TechSpot

Battlefield 3 Beta GPU & CPU Performance

By Julio Franco
Oct 4, 2011
Post New Reply
  1. Medium are the highest settings you will get in this beta. Only AA, AF ambient occlusion, motion blur etc can be set to the highest setting.

    I think it has something to do with the file size of the beta and the comparison between console and PC versions.
     
  2. Steve

    Steve TechSpot Staff Posts: 1,369   +467 Staff Member

    First of all you bash us by saying were
    You then go on to say
    Okay well if you look at our 1920x1200 results our average for the 6950 was 41fps. You do realize that 1920x1200 is pretty much the same thing as 1080p and that the number 41 lies between 40 and 45 right?

    In any case here are some other performance reviews that found the 6970/6950 to be even slower…

    http://www.hitechlegion.com/images/stories/bf3beta/intelmax.jpg
    http://www.hitechlegion.com/reviews/gaming-software/13498-bf3-beta-performance-benchmarks?start=2

    http://static.cms.netshare.pl/cms/2...dajnosc_-_ustawienie_antialiasing-596x539.png
    http://www.frazpc.pl/artykuly/918822/Battlefield_3_BETA_-_jakosc_grafiki_i_wydajnosc

    Finally I thought you weren’t coming back?

    Utilization does not tell the whole story, the dual-core can have 100% utilization while the quad can have 50%, doesn’t mean there will be a huge difference in frame rate performance. Again we simply report the results as we find them, I am not trying to trick anyone.

    As for your findings on the Caspian Border map with dual-cores I was not able to test here so I will have to wait for the full version for more conclusive testing. Again we did mention in the article that this was a preview and the results are based on a beta version of the game so a lot could change when the final version is released.
     
  3. I wasn't trying to imply that you guys tried to trick people with this review! I have always found techspot to be one of the best of the crop of on-line sources of computer related articles and reviews.

    I should have mentioned that there are other sites on the internet saying that the beta is extremely well optimized and that people without CPU bottlenecks see their GPU utilisation pretty much sitting at 100% all the time on Operation Metro. This coupled with the fact that my GPU was utilised at 75-80% rate while the CPU sat at 100% led me to say that dual cores are a performance bottleneck in this game. This seems to be roughly in line with what you say - if my GPU goes from 80% to 100% utilisation I should see about 25% increase in frame rates... Anyway, I've bought a Phenom II x4 965 yesterday and it arrive at my house later today, so I'll pop back in and let you know if it changes anything.

    As for Caspian Border, many people didn't even manage to get into one of the servers when they were up, certainly there are no performance benchmarks on that map on the net. When I played on that map I was so disappointed with performance that I decided to find out what is bottlenecking my system (I was pretty sure it wasn't my GTX 580 overclocked by 15%) so I've configured the overlay that comes with MSI Afterburner to show GPU utilistaion, video ram usage and frame rates and guess what - CPU usage was constantly at 100% while my GPU use was at about 50 and this produced framerates in the low 20's... My conversations with people on the server confirmed that dual cores are the bottleneck in this game - there were people with entry level quads, and worse graphics cards who enjoyed far better performance than I did.

    To the people saying that this game is CPU heavy because it's not a console port all I can say is: get a grip, man! A. If it were true that the CPU's in consoles were much weaker than the ones in PC's and the game was optimised for PC hardware then consoles wouldn't be able to run this game, no? I am aware that console CPU's are much slower at single threaded tasks, but they ware designed to be very good at multi threaded tasks, like physics calculations and simulating bullet ballistics, which I'm almost sure is what makes PC CPUs sweat so hard in this game. On consoles, I'm almost certain that the GPU is the limiting side.
     
  4. Caspian Boarder has been opened to the public again and will stay open until the end of the open beta, so you guys have a chance for a quick performance review. I assure you, you will see collosal differences between duals and quads. I'm still playing with slide show quality frame rates, my x4 965 should arrive within the hour though. Maaaan, I can't wait to play this with smooth frame rates, this map is SO much fun!
     
  5. I've installed the 965 and clocked it at 4Ghz. I can confirm that my frame rates on Caspian essentially doubled. I now rarely see a dip below 40 fps with everything on ultra + 19x anisotropic filtering + 2x MSAA and no post processing AA. My CPU usage is around 95%, so even a quad core Phenom x2 clocked at 4Ghz is on the edge of bottlenecking the graphics card O_O. Makes me wonder what's going to happen once full environmental destruction is enabled... Then again tesselaation will also be enabled and that hits the graphics cards pretty hard.

    Anyway, playing Caspian now is a joy. Frostbite 2 at low to mid 40's feels smooth as butter. Hell, even high 30's are very playable.
     
  6. *16x Aniso and quad core Phenom II . I should have created an account to be able to edit my mistakes lol.
     
