Win8.1 BF4 performance

[96Firebird]

Why is everyone discussing Mantle in this thread? I thought we were discussing the benefits of going from Win7 to Win8.1 in BF4?

Just because the Mantle thread got closed doesn't mean we need to drag it into every other thread...

 

[VulgarDisplay]

Why is everyone discussing Mantle in this thread? I thought we were discussing the benefits of going from Win7 to Win8.1 in BF4?

Just because the Mantle thread got closed doesn't mean we need to drag it into every other thread...

I honestly forgot this wasn't the mantle thread. Posting on a small phone screen gets me disoriented a lot.

 

[cmdrdredd]

Oh, my bad. Nevermind my posts in this thread then. I thought the first pictures in this thread was from DX11.2.

Well then its settled. Mantle is better. Especially with CF like parvadomus linked to


When can we expect games that use DX11.2? I`m curious to see what that brings since its being used by Xbox One

Well, the XB1 isn't using the same DX as you have in windows. It's using the same feature set but it's more specifically optimized for the hardware in the system. The DX running on the Xbox One is much more low level.

I haven't researched DX11.2 but does it offer anything significant for games? Performance gains expected due to efficiency, or extra IQ stuff?

As for Win 8.1 being faster it's likely it's just more efficient overall as compared to 7 or even 8. There's a lot of things under the hood of Windows 8 to reduce overhead. Aside from the server issues I think Battlefield is very well coded. It runs pretty good across a variety of hardware and takes advantage of x64 and multi-core.

 

[wand3r3r]

Mythical multithreading? What on Earth are you babbling about? It's a fact that NVidia has supported multithreading in their drivers ever since dual core processors first became available.

NVidia drivers multithreaded since 2005.

The extent of those optimizations has certainly increased over the years, as more and more cores/threads have been added to CPUs, and via refinement..

I should have worded it differently, mythical benefits since there is no major performance difference, e.g. the 770/280x are neck and neck. If it was as wonderful as some pretend there would be a difference. To me it's merely a marketing bullet point until the cards start actually leading.

Oh, my bad. Nevermind my posts in this thread then. I thought the first pictures in this thread was from DX11.2.

Well then its settled. Mantle is better. Especially with CF like parvadomus linked to

When can we expect games that use DX11.2? I`m curious to see what that brings since its being used by Xbox One

Please link to the source when you add external images. (At least give them some traffic since you are using their bandwidth to post images elsewhere) I'd like to read the review but don't feel like tracking down the source. Also, this is going off topic and belongs in the mantle thread.

Well, the XB1 isn't using the same DX as you have in windows. It's using the same feature set but it's more specifically optimized for the hardware in the system. The DX running on the Xbox One is much more low level.

I haven't researched DX11.2 but does it offer anything significant for games? Performance gains expected due to efficiency, or extra IQ stuff?

As for Win 8.1 being faster it's likely it's just more efficient overall as compared to 7 or even 8. There's a lot of things under the hood of Windows 8 to reduce overhead. Aside from the server issues I think Battlefield is very well coded. It runs pretty good across a variety of hardware and takes advantage of x64 and multi-core.

I think it's forward thinking and is technically advanced. I wasn't playing it at launch so I haven't ran into the numerous issues (that they are even being sued for) but the server lag = rubber band effect is annoying. It has crashed a few times, but is somewhat rare.




--

Has anyone attempted to bench 7 vs. 8.1 yet? How's it going?

 

[Carfax83]

I should have worded it differently, mythical benefits since there is no major performance difference, e.g. the 770/280x are neck and neck. If it was as wonderful as some pretend there would be a difference. To me it's merely a marketing bullet point until the cards start actually leading.

The GTX 770 and the 280x aren't what I'd call neck and neck. It's more appropriate to say they trade blows, with NVidia leading in some games and AMD in others.

One thing I believe though, is that NVidia has an advantage the more CPU bound a game is (plus the more threaded the engine is), and I think this is because of how well the drivers make use of multicore processors. As you ratchet up the resolution and AA though, things start to swing back in AMD's favor.

Take BF4 for instance. Until Mantle arrived, NVidia had completely caught up with AMD's early performance lead in the DX11.1 path (due to driver updates and patches), and was actually outperforming AMD; particularly in multiplayer.

Frostbite 3 engine can use as many as 8 threads, just like CryEngine 3. Those are the only two engines I'm aware of that can use that many threads, and NVidia has the edge in both games. Look at this graph:

NVidia has superior scaling at lower, less GPU bound resolutions than AMD. From the GTX 770, to the GTX 780, to the GTX 780 Ti, there is a clear performance increase.

With AMD, there's barely any increase at all going from a 290 to a 290x. That's a CPU limitation.

Mantle is helping AMD level the playing field in this regard, but it's a shame they had to resort to such a drastic tactic as creating a whole new API, rather than just using more aggressive multicore optimizations in their drivers like NVidia has done.

Windows 8/8.1 helps NVidia even more, as the OS is fine tuned for CPU performance...

