Win8.1 BF4 performance

[Hitman928]

HardOCP actually shows ~6 fps increase from win7 to win8.1 across three different tests. Two of which are single player and the third is multiplayer. Using the same graphics settings for single player and 64-player multiplayer showed almost the exact same increase from win7 to win8.1 (6.3% to 6.2%) as shown here: http://www.hardocp.com/images/articles/13849867120D7cnkMfuM_3_1.gif. Even with the high graphical settings, there is a huge cpu load difference going from single player to a 64 player map, surely if there was a significant cpu load difference improvement from the OS change it would sure SOME change in this comparison rather than basically 0 change in performance increase from win7 to win8.1 going from single player to multiplayer.

As for the pclab results, maybe someone who doesn't have to rely on google to translate from Polish, but I see zero explanation in their latest result in comparing 8.1 to 7. No test conditions, no system settings, nothing. They throw up one comparison at the end of the mantle review using a GTX770 when trying to say that dx11.2 brings large improvements in API overhead as well, which doesn't make sense to me as Nvidia doesn't even run the dx11.2 render path as far as I know in BF4. They easily could have pulled that conclusion from their earlier testing with windows 7 without re-testing windows 7 with the new patch. Like I said, I don't speak Polish so it's hard to know for sure, but even just at the surface that comparison doesn't quite pass the smell test, not when there's nothing from any other site that I've found to collaborate their findings, on contradictory evidence.

 

[Cloudfire777]

I found a benchmark of the finished game, not the BETA which was linked to in the first post. Sadly no multiplayer benches or Crossfire/SLI since thats where Mantle gave most improvement, but even in single player you get pretty good boost.

Im upgrading to Windows 8 for sure if they just get their behinds in gear and giving me my start menu.

Windows 8:

 

[omeds]

Well I installed BF4 into win7 and compared myself. These preliminary results are not very scientific and were not in a populated server, because I wanted a straight up apples to apples comparison with repeatable results. I did try to include the largest views with most objects to raise draw call count, and used 1080p with no AA to reduce GPU load and promote CPU bound performance.

I agree further testing with 64 players is whats needed, but these early results look promising:

Win7

Win 8.1

I had to double check the win7 results, I thought perhaps SLI wasn't working, but it 100% was, and even when I enabled 4xAA+TRSSAA frame rate didn't change because it was completely CPU bound.

 

[Cloudfire777]

wow pretty nice results you got. That is like 50-60% increase in FPS when gaming in Multiplayer.

What GPUs are you running?

I wish I could rep you.

 

[omeds]

4770k at 4.8 + 780 SLI

And thanks :thumbsup:

 

[Cloudfire777]

I think this goes to show that like Mantle, the more GPU power you have, you eventually get to CPU bottleneck and you need to optimize the software to squeeze out more power from the CPU.

You have 2x780 which is a good amount of GPU power just like the Mantle benches where they ran 2xR9 290x in CF and saw the biggest gains
Yeah I`m def upgrading to Windows 8 when I get my next SLI build

 

[Erenhardt]

^^ You sure those are the same settings? There are some differences between those screenshots.

wow pretty nice results you got. That is like 50-60% increase in FPS when gaming in Multiplayer.

What GPUs are you running?

I wish I could rep you.

By definition multiplayer requires more than 1 player on the server.

 

[Cloudfire777]

And MP should give even better results because you need the CPU even more.

 

[omeds]

^^ You sure those are the same settings? There are some differences between those screenshots.



By definition multiplayer requires more than 1 player on the server.

Correct, however those are still CPU bound performance figures. Yes, 64 player testing is needed, and I would think more CPU load will only show greater increases.

 

[Skurge]

Yeah,

You won't getting those numbers in a full server even on a 4770K. And if someone thinks the gap will be bigger when the server is full, then BF4 won't be playable on a midrange CPU on win7.

 

[omeds]

I run the game with a 147fps cap at 144hz, and it stays at that cap ~90% of the time on most maps, some are worse than others. I was using a 125 cap at 120hz at release, but the game runs a little faster now. Those screencaps were taken where frame rate was lowest.

 

[BrightCandle]

I also see around 110-177 fps in the game with my two 680's. Admittedly I turn motion blur off and I turn the MSAA setting down to 2x but otherwise its on ultra settings. I see these sorts of FPS' all the time in real multiplayer games.

 

[Hitman928]

I also see around 110-177 fps in the game with my two 680's. Admittedly I turn motion blur off and I turn the MSAA setting down to 2x but otherwise its on ultra settings. I see these sorts of FPS' all the time in real multiplayer games.

win7 or win8?

omeds, I'm sorry but screenshots of instantaneous frame rate in an empty server doesn't really show me much.

