Battlefield 4 Benchmark – Multiplayer CPU and GPU W7 vs W8.1 Milos Milosevic November 4, 2013 Benchmarks The long wait is over, Battlefield 4 is here! (and our BF4 Benchmark that is) . DICE has managed to pull off another miracle with this title and delivered probably the best FPS game ever, with server crashes, bad net code and glitches that came with it. We must apologize for the delay, but there were more than a few issues that stopped us from delivering this benchmark on time. First of all after the launch DICE released a 677 MB update which killed the frames per second by around 5-10 on average, which forced us to redo the whole thing again although we finished half of the job. Also after the official launch for Europe it was impossible to enter any server with AMD cards. AMD reacted promptly and delivered a new driver which fixed those issues, but in all good conscience we had to repeat the benchmarks again since it was a new driver. However server crashing continued even after the new driver , but this wasn’t AMD’s fault it was EA and DICE’s and until now server crashing has not been fixed completely, although there have been at least 5 patches that took place. We could have just done Single Player benchmarks (which all of the other sites do) but we as Battlefield players know that SP benchmarks don’t mean squat when it comes to real world multiplayer performance, and lets face it people buy Battlefield because of the multiplayer, not singleplayer and we refused to go along with this meaningless trend that is being forced by other sites. But then again that is why we started HardwarePal. Since alot of readers asked why we didn’t include the i3 CPU in our BF4 Beta Benchmark we decided to go along with it (although honestly no one likes i3 CPU’s), since DICE recommends the GTX 660 and AMD 7870 on there pc requirements chart we added them as well. Win7 vs Win8.1 was one of the biggest questions in BF4 Beta, since we had reports of Windows 8 actually performing much better than Windows 7. At that point , as Windows 8.1 wasn’t officially released and since we had issues earlier with Windows 8 we decided to leave it out of the beta benchmark. Windows 8 brings more features that utilizes CPUs with multiple cores more effectively, especially for AMD’s FX CPUs but for Intel CPUs also to some point. Since Win 8.1 has been officially released we added one more category which doubled the amount of work, add to that 2 more GPU’s and one more CPU and get the idea of how enormous this task of actually completing the benchmark was. Compared to BF4 Beta this was almost 5 times more work for us in order to get the benchmarks out. All this in order to provide one of the most complete game benchmarks that we have ever seen and done. On their blog, DICE has acknowledged the issues we reported in BF4 Beta concerning the FX 6300 and FX 4300, we expected that they will fix the issues with these AMD CPUs but also expected that issues with the i7 CPU will be resolved as well. Just to remind you the i7 4770K in our beta benchmark was performing the same as a i5 4670k most of the time and sometimes even worse. High system memory usage was one more thing we expected to be fixed by launch, in Beta the system memory usage was around 9-10GB, 25% more than DICE recommended. With that in mind we decided to directly compare Bf4 Beta W7 performance with BF4 Official Release on W7 and W8.1 operating systems. Make sure to read all of the pages from this benchmark, lots of weird things will occur again so reading all the material is a must in order to understand what’s going on in Battlefield 4 now, when it comes to system requirements and performance. Short disclaimer: We tested the CPUs, GPUs and operating systems on the same map where we had already with our BF4 Beta benchmark (The Siege of Shanghai) at the exact same location in order to get the most accurate results when we compared data. The benchmark was done on an empty server, since when you do testing of this scale you need absolutely the most stable surroundings as possible as every little variable can affect the results. System memory usage was done on a full 64 man server. You might get better or even worse performance on similar setups (not by alot though), so take this benchmark only as general guideline when it comes to comparing performance of specific pieces of hardware on specific settings, and this is not the actual result that you might find in the game it self on a 64 man full server. PC Spec Requirements by DICE Battlefield 4 Benchmark Rigs and Components BF4 AMD Platform BF4 Intel Platform CPUFX 8350 @ 4GHz (4.2GHZ Turbo) - FX 6300 @ 3,5 GHz (4GHz Turbo) - FX 4300 @ 3.5 Ghz (4.1GHz Turbo)i7 4770k 3.5GHz (3.9GHz Turbo) - i5 4670k 3.4GHz (3.8GHz Turbo) - i3 3430 MOTHERBOARDAsus Sabertooth 990FX Rev. 2.0Asus Gryphon Z87 RAM2x 8GB@ 2133 Mhz Kingston HyperX Beast : KHX21C11T3K2/16XRam: 2x 8GB@ 2133 Mhz Kingston HyperX Beast : KHX21C11T3K2/16X SSDIntels 520 Series 240 GBIntels 520 Series 240 GB GPU7970GHZ Edition 3GB - 7870 GHZ Edition 2GB - AMD Catalyst 13.11 Beta v8 GTX770 4GB - GTX660 2GB - ForceWare 331.657970GHZ Edition 3GB - 7870 GHZ Edition 2GB - AMD Catalyst 13.11 Beta v8 GTX770 4GB - GTX660 2GB - ForceWare 331.65 MONITORShimian QH270-IPSMS 2560x1440p 27"Shimian QH270-IPSMS 2560x1440p 27" PSUThermaltake ToughPower XT 775WThermaltake ToughPower XT 775W OSWindows 7 64-bit Ultimate SP1 - Win 8.1 Pro x64 bitWindows 7 64-bit Ultimate SP1 , Windows 8.1 Pro x64 gingababobs Good thing I didn’t buy bf4 day 1. Having 1gb of vram sucks