Bad Company 2 Beta Benchmark

[evolucion8]

as for Crysis it usually is a little bit faster on a quad then on a dual core of the same architecture. in other words a Q9550 at 2.83 will usually match or beat a 3.16 E8500. of course some of that could be the additional cache.

Doubtfull, the Q9550 has the same 6MB of cache shared between two cores as the E8500, like two 8500 glued together, and if data has to be shared across the two different sets of dual cores, the performance penalty would be great because it would have to travel through the memory bus. The performance gains would be thanks to the additional cores, but is just too little of a gain though.

 

[Cogman]

That's the thing. You don't design for a specific core count, you design around a paradigm. If you go multithreaded, then inherently having additional cores will speed things up - if your code is designed properly. But you also have to take the least-common-denominator in terms of hardware into play. A single core can handle multiple threads; it does so as I've described - serially. Having multiple cores on the other hand can be just as bad, for instance, say you have one thread waiting on another. You've just stalled a core waiting for input. There are many reasons for an application to simply release it's slice of CPU time back to the operating system, which in turn means that application isn't utilizing 100% of a core. That doesn't mean it's not efficient. Quite the contrary - it means the code is doing what it is supposed to do in the minimal amount of time necessary. That is what is called efficiency. Running a CPU up to 100% is simply brute forcing your way through, but it doesn't mean it's doing so efficiently.

There are cases where 100% is better then, say 50%, but they are rare. In games, they are pretty much non-existent (because, as you say, if the game plays at 10% CPU utilization, that means it was efficiently using its time slice to get everything it needs done.)

100% utilization is generally considered efficient when you have a large chuck of data that you are processing (and you aren't overloading the OS with too many threads). The goal then is to get through the data as fast as possible, you don't want the threads having to wait for each other, that would be inefficient as it would take a longer amount of time to complete the processing (thus wasting clock cycles).

But you know that already, I guess this is more for everyone else . Games don't need 100% utilization to be considered efficient, in fact, less then 100% is a good thing. (IE, why should a game like starcraft use its entire time slice when all its calculations can be done in a minuscule portion of the time slice.)

 

[Skurge]

For what its worth, I got bout 60% CPU usage on the rig in the sig. Max settings.Dnt know what the FPS was, but it was defo playable.

 

[Silverforce11]

If you are using a duo core system, you can edit the ini file to have low sound quality, it drastically improves FPS.

DICE have said they are tweaking the engine to be less stressful on duo cores, but realistically i don't think there is much they can do since havok is tough on CPUs. Pretty much everything is destructible, with big fire fights and explosions going all over the place, a quad core is required for 60 fps+.

The PC beta has dx11 but only result in softer shadows at a fps drop, not worth it imo. Dunno what other dx11 features retail will have.. but so far, all the dx11 titles are a letdown. But i guess thats how it is since these games were designed with dx9 in mind, and other stuff just add minor things. It will be several years before we get games fully designed for dx11.

 

[toyota]

so is a sound card basically useless for this game?

 

[RussianSensation]

so is a sound card basically useless for this game?

More or less. The main benefit of a dedicated sound card is better DACs for this game.

 

[AzN]

Because sometimes there's no work to do. It's that simple. If you're waiting for the user to shoot the bad guy in the head there are no computations to make. Once the user does pull the trigger only THEN do you need to compute impact, physics of blood splatter and body recoil, sound, etc.

Seeing where this is going? You can't simply store up computations and use them later. (People bringing up CPU out-of-order execution just to be difficult will be ignored.)

Old games running on single cores did have busywait event loops. They'd peg any single core 100%. That's not efficiency, that's poor design.

There are plenty of engines that scale well with dual core and quad core respectively. Why is that you are giving excuses that somehow Frostbite engine is optimized for dual core when quad core uses same amount of CPU load?

If you do not stop your confrontational/antagonistic posts, you may lose your posting ability for a while in the video forum. Have a civil, NON accusational, conversation. Nothing less.
You've just come off 3 days vacation and your very first two posts are to tell people to stop making excuses. This tells me you do not have any interests in keeping a conversation civil. If you need more time off, let me or another moderator know.

Anandtech Moderator - Keysplayr

 

[AzN]

That's the thing. You don't design for a specific core count, you design around a paradigm. If you go multithreaded, then inherently having additional cores will speed things up - if your code is designed properly. But you also have to take the least-common-denominator in terms of hardware into play. A single core can handle multiple threads; it does so as I've described - serially. Having multiple cores on the other hand can be just as bad, for instance, say you have one thread waiting on another. You've just stalled a core waiting for input. There are many reasons for an application to simply release it's slice of CPU time back to the operating system, which in turn means that application isn't utilizing 100% of a core. That doesn't mean it's not efficient. Quite the contrary - it means the code is doing what it is supposed to do in the minimal amount of time necessary. That is what is called efficiency. Running a CPU up to 100% is simply brute forcing your way through, but it doesn't mean it's doing so efficiently.

