Civ: Beyond Earth with Mantle aims to end multi-GPU microstuttering

The next installment in Sid Meier's Civilization series, Civilization: Beyond Earth, comes out tomorrow. The folks at AMD have been working with its developer, Firaxis, to optimize the game for Radeon graphics cards. Most notably, Firaxis and AMD have ported the game to work with AMD"s lightweight Mantle graphics API.

Predictably, AMD and Firaxis report that Mantle lowers the game's CPU overhead, allowing Beyond Earth to play smoother and deliver higher frame rates on many systems. They've even provided a nice bar graph with average FPS showing AMD in the lead, like so:

That's all well and good, I suppose (although *ahem* the R9 290X they used has 8GB of RAM). But average FPS numbers won't tell you about gameplay smoothness or responsiveness. What's more interesting is how AMD and Firaxis have tackled the thorny problem of multi-GPU rendering in Beyond Earth.

Both CrossFire and SLI, the multi-GPU schemes from AMD and Nvidia, handle the vast majority of today's games by divvying up frames between GPUs in interleaved fashion. Frame one goes to GPU one, frame two to GPU two, frame three back to GPU one, and so on. This technique is known as alternate-frame rendering (AFR). AFR does a nice job of dividing the workload between GPUs so that everything scales well for the benchmarks. Both triangle throughput and pixel processing benefit from giving each GPU its own frame.

Unfortunately, AFR doesn't always do as good a job of improving the user experience as it does of improving—or perhaps inflating— average FPS scores. The timing of frames processed on different GPUs can go out of sync, causing a phenomenon known as multi-GPU micro-stuttering. We've chronicled this problem in our initial FCAT article and, most extensively, in our epic Radeon HD 7990 review. AMD has attempted to fix this problem by pacing the delivery of frames to the display, much as Nvidia has done for years with its frame metering tech. But frame pacing is imperfect and, depending on how a game's internal simulation timing works, may lead to perfectly spaced frames that contain out-of-sync visuals.

Making AFR work well is a Hard Problem. It's further complicated by variable display refresh schemes like G-Sync and FreeSync that attempt to paint a new frame on the screen as soon as it's ready. Pacing those frames could be a hot mess.

In a similar vein, virtual reality headsets like the Oculus Rift are extremely sensitive to input lag, the delay between when a user's head turns and when a visual response shows up on the headset's display. If that process takes too long, the user may get vertigo and go all a-chunder. Inserting a rendering scheme like AFR with frame metering into the middle of that feedback loop is a bad proposition. Frame metering intentionally adds latency to some frames in order to smooth out delivery, and AFR itself requires deeper queuing of frames, which also adds latency.

At the end of the day, this collection of problems has conspired to make AFR—and multi-GPU schemes in general—look pretty shaky. AFR is fragile, requires tuning and driver support for each and every game, and doesn't always deliver the experience that its FPS results seem to promise. AMD and Nvidia have worked hard to keep CrossFire and SLI working well for their users, but we at TR only recommend buying multi-GPU solutions when no single GPU is fast enough for your purposes.

Happily, game developers and the GPU companies seem to be considering other approaches to delivering an improved experience with multi-GPU solutions, even if they don't over-inflate FPS averages. Nvidia vaguely hinted at a change of approach during its GeForce GTX 970 and 980 launch when talking about VR Direct, its collection of features aimed at the Oculus Rift and similar devices. Now, AMD and Firaxis have gone one better, throwing out AFR and implementing split-frame rendering (SFR) instead in the Mantle version of Beyond Earth.

AMD provided us with an explanation of their approach that's worth reading in its entirety, so here it is:

With a traditional graphics API, multi-GPU arrays like AMD CrossFire™ are typically utilized with a rendering method called "alternate-frame rendering" (AFR). AFR renders odd frames on the first GPU, and even frames on the second GPU. Parallelizing a game's workload across two GPUs working in tandem has obvious performance benefits.

