Win8.1 BF4 performance

[Silverforce11]

PCLab also notes they entered a 64 player server. If your implication is that they entered an empty server, their testing methodology does not state that. My translation isn't perfect but the wording to me indicates a full 60+ player game as tested by PCLab, some AMD users were spreading misinformation about this to no end. As far as the other benchmark from PCGE, I do not know what their testing method for MP is..

There is no misinformation, a lot of early comparisons were done using GameGPU.ru's results for BF4 as well as the xpac.. until it turns out, they posted a video of their actual testing, it was them running around on 0/4 server and driving an APC shooting at the ground. Likewise for other sites so far. If PCLab.pl actually did 64 player MP benches, then kudos to them, because we as gamers certainly need to demand better standards from review sites.

The real point here is testing a game intended for MP in SP/fake MP is doing a disservice to readers. As for win7 vs 8, DICE supposedly patch up the poor win7 performance awhile ago and it IS due to multicore handling.

 

[blackened23]

What websites do simulated MP on empty servers?

The main reason for not doing MP testing is because the results are not consistent or repeatable. Now some websites still test MP, but the testing process is still questionable at best. The results simply aren't consistent, but what a lot of sites do is average a lot of data over a long time span so it averages out. Anyone with a free thinking mind could easily see that repeating results in MP is a problem - When you're testing 10+ video cards for a review that is a problem.

That said, some sites are doing MP tests. There isn't a good answer on HOW to do it. HardOCP doesn't even attempt to get repeatable results, so I find the notion of HardOCP's teseting for MP being "great" to be laughable at best. If you read their teseting method, none of it is repeatable. Because of the consistency issue - every run in MP is different. Different number of explosions, etc.. It's an obvious issue for anyone testing MP.

Please note. I'm not saying MP testing shouldn't be done. It should be done, as it *IS* good information. What should be done is finding a better testing method for repeatable results. Quite frankly, HardOCP's testing method for MP is questionable at best. Read their testing method.

In Battlefield 4, we selected the Siege of Shanghai as our map for testing.Siege of Shanghai is presented as a large conquest map with up to 64 players able to be connected at one time. The map provides a good representation of all of the eye candy that can be found throughout the game, including devolution of the sky scraper, water effects, tight city streets, and of course plenty of tanks. For our testing procedures, we searched for a server with at least 60 players in a match and where the ping was under 50ms for a server.

We then played within the server performing various tasks, including capturing flags, blowing up tanks, and causing general mayhem as part of the team that we were placed on. We recorded the frame rates via FRAPS for five minutes for presentation below, even though we spent significantly more time playing at each graphical setting. Keep in mind that because this is a multiplayer map where the actions of the other 63 players can be unpredictable, it is very difficult to look at a single FPS graph and discern the gameplay experience in the multiplayer environment. Performance or is determined through more time spent playing with the cards than represented on the graph.

Read through that mess, when they themselves admit that the action is unpredictable and not repeatable. Every run is different. I think there should be a consensus developed for better testing methods than this.

 

[96Firebird]

Hmm, I might have to look into switching over to Win 8.1 again. It didn't seem worth it a while ago, and I am still happy with the way things perform on my end. It is always a pain having to install everything again...

 

[Grooveriding]

Some of those benches are under a week old and still show similar improvements on Nv hardware. That win7 vs 8 [H] review is from November and at rather GPU limited settings. Their recent GHz review is showing larger gains than expected against a 290 imho.

Again if you were a Battlefield 4 player you'd know this was a bug that was fixed. You're using beta benchmarks from October that exhibited this bug to spread a FUD theory and adding in a bench that is unrelated and manipulating core counts and clock speeds.....

We even discussed this on this forum with that exact bench from October in the beta....

http://forums.anandtech.com/showthread.php?t=2346906


FUD debunked

This is just a repost of a troll/fud thread reposted with a new title to try and spread the FUD again...

The angst on this forum is unbelievable now. A video card company released a new feature.... who cares, if it's so hard to handle just don't post imo.

 

[omeds]

Quite the hostile reply indeed.

First of all, those results are at GPU limited settings and from November. There have been quite a number of updates since then, and testing at resolutions more likely to expose CPU bound performance are still showing improvements - this week.

 

[Grooveriding]

You seem awful hostile for someone bringing up a valid point there buddy. The latest beta does in fact have better performance under 8.1 in MP, it's just a rational discussion of that. Not sure why you're getting so angry/hostile.....

Back that statement up with some links to benchmarks that are not from October 2013 before the game launched.

