First Fallout 4 performance results

[Misanthrope]

@Kloaked posted this one on the xbone thread so quoting:

 

So Nvidia posted a performance guide for Fallout 4, but it was taken down? I found a cached copy, though: http://webcache.googleusercontent.com/search?q=cache:http://www.geforce.com/whats-new/guides/fallout-4-graphics-performance-and-tweaking-guide

 

As I thought, God rays affect it the most, using 4x Tessellation should solve performance issues for AMD users.

[Monarch]

Except the amount of draw calls in a real game almost never reach the amount that are showing in a draw call test. It's like saying that Fire Strike is an accurate measurement for how intensive a game is. It's completely unrealistic and should never be compared to real world scenarios. 

As a matter of fact, how many times has Linus and Luke said that synthetic benchmarks are completely unreliable to use for real world performance? Like, what, every benchmark video?

 

Secondly, every single game that has had gameworks in it in the past half year has run horribly on all hardware that isn't Nvidia's 9xx series. But every other game that doesn't have gameworks in it has been reasonably optimized for both side, strange isn't it?

 

That's not entirely true.

Some games require even more draw calls than you can get with an Nvidia GPU. Batman Arkham Knight, for example. People blame GW, but it's a victim of D3D11 API. Devs are currently struggling with CPU intensive games, as they have reached the limitation of DX11. If you look at what happens when the game stutters, GPU usage drops, which indicates a CPU bottleneck. There's no reason for high-end CPUs like 5930K to be bottlenecking this game, unless it's the (single) rendering thread that's bottlenecking. Devs decided to display too much stuff on screen for single CPU core to handle. 

 

Benchmark tests can be indicators of real world performance, but you have to understand in what situations and scenarios. Graphics benchmarks like Fire Strike are meant to test your GPU to its limits and peg it at 100%. While games usually won't do that 100% of the time, depending on the game and what's going on in the game, you will encounter a scene that is as demanding, or close to Fire Strike. I've never had temps higher than 70 on my GTX 970 until I played Witcher 3. That's because other games I play are simply not that demanding. But even Witcher 3 is not always that demanding. BF4 is only that demanding when there's a huge explosion in front of you. In Witcher 3 it's when you go to the woods with a lot of foliage. 

FS is not to be used as an indicator of how much FPS you will get in games, but it does show how your GPU will handle heavy loads like its own. 

After all, we don't really use FS as an indicator of game performance, but rather compare different GPUs.

 

API Overhead tests the API and its efficiency. Since it's CPU that's making draw calls, and given that it's not MT, single core performance does matter, but that's not the point. It's again, the difference that matters. You can take any CPU you want, and obviously numbers will be lower/higher depending on the CPU, but you have to test both an AMD and an Nvidia card with the same CPU and system. API overhead test is not GPU taxing at all, it even renders at 720p. So if you get more draw calls with an Nvidia GPU in a CPU-bound scenario, then it's obvious that Nvidia's drivers have higher draw call throughput.

 

And about all GW games being broken, yes, the logical conclusion may seem to be that it's GW's fault. But it's been proven time and time again that the real reasons were rushed releases, tessellation, and driver overhead. I think Nvidia made really bad choices and partnered with bad companies. AMD were either smart or lucky to mostly pick those that didn't rush their releases. BF4, for example, was a mess, and it's a GE title. 

[ivan134]

That's not entirely true.

Some games require even more draw calls than you can get with an Nvidia GPU. Batman Arkham Knight, for example. People blame GW, but it's a victim of D3D11 API. Devs are currently struggling with CPU intensive games, as they have reached the limitation of DX11. If you look at what happens when the game stutters, GPU usage drops, which indicates a CPU bottleneck. There's no reason for high-end CPUs like 5930K to be bottlenecking this game, unless it's the (single) rendering thread that's bottlenecking. Devs decided to display too much stuff on screen for single CPU core to handle.

Benchmark tests can be indicators of real world performance, but you have to understand in what situations and scenarios. Graphics benchmarks like Fire Strike are meant to test your GPU to its limits and peg it at 100%. While games usually won't do that 100% of the time, depending on the game and what's going on in the game, you will encounter a scene that is as demanding, or close to Fire Strike. I've never had temps higher than 70 on my GTX 970 until I played Witcher 3. That's because other games I play are simply not that demanding. But even Witcher 3 is not always that demanding. BF4 is only that demanding when there's a huge explosion in front of you. In Witcher 3 it's when you go to the woods with a lot of foliage.