As AFR requires frames to be rendered in advance, this approach can occasionally suffer from some issues:

·         Large queue depths can reduce the responsiveness of the user's mouse input

·         The game's design might not accommodate a queue sufficient for good mGPU scaling

·         Predicted frames in the queue may not be useful to the current state of the user’s movement or camera

Thankfully, AFR is not the only approach to multi-GPU. Mantle empowers game developers with full control of a multi-GPU array and the ability to create or implement unique mGPU solutions that fit the needs of the game engine. In Civilization: Beyond Earth, Firaxis designed a "split-frame rendering" (SFR) subsystem. SFR divides each frame of a scene into proportional sections, and assigns a rendering slice to each GPU in AMD CrossFire™ configuration. The "master" GPU quickly receives the work of each GPU and composites the final scene for the user to see on his or her monitor.

If you don’t see 70-100% GPU scaling, that is working as intended, according to Firaxis. Civilization: Beyond Earth’s GPU-oriented workloads are not as demanding as other recent PC titles. However, Beyond Earth’s design generates a considerable amount of work in the producer thread. The producer thread tracks API calls from the game and lines them up, through the CPU, for the GPU's consumer thread to do graphics work. This producer thread vs. consumer thread workload balance is what establishes Civilization as a CPU-sensitive title (vs. a GPU-sensitive one).

Because the game emphasizes CPU performance, the rendering workloads may not fully utilize the capacity of a high-end GPU. In essence, there is no work leftover for the second GPU. However, in cases where the GPU workload is high and a frame might take a while to render (affecting user input latency), the decision to use SFR cuts input latency in half, because there is no long AFR queue to work through. The queue is essentially one frame, each GPU handling a half. This will keep the game smooth and responsive, emphasizing playability, vs. raw frame rates.

Let me provide an example. Let's say a frame takes 60 milliseconds to render, and you have an AFR queue depth of two frames. That means the user will experience 120ms of lag between the time they move the map and that movement is reflected on-screen. Firaxis' decision to use SFR halves the queue down to one frame, reducing the input latency to 60ms. And because each GPU is working on half the frame, the queue is reduced by half again to just 30ms.

In this way the game will feel very smooth and responsive, because raw frame-rate scaling was not the goal of this title. Smooth, playable performance was the goal. This is one of the unique approaches to mGPU that AMD has been extolling in the era of Mantle and other similar APIs.

All I can say is: thank goodness. Let's hope we see more of this kind of thing from AMD and major game studios in the coming months and years. Multi-GPU solutions don't have to double their FPS averages in order to achieve smoother animations or improved responsiveness. I'd much rather see a multi-GPU team producing more modest increases that the user can actually feel and experience.

Of course, while we're at it, I'll note that if you measure frame times instead of FPS averages, you can more often capture the true improvement offered by mGPU solutions. AMD has been a little slower than Nvidia to adopt a frame-time-sensitive approach to testing, but it's clearly a better way to quantify the benefits of this sort of work.

Fortunately, AMD and Firaxis have built tools into Beyond Earth to capture frame times. I have been working on other things behind the scenes this week and haven't yet had the time to make use of these tools, but I'm pleased to see them there. You can bet they'll figure prominently into our future GPU articles and reviews.

Comments closed
    • BIF
    • 5 years ago

    Dammit Scott, as soon as I read this article, I marched right off and bought the game!

    Did you hypnotize me or something?

    Haha, well…looking forward to playing it sometime this weekend… 🙂

    • brucethemoose
    • 5 years ago

    Turn times are much longer in DX11 compared to Mantle, even on my OC’d 4670k. In a game like Civ:BE, that’s arguably more important than small differences in frame times.

      • Klimax
      • 5 years ago

      sleep(10)

    • Krogoth
    • 5 years ago

    Interesting article, but unfortunately modern strategy games are typically CPU-bound (clockspeed) than anything else at the late-game stages.

    You don’t even need super-fast reflexes to be effective at RTS games. It is pointless for turn-based stuff (unless you are playing with a time limit).

      • tipoo
      • 5 years ago

      Microstutter isn’t just about “super-fast reflexes”, it’s a visual annoyance.

        • Krogoth
        • 5 years ago

        Why are you using SLI/CF with a game that doesn’t benefit from it and it actually hurts performance(CPU overhead from SLI/CF)?