 

[omeds]

There you go:

 

[Grooveriding]

There you go:

Should of linked to the bench for transparency, since it is from:

October 29th 2013

http://pclab.pl/art55318-10.html


DICE on the issue in November: http://www.reddit.com/r/battlefield...pu_lag_in_battlefield_4_i_found_a_fix/cd6nuha

November 24th, 2013

http://www.hardocp.com/article/2013...ows_7_vs_81_performance_review/4#.Uu-dWrRv3Ho

 

[rancherlee]

So many variables here. Just the newest BF4 patch (which includes Mantle) seems to give a pretty decent boost to FPS. I noticed ~5fps boost on my system and it plays smoother while still under DX11.x in Win 8.1. I haven't even tried Mantle yet as the 79xx/280x cards don't seem to gain much at this point. To really dig deep there needs to be a Pre patch, post patch, and Mantle test done in both 7 and 8.1 to get the real numbers but getting the real numbers with any accuracy is a royal pain with no scripted 64player demo/benchmark.

 

[omeds]

Should of linked to the bench for transparency, since it is from:

The second graph clearly states 2014.02.

Sure, I wasn't aware of the date of the first graph as it was included in their review which is only 2 days old, to explain this -

[H]'s GHz review shows a larger than expected lead over the 290 imho, and is consistent with Pclabs GHz results under win8.1, but at a higher resolution

Read the Pclab explanation - are you saying this is not true?

 

[parvadomus]

There you go:

Dont you have another source different than NVLab.pl? :whiste:

 

[omeds]

That's the point of the thread, to find more information on the subject. If you have some current 8.1 results please share.

 

[96Firebird]

Dont you have another source different than NVLab.pl? :whiste:

If you could find another recent review that shows a comparison between Win 7 and Win 8.1 on a recent BF4 build, I'd appreciate it.

 

[Grooveriding]

The second graph clearly states 2014.02.

Sure, I wasn't aware of the date of the first graph as it was included in their review which is only 2 days old, to explain this -

[H]'s GHz review shows a larger than expected lead over the 290 imho, and is consistent with Pclabs GHz results under win8.1, but at a higher resolution

Read the Pclab explanation - are you saying this is not true?

Well now that your first three graphs have been debunked as from a beta state exhibiting a bug, you've added a third graph that shows what exactly ? A different overclocked SKU performing faster than a 290X. Now you offer a new graph from pclab.pl showing different cards in a bench that is days old and is obviously not even the same bench run from back in the beta of October.

Then a graph from [H] using a reference 290, not a 290X, vs a custom OC 780 Ghz model. A custom 780Ghz is faster than a reference 290, we all know this.

The most odd thing about the original beta benches from pclab showing win 7 vs 8 performance is that they are only using an nvidia card to demonstrate the difference. Anyone know why they didn't include both AMD and nvidia ? Seems odd that they would make a claim nvidia performs better from moving from Win 7 to 8 without including an AMD bench showing that they don't see the same benefits with the OS switch.


Well so far you have 3 debunked benches showing beta issues with the game. Two completely irrelevant benches thrown into the mix. Looks like what is left is just your opinion that this is the case with zero to back it up and benchmarks as evidence against your opinion.


Keep trying I guess :whiste:


Your "Keep trying" comment was unnecessary. Now you have brought my attention to this thread. Keep it civil, please.

-Rvenger

 

[Techhog]

see the word beta..... yeah... this thread needs to closed..........

To be fair, the final game definitely feels like a beta.

 

[Cloudfire777]

I`m not sure if you guys noticed, but just like Mantle, DX 11.2 (Win8) give better results with AMD APUs than with Intel CPUs.
So it looks like DX11.2 works kinda in the same way as Mantle, just not as optimized.

About 10% improvements with i7 and 20% with APUs.

EDIT: Its actually +50% with FX 8350....What the hell?

 

[omeds]

Well now that your first three graphs have been debunked as from a beta state exhibiting a bug, you've added a third graph that shows what exactly ? A different overclocked SKU performing faster than a 290X. Now you offer a new graph from pclab.pl showing different cards in a bench that is days old and is obviously not even the same bench run from back in the beta of October.

Then a graph from [H] using a 290, not a 290X, vs a 780 Ghz. The 780Ghz is faster than a 290, we all knows this.

The most odd thing about the original beta benches from pclab showing win 7 vs 8 performance is that they only using an nvidia card to demonstrate the difference. Anyone know why they didn't include both AMD and nvidia ? Seems odd that they would make a claim nvidia performs better from moving from Win 7 to 8 without including an AMD bench showing that they don't see the same benefits with the OS switch.


Well so far you have 3 debunked benches showing beta issues with the game. Two completely irrelevant benches thrown into the mix. Looks like what is left is just your opinion that this is the case with zero to back it up and benchmarks as evidence against your opinion.


