First Fallout 4 performance results

[Agosto]

285 beating the 960, AMDs results hardly suck.

Beating? I wouldn't say this, since minimum framerates are lower - and minimums are important

I can only see 390X losing to a 970...

Hoping for a driver release

[Sakkura]

Beating? I wouldn't say this, since minimum framerates are lower - and minimums are important

I can only see 390X losing to a 970...

Hoping for a driver release

 

1st percentile performance is higher too. Only 0.1 is a little lower, and that's subject to much more statistical uncertainty. Overall, the 285 outperforms the 960.

 

Still, AMD can probably improve things with a driver update.

[Humbug]

For GPU performance I will have to wait until AMD's driver to get an idea...

But the CPU graph is fascinating. Never seen such a gap between i5 and i7!

[Misanthrope]

Fallout 4 seems to favor moar corez also. Good showing for the FX-8350 here. Even Faildozer is beating the i5-2500k here.

 

 

Between this and Witcher 3 it seems like it's time to retire the gamer religion commandment that says i7s are the same as i5s in gaming.

 

This game perhaps, but Witcher 3? Completely GPU bound:

 

[Jamdude]

How is 30 fps+ not playable

that's how much I get on skyrim with ENB and it still plays smooth ;p

"smooth"...yea ok. You play with a mouse and keyboard, right? 30 fps is horrible and is pretty much unplayable.

[MercianUK]

How is 30 fps+ not playable

 

It absolutely is, especially with adaptive refresh. It might not be ideal without it, but it's certainly still playable.

 

4K >=30fps w/ adaptive refresh is what I was aiming for when I picked up my 2nd 970; still waiting to bite the bullet on an adaptive refresh monitor but for the meantime I'm fairly happy alternating between 1080p @ 144Hz on my monitor and 2160p/1440p @ 60Hz on my TV -- 30 fps might be playable but I'll still take 60fps 1440p over 30fps 4K for a lot of games.

[Monarch]

You have fundamentally misunderstood the feature test. They specifically say it's not a benchmark, it's a feature test to test the API differences on a single system. That is why they talk about relative performance on a single system:

 

 

As such you cannot compare any score (pr the performance increases) between AMD and NVidia (or Intel for that matter). The score is only there to show the benefits of DX12/Mantle over DX11. Nothing more.

 

I understand that. But it also reveals differences between AMD's and Nvidia's overhead when you test an AMD and an equivalent Nvidia GPU on the same system, with the same CPU. That's because this test depends on CPU single-core performance and driver overhead. The test is not GPU-bound, so it doesn't matter which GPU you're using, as long as the rest of the system is the same.

Just because it's meant to show difference between APIs on a single system doesn't mean it can't be used for other purposes.

 

 

Untrue Fallout 4 should be even less demanding as an offline game it doesn't need to calculate as much in the background normally you only need to calculate what is in direct view or surrounding not to mention that the draw distance of a Battlefield map is very comparable with that of Fallout 4.

Also in Battlefield and Battlefront every shot is a physical object with travel time and physics and it's calculated even outside of your view so you can get shot from the other side of the map while at the same time people, vehicles, and destruction are tracked in the background while displaying good graphics.

The reason why Battlefield and Battlefront run so much better than Fallout 4 is because they are amazingly optimized Frostbite even uses DX11.1 in Windows 8/8.1/10 and it takes full use of 8 cores as well as hyper threading.

 

Game logic =/= rendering. Game logic can be multi-threaded. So it doesn't matter if the game is online or offline.  Draw distance also doesn't matter. It's the amount of draw calls that matters. Battlefield's draw distance and Fallout's draw distance aren't equal in terms of draw calls, even though they might be in terms of distance itself.

[Prysin]

You do realise that the 2500K is just as old, and its taken 4 years for it to actually become semi relevant against the 2500K in games?

you do realise the 2500k lost to THE faildozer (aka 81xx series, which uses THE shitties version of AMDs CMT architecture) ???

 

we are talkin about faildozer, that on average is 40% faster in single core alone.....

 

yeah...!!!

[ACatWithThumbs]

Game logic =/= rendering. Game logic can be multi-threaded. So it doesn't matter if the game is online or offline.  Draw distance also doesn't matter. It's the amount of draw calls that matters. Battlefield's draw distance and Fallout's draw distance aren't equal in terms of draw calls, even though they might be in terms of distance itself.

