Mass Effect: Andromeda Benchmarks [PCGH]

[Bacon1]

http://www.pcgameshardware.de/Mass-...55712/Specials/Mass-Effect-Andromeda-1223325/

970, 1060 and 480 all running the exact same.

Looks like they went nuts with Tessellation, PCGH found that turning it to AMD Recommended bumped FPS 13-18% and turning it all the way off gave upto 26% more performance.

It can be worthwhile for Radeon owners to play around with the "Tessellation Mode" option because the effects on the frame rate are quite remarkable. In our test scene, the tessellation profile in the driver (setting "AMD optimized") leads to an image rate of 44 / 53.2 Fps (min./avg.) Instead of 40 / 47.2 - an increase of almost 13 percent. If the driver completely disables the Tessellation (setting "Off"), the picture rate even increases to 51 / 59.5 Fps, +12% compared to the standard driver and +26% to the Tessellation required by the game .

On a Radeon R9 390 (Asus Strix, 1.050 MHz Kerntakt) the Tessellationoptimierung in the driver has a slightly stronger effect, we measure in Full HD at our benchmark the following values: 34 / 41.6 Fps (min./avg.) Instead Of 30 / 35.2 Fps. This corresponds to a gain of 18 percent. If you are using a graphics card with Hawaii GPU (R9 290 [X], 390 [X]), you should definitely pay attention to the tessellation switch in the driver to get higher picture rates.

I don't think the changes are noticeable, anyone see anything major different?

http://www.pcgameshardware.de/scree...a-Tessellation-AUS-via-Radeon-Driver-pcgh.png

http://www.pcgameshardware.de/scree...meda-Tessellation-Default-Maximumpng-pcgh.png

http://www.pcgameshardware.de/scree...ation-Optimization-via-Radeon-Driver-pcgh.png

 

[Yakk]

It's really negligible, or sometimes even sub-pixel difference... or COMPLETELY game changing, makes my eyes bleed without 64x tessellation MINIMUM... depending on who you ask.

I find even AMD's optimized tessellation settings are generally too high, 16x or sometimes even 8x are usually more than enough.

Being an nvidia title it's standard they default to maximum tessellation.

 

[Carfax83]

Looks like they went nuts with Tessellation, PCGH found that turning it to AMD Recommended bumped FPS 13-18% and turning it all the way off gave upto 26% more performance.

From the sample images they uploaded, "nuts with tessellation" is clearly overstating things a bit. I can't even see any tessellation in those pics to be honest, and tessellation is usually very easy to spot.

On another note, looks like my GTX 1080 should be able to deliver 60 FPS at 1440p using almost maxed out quality. I'll be selling it soon though as soon as my aftermarket GTX 1080 Ti of choice becomes available.

 

[Jaydip]

For a frostbite engine game it looks extremely demanding.

 

[Carfax83]

GameGPU just uploaded their results:

Doesn't seem to be CPU heavy at all, or maybe it's just the area where they benchmarked. But it definitely is heavy on the RAM. I can see why they recommended 16GB:

 

[Mercennarius]

Seems like with the latest drivers the RX 480 has finally passed up Hawaii at least in the newest games. For a while the 480 routinely fell behind the 390(X) in benchmarks but in the last few games coming out the 480 has matched or beaten Hawaii.

 

[Bacon1]

From the sample images they uploaded, "nuts with tessellation" is clearly overstating things a bit. I can't even see any tessellation in those pics to be honest, and tessellation is usually very easy to spot.

Nothing changed IQ wise yet performance went up by 26% by turning off Tessellation. So something is being way over tessellated and killing performance. I dunno how they let things like that through testing :\

 

[Carfax83]

For a frostbite engine game it looks extremely demanding.

Frostbite can be a pretty demanding engine depending on how it is implemented. The reason why most people consider it to be a more performant engine is probably because it's been used by DICE for multiplayer games where performance takes precedence over image quality.

From the PCgameshardware.de article, they state that ME.A using very high quality DoF, global illumination, volumetric lighting, and very high resolution screen space reflections that are far superior to what was found in Dragon Age Inquisition. Looking at the trailers I could also see a lot of dynamic lighting, which is usually turned off in Frostbite 3 MP games.

Plus the game is using a more up to date and powerful streaming system, which explains the much greater RAM usage compared to previous Frostbite 3 titles.

 

[tamz_msc]

Back in the day when DX11 was first launched Stalker: Call of Pripyat was one one the first games supporting it and showed real differences when tessellation was enabled. The way it is done now makes me sad.

This is another example of tessellation being overdone simply for the sake of it. Though I have seen worse - Crysis 2 rendered tessellated water hidden underneath solid ground.

 

[Carfax83]

Crysis 2 rendered tessellated water hidden underneath solid ground.