Keep trying I guess :whiste:

That's great buddy, but would you kindly stop posting? You're turning into an AMD vs Nv debate, when we are discussing possible win8.1 DX11.2 CPU bound performance improvements.

I may have to investigate this myself, I have both win7 and win8.1 installed, but BF4 is only installed under win8.1 atm.

Next time hit report post and let the moderators handle this situation.

-Rvenger

 

[el etro]

R9 290 at [H] test runs at ~930Mhz. GTX 780 at this same test runs over ~1100Mhz(yes, all Nv cards override first boost limit and Nvidia don't disclose the upper boost limit the cards have).

Knowing how much a chip of this size scales with each additional Mhz on core clock, you will find that is not fair(unless we are talking about pricing of the cards) compare the Gigabyte premium GTX 780 card with the flawed reference card from AMD.

In their original prices, the real duel should be R9 290 Lightning vs GTX 780 Ghz Edition.

 

[Abwx]

The most odd thing about the original beta benches from pclab showing win 7 vs 8 performance is that they only using an nvidia card to demonstrate the difference. Anyone know why they didn't include both AMD and nvidia ? Seems odd that they would make a claim nvidia performs better from moving from Win 7 to 8 without including an AMD bench showing that they don't see the same benefits with the OS switch.

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.

 

[Grooveriding]

That's great buddy, but would you kindly stop posting?

First off, you wanted to 'discuss' this. You'd gotten the wrong idea from benchmarks from the beta of the game displaying a bug that had nothing to do with what you thought it did. That is now corrected.

You're turning into an AMD vs Nv debate,

Secondly, I've done nothing of the sort. I've shown that your assumptions are incorrect and related to a bug with the beta of the game and that you were using beta benchmarks to make that assumption. If anything your OP predicated an idea of this being AMD vs nvidia:

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)

when we are discussing possible win8.1 DX11.2 CPU bound performance improvements.

There is no possible at this point beyond your opinion. Your benchmarks were from beta, showing a bug in beta, as I said. So what you are trying to discuss at this point is just your statement, with nothing to back it up, and evidence to the contrary.

I'll gladly continue to post and it will be interesting to see if you can offer some evidence that there is even one benchmark that backs up your assumption. So far what you've offered has shown a game bug from the pre-release beta of BF4 and then an unrelated benchmark of different cards showing raw horsepower differences.

 

[omeds]

[redacted]

No I think there are improvements to be seen, as explained by Pclab.


That will be enough with the sarcasm.

-Rvenger

 

[Hitman928]

[redacted]

No I think there are improvements to be seen, as explained by Pclab.

Are you really insistent to claim AMD bias on a guy running 780 SLI? All I've seen is him point out the flaws in the original argument to show there's no proof of anything discussion worthy. I happen to agree with him unless you have some new evidence to provide?

 

[omeds]

No one has shown any other data, except [H] at GPU limited settings, and they still recorded an increase. I have no reason to doubt Pclab, their review is very comprehensive.

I don't believe 780GHz is that much faster than Hawaii in BF4 based on GPU performance alone, either.

 

[Rvenger]

Omeds and Grooveriding - If I see anymore arguing this thread will go down the same as the other. I am not kidding, get back on topic and cut the personal attacks or I am locking this discussion and handing out infractions.

To me, I see this as a Mantle thread in disguise and this discussion needs to be about BF4 performance as a whole and not a Mantle argue thread.

-Rvenger

 

[Grooveriding]

No one has shown any other data, except [H] at GPU limited settings, and they still recorded an increase. I have no reason to doubt Pclab, their review is very comprehensive.

Yes there is no reason to doubt pclab's benchmarks from October of the beta game showing a bug in the beta. Although it is dubious that they tried to show advantages in going from Win 7 to Win 8.1 on nvidia, but not for AMD, by only using a Win 7 to Win 8.1 benchmark for nvidia.

I don't believe 780GHz is that much faster than Hawaii in BF4 based on GPU performance alone, either.

A custom Gigabyte 780ghz OC, with non-reference board and cooler that performs similarly to a 780ti is not going to be faster than a reference 290 with stock PCB and crappy reference cooler...

Never mind the benches we've seen here are more GPU limited with 2560x1600 4xMSAA Ultra settings used

Let's take a look at their less GPU limited bench of 2560x1600 2xMSAA Ultra, that should highlight this assumed CPU advantage:

Looks like it doesn't actually. They actually perform closer to one another than they do in the more GPU demanding benchmark. Still no evidence whatsoever presented in this thread to back up the original claim in the OP.