 

[Gloomy]

omg you are still using empty server benches

 

[Carfax83]

omg you are still using empty server benches

I and others have already told you several times already. For testing GPU performance, an empty server is best because you can get more reliable test data, and the GPU only renders what's on the individual player's screen so overall player count has a minimal or negligible impact on frame rate, as the majority of the battle isn't being witnessed.

Server load is only important for measuring CPU performance, as the CPU workload increases dramatically alongside player count.

Obviously gamegpu understands this simple concept as their CPU test was conducted on a populated server..

 

[Braxos]

Since when 770 can dx11.2? I thought it supports 11.1 only.

 

[wand3r3r]

I and others have already told you several times already. For testing GPU performance, an empty server is best because you can get more reliable test data, and the GPU only renders what's on the individual player's screen so overall player count has a minimal or negligible impact on frame rate, as the majority of the battle isn't being witnessed.

Server load is only important for measuring CPU performance, as the CPU workload increases dramatically alongside player count.

Obviously gamegpu understands this simple concept as their CPU test was conducted on a populated server..

Hahaha now I suspect you are just joking. So 64 players have no affect on FPS?

Yeah you get more consistent data on an empty server, with much higher FPS than a full server. Strange isn't it?

Since when 770 can dx11.2? I thought it supports 11.1 only.

It doesn't. Iirc It doesn't even support 11.1 fully either, just a subset.

 

[Carfax83]

Hahaha now I suspect you are just joking. So 64 players have no affect on FPS?

I never said it had no impact, only that it was negligible and not worth going over the hassle for as getting reliable data is very difficult..

Really, I feel like a broken recorder or something as I'm having to repeat myself over and over again on this issue.

It's if it's not being rendered, the GPU doesn't care about it.

If the 64 player match took place in a much smaller environment where you could see damn near everything that was happening at once, then you'd have a point.

Yeah you get more consistent data on an empty server, with much higher FPS than a full server. Strange isn't it?

Yes, and I'll give you three guesses as to why that is. Clue, it has nothing to do with the GPU..

 

[Carfax83]

Since when 770 can dx11.2? I thought it supports 11.1 only.

The GTX 770 and all Kepler based cards supports all of the gaming functions of DX11.1, and most of the major functions of DX11.2 to a degree I believe.

It doesn't have full support for either though, but it doesn't really matter.

 

[MeldarthX]

The GTX 770 and all Kepler based cards supports all of the gaming functions of DX11.1, and most of the major functions of DX11.2 to a degree I believe.

It doesn't have full support for either though, but it doesn't really matter.

No its not fully 11.1; nor it 11.2.......while it handles gaming functions through 11...

It could of been fully 11.1 but Nvidia decided not to do it.

http://www.guru3d.com/news_story/nvidia_kepler_not_fully_compliant_with_directx_11_1.html

It uses backwards compatiblity to do it.....

http://www.hardwarecanucks.com/news...-and-amd-over-directx-11-1-support-continues/

 

[Racan]

Mantle is helping AMD level the playing field in this regard, but it's a shame they had to resort to such a drastic tactic as creating a whole new API, rather than just using more aggressive multicore optimizations in their drivers like NVidia has done.

No it's definitely not a shame and we can only hope that this will provide an impetus for future DirectX and OpenGl versions to be made to work more like Mantle.

You could never accomplish something like this with DX11.

 

[cmdrdredd]

No its not fully 11.1; nor it 11.2.......while it handles gaming functions through 11...

It could of been fully 11.1 but Nvidia decided not to do it.

http://www.guru3d.com/news_story/nvidia_kepler_not_fully_compliant_with_directx_11_1.html

It uses backwards compatiblity to do it.....

http://www.hardwarecanucks.com/news...-and-amd-over-directx-11-1-support-continues/

Supposedly the features that matter in games can be done through the driver.

 

[el etro]

I and others have already told you several times already. For testing GPU performance, an empty server is best because you can get more reliable test data, and the GPU only renders what's on the individual player's screen so overall player count has a minimal or negligible impact on frame rate, as the majority of the battle isn't being witnessed.

What you stated basically says that the GameGPU test is valid.

 

[Erenhardt]

NVidia has superior scaling at lower, less GPU bound resolutions than AMD. From the GTX 770, to the GTX 780, to the GTX 780 Ti, there is a clear performance increase.

With AMD, there's barely any increase at all going from a 290 to a 290x. That's a CPU limitation.

Well this have more to do with graphics cores than CPUs. 290 to 290X is berely any difference in shaders/tmu/rops/memory. Both are TDP constrained and are hitting thermal wall.

780 to 780Ti on the other hand is a massive difference in die resources. 780 have 20% less active die than 780ti. Both are under 290/290X TDP

 

[Carfax83]

No it's definitely not a shame and we can only hope that this will provide an impetus for future DirectX and OpenGl versions to be made to work more like Mantle.

You could never accomplish something like this with DX11.

If BF4 implemented DX11 multithreading, you could see similar performance increases, although probably not as high.