 

It matters because an offline game can cut out tons of things rendered in the background including objects and entire level sections which you can't do in a game like Battlefield due to the nature of the game it's not just game logic.

In an open world game everything should be streamed in as needed this is very resource light there shouldn't be a high draw call output unless they didn't optimize well or have unnecessary high draw distances with high detailed objects and shadows loading in from miles away.

The overall poly count should also be low when they implemented a good LOD especially if they implement tessellation based LOD which was the main point of tessellation in the first place.

There is literally 0 excuse for the performance/graphics there are several open world games that came out recently that did it far better and if draw calls were an actual problem they should've implemented DX11.1 or DX12 which they didn't even though they have more than enough resources for an implementation.

 

[ivan134]

It matters because an offline game can cut out tons of things rendered in the background including objects and entire level sections which you can't do in a game like Battlefield due to the nature of the game it's not just game logic.

In an open world game everything should be streamed in as needed this is very resource light there shouldn't be a high draw call output unless they didn't optimize well or have unnecessary high draw distances with high detailed objects and shadows loading in from miles away.

The overall poly count should also be low when they implemented a good LOD especially if they implement tessellation based LOD which was the main point of tessellation in the first place.

There is literally 0 excuse for the performance/graphics there are several open world games that came out recently that did it far better and if draw calls were an actual problem they should've implemented DX11.1 or DX12 which they didn't even though they have more than enough resources for an implementation.

 

Stop wasting your time. I knew he was a troll when he said Arkham Knight was an example of high CPU overhead in AMD's drivers. Lets ignore all the other open world games that came out since last year that look and run better like SOM, FC4, DAI, GTA5, dying light, witcher 3 and point to the fringe game as evidence that AMD cant optimize fallout 4. Arkham knight runs at 30 fps on a 980 ti because its the best looking game ever made.

[ACatWithThumbs]

Stop wasting your time. I knew he was a troll when he said arkham knight was an example of high cpu overhead in amd's drivers. Lets ignore al the other open workd games that came out since last year that lok and run better like SOM, FC4, DAI, GTA5, dying light, witcher 3 and point to the fringe game as evidence that amd cant optimize fallout 4. Arkham knight runs at 30 fps on a 980 ti because its the best looking game ever made.

Agree it's neither AMD's nor Nvidia's fault the game doesn't run much better on Nvidia either.

And if Nvidia caused the issue then the performance on consoles would be better but the game literally freezes on XboxOne.

Next someone is going to tell me the textures are that crappy because it would be other wise to demanding...

 

[App4that]

Agree it's neither AMD's nor Nvidia's fault the game doesn't run much better on Nvidia either.

And if Nvidia caused the issue then the performance on consoles would be better but the game literally freezes on XboxOne.

Next someone is going to tell me the textures are that crappy because it would be other wise to demanding...

 

It runs fine on Nvidia, the 970 never leaves 60fps with ultra settings. The game runs like shit on AMD, and AMD needs to get off their ass and do something about it.

[ivan134]

It runs fine on Nvidia, the 970 never leaves 60fps with ultra settings. The game runs like shit on AMD, and AMD needs to get off their ass and do something about it.

No, it doesn't run fine on Nvidia. Having the same performance as other open world games which look a million times better means it runs like shit. Running at 60 fps doesnt mean it runs great.

[App4that]

No, it doesn't run fine on Nvidia. Having the same performance as other open world games which look a million times better means it runs like shit. Running at 60 fps doesnt mean it runs great.

I'm talking about fps here, not your take on the game itself. Good game or bad the Nvidea cards run it more smoothly and at much higher fps. A 970 beats a Fury! That's not normal.

[ivan134]

I'm talking about fps here, not your take on the game itself. Good game or bad the Nvidea cards run it more smoothly and at much higher fps. A 970 beats a Fury! That's not normal.

I never said anything about that. I was pointing that you saying it runs fine on Nvidia is false. I never said it runs good on AMD.

[App4that]

I never said anything about that. I was pointing that you saying it runs fine on Nvidia is false. I never said it runs good on AMD.

We're arguing semantics I think. In the context of the game, Nvidia GPUs offer the experience intended by the developers. AMD does not.