  7. http://static.techspot.com/articles-info/448/bench/Ultra_1920.png
    that's your ultra results, and you're posting the 6950 at 28 fps - now that's fine if you've gotten access to final build settings which we aren't seeing in the beta, but you need to make that clear, because I ain't running any settings on the high that sees you guys get 41 fps

    I know that there are other 'benches' showing the 69xx cards doing apallingly in the BF3 beta (though I hadn't seen those two before), but I trust only a handful of sites for this kind of info, and of those only [H] and you guys have taken a proper look at BF3, but your results are so divergent from what myself and [H] have seen that I'm just a tad pissed off

    and of course I'm coming back, TS is the easiest techsite to read with a decent, but not overeaching array of news and clear presentation for benches and written jazz

    .

    and re:Metro benches - TS give the 6950 47 fps (iirc) maxxed @1200p, but if anyone can reproduce that with the in-game bench please post screenies, 'cause I'm hitting about 35, and so to [H], 3dguru, toms et al so... :/
     
  8. Steve

    Steve TechSpot Staff Posts: 1,369   +467 Staff Member

    I am surprised such an article could piss anyone off. Based on what I have seen around the net our results are very similar. I have not looked at the Hard|OCP results but if they differ greatly from mine it is because they tested differently. Again all the articles I have looked at so far align with mine.

    As for Metro you do realize we test with the DX11 effects disabled right?

    "Metro 2033 has no built-in benchmark tool so we have relied on Fraps to gauge GPU performance. The “Chase” level was used where we recorded the first 60 seconds of the cart ride. The game was tested using DirectX 11 though Tessellation was disabled as it hammered performance too severely. Anti-aliasing was disabled and instead we used the default analytical anti-aliasing with 4xAF."
     
  9. DAOWAce

    DAOWAce TS Rookie Posts: 68

    He's right about hit detection; it's very bad. I experienced all the problems for myself (on multiple server clusters that I all pinged for 12-20ms). Instant deaths, getting shot by people not even on your screen, dying well behind cover and hits failing to register despite the game confirming them. Bluntly put: Caused me to quit.

    I just checked the servers now and see they finally added Caspian Border. About damn time; may go back and try it on a higher ping server to see if I get a negative effect of the hit registration issues. If not, the game will be uninstalled (as I finally got to experience an actual Battlefield map) and I will not be buying it (until I hear about the netcode being improved).

    For anyone who's interested in the details: http://battlelog.battlefield.com/bf3/forum/threadview/2826551139123655506/

    Also: A multiplatform release is ruining the full potential of this game. Very disappointed.
     
  10. Why in the world would you conduct a CPU test at 1080p? At that resolution the GPU, not the CPU, is the limiting factor. You're supposed to benchmark a CPU by turning down graphics settings and thus eliminating the GPU bottleneck, at which point the CPU becomes the limiting factor.

    Your CPU chart is very misleading because at lower resolutions, AMD CPU's struggle against Intel.
     
  11. Steve

    Steve TechSpot Staff Posts: 1,369   +467 Staff Member

    Yes thank you I am very familiar with that technique. If this were a CPU review I would have tested at a lower resolution and disabled AA but it’s not. It’s a performance review on BF3 that is designed to show gamers what it takes to play the game using maximum/high quality settings. Using the high preset ay 800x600 isn’t exactly going to do that.
     
     
  12. To the guy claiming that the CPU isn't the limiting factor at 1080p: play some Caspian Border on a dual core and watch your CPU and GPU utilisation.
     
  13. I think CPU demands have a lot to do with the number of people on the server.

    I have a 560 TI & Core 2Duo @ 4GHz. I've noticed when I play Caspian Border servers with 48 players, my framerate is significantly better than on servers with 64 players... and I notice my card stays much closer to 100% usage on the 48 player servers whereas on the 64 player servers it tends to be in the 70-80% range.

    Caspian is a load of fun... too bad about the lack of game menus/browser, but I suppose I'll get used to it and not think about it that much about it after a while.
     
  14. @Guest about CPU utilization:

    I don't disagree that CPU usage may be a bottleneck for some people (see post I just made on this topic), but I don't think CPU utilization is necessarily a meaningful stat.

    I currently have my C2Duo @ 4Ghz... but for a couple years I had it at 3.17. When playing BC2, my CPU usage was generally about maxed out at 3.17. After I bumped it to 4GHz.... still generally maxed out while FPS stayed about the same.

    I suspect the physics engine scales to use left over power for lower priority objects that otherwise would not be rendered/animated.

    So basically my guess is that there is a certain amount of CPU power that is required to do the mandatory physics + general game code, and then a certain amount of additional CPU power that can be taken advantage of to do lower priority physics (if that CPU power is available).

    Therefore, if I'm right, just because your CPU is maxed out, it doesn't mean its a bottleneck. OTOH, I think if your GPU usage is not maxed out, that is much more indicative that the CPU is a bottleneck. (In my case, GPU usage is basically 100% on 32 player metro, but only 70-80% on 64 player Caspian... and on 64 player Caspian, my FPS is 20-30% lower).
     
  15. That is a very good point about CPU use you make. I've also noticed that I get better frame rates on servers with fewer people, which is why I said earlier that I suspect that physics calculations and modelling bullet ballistics is what chokes CPU's in this game.