        Microstuttering is the well understood cost of performing AFR. The problem is mostly masked when you multi-GPU solution can render a high enough framerate (80FPS). It becomes noticeable when your multi-GPU solution is only able to handle 60FPS or less and it becomes painfully obvious with 30FPS or less. CF/SLI users have known all of this for years.

        GPU marketing guys love ARF because it allow GPU solutions to pump out larger numbers in games and synthetic benchmarks. It happens to be easier path to code for as well.

        It has been the GPU industry’s dirty little secret for almost a decade. It is only becoming more noticeable to the masses as reviewers are paying more attention to “framerate smoothness” and graphing framerates over time.

      • Prestige Worldwide
      • 5 years ago

      Underestimate the power of uber micro at your own risk.

      [url<]https://www.youtube.com/watch?v=ShOusaHBSdI[/url<]

        • Krogoth
        • 5 years ago

        Micromanagement only works well in a few RTS games at higher levels of gameplay (C&C and Bilzzard-style stuff). It is typically with specialist/caster units.

        Macro-management is a far more important factor in all RTS games.

          • dragontamer5788
          • 5 years ago

          [quote<]Micromanagement only works well in a few RTS games at higher levels of gameplay (C&C and Bilzzard-style stuff). It is typically with specialist/caster units.[/quote<] Right... [url<]http://www.youtube.com/watch?v=WJp0t9n8DWk[/url<] So many "caster" units in this crazy micro-fest. </sarcasm>

      • BIF
      • 5 years ago

      So I’ve been playing Beyond Earth now for “a few” (ahem) hours. I don’t have Mantle, but I have all settings as high as they’ll go. Looks fine.

      When I’m gaming, I usually set folding to “light” which stops the GPU slot (wait for idle) and limits the CPU (12 threads) to something around 75% utilization, leaving about 25% headroom for OS and other work.

      I’m up to about turn 300 and game play is beginning to noticeably slow down. I may soon have to stop folding entirely to see if Civ:BE can really benefit from a greater share of my 3930.

    • Klimax
    • 5 years ago

    [url<]http://www.pcper.com/reviews/Graphics-Cards/Civilization-Beyond-Earth-Performance-Maxwell-vs-Hawaii-DX11-vs-Mantle/2560x1[/url<] PCPer shows AMD already skipping DX11 code path for Mantle using games. Very bad idea, because DX11 is for them important fallback path when devs don't have code ready for new cards. (If they will bother to begin with) 285 should have warned them, but apparently they failed to listen. They are about to prove once more why API like Mantle are bad idea.

    • Theolendras
    • 5 years ago

    That’s great news ! Somewhat ironic to see this rendering technique introduced by a game that is pretty much unaffected by latency at all… I mean this isn’t a real-time game, so input lag would have to be extremely bad to be an issue.

    Anyway good job Firaxis/AMD !

    • Ninjitsu
    • 5 years ago

    AnandTech wasn’t seeing any additional benefit from CF with Mantle, relative to a single card.

    AT also mentions that the AMD slide (also presented in this post by Scott) pre-dates Nvidia’s 344.48 drivers, which included support for the game.

    [quote<] The story with CrossFire is a bit different this time. We mentioned that minimum frame rates didn't tend to be too inconsistent – that there wasn't a lot of "jitter" – but that's not true when we talk about CrossFire. Average frame rates are much higher, but the reported minimums tend to be quite a bit lower than our single GPU results. [b<]Mantle on the other hand "fixes" that problem, inasmuch as delivering performance essentially identical to a single GPU with Mantle is a "fix".[/b<] [/quote<] Update: AT's managed a fix, also has frame times: [url<]http://www.anandtech.com/show/8643/civilization-beyond-earth-crossfire-with-mantle-sfr-not-actually-broken[/url<]

      • Damage
      • 5 years ago

      Yeah, after I wrote what I did yesterday, I decided to post an update with performance results from Mantle with CrossFire. Unfortunately, despite everything that AMD said, I am seeing zero impact from enabling a second GPU. The frame times are identical:

      [url<]https://dl.dropboxusercontent.com/u/68992780/290x-mantle.gif[/url<] That ain't right. I am looking into it, but it looks like Mantle w/SFR in Civ:BE is just broken right now.

