Win8.1 BF4 performance

[96Firebird]

BF4 can output the perf overlay data into a CSV. use that for everything.

then it will be simple to graph in excel.......

Thanks, I did not know this... Looks like the command is "PerfOverlay.FrameFileLogEnable".

That should give me all the info I need.

 

[Carfax83]

Quite possible that the radeons are more heavily loaded by BF4
since it use DX11.1 features wich are supported only by the radeons,
Nvidia has not hardware support of theses features, on PClabs
graph it is clearly stated 780 dx11 and radeon dx11.2 , they should
force the radeons to dx11.0 if they really want to do an accurate
bench otherwise it s like comparing a gpu under dx10 with one on dx9.

Can you stop repeating this nonsense?

NVidia does have hardware support for DX11.1, as seen by NVidia's response

The reason why it shows DX11 instead of DX11.1 in the screenshots is because NVidia does not fully support the DX11.1 spec. It's supports the gaming functions only, but that doesn't mean it's not benefiting from running the DX11.1 pathway.

 

[wand3r3r]

I think Windows 8.1 is faster for several reasons:

Basically, there was very little difference between 3.5ghz and 4.5ghz, which kind of surprised me. But it just goes to show how efficient NVidia drivers are when it comes to exploiting multithreaded CPUs, and how well optimized the FB 3 engine is for multicore processors.

I'm sure it just proves how efficient and optimized they are.

In your own words.

Can you stop repeating this nonsense?

It's pointless to cling to that point and put it in every thread until you demonstrate proof beyond mere speculation.

You'd think your VC's have something else going for them beyond this mythical multithreading which doesn't even result in any concrete gains vs. the competing 280x (3% difference?).

The difficulty here is not only testing win 7 vs. 8.1, but also comparing that across both brands. I actually am curious if one or the other scales better and the reasons if so.

 

[Grooveriding]

I'm sure it just proves how efficient and optimized they are.

In your own words.


It's pointless to cling to that point and put it in every thread until you demonstrate proof beyond mere speculation.

Basically my point in having to debunk the OP's initial claims. There is no substantiation to any of it beyond broken benchmarks run on the bugged beta client of the game.

I'm going to do some straight-forward benches using fresh installs of Win 7 and Win 8.1 using a 4.5Ghz 4670K, a 7950 and then a 780ti. I'll run the game without any MSAA, no HBAO and everything on low except for the terrain and mesh settings on ultra (the CPU component). Then do the benches on a full game with 64 players on Shanghai, spawn in a tank at the start of the round and head full tilt for the second bridge with a few tank shots at terrain.

This is an easily repeatable bench of 64 multiplayer and allows consistency with taking the tank shots at the same time and stopping the benchmark near the second bridge. This allows you to avoid enemy contact, but still induce the high CPU stress of the constant calculations being done tracking every other player's activity in the game. The only hassle of it is needing a fresh match to do the run each time and the odd chance of a friendly blasting something with a vehicle for no reason at the start.

I'll add a youtube of the initial run as well, what is also nice is with the graph overlays provided in BF4 you can have the overlay indicate the game settings used, CPU used, GPU used, operating system as well as clock speeds everything is running at, helps for transparency. Put this heavy FUD being spread to rest. The irony is that this FUD is forum poster genesis based. Not really clear what the agenda is behind it... There are no benchmarks other than the broken beta benches done at pclab.pl There are even legitimate benches done on the shipped game at [H] debunking this nonsense.

 

[Abwx]

Can you stop repeating this nonsense?

NVidia does have hardware support for DX11.1, as seen by NVidia's response

The reason why it shows DX11 instead of DX11.1 in the screenshots is because NVidia does not fully support the DX11.1 spec. It's supports the gaming functions only, but that doesn't mean it's not benefiting from running the DX11.1 pathway.

My point was about conditions of an experience to yield
credible results , namely using the same API for any comparison.

Anyway i like how in your link they say that they have a feature but do not support it in dx11.1 even if this is a 3D feature that may be used in compatible cards so an apple/apple comparison should require forcing the
devices at the lowest common denominator otherwise one hardware
could be given an advantage if ever the feature is used from time to time.

The third feature…
· UAVs in the vertex, geometry and tessellation shaders
is supported by Fermi and Kepler GPUs however because this is only exposed through the hardware feature level 11_1, as a group of three features, we currently do not support it via the DX11.1 interfaces. We may expose support for the UAVs in the vertex, geometry and tessellation shaders feature on an app-specific basis in the future.