    Your point about the game using leftover CPU time for less vital things made me think. After playing a few rounds on servers with varying player counts it seems to me that there is higher quality building destruction on less populated servers - on packed ones, when I shoot an RPG at the wall it seems that a chunk of the wall vanishes in a puff of dust, whereas on smaller servers I can clearly see the wall break up to many chunks of various sizes flying off in different directions... Perhaps this is the kind of extra detail the game uses leftover CPU time to calculate. Anyway it would be nice if DICE entered the XXI century with their physics calculations and offloaded some of them to the GPU...
     
  16. "Anyway it would be nice if DICE entered the XXI century with their physics calculations and offloaded some of them to the GPU..."

    I whole-heartedly agree. Such a shame that this many years after GPU assisted physics became available that most major games still don't use it. But I think the problem is less with DICE (or the other developers) than it is the proprietary nature of the physics engines out there.

    PhysX support would be great for those of us with Nvidia cards... but multi-platform support is not as good for PhysX and using it would likely leave the AMD crowd up in arms and possibly hurt BF3 sales as a result. Havok, which BF3 uses and which has excellent cross-platform suppport, was on track to support GPU off-loaded calculations until Intel bought it in 2007 (... hmm, I wonder why Intel bought Havok...)

    Given the options, I can't wholly blame DICE for going with Havok. What we need is a major physics engine not owned by a hardware company. I'm thinking Bullet may become our knight and shining armor some day. Maybe for BF4....
     
  17. "Frankly, it's disappointing to see the lack of polish in current builds of BF3"

    It's not a current build, the beta is more aimed at having the online backend ready and able for the end of the month to help them not crash and burn at launch. People seem to misunderstand the difference between Beta and Demo. Many of the bugs and other problems have already been addressed in the final build. That being said, I am sure it will take 2 or 3 updates to fine tune the balance and other issues.
     
  18. I played about 20 hours on the Caspian Border map, and a lot of that flying in the jets and my framerate was great. I never measured it, because I don't really care... for me, framerate is one of those things that you shouldn't ever notice, because when you do, it's usually a bad thing... like you don't notice the tires on your car, until one of them explodes at 70 mph...

    Anyway, here's my rig:

    i5 2500k OC to 4.5 Ghz
    560 Ti (900 mhz core)
    8 Gigs DDR 1600
    Win 7 64-bit Ultimate

    I played on max settings minus AA... I simply cannot afford a rig (at the moment) that can acceptably run the game with AA enabled. Had I noticed any choppiness or the need for AA, I would be talking with the missus about saving another $220 for another 560ti. As it is, I am completely happy with my setup and can't wait for the game to come out.

    Also, I have to point out that the whole "I'm worried about the final product" crap needs to go away, especially from a website that is fantastic about providing empirical, objective data... we don't need your speculation, and all it does is fuel the trolls and pointless debates.
     
  19. Steve

    Steve TechSpot Staff Posts: 1,369   +467 Staff Member

    Yes our point exactly, I think you were agreeing with us without knowing it ;)

    Umm what are you doing here then?

    Right but if you want to play with maximum quality settings which a lot of gamers do then you need to know what cards can handle it. Also I pay close attention to the tires on my car to avoid them exploding at 70mph, things like wear indicators and pressure are good things to keep an eye on, in fact I would say its far more important to keep tabs on your tires rather than your frame rate – that’s my road side tip for the day.

    Using AA is certainly a personal preference but even at 2560x1600 I cannot play games without it. AA does not reduce “choppiness”, it removes jagged edges by smoothing them and games look significantly better with it enabled.

    Circling back to the original guest post, I am not sure how much attention you pay to game releases but there is a very real need to be worried about the final product. This year alone there have been tens of thousands of gamers that have felt cheated by their pre-ordered games as they have not been able to properly enjoy them weeks or even months after the release date.

    If you look at the last dozen or so performance articles we have written you will see exactly what I am talking about.
     
  20. red1776

    red1776 Omnipotent Ruler of the Universe Posts: 5,897   +88

    .....
    Exactly. the Crysis2 debacle is a perfect example of why the "worried about the final product" should be out front. I would submit that the later arriving DX11 patch and Hi res package was a direct and proportional response to the voicing of anger and disappointment over "the final product". In other words, if everyone had just shut up and been quiet, C2 would still be a ****** looking DX9 affair. You can bet other software companies were watching that situation closely to see what they could get away with. Like Steve said, look at the last dozen or so releases. Maybe all but two or three looked/played like they came out of the oven too soon.
     
  21. Hey Guys will i be able to run BF3 on my Alienware rig at the max settings..here is the config

    Intel Core i7 3.4GHz
    AMD RADEON HD 6870
    1GB Video Memory
    8GB RAM
    1TB HD, 7200 RPM
    875 Watt Power Supply
    Liquid Cooling
     


Add New Comment

TechSpot Members
Login or sign up for free,
it takes about 30 seconds.
You may also...
Get complete access to the TechSpot community. Join thousands of technology enthusiasts that contribute and share knowledge in our forum. Get a private inbox, upload your own photo gallery and more.