        • ptsant
        • 5 years ago

        Although I imagine you have already done that, I would like to point out that the Mantle w/SFR mode requires modification of an .ini file.

        See for example:
        [url<]http://www.anandtech.com/show/8643/civilization-beyond-earth-crossfire-with-mantle-sfr-not-actually-broken[/url<]

          • Damage
          • 5 years ago

          Yeah, I’ve been talking w/AMD and with Jarred. I set the INI, but it didn’t help.

          • Ninjitsu
          • 5 years ago

          Yup, just saw this and came to edit.

    • Chrispy_
    • 5 years ago

    Shame on Jarred Walton at Anandtech:

    [quote<]Firaxis claims that they're working on reducing input latency rather than just improving raw frame rates, but testing for latency reductions can be difficult and in a game like Civilization it's questionable whether it will really matter much. More to the point, I'm not sure how you would get slightly lower FPS but still improve latency, so most likely the SFR rendering needs a bit more TLC.[/quote<] It's this sort of mentality that allowed choppy gameplay through falsely inflated framerates in the first place. We don't want over 9000 frames a second, we want 60-144fps at most, with the lowest amount of lag and stutter possible, please.

    • Klimax
    • 5 years ago

    Just simple yet important question: Does this Civilization still use Driver Command Lists? (As it was used by C5?)

    • LordVTP
    • 5 years ago

    The endgame for multi-gpu might be a “distributed die” approach, where thru silicon vias and silicon photonics combine to enable direct gpu die to gpu die access. Resources on slave gpu’s would be used as proper cores augmenting the effective counts and being assigned work threads via the master gpu with the results mapped back to it, rather then trying to render ahead or split frame rendering. Though, perhaps a much finer tiled render approach might yield better results (256x blocks of screen space evenly distributed between cores).

      • willmore
      • 5 years ago

      Bingo!

      Wait, we’re not playing buzzword bingo? Drat!

    • Wildchild
    • 5 years ago

    To me, it’s not even really about FPS. It’s about how long it takes the CPU to calculate turns for the AI. On huge maps with lots of other civilizationss and city-states, the turns only can take anywhere from 1 to 2 minutes with my 2500K.

    It would be nice if they could benchmark this to see if there was any difference.

      • UnfriendlyFire
      • 5 years ago

      On mobile dual-core chips… Good luck getting reasonable turn times when you’re playing on huge map with all 22 civs.

      • Ninjitsu
      • 5 years ago

      I played about 100 hours of TW: Rome II, with the game running at between 15 to 30 fps.

      Trust me, it’s about the FPS too. And i used to wait almost 5 minutes for Rome II to calculate one turn.

    • christos_thski
    • 5 years ago

    I’ve read developers practically rave about the performance benefits of DX12, but Mantle, which is supposedly built upon similar premises, fails to impress each and every time. What’s up with that?

    PS: Speaking about single gpu performance only, as sli is an irrelevant marginal choice even amongst the hardcore enthusiasts .

      • willmore
      • 5 years ago

      Are you a registerd shill?

        • christos_thski
        • 5 years ago

        fanboy, much?

          • willmore
          • 5 years ago

          Did I express an opinion towards or against any side in this discussion? No. Learn what a fanboy is before throwing around words you don’t know the meaning of.

          Try to stick with facts. Any facts. Any at all.

            • christos_thski
            • 5 years ago

            I’ve called you a fanboy because you jumped to the conclusion I was “shilling” for something or other just because I called multigpu setups a marginal choice. How did you come to that conclusion?

            As for facts, I base my opinion on multi-gpu setups on this very site’s recent poll (a hardware enthusiast site, if there ever was one) ; even amongst Techreport’s registered users (a specifically enthusiast crowd) multi-gpu setup owners were a small minority.

            Ever since the upvoting/downvoting system was boosted to 11, people are quick to bite each other’s head over here, the atmosphere is nearly toxic. How about calming down and telling me where I’m wrong, instead of accusing me of “shilling” and acting all insulted when I return the favour?