balls. Will probably get it when few more patches are out and i get a new gpu. Kyle Jackson Nvidia Maxwell next year, that’s what I’m waiting to upgrade for. carol argo its always the same issue on various system ! and it is always ignored or forgotten ! IRQ ! MSI ! MSIX ! not many hardware vendor bother setting up msix .oh they all support it , its a requirement . but not many bother setting it . and when its left to the os to decide , ms tend to go on the conservative side if it doesn’t compute something . Kajot important question for the reviewers. does it make a difference with gtx 770 to have a 4gb ver over 2gb? or is performance the same on these? Matt I used a 2gb 680 for the Beta and I had to set textures to High and MSAA to 2x for stable-ish frame rates (around 60fps). I could play it with the ultra presets but there was noticeable lag when looking around corners at speed… not sure if it was a beta or memory problem though. Milos M It depends on resolution of your monitor, if its 1920×1080 I guess you are safe with the 2GB version, if the resolution is higher, 2GB wont cut it. I presume the question was concerning BF4, right? kajot yes milos, its for 1440p bf4. i want to get 2 770sli, but i wonder if 2gb will hiccup when vram goes 2.5gb. I can get 4gb versions so that will be ok? i know performance wise they wont make much, but i’m afraid i will get end of vram hiccups. Milos M In that case buy one gtx 770 4GB and one GTX 770 2GB, no need for anothoer one to have 4GB as well. Although I would suggest that you buy just one GTX770 and then if you really need another one buy it. Multi GPu configurations are a pain in the ass when it comes to driver support and game optimization. Guest ^^^^^^Milos you did not just write that^^^^^^^ Vlasov_581 ^^^^^^^^^^^Milos you did not just write that^^^^^^^^^^ Milos M Thank you for the feedback, that’s what i get when i tell my to write messages for me. Edited. kajot I heard you need to mirror hardware. I mean that’s how it is in amd. When i crossfire, I need to use equal ram/gpu type. If I put same gpu different ram it will just downgrade the bigger one to mirror the smaller one and not use the other. Milos M You are right there has been a miscommunication between me and the staff that responds in the comments. Sorry for misleading you it wasn’t out of any bad intention. The post has been edited. Dennis Juchems An interesting benchmark, especially since you included the CPU and GPU usage. Regarding the 6300 being more powerful than the 8350 in this case, could you disable two cores of the 8350 and by that get a performance boost? Or doesn’t it work like that? Also, what’s your opinion on mantle? Will it put the combination of AMD CPU and card ahead or do you still expect that combining AMD with the 770 would be better than with the r9 280x? Milos M Yeah you can do that, you can lock the cores in BIOS and you got yourself a hexa core. When it comes to Mantle I am not sure at this point in time whether it will affect AMD CPUs it is all a speculation at this point in time and AMD hasnt said anything about CPUs, at least from what I know. When it comes cards at this point in time GTX770 is better but also more expensive (specially the 4GB version) if you are on the market for a new card, and its not the matter of life and death, you should wait for sales as we approach Christmas and New Year, by then you will know whether the Mantle works since we will be doing some extensive benchmarking as well. Dennis Juchems Thanks, Milos. My girlfriend and I are thinking about upgrading now to the end of November. Hopefully there’ll soon be some more information about Mantle, even if there won’t be any proper benchmarks yet by that time. From how I understood it so far I could imagine that AMD’s cards catch up with their nvidia counterparts, but obviously it’s not sure it’ll be that way. With our current budget it somewhat boils down to a choice of 6300 and 770 (4GB) or 8350 and r9 280x. Russell Collins “When it comes to Mantle I am not sure at this point in time whether it will affect AMD CPUs it is all a speculation at this point in time and AMD hasn’t said anything about CPUs, at least from what I know.” According to AMD using Mantle will lower the overhead of the CPU, thus increasing performance as it allows for the CPU to be focused on other things. Also in the Dice presentation they said that Mantle allows for perfect parallel rendering, ie using all 8 cores+. Dennis Juchems Do you think that will alleviate the problem of the cores not being used to their full potential? What’s your opinion on whether the AMD cards will come up to nvidia’s performance in multiplayer, especially combined with the AMD CPUs? Russell Collins I answered your question and gave a link supporting my claims but it has yet to be approved. Milos M My point was that I am not buying it till I see it and test it. AMD and DICE have been awfully quiet about the performance benefits that will come with Mantle and if you look back and remebr how Buldozer was marketed and at the end what did we get? AMD (with Crytek) has been complaining for a long time now that Direct X is limiting today’s GPU’s, Mantle has been announced as a substitute for DirectX and if I expect improvements in any field, that’s GPU’s, but again until I test it I cant say much about it. When it comes to CPUs I am skeptical, but we will see… Russell Collins I agree, proof is in the pudding so to speak. I don’t see how it’s possible to be any slower than DirectX or OpenGL though. Matt The stuttering issue will