Also, who on Earth would pair high end GPUs like that with such a low end CPU?

Answer, no one. That pairing superficially makes Mantle look better than it is.

 

[Carfax83]

What you stated basically says that the GameGPU test is valid.

Yes, of course it's valid. No reason why it shouldn't be.

 

[Carfax83]

Well this have more to do with graphics cores than CPUs. 290 to 290X is berely any difference in shaders/tmu/rops/memory. Both are TDP constrained and are hitting thermal wall.

That's not really true.

The 290 and 290x have a higher thermal threshold than any of the NVidia cards, and the 290x has 10% more GCN core and texture units than the 290, plus higher clock speed. And yet it's only a measly 3% faster.

Also, look at the Sapphire Tri-X OC. It has no trouble staying under it's thermal threshold, so the full 1ghz clock speed is easily maintained, but yet it's only 2% faster than the stock 290.

Face it, the AMD cards in that test have run into a wall.

780 to 780Ti on the other hand is a massive difference in die resources. 780 have 20% less active die than 780ti. Both are under 290/290X TDP

Yes, the GTX 780 Ti is has more hardware, but it's just as reliant on the CPU as the GTX 780 for rendering. The GTX 780 Ti is nearly 20% faster than the GTX 780 in this test, which is what you'd expect..

The point of what I'm saying, is that the CPU is less of a limiting factor with NVidia hardware than AMD hardware, as shown by countless benchmarks.

How does this pertain to Windows 8.1 BF4 performance? Let's get back on topic please.

-Rvenger

 

[el etro]

Also, who on Earth would pair high end GPUs like that with such a low end CPU?

Answer, no one. That pairing superficially makes Mantle look better than it is.

R9 290 costing $400 fill the budget of many hardcore gamers. 290x not.


R9 290 costs $400, can be paired with a $200 processor. It makes a gaming system than can match ~90% of the framerates the most powerful system(i7 4690x + 780Ti) can do.
A friend of mine uses his 290(the only one 290 i saw personally) with a i5 here in Brazil.

Hardcore gamers sweet spot is $150-$300 gaming cards and equally priced CPUs. We all know most of times a processor(especially intel's and FX-6300/8300) can drive cards that cost double with relatively low bottleneck. And mantle means much for all gamers that Go with these configurations, even in fixing the framedrops, frametimes and other badness these systems exibit in playing.

 

[sontin]

First look from Techreport in the SIngle-Player mode with Windows 8.1:

http://techreport.com/review/25995/first-look-amd-mantle-cpu-performance-in-battlefield-4/2

DX11.1 gives nVidia user the same performance increase like Mantle over DX11.0.
AMD had had luck that noboby looked at the MP for the last 3 months.



Next troll attempt you will be on vacation. I have had enough of you.

-Rvenger

 

[Erenhardt]

That's not really true.

The 290 and 290x have a higher thermal threshold than any of the NVidia cards, and the 290x has 10% more GCN core and texture units than the 290, plus higher clock speed. And yet it's only a measly 3% faster.

Also, look at the Sapphire Tri-X OC. It has no trouble staying under it's thermal threshold, so the full 1ghz clock speed is easily maintained, but yet it's only 2% faster than the stock 290.

290 vs 290X was a fan speed battle. 290X result is in quiet mode, while 290 is running higher fan speeds. Bigger 290X is running lower frequency due to thermal throttling, while smaller 290 is enjoying almost no throttling at all. If you read through 290 reviews, you will find out that often it was as fast or even faster than 290X (within "out of the box" testing policy:whiste

Sapphire 290 Tri-X is running on 1000MHz while stock 290 clocks is 950MHz - hardly any difference if you ask me.

Get back on topic, now.

-Rvenger

 

[Rvenger]

If I see anymore off topic discussion I will be locking and cleaning this thread out.


-Rvenger

 

[Fastx]

GameGPU test:


i7 3970X@5GHZ (Zero bottleneck!)
Windows 8.1
Mantle vs DX11.2(Or whatever cards can use)
Resolution 1920x1080p 4xMSAA Max Settings(No supersampling)
290x on default mode.

Same test with all cards on DX only:

Spoiler

The test lacks the use of Xfire 290x.

http://gamegpu.ru/action-/-fps-/-tps/battlefield-4-mantle-vs-directx-11-test-gpu.html

Thanks for posting the charts/link, for me this still keeps the 780 open as a possible bang for the buck option should I decide to upgrade.

 

[omeds]

First look from Techreport in the SIngle-Player mode with Windows 8.1:

http://techreport.com/review/25995/first-look-amd-mantle-cpu-performance-in-battlefield-4/2

DX11.1 gives nVidia user the same performance increase like Mantle over DX11.0.
AMD had had luck that noboby looked at the MP for the last 3 months.



Next troll attempt you will be on vacation. I have had enough of you.

-Rvenger

Interesting results, thanks for the link. It seems Nv hardware/drivers show larger gains with faster CPU's, perhaps IPC speed is playing a part for Nv drivers?