            • anubis44
            • 5 years ago

            No, your primary point was that somehow Mantle isn’t getting support from developers (an unsupported and likely false assertion) which is a (not so) veiled jab at AMD, hence the suspicion that you may be an nVidia shill.

            • willmore
            • 5 years ago

            Score one cookie for anubis44. That’s exactly why.

            • christos_thski
            • 5 years ago

            It is getting support from developers, but the benefits seem to be marginal. Is that a questionable assertion too? Does it make me a paid nVidia shill?

            If that makes the fanboys sleep better at night, I’m perfectly happy to “concede” that Radeon GPUs are perfectly competitive in and of their own, considering I use one on my main rig (a 7870).

            The main reason I referred to this was because I’d read several developer comments that made a big deal out of DX12’s optimizations, and I was wondering whether Mantle’s limited performance boost was due to inefficient optimization of similar functionality, or it is as far as these kinds of optimizations will get you (ie we don’t expect anything significantly better from DX12).

            sorry to have stepped on your pathetic brand loyalty sensivities by asking…. won’t do it again.

            • LostCat
            • 5 years ago

            I’ll happily ignore the catfight – The benefits may be marginal with a shiny new $300 CPU, but with many other CPUs they increase depending on how much the game is doing with that CPU. Also good for laptop builds which probably trend cheaper than desktops.

            I don’t really care anymore. My budget has been strained so much I’m moving to consoles for now.

            • willmore
            • 5 years ago

            I buy both AMD and nVidia cards depending on which is best for any specific use I may have, so your fanboy assertion is baseless.

            I think the improvements we have seen from early adopters of Mantle have shown a meaningful benefit. We are quite early in developers using Mantle and I wouldn’t expect it to have shown its full impact just yet.

            The question as to DX12 and any possible performance improvements it may some day offer is interesting. Will anything that benefits from Mantle also benefit from DX12? Will the DX12 implementation benefit from the work the developers of that engine have already put into adapting to Mantle?

            We don’t know. Until more is known about DX12 (and Mantle for that matter), there isn’t much we can say for sure. We can speculate, but I don’t see much point to that.

            If you want to discuss these issues, that would be fine, but if you start the conversation by making false assertions and rude characterizations of anyone who might disagree with you, don’t expect civil discourse. Expect to be thumbed down and ignored like any other troll/flamebater.

      • LordVTP
      • 5 years ago

      Any single GPU build, regardless of whether it’s a flagship card, is an entry level build.

        • Airmantharp
        • 5 years ago

        No. Even if you add ‘for gaming’ on the end.

        • Waco
        • 5 years ago

        Wrong.

        • Chrispy_
        • 5 years ago

        LOLWHUT?

        • Ninjitsu
        • 5 years ago

        So… GTX 450 SLI > GTX 980?

        • Theolendras
        • 5 years ago

        Tell that to the ones who bought the Titan Z !

    • drfish
    • 5 years ago

    Not trying to start a war, but meanwhile over on [url=http://anandtech.com/show/8640/benchmarked-civilization-beyond-earth<]Anandtech[/url<]... [quote<]Firaxis claims that they're working on reducing input latency rather than just improving raw frame rates, but testing for latency reductions can be difficult and in a game like Civilization it's questionable whether it will really matter much.[/quote<] The ignorance is staggering... [i<][url=http://www.pcper.com/reviews/Graphics-Cards/Civilization-Beyond-Earth-Performance-Maxwell-vs-Hawaii-DX11-vs-Mantle<]PC Perspective[/url<] gets it.[/i<]

      • Damage
      • 5 years ago

      I’d say he’s right on both counts. We can test for frame rendering times with Fraps and even frame delivery times with FCAT, but we can’t test the entire input-output loop with either tool.

      You’d need to point a high-speed camera at the screen with a mouse overlaid. Kind of a PITA, though not impossible.

      And… whether you get 20 ms or 30 ms of input lag on a game like this one, which isn’t exactly a twitch shooter, isn’t gonna matter a lot.