[ivan134]

We're arguing semantics I think. In the context of the game, Nvidia GPUs offer the experience intended by the developers. AMD does not.

I'm not arguing any semantics. I was quite clear. it doesn't run "great" on either vendor's GPUs. 60 fps is not the measure of whether a game runs great, especially not a game whose textures make me think of CSGO (yes, i know this is an exaggeration but you get the point).

[App4that]

I'm not arguing any semantics. I was quite clear. it doesn't run "great" on either vendor's GPUs. 60 fps is not the measure of whether a game runs great, especially not a game whose textures make me think of CSGO (yes, i know this is an exaggeration but you get the point).

OK. I will correct my statement by saying that Fallout 4 runs much faster on Nvidia graphics cards.

 

Satisfied?

[Sakkura]

OK. I will correct my statement by saying that Fallout 4 runs much faster on Nvidia graphics cards.

 

Satisfied?

 

It doesn't though. It runs a little faster on Nvidia graphics cards. The difference is not big, and it's nothing special compared to the usual differences from game to game.

[App4that]

It doesn't though. It runs a little faster on Nvidia graphics cards. The difference is not big, and it's nothing special compared to the usual differences from game to game.

30 fps between the 390 and 970 is a big difference.

 

[Monarch]

It matters because an offline game can cut out tons of things rendered in the background including objects and entire level sections which you can't do in a game like Battlefield due to the nature of the game it's not just game logic.

In an open world game everything should be streamed in as needed this is very resource light there shouldn't be a high draw call output unless they didn't optimize well or have unnecessary high draw distances with high detailed objects and shadows loading in from miles away.

The overall poly count should also be low when they implemented a good LOD especially if they implement tessellation based LOD which was the main point of tessellation in the first place.

There is literally 0 excuse for the performance/graphics there are several open world games that came out recently that did it far better and if draw calls were an actual problem they should've implemented DX11.1 or DX12 which they didn't even though they have more than enough resources for an implementation.

 

So you think they're rendering useless things we don't see in the background? No, graphics have come a long way. We've reached a point where there's so much detail on screen that surpass the capability of D3D11. Online games don't render things in the background either. They do various calculations that aren't there in offline games, but they don't have an impact on draw calls. Keep in mind that only 1 CPU core can talk to the GPU at a time. Do you realize how easy it is to reach the draw call limit because of this? We've had multi-core CPUs for years, and yet rendering is still single-threaded. 

AMD know this. They know games are already limited by DX11, and since they have even higher driver overhead than Nvidia they know their performance suffers more. Why do you think they created Mantle? They couldn't wait for DX12. Some developers like Firaxis (guys that made Civilization:Beyond Earth) couldn't either, because they knew they're going to be draw call limited, so they used Mantle. And why do you think AMD is so excited about DX12 so much?

You said

 

 

In an open world game everything should be streamed in as needed this is very resource light there shouldn't be a high draw call output unless they didn't optimize well or have unnecessary high draw distances with high detailed objects and shadows loading in from miles away.

 

It's not that the draw call output is high, it's that DX11 has really low draw call throughput. I get 1.3m draw calls under DX11, 10m+ under DX12/Mantle.

[Nineshadow]

Well damn, this game runs pretty bad.

Heck, I managed to play Witcher 3 at 50-60 fps on low-medium settings on my GTX 560 . I can't get more than avg. 30fps on Fallout 4 at any settings (at native res, 1680x1050).

Same way with my laptop , which has an 860M.

I get avg 30fps on both low and high settings (1080p) . Haven't tested ultra. And yes, I unlocked the framerate.

[Dabombinable]

you do realise the 2500k lost to THE faildozer (aka 81xx series, which uses THE shitties version of AMDs CMT architecture) ???

 

we are talkin about faildozer, that on average is 40% faster in single core alone.....

 

yeah...!!!

If your being sarcastic then fine, otherwise Bulldozer has an far IPC worse than that of K10 and Conroe, and maybe Dothan/Yonah.

[Prysin]

If your being sarcastic then fine, otherwise Bulldozer has an far IPC worse than that of K10 and Conroe, and maybe Dothan/Yonah.

opps, typo. meant slower.

[Brennan Price]

Do you even snip bro? Don't quote images please!

Why should I not quote images. BTW I didn't know I was quoting an image.