be worked out by dice for Windows 7… There is absolutely no reason to switch to Windows 8. Milos M Its not just about Windows 8,1 being stutter free, it does bring performance increase specially when you get a load of frames. Personally, I don’t like W8 either but I sense the transition is in the air. Usman Khan it’s not just battlefield alot of games run better in windows 8.1, most of old games have minor shutter in windows 7 on my pc but on windows 8.1 its silky smooth. WinSomeLoseNone How about direct X 11.1/2 and the added performance that comes with it? GTX 7XX cards are 11.1 compatible. R7/R9 cards are 11.2 compatible. Part of the performance boost for 8.1 is due to DX 11.1/2 which will never be coming to W7 as per MS. Archie Paras Do you know if Mantle is going to be implemented in the Frostbyte 3 engine so all games that use it will benefit from it or is Dice just having Battlefield 4 support it via the patch in December. If all games using Frostbyte 3 will support it the future that would be an amazing advantage for AMD. Bioware’s Mass Effect and Dragon Age Inquisition and DICE’s own Mirror’s Edge and Star Wars Battlefront are huge and having them support it will be incredible. Milos M DICE has announced that Mantle will be a part of Frostbite Engine 3.0 and that it will be implemented in all future games that they make with that engine including DA:Inquisition, Battlefront etc…. However we have yet to see mantle in Action. Hans BF4 never used more than 2GB vom RAM on my Win 8.1 64Bit System. I go 16GB installed. Its set to start the x64 Version. How did you manage to use over 6GB? Milos M This is MP benchmark not SP… If you are on an empty Large conquest map it eats around 4,5-5Gb in average Max Gockel One can fix the stuttering on windows 7 by disabling “core parking” Milos M It depends, core parking is a windows measure to save power. If you tweak power options in Windows to High Performance you really don’t need that. But for the sake of objectivity I have tried that Core Unparking thing and got 0.0% performance improvement under W7. Stutter continued as before. I see people are saying all over the internet that it helped them, we didnt see performance gain what so ever. EpitaxialOne Many people quote work like this to justify the erroneous position that multi-threading doesn’t work in gaming – and BF4 specifically – but don’t understand the context necessary to draw the right conclusions. A “benchmark” on an empty server will not provide tangible benefit at the CPU level relative to a fully-loaded server. I’m not saying misconception is the fault of the reviewer, but an even more explicit disclaimer is probably required to help the layman. Milos M Hyper threading works very well, you can see that with i3 performance but bad i7 performance has nothing to do with HT. We have tried benching on a full 64 player server but that is truly impossible. First of all, lots of people (the ones that want to skill you) do not have any understanding for the higher purpose when it comes to benching so, we skipped that after a few tries. When you die frame rate goes all over the place and the results are useless. From what I saw while gaming the performance in terms of average fps is pretty much the same, same thing goes for min fps but max fps varies from time to time leaving us with no clear conclusion. As we said this benchmark is for comparing CPUs only, with the current game optimization things are as you see them in benchmark, DICE might patch this but they have bigger things to think about now. Also we wanted to answer some questions: Is it worth switching to W8.1 and do cheaper CPUs bottleneck high end GPUs? We think we answered all of those questions. Tommaso Re Have you got any answer from AMD about their video cards misbehavior? Some drivers update solve this issue? Milos M No feedback from AMD, they did release a new driver that fixed performance in COD:Ghosts, but to be honest with you we didn’t really have time to test it, AMD in release notes hasnt mentioned anything about performance improvement in BF4, so we thought: “why bother?”. Jynn Milos I have seen other people reporting 60 – 80% cpu usage on all 8 cores of the FX 8350 and getting great performance from it in MP. So I’m rather confused by your results. Probably a stupid question but did you use the 64 bit executable for BF4? I’ve heard the 32 bit one will only support 4 cores which would explain your results if you were using it. georgieboy Milos, as someone looking to upgrade his CPU what would you recommend me go for a 6300? or a 4670k Bruh I know im late but 4670k is way better in general Dave4321 Does the 770 have 2 or 4 GB of ram? Nikolas Nikolaou On the Benchmark Rigs and Components page you can see the VRAM. It’s a 4GB. Dave4321 Thanks! would be interesting to see how a 2 GB 770 fares against a 3GB 7970. In other words, a faster card vs a card with more ram as they are also closer in price. Nikolas Nikolaou Well unless you are using settings that use more than 2GB VRAM (which would bottleneck a 2GB 770) the benchmarks should be identical as the horsepower is exactly the same. sikkkkka “1080/1200p 2GB card MIGHT do the job.” Seriously? I play bf4 with core i5-3570K, 1440p, 70 to 80 fps on ULTRA PRESET with GTX 770. David Miller Did you guys have hyperthreading on the CPU Nikolas Nikolaou Yes David hyperthreading was enabled. The benchmark was done before the 17 Update releases just so you know. Chandrian talking about total system ram usage, were the results the same both with x64 and x86 exe files?