From the link to MS :

Feature level 11.1 provides the following additional features:

• UAVs at every shader stage with 64 UAV bind slots instead of 8.
• Target-independent rasterization, which enables you to set the ForcedSampleCount member of D3D11_RASTERIZER_DESC1 to 1, 4, 8, or 16 and to render to RTVs with a single sample.
• UAV-only rendering with the ForcedSampleCount member of D3D11_RASTERIZER_DESC1 set to up to 16 (only up to 8 for feature level 11).

 

[omeds]

I'm going to do some straight-forward benches using fresh installs of Win 7 and Win 8.1

You do that, I look forward to your results.

No, Pclab 290x Mantle results are not broken.

 

[wand3r3r]

I don't know anything about pclab, however with a feud between a manufacturer and the site I doubt their credibility and their integrity.

Why is it that certain posters (not just you) cling to one foreign sites review which isn't even a major tech site such as AT, [H], Toms, Guru3D, TPU, TR, Bit-Tech, etc. etc. etc. Their results don't even line up with any other sites that I'm aware of. Who would even look twice unless you are or speak Polish? Perhaps there aren't good sites investigating it, but at least pick a credible site and not one allegedly in a conflict with a gpu manufacturer (I don't speak Polish so I don't know about the controversy surrounding the site, I'm just going off posters remarks here in AT).

 

[Carfax83]

You'd think your VC's have something else going for them beyond this mythical multithreading which doesn't even result in any concrete gains vs. the competing 280x (3% difference?).

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..

AMD has done so as well, but not to the same extent that NVidia has.

The difficulty here is not only testing win 7 vs. 8.1, but also comparing that across both brands. I actually am curious if one or the other scales better and the reasons if so.

There's plenty of anecdotal evidence (from tons of gamers across the net), as well as written testimony from the developers themselves supporting why BF4 is faster under Windows 8/8.1.

Testing is difficult though I agree, because multiplayer is so dynamic.

I would suggest to anyone that has the means and the desire to test this, look at the area in the SP campaign, the South China Sea. According to repi, it's the most CPU intensive part in the SP campaign; particularly when you get out topside on the Titan.

That would probably be the best way to test Windows 7 and Windows 8/8.1, without resorting to the unreliability of a 64 player server.

 

[el etro]

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

 

[Carfax83]

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)

Those benches are useless, because they are in a not so very CPU dependent part of the SP campaign.

 

[PPB]

I don't know anything about pclab, however with a feud between a manufacturer and the site I doubt their credibility and their integrity.

Why is it that certain posters (not just you) cling to one foreign sites review which isn't even a major tech site such as AT, [H], Toms, Guru3D, TPU, TR, Bit-Tech, etc. etc. etc. Their results don't even line up with any other sites that I'm aware of. Who would even look twice unless you are or speak Polish? Perhaps there aren't good sites investigating it, but at least pick a credible site and not one allegedly in a conflict with a gpu manufacturer (I don't speak Polish so I don't know about the controversy surrounding the site, I'm just going off posters remarks here in AT).

We have a Polish fellow who told the story with pclab.pl. Basically since they stopped getting AMD hardware for reviews they have been pulling a OriginPC on AMD.

So the credibility of those guys MIGHT be compromised.

 

[omeds]

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.

Those are GPU limited SP results, Mantle only gains 5 fps yet we know when CPU bound its gains are much higher.

 

[el etro]

Those are GPU limited SP results, Mantle only gains 5 fps yet we know when CPU bound its gains are much higher.

It is intended to show DX11 performance difference of 290x and 780 cards on Battlefield 4. 3970x@5Ghz interfere on Mantle functioning, even at SP. Is intended too to show how PClab tests are unreasonable.

That's why i highlighted system settings on post.

 

[boxleitnerb]

Different test scenes may display different results. It is not possible to directly compare different reviews. It's clear that there are scenes where DX11.1 is beneficial and lessens the impact of Mantle while on others it apparently doesn't have any effect. Instead of hyping Mantle and dismissing everything that distracts from its glory, maybe some of you should acknowledge that.

 

[omeds]

It is intended to show DX11 performance difference of 290x and 780 cards on Battlefield 4. 3970x@5Ghz interfere on Mantle functioning, even at SP. Is intended too to show how PClab tests are unreasonable.

That's why i highlighted system settings on post.

But those tests are GPU limited, in single player. Pclab tested on a 64 player server.

 

[Cloudfire777]

There is something seriously flawed about pretty much all reviews that have happend with Mantle.
No one of the sites that tested with 290x CF which shows the greatest boost with i7 processors, have tested the DirectX systems with DirectX 11.2.

Nobody. Not even Anandtech. They tested with 11.1.
11.2 is part of Windows 8.1. 11.1 is with Windows 8.