FS is not to be used as an indicator of how much FPS you will get in games, but it does show how your GPU will handle heavy loads like its own.

After all, we don't really use FS as an indicator of game performance, but rather compare different GPUs.

API Overhead tests the API and its efficiency. Since it's CPU that's making draw calls, and given that it's not MT, single core performance does matter, but that's not the point. It's again, the difference that matters. You can take any CPU you want, and obviously numbers will be lower/higher depending on the CPU, but you have to test both an AMD and an Nvidia card with the same CPU and system. API overhead test is not GPU taxing at all, it even renders at 720p. So if you get more draw calls with an Nvidia GPU in a CPU-bound scenario, then it's obvious that Nvidia's drivers have higher draw call throughput.

And about all GW games being broken, yes, the logical conclusion may seem to be that it's GW's fault. But it's been proven time and time again that the real reasons were rushed releases, tessellation, and driver overhead. I think Nvidia made really bad choices and partnered with bad companies. AMD were either smart or lucky to mostly pick those that didn't rush their releases. BF4, for example, was a mess, and it's a GE title.

At this point I'm convinced youre a troll.

[bcredeur97]

so how much would my 6970 get? 

[Atmos]

"Game works does not affect performance negatively on other vender's cards!" -Nvidia

- Fallout 4, 390X performs significantly worse than a 970

 

 I smell bullshit bethesda, fix you're game, and help stop this fucking vender bias.

[App4that]

"Game works does not affect performance negatively on other vender's cards!" -Nvidia

- Fallout 4, 390X performs significantly worse than a 970

 

 I smell bullshit bethesda, fix you're game, and help stop this fucking vender bias.

Anyone else notice the release of the Fallout 4 edition 970 for 385us at Best Buy? Is Radeon asleep behind the wheel? Already had to prepare myself that I might be returning Fallout 4, the game I built my PC to run. Oh what a fanboy they'll lose in me.

[Memories4K]

Anyone else notice the release of the Fallout 4 edition 970 for 385us at Best Buy? Is Radeon asleep behind the wheel? Already had to prepare myself that I might be returning Fallout 4, the game I built my PC to run. Oh what a fanboy they'll lose in me.

Earlier you talked about a new driver AMD released recently and your crossfire setup.

Driver update not good enough?

This whole release was pretty sketch from the start with the minimum requirements.

A 7870 is a whole other level better than a 550 ti.

[JerkyMcDilerino]

Hey, the 970 perform not bad in 1440P. But the 970 beat R9 390 in 1080P? Must be NVIDIA driver increase the performance.

[Humbug]

And about all GW games being broken, yes, the logical conclusion may seem to be that it's GW's fault. But it's been proven time and time again that the real reasons were rushed releases, tessellation, and driver overhead. I think Nvidia made really bad choices and partnered with bad companies. AMD were either smart or lucky to mostly pick those that didn't rush their releases. BF4, for example, was a mess, and it's a GE title.

Thing is it's not just about gameworks features. Most of these are Nvidia the way it's meant to be played titles and NVIDIA boasts about working closely with the game devs. So when things happen such as fallout4 not supporting enough resolutions or fps lock or games being over-tessellated it just looks bad on NVIDIA because they are supposed to be working closely with the game devs.

this is about more than AMD vs Nvidia. This is about best practices in game development, period.

[App4that]

Earlier you talked about a new driver AMD released recently and your crossfire setup.

Driver update not good enough?

This whole release was pretty sketch from the start with the minimum requirements.

A 7870 is a whole other level better than a 550 ti.

Guess not, pretty salty about it. Way too late to go Nvidia now, really sucks.

[Memories4K]

Guess not, pretty salty about it. Way too late to go Nvidia now, really sucks.

I'm not planning on upgrading until Arctic Islands/Pascal so i still have time.

Jesus, i hope something can be done on current AMD cards or else my next card will be Pascal.

Even if this issue would be resolved with Arctic Islands i would have lost faith in AMD's product by then.

[TOMPPIX]

YES THIS IS SOME GREAT NVIDIA TECH RIGHT HERE.

[Monarch]

At this point I'm convinced youre a troll.

 

I'm convinced you're a troll. You have posted 0 evidence to back up your claims.

[SteveGrabowski0]

Anyone else notice the release of the Fallout 4 edition 970 for 385us at Best Buy? Is Radeon asleep behind the wheel? Already had to prepare myself that I might be returning Fallout 4, the game I built my PC to run. Oh what a fanboy they'll lose in me.