There are plenty of engines that scale with quad, dual core, and clock speeds. Stop giving excuses that somehow it is optimized when we have fast dual core users getting crappy results compared to slow quad users.

If you do not stop your confrontational/antagonistic posts, you may lose your posting ability for a while in the video forum. Have a civil, NON accusational, conversation. Nothing less.
You've just come off 3 days vacation and your very first two posts are to tell people to stop making excuses. This tells me you do not have any interests in keeping a conversation civil. If you need more time off, let me or another moderator know.

Anandtech Moderator - Keysplayr

 

[felang]

Competely trying to avoid the whole "what is or isn´t multithreaded" argument, all i can say is that this is the first game that has loaded my q6600 more than 70%. I actually had to take my less than stable oc down from 3.6ghz to 3.2, this is the first game I´ve ever had crash because of having my cpu at 3.6. I can run mass effect or cod6 all day at 3.6 without problems, but my cpu usage is never over 50% in these titles...

 

[schneiderguy]

Azn, have you tried running the game in DX9 mode?

 

[AzN]

It crashes to desktop when running dx9

It's funny how Keys says it's confrontational but the first responses from some of these guys were confrontational from the start. I don't see Keys saying anything to them.

You don't know what you're talking about. I see other people already addressed that.

Does the term beta mean anything to you? :hmm:

This thread is seriously lulz @ all the people thinking they know what multithreaded means, especially AzN thinking that Vista or 7 can magically turn a single-threaded application into something that uses multiple cores. Here's a clue: The Vista and Win7 scheduler routinely bounces threads from core to core - doesn't mean it's multithreaded.

Personally, if the developers at Dice say they're using multiple threads, it'd be kind of stupid to debate that because you're looking at task manager and not seeing 100% CPU load.

Here's a suggestion - finish high school and go to college studying under a compsci degree. Pay particular attention in your systems architecture and operating systems classes. Then you might have an idea of what's going on inside of a computer and application.

Regarding Dice's comments - DUH. But then again you seem to have never developed software, so you probably wouldn't have any clue how much overhead something like simple debug code can affect an application. Their comment still has no bearing on your "100% CPU usage" fallacy.

 

[schneiderguy]

It crashes to desktop when running dx9

The reason I'm asking is because on my computer in DX10 mode it appears to be completely GPU limited to around 30-35 fps even on a GTX 260. My CPU usage falls down to about 35% on a quad core compared to 60-70% in DX9 mode.

I get around 35 FPS in dx10 and over 70fps in DX9.

And my first post in this thread was this, by the way:

The game gets 50%+ usage on my quad core. Unless Windows is having the same thread run concurrently on two of my cores just "for the lulz" then the game is multithreaded

When I say "you don't know what you're talking about", it's not meant as an insult. There's nothing wrong with not knowing something, just don't go on and on saying false things.

 

[AzN]

You were antagonizing from the start schneiderguy.

Somehow "Stop giving excuses" is accusational? It's silly really.

See you in a week.
Anandtech Moderator - Keysplayr

 

[Tab]

The reason I'm asking is because on my computer in DX10 mode it appears to be completely GPU limited to around 30-35 fps even on a GTX 260. My CPU usage falls down to about 35% on a quad core compared to 60-70% in DX9 mode.

I get around 35 FPS in dx10 and over 70fps in DX9

Damn.

Have you tried using the nvidia beta drivers?

 

[SunnyD]

There are plenty of engines that scale with quad, dual core, and clock speeds. Stop giving excuses that somehow it is optimized when we have fast dual core users getting crappy results compared to slow quad users.

I think that right there is a pretty darn good indication that it is optimized for multiple cores/threads, unless you've mixed up your statement or I'm understanding your statement wrong again. If it's multi-core optimized, why wouldn't it get better performance out of more cores? 4 > 2 right?

Random example: A thread takes X amount of time at clock speed Y. If you have 4 threads running on 4 cores (1 per core), then each thread is able to complete in time X at clock speed Y. Trivial math.

Now say you only have 2 cores to execute those 4 threads, meaning each core has to do twice the work (handle 2 threads each). So at clock speed Y, obviously each thread is going to take 2X the time to complete as each thread has essentially only half the resources (in reality, it's slightly more than 2X because of overhead and whatnot from switching tasks between threads).