I`m also dissappointed that no reviewer site have done a DX11.2 vs DXxx.x to compare. Especially multiplayer in BF4 would be interesting. Or CPU bound games like Starcraft or that Star swarm demo.

 

[el etro]

Different test scenes may display different results. It is not possible to directly compare different reviews. It's clear that there are scenes where DX11.1 is beneficial and lessens the impact of Mantle while on others it apparently doesn't have any effect.

GameGPu shows the same as HardOcp test.

Instead of hyping Mantle and dismissing everything that distracts from its glory, maybe some of you should acknowledge that.

Hyping Mantle? When no test suggests Mantle is a hype?
Test was made to make Mantle look like a fail. And GameGPU shows the inverse.

Nobody. Not even Anandtech. They tested with 11.1.
11.2 is part of Windows 8.1. 11.1 is with Windows 8.

Anandtech test is made with Win8.1 Pro. I'm wrong if they did something to disable DX11.2.

GameGPU uses Win 8.1Pro.

 

[cmdrdredd]

There is something seriously flawed about pretty much all reviews that have happend with Mantle.
No one of the sites that tested with 290x CF which shows the greatest boost with i7 processors, have tested the DirectX systems with DirectX 11.2.

Nobody. Not even Anandtech. They tested with 11.1.
11.2 is part of Windows 8.1. 11.1 is with Windows 8.

I`m also dissappointed that no reviewer site have done a DX11.2 vs DXxx.x to compare. Especially multiplayer in BF4 would be interesting. Or CPU bound games like Starcraft or that Star swarm demo.

How do you force a game to use a DX version it was not made for?

 

[Cloudfire777]

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

Please dont post results from the biggest BS reviewer site out there.

The AMD cards barely gained anything over DX11.2...
But its a useless test because Mantle doesn`t give anything unless you are CPU bottlenecked. Which you are not when running at 5GHz with that beastly CPU.

Anandtech test is made with Win8.1 Pro. I'm wrong if they did something to disable DX11.2.

GameGPU uses Win 8.1Pro.

They may have used 8.1 but it say in the graphs that they use DirectX 11.1 and not 11.2 which Battlefield 4 actually support.

Did AMD request this?

 

[parvadomus]

Here's another win 8.1 benchmark. It only has a gtx770 and a 290X, but its difference is the expected, nothing special in favor of nvidia. Link

 

[cmdrdredd]

cloudfire... Battlefield 4 uses DX11.1 not 11.2. Just because Windows 8.1 has that version of DirectX does not mean the game will be using any features specific to that API version.

The Frostbite Engine Technical Director over at DICE, Johan Andersson, has stated that Battlefield 4 will completely utilise Windows Direct X 11.1 and that in doing so will actually improve BF4 gaming performance...

Battlefield 4 Supporting DX11.1 Will Improve PC Game Performance
With any Battlefield title usually requiring a huge amount of hardware at its time of release gamers will be looking for all sorts of methods to get that extra bit of graphical goodness we all know Frostbite 3 can deliver. Some of our machines may be a couple of years old but we are going to be exploring ways to get BF4 looking top notch.

Johan Andersson offers up one such way to enhance BF4 on our rigs, when he recently said

"We use DX11.1, there are some optimizations in it (constant buffer offsets, dynamic buffers as SRVs) that we got in to the API that improves CPU performance in our rendering when one runs with DX11.1. This will be in BF4"

So BF4 will have CPU performance increased when using DX11.1

http://www.game-debate.com/news/?ne...rting DX11.1 Will Improve PC Game Performance
http://www.techpowerup.com/187812/b...x-11-1-could-give-radeon-hd-7000-an-edge.html

 

[el etro]

They may have used 8.1 but it say in the graphs that they use DirectX 11.1 and not 11.2 which Battlefield 4 actually support.

Did AMD request this?

How to disable DX11.2 if every driver newer than cat13.12 makes all GCN cards activate DX11.2?? (I know it has a way, i'm only talking about the test).

 

[Cloudfire777]

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

 

[Silverforce11]

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

Nothing has changed since DICE themselves release their numbers, review sites show the same scenario, you get a great performance improvement anytime you are CPU limited, ie. 64MP, but a low increase when its GPU limited (maxed settings, SP, empty servers etc). This helps those on weaker CPUs as well as helping those on multi-GPU setups (too much GPU grunt, bottleneck shifts to CPU, esp in 64MP).

Is it good? Yes, under certain conditions it is amazing, but its not for all setups. Typically, someone with an OC i5/i7 and one R290/X isn't going to see much gains, 5-10% tops even in MP. What this does help, is AMD CPU/APU systems since they are so much weaker than Intel CPUs.

 

[VulgarDisplay]

Sorry off topic.