That was debunked a long time ago. The tessellated water was culled so it wasn't being rendered at all. That said, I think we should wait for more examples of the game's use of tessellation to determine whether it's overdone or worse, not effective.

 

[Carfax83]

Also this is the first Frostbite 3 title I've ever seen with a dynamic skybox! Every single Frostbite 3 game I've ever seen before this one had a static skybox where the clouds didn't move at all.

 

[tamz_msc]

That was debunked a long time ago. The tessellated water was culled so it wasn't being rendered at all. That said, I think we should wait for more examples of the game's use of tessellation to determine whether it's overdone or worse, not effective.

Perhaps it was, but it was utterly wasteful. Turning it off gave you much better performance and you had to go look for the objects where the extra geometry detail was noticeable.

 

[Sweepr]

GameGPU just uploaded their results:

GameGPU just uploaded their results:

Doesn't seem to be CPU heavy at all, or maybe it's just the area where they benchmarked. But it definitely is heavy on the RAM. I can see why they recommended 16GB:

So GTX 1060 6GB ~15-20% faster than RX 480 8GB at 1080P. GTX 980 Ti and GTX 1070 beating Fury X by ~15-22% at 1440P, likely extending the lead when OCed. Doesn't look like running exactly the same like the OP says.

 

[tamz_msc]

So GTX 1060 6GB ~15-20% faster than RX 480 8GB at 1080P. GTX 980 Ti and GTX 1070 beating Fury X by ~15-22% at 1440P, likely extending the lead when OCed. Doesn't look like running exactly the same like the OP says.

You sire, are indulging in a bad practice - favoritism in benchmarks.

 

[Carfax83]

So GTX 1060 6GB ~15-20% faster than RX 480 8GB at 1080P. GTX 980 Ti and GTX 1070 beating Fury X by ~15-22% at 1440P, likely extending the lead when OCed. Doesn't look like running exactly the same like the OP says.

Good catch. I just noticed that the GTX 1060 in the PCGH.de preview was the 3GB model and not the 6GB model. So the 6GB model should be significantly faster than the RX480. Games with lots of geometry typically run faster on NVidia hardware I suppose, but we'll have to wait for patches and driver updates to really come to the final conclusion, and that will take around 6 months or so.

 

[Sweepr]

The GTX 1060 tested by PCGH is a 3GB model (no mention of this in the OP). And indeed this model is equivalent to RX 480 8GB @ 1080P according to both GameGPU and PCGH.

 

[tamz_msc]

The 1060 6GB is 2-3 FPS faster than the RX 480 8GB OC vs OC.

Hardly "significant".

 

[Carfax83]

The 1060 6GB is 2-3 FPS faster than the RX 480 8GB OC vs OC.

Hardly "significant".

He's not using the Mass Effect Andromeda optimized drivers from NVidia though. They should be released today from what I hear, but some reviewers already have them, like PCGH.de.

 

[Sweepr]

He's not using the Mass Effect Andromeda optimized drivers from NVidia though. They should be released today from what I hear, but some reviewers already have them, like PCGH.de.

PCGH's Palit GTX 980 Ti Super Jetstream is anywhere from 20-50% faster than Fury X. Not sure about the driver situation on the AMD side.

 

[Bacon1]

So GTX 1060 6GB ~15-20% faster than RX 480 8GB at 1080P

Reference 480 that throttles at stock settings?

Welp disable that overdone tessellation and right back to tied... amazing :O

 

[tamz_msc]

He's not using the Mass Effect Andromeda optimized drivers from NVidia though. They should be released today from what I hear, but some reviewers already have them, like PCGH.de.

I doubt how much further they can optimize a DX11 title, going by how the 378.78 version turned out.

 

[Headfoot]

Normally Frostbite games are one of the few games I can count on having CF on day one out of the box. Looks like even mighty Frostbite is slipping into the CF/SLI graveyard

 

[Krteq]

...I can't even see any tessellation in those pics to be honest, and tessellation is usually very easy to spot...

Can you please enlighten us? How can you "see/spot tessellation" without wireframe?

 

[pj-]

Can you please enlighten us? How can you "see/spot tessellation" without wireframe?

If you can't see the effects of tessellation without wireframe then what is the point of it?

 

[Yakk]

If you can't see the effects of tessellation without wireframe then what is the point of it?

LOL... Good point, tessellation effects are usually pretty subtle. Things like ears, water, and rocks can be a little rounder.

It can be hard to see sometimes also because the texture maps usually are made to fool the eye into seeing something which it a lot rounder than what the wireframe really is. So applying tessellation to an item which already has a rounded effect texture map may not actually show any difference outside of a wireframe, yet can consume a significant amount gpu resources.