 

[Grooveriding]

win7 or win8?

omeds, I'm sorry but screenshots of instantaneous frame rate in an empty server doesn't really show me much.

Agreed. Providing confirmation bias for your own already debunked argument on the basis of what was initially provided substantiates nothing. I also like the previous poster see some clear IQ differences between the Win 7 and Win 8.1 screenshots

I'll gladly run some win 7 vs win 8.1 benchmarks using a 7950 and a 780ti when I have some time. Still provides no evidence because I'm just one person on the forums that doesn't do professional reviews or run a website that does.

 

[PPB]

Judging by the weapons you are using seems that either that was an unranked server/server without connection to Blaze or you are just new to the game.

If the latter is the case, I recommend you to try 64p test. Empy servers really dont mean anything, any of us playing this game for already a few months can show totally different perf on empty vs full 64p server

 

[BrightCandle]

win7 or win8?

Windows 8. Its actually 8 not 8.1, haven't run the update yet as last time it didn't go well.

 

[omeds]

Tough crowd. Those screencaps clearly prove there are large gains under win8.1. You are all more than welcome to try it yourself.

I would like to see 64 player testing too.

 

[itsmydamnation]

Tough crowd. Those screencaps clearly prove there are large gains under win8.1. You are all more than welcome to try it yourself.

I would like to see 64 player testing too.

here you go:

http://forum.beyond3d.com/showpost.php?p=1824372&postcount=990

Never ever compare frame rate in absolute 'fps' numbers, it has a non-linear relationship. Use milliseconds / frame instead for comparisons or percentages. This is your first warning

Johan Andersson.
ie use BF4 perfoverlay commands.
the poor representation of data means we cant tell the quality of data.

 

[96Firebird]

So, if I were to try to compare the differences, how would you guys like me to get numbers?

 

[wand3r3r]

At the very least try do something with a min-max-avg run for a while with the stats being recorded. I don't think it's very scientific, however you could try run though a map going in the same direction and doing something to ensure the run is as close as possible to being repeatable. (Note, I wouldn't be convinced of much, however it could reveal the need to investigate further)

The single screenshots are beyond ridiculous for proof, you can look in a different direction and have very different results. I haven't tried benching much but in BL2 when I had issues with physx I could look in one direction and get 150 FPS and look in the other and get 75 FPS.

I think Grooveriding is close with his statements, and has debunked several myths in this thread already.

 

[96Firebird]

How should I record stats? As far as I know, MSI Afterburner doesn't work in BF4 unless you run in x86, but I don't want to do that. Was there an update so it works now?

I was thinking about just playing a 32-player game as I normally would, and then do it on the same map after I upgrade. Maybe a couple different maps. I don't enjoy 64-player servers, so I'm sticking to 32.

 

[wand3r3r]

Fraps?

I'll admit I'm already skeptical but interested anyway.

Try do it as repeatable as possible, for as long as possible, as many times as possible.

 

[omeds]

The single screenshots are beyond ridiculous for proof, you can look in a different direction and have very different results. I haven't tried benching much but in BL2 when I had issues with physx I could look in one direction and get 150 FPS and look in the other and get 75 FPS.

I think Grooveriding is close with his statements, and has debunked several myths in this thread already.

That's why I took the screenshots from the same locations with the crosshair pointed at the same thing. He didn't debunk anything, there are clearly performance gains under win8.1.

I would also like to see how this translates to 64 player performance. Judging by Pclab 290x Mantle vs 780 win8.1 results, its looking good.

 

[itsmydamnation]

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

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

 

[Carfax83]

As some of you may or may not know, Nvidia seems to be leveraging DX features under win8.1 which offer some similar benefits as Mantle in BF4 - reduced draw call and CPU overhead. For whatever reason, AMD is not doing the same (probably because they figure everyone will use Mantle, and they've always had trouble with DX multithreading)

I think Windows 8.1 is faster for several reasons:

1) Improved thread scheduler and lower kernel latency. Windows 8.1 uses the CPU more effectively than Windows 7.

2) The DirectX 11.1 renderer has some optimizations in it that improve CPU performance. From repi himself, "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"

3) NVidia has superior multicore optimizations in it's drivers, which when combined with the above, result in large gains in BF4 when running under 8/8.1 in CPU limited scenarios.

I game at 1440p, but when I was testing frame rates and frame latency in the Mantle thread, I ran my CPU at 3.5ghz instead of the regular 4.5ghz and I lowered my resolution to 1080p in the same area used by AMD for their screenshots. This area supposedly is the most intensive part of the SP campaign.

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.