      I think low overall input latency and the other benefits of SFR have much to recommend them, but it’s hard to make a case that the Civilization series will be their ultimate use. 🙂

        • drfish
        • 5 years ago

        I guess I was less than impressed that he seemed dismissive of the effort instead of appreciating the progress even if it might be slightly misplaced.

        • Flapdrol
        • 5 years ago

        blurbusters guy did some latency testing, he hotwired an led to his mouse, then filmed the game with a high speed camera, then looking at the nr of frames between the led on and the shot to show.
        [url<]http://www.blurbusters.com/gsync/preview2/[/url<]

      • kuraegomon
      • 5 years ago

      That Anandtech review is basically worthless without frame time analysis. The PCPer review is signifcantly stronger, as you’d expect from Ryan S.

    • CrazyElf
    • 5 years ago

    Out of curiosity, does anyone know if split frame rendering works for 3 or 4 GPU setups?

    I know it is available now for 2 GPU setups (and imo, whenever possible should be favored over AFR which is problem plagued), but is it possible to support it where each GPU renders 1/3 or 1/4 of the screen?

    Like the article, personally I’d rather see more modest gains per GPU, in exchange for the kind of frame times that single GPU solutions have.

    I just wish that the frame rate hype was not so big and that we looked at things like frame time more. To me, what matters most is minimum frame rates – that’s the most demanding scene in the game (ex: late game in an RTS with the most units out).

    Edit:
    Perhaps the best solution is the one that offers the “lowest maximum frame latency”. As in, the lowest latency at the 99th percentile of frame.

      • Airmantharp
      • 5 years ago

      Nvidia maintains two ‘golden finger’ SLI connectors on their higher-end products, which facilitates SLI connections for three and four card setups, so it’s more possible there; but even then, given how little PCIe bandwidth is usually needed for GPUs, AMD’s over the bus setup for Crossfire shouldn’t really have a problem with it either.

      • Pwnstar
      • 5 years ago

      [quote<]anyone know if split frame rendering works for 3 or 4 GPU setups[/quote<] It does. Each GPU gets a proportional part of the frame to render then transfers it to the first GPU for display.

    • sschaem
    • 5 years ago

    Considering the R9-290x is now $299 (with Alien & star citizen)
    [url<]http://www.newegg.com/Product/Product.aspx?Item=N82E16814150696&cm_re=r9-290x-_-14-150-696-_-Product[/url<] the performance chart plot would look ridiculous if you include games like Sniper elite3 and shadow of mordor. Almost $250 cheaper, and even faster in newer titles... (And we have yet to see the Audio DSP cores in action) (compared to the $550 GTX 980, cheapest model on newegg)

      • Pwnstar
      • 5 years ago

      All true, but this is the 8GB version, so it is a bit more expensive. Still a lot cheaper than nVidia though, if you only played those games.

    • rechicero
    • 5 years ago

    The funny thing about all this is when we remember the origin of SLI. No, it wasn’t Scalable Link Interface. It was Scan Line Interleave and what it did (in 1998) was similar to this “new” approach (both are SFR).

    NVidia chose the other approach, even when they had the IP from the original SLI, because of the FPS craze. AFR was worse for the users, but better for the marketing department.

    Now we can say: Thanks TechReport and the rest of people that took the time to actually look for more than mere numbers. This SFR revival wouldn’t be without your work.

    [b<]Thanks[/b<]

      • Thrashdog
      • 5 years ago

      My recollection from that initial re-launch of SLI was that AFR and SFR modes were supported, and that NVidia’s driver would select one based on game profiles. I haven’t paid too much attention to multi-GPU in the years since, but has NVidia’s SFR approach been set aside since?

        • Flapdrol
        • 5 years ago

        You can still select sfr in nvidia inspector afaik, saw someone on the arma forums claim it fixed the render to texture stuff. Scaling was pretty bad though, only 30% fps or something.

      • Klyith
      • 5 years ago

      Scan line alternation IIRC doesn’t work at all with modern rendering because so many things use information from more than one pixel to make the final output. I think that by the time nvidia acquired 3dfx that writing was already on the wall.