 

I see it for $370. That's ridiculous since it doesn't come with the game. And because I'm a Bethesda fanboy I'll probably go drop $15 on ebay for the case badge if it has Vault Boy on it. 

[SteveGrabowski0]

Anyone else notice the release of the Fallout 4 edition 970 for 385us at Best Buy? Is Radeon asleep behind the wheel? Already had to prepare myself that I might be returning Fallout 4, the game I built my PC to run. Oh what a fanboy they'll lose in me.

 

Fuck me, it comes with a backplate with Vault Boy on it. I think I'm not retarded enough to go SLI just to get a Vault Boy backplate.

 

 

EDIT: Fuck that, I think I could paint that myself, I'm pretty good with an airbrush and there are so few colors. I'll just buy a regular backplate and see if I can do that shit.

[MEC-777]

Wonder if it will play nice with crossfire... 

[Kinda Bottlenecked]

YES THIS IS SOME GREAT NVIDIA TECH RIGHT HERE.

 

That looks to be very tessellation heavy...

[SteveGrabowski0]

Nah, nevermind, the backplate that fits my 970 has way too many holes in the wrong places to paint that shit on.

 

[Misanthrope]

That looks to be very tessellation heavy...

 

Yep, called it. In fact everyone did days ago when Beth published that article and mentioned tessellation. It's just ONE setting that kills performance. Instead of hair now it's god's rays....so in a metaphorical way, kinda hair again.

[Humbug]

That looks to be very tessellation heavy...

Would have hoped that 980ti stays well above 60fps even on ultra. Since this isn't witcher 3 quality visuals...

[xAcid9]

without Gamewreck. 

 

390x gained 37% average fps and 55% min fps.
980 gained 22% average and 44% min fps.

 

an old quote..

 

"Gameworks hit performance for both vendors but it hit more on AMD hardwares." - crazy mofo

[Derangel]

without Gamewreck. 

 

390x gained 37% average fps and 55% min fps.

980 gained 22% average and 44% min fps.

 

an old quote..

 

"Gameworks hit performance for both vendors but it hit more on AMD hardwares." - crazy mofo

 

The percentages look bad but look at the average numbers as well. Nvidia gained 22.4 FPS while AMD gained 22.6. Factoring in a margin of error they gained exactly the same amount of FPS.

 

For fun: Nvidia gained 16 FPS for the minimum while AMD gained 15, again, close enough to call exactly the same.

[You_are_a_cunt]

Played for about 8 hours last night. Here's the tl;dr version:

 

1680x1050, everything maxed out, rock solid 30 FPS (VSync is ON). Will take a look at framerate with VSync OFF later today (after I get home).

 

0 crashes, minimal load times (although they're few and far between)

 

More than happy with the performance

[Notional]

No, they haven't. Have you been living under a rock? Why the hell do you think AMD performs bad in CPU-bound games? This is why:

 

AMD:

 

Nvidia:

 

That's 32% difference.

 

The test LITERALLY says you cannot compare the numbers between systems OR vendors. Your entire point is moot.

 

YES THIS IS SOME GREAT NVIDIA TECH RIGHT HERE.

 

Yup. @Kloaked made a sarcastic comment that it was NVidia's fault the game runs like crap. Ironically he wasn't wrong. Sure the game engine might not be very efficient or good per se, but this is the most taxing setting in the game, officially made in cooperation with NVidia. Odd to see NVidia delete their own performance guide. I wonder what they leaked out that they should not have.

 

without Gamewreck. 

 

390x gained 37% average fps and 55% min fps.

980 gained 22% average and 44% min fps.

an old quote..

 

"Gameworks hit performance for both vendors but it hit more on AMD hardwares." - crazy mofo

 

It's astounding that people still cannot see it. NVidia tech like GameWorks, god rays, etc. hurt performance for everyone, but especially AMD. With source code being unavailable for AMD, they will always suffer in these games, with the resulting blame on AMD drivers. It's getting old fast and people should know better by now.

[Humbug]

It's astounding that people still cannot see it. NVidia tech like GameWorks, god rays, etc. hurt performance for everyone, but especially AMD.

A lot of NVIDIA users look at TWIMTBP titles on launch day and feel good about things like 15% better performance than AMD. They don't stop to think about how much absolute performance has been sacrificed in order to achieve that goal, I.e. Is this game made in the optimal way with best possible performance in general.