on the right.
Both are running 60fps

The Xbox is clearer because of the higher resolution, but it's missing a lot of image depth and detail due to missing features - like shadows.