      The main reason most SLI defaults to AFR is not just FPS, but because that method requires the least amount of effort to implement in the game engine. SLI users are such a tiny percent of the market, spending a bunch of developer time to improve their experience is pretty marginal compared to improving the game for anyone else. AMD surely threw Firaxis a chunk of money / co-marketing to get the game on Mantle + support their tech. (Maybe SFR is also easier to do through mantle?)

        • WaltC
        • 5 years ago

        [quote<]AMD surely threw Firaxis a chunk of money / co-marketing to get the game on Mantle + support their tech. (Maybe SFR is also easier to do through mantle?)[/quote<] I wish people would stop saying things like this off the cuff without references or corroborating links. It's been stated repeatedly by AMD that use of Mantle is a developer prerogative as opposed to some kind of AMD paid-for service. What doesn't seem generally understood at all, for some reason, is the idea that developers might like to [i<]make use of something useful[/i<] when developing a game, especially when that something useful is provided to them gratis. I mean, d'you think that Microsoft has to throw money at game developers to get them to use D3d?....;) Hardly...it's all the tools and free development help Microsoft provides that keeps D3d on top--it's no accident. There would be little to no point in Mantle at all if AMD wasn't willing to do the same because Mantle use would go nowhere fast. It's obvious that Firaxis saw something useful that it liked in Mantle, and something it thought worth implementing--as has been true with many developers. I don't know if you've looked at AMD's numbers lately, but the company cannot afford to lavish large sums for the purpose of pushing its own custom 3d API, [i<]in addition[/i<] to what it spends developing Mantle and the high-level Mantle tools developers like--as I said, as rich as Microsoft is, not even Microsoft does that--doesn't have to. I'll wager that most of what AMD is doing at this point is lighting a fire underneath Microsoft's rear end to get the company to start developing the D3d API again in earnest...Microsoft has sort of languished since D3d10, if you haven't noticed. It's no accident that Microsoft's D3d 12 announcement came post AMD's Mantle announcement, and at a time when people were wondering if there'd ever be a D3d 12. . I'm absolutely sure that at the point at which Microsoft gets back up to speed with respect to AMD GPU support inside D3d that we'll see AMD much less interested in its own Windows API. I'm waiting for Scott's evaluation, of course, but in the meantime I thought this was a worthy read: [url<]http://www.pcper.com/reviews/Graphics-Cards/Civilization-Beyond-Earth-Performance-Maxwell-vs-Hawaii-DX11-vs-Mantle[/url<] And, yes, it involves bringing Crossfire into the API itself so that games can directly support it.

      • willmore
      • 5 years ago

      Edited because reading is fundimental.

      • sschaem
      • 5 years ago

      AFR come out of necessity with modern game engine that leverage heavily render targets.
      Mainly because AFR requires very little cross GPU synchronization.

      I personally think that sli / crossfire to this day is borked, its broken by design.

      And I think if AMD/nvidia fixed the problem, it might be some real “secret sauce” full of ‘patents’.
      Because under the hood, the idea of driver managed SLI/Cross fire is a complete nightmare.

      • jihadjoe
      • 5 years ago

      With GPUs turning more and more into compute cards, I imagine the next step in mGPU will be a pure compute solution, with the drivers utilizing all cards to solve shaders, and then dispatching compute results to a second pool consisting of the render resources. No more problems with scaling, and/or stuttering.

      • HisDivineOrder
      • 5 years ago

      Impossible to do what the original SLI did in terms of alternating horizontal lines. It just won’t work with today’s code.

      So complaining (again) that nVidia bought up 3dfx’s assets without buying up their hardware or hardware support, well, that’s just sour grapes, ain’t it? I remember reading the wailing and the tears and the gnashing of teeth as thousands upon thousands of 3dfx owners were left in the cold by 3dfx’s mismanagement and bankruptcy (and I did laugh, I admit, because the writing had been on the wall much as it’s on AMD’s wall now), but that moment made thousands upon thousands of anti-nVidia fanboys that continue to plague the internet to this day.

      The funny thing is how stupid it all is. I bought 3dfx back in the day when it was the best. I bought nVidia and AMD in alternating patterns until AMD stopped competing. Then I bought only nVidia. Same as I did with Intel and AMD when they were alternating in terms of awesome until one day–surprise, surprise–AMD stopped competing there.