Obviously then, in order to compensate a two core processor would need twice the clock frequency to keep up (simplifying out a lot of things, in reality clock speeds aren't the end-all be-all of an architecture). Ergo the reverse is true, a quad core processor could run at half the clock speed of a dual core processor to achieve the same thread throughput. Again, it's not QUITE that trivial, but the math is fundamental.

Now given something running, lets say only two threads, then the dual and quad would be on equal footing, making clock speed a factor.

So taking your statement, it shouldn't be difficult to see what's going on here: In a multithreaded optimized application (using more than two threads), a multi-core processor is obviously going to perform better than a dual core processor. Again, this is an extremely simplified view, as not all threads in this title are going to be doing the exact same thing, thereby weighting thread performance differently. But still, more cores (up to whatever the application was designed for) will make a performance difference. In essence, you just disproved your own argument.

 

[schneiderguy]

Damn.

Have you tried using the nvidia beta drivers?

I'm using the latest ones nvidia has on their site... 196.21? I can't figure out why performance is halved in DX10 mode because theres very little (if any) quality difference that I can tell.

 

[tweakboy]

That's why the name of the game starts with Bad ,

no all kidding aside. Its a beta, and that means its test version. It is not optimized yet Im sure final version will play much better,,,()

 

[Phil1977]

I doubt they will improve performance much. I mean the game will be out next month. I think in 3 weeks time...

Anyway I have enjoyed the game heaps. Got all the unlocks (well all the ones in the Beta).

I admit this game was the reason why I upgraded to an i7 860 system. I know I will be playing this games for hours and hours in 2010, maybe even longer depending on when BF3 comes out. Performance is outstanding now. At least double the frames compared to my Athlon II 250 system before. Same 5750 video card though.

I am really amazed everytime I see the Xbox360 version on you tube. It's super smooth and looks just as good.

The PC Beta version didn't have the final Hi res textures yet. So IQ should improve a little bit more.

Also this game is full of jaggies. You need a lot of AA to get rid of it.

 

[zerocool84]

I doubt they will improve performance much. I mean the game will be out next month. I think in 3 weeks time...

Anyway I have enjoyed the game heaps. Got all the unlocks (well all the ones in the Beta).

I admit this game was the reason why I upgraded to an i7 860 system. I know I will be playing this games for hours and hours in 2010, maybe even longer depending on when BF3 comes out. Performance is outstanding now. At least double the frames compared to my Athlon II 250 system before. Same 5750 video card though.

I am really amazed everytime I see the Xbox360 version on you tube. It's super smooth and looks just as good.

The PC Beta version didn't have the final Hi res textures yet. So IQ should improve a little bit more.

Also this game is full of jaggies. You need a lot of AA to get rid of it.

Yea it's jag city. 2xAA doesn't help much and neither does 4x so I'll turn it up to 8x and deal with the fps drops I get when there's a lot of smoke everywhere cus it looks so much better.

 

[Phil1977]

Also the way mipmaps and textures appear on the screen (viewing distance) is exactly like on the Xbox 360.

The appear quite late, which you notice a lot when you ride the quad as its very fast.

It looks like PC gamers really want BF3 and BC2 is the console port version.

 

[zerocool84]

Also the way mipmaps and textures appear on the screen (viewing distance) is exactly like on the Xbox 360.

The appear quite late, which you notice a lot when you ride the quad as its very fast.

It looks like PC gamers really want BF3 and BC2 is the console port version.

Well like said multiple times, this is the Beta version and they are adding all the visual goodies in the retail version. It runs fine for me and not going to comment on anything until it comes out or there are reviews out on the retail game.

 

[Phil1977]

Well like said multiple times, this is the Beta version and they are adding all the visual goodies in the retail version. It runs fine for me and not going to comment on anything until it comes out or there are reviews out on the retail game.

Fair comment and I hope you are right!

But with only 3 weeks to go I don't expect miracles...

Edit: 3 weeks, 2 days and 3 hours :-D

 

[zerocool84]

Fair comment and I hope you are right!

But with only 3 weeks to go I don't expect miracles...

lol I hope so too cus I pre-ordered it on Steam and can't get a refund if the reviews suck.

 

[scooterlibby]

Continuing on that point - they are adding features not to the beta version, but to a later incarnation of the build. In other words, the build that they are adding features to isn't even the beta, so could already at base be better than the beta already without the added features.

 

[schneiderguy]

The beta build was around 2 months old when it was released according to the devs. I think their focus was probably on making it stable rather than having it run well.