      Don’t be a fanboy. Just buy what’s best at any given time. I know it’s hard to support a company that left you crying and wailing on some 3dfx forum way back when, but it’s been a long time.

      SLI was reinvented because AFR worked across more games with larger gains and with greater ease than SFR. If they had pursued SFR, you would have seen far fewer games supported and with less support would have meant less people investing in it, which would have meant even less support until there’d have been none at all.

      At least nVidia gave you the OPTION to try SFR if you liked and if the game would support it. Most didn’t. Not because they were all chasing highest FPS, but mostly because today’s tech just doesn’t work with it very well.

      Not without the developer going in and MAKING it work, which is rare. That’s what Firaxis did here. They could do it with DirectX, too, but AMD paid them the big bucks and/or development support, so they’re doing it with Mantle instead.

      Let’s all hope this is them setting up to support DX12’s low level access in a similar fashion when it arrives next year, right? Otherwise, we’re see a lot of effort being tied into an AMD-only, certain GCN GPU’s-only (especially if GCN 1.2 vs 1.0 incompatibilities are anything to go by) initiative that’ll be a blip on a timeline of custom API support, which won’t be nearly as useful or interesting.

      But you still gnashing your teeth about what SLI used to mean versus what it means today sounds like an old 3dfx fanboy still waging war like an old land mine waiting to blow some poor sod’s foot off after the wars that birthed it are long over.

    • anotherengineer
    • 5 years ago

    Out of curiosity, what game engine does civ 5 use?

      • Narishma
      • 5 years ago

      Probably a custom one.

    • Bauxite
    • 5 years ago

    With Occulus type VR, I see the opportunity for “synchronized multi-frame rendering” or whatever it ends up being called.

    Basically GPU1 for one eye, GPU 2 for the other. All things being equal, the perspectives will have roughly the same rendering time per frame as they are fairly close together.

    The minor differences can be budgeted for, as in max fps goes down [i<]slightly[/i<] to guarantee both eyes are delivered a frame with each refresh, and at the same time. You can have a fallback for the extreme edge cases where one eye is looking around a corner while the other only sees the wall or one above water when the other is not.

      • UnfriendlyFire
      • 5 years ago

      I wonder what would a vertical screen tear look like in VR?

      • LordVTP
      • 5 years ago

      I actually talked about it to some Nvidia driver team folks back when DK1 shipped (got mine on launch day, mine was in the two digit unit # range). Your exactly right about the render delta being much more predictable and manageable with split frame/split eye rendering there.

      Mentioned it here – [url<]https://developer.oculusvr.com/forums/viewtopic.php?f=20&t=331&p=95779&hilit=lordvtp#p8070[/url<] "I plan to drop a line to some contacts I have with the nvidia driver team to see if they are interested in making HMD specific tweaks to the 3dvision system...In addition to checking viability about a new 2 gpu side by side split frame sli rendering mode specific to low latency VR usage. :D"

    • internetsandman
    • 5 years ago

    If Civ V is anything to go by, this will the only game for which my CPU is actually the bottleneck. I don’t know how well Civ BE will handle multithreading, but if the old performance reviews that incorporated Civ V are anything to go by, I’m sorely tempted to upgrade to an LGA 2011 6-core CPU…

      • dWind
      • 5 years ago

      LOL why? unless you already have a super low hz lga2011 cpu there is no point.
      Civ5 like all other games benefit most from MHZ boost. Unleass you are looking to buy into the top tier lga2011 cpu, get a 1150 i5-4***k and OC it. IIRC civ5 benchmarks showed a tenuous 4 core support – as in 2core+HT cpu with higher Hz out-perform a lower clocked quad. And since civBE uses civ5 engine i doubt it will change.
      This is not surprising since while, many games developers separate major distinct functional parts like graphics, sound and network handling to separate threads, they leave core game mechanics/physics/AI blob handled by 1 thread.
      IIRC only GalCiv3 devs are going to separate each AI player to its own thread…. Well i guess new console game devs are/will be forced to do it with their wonderful 1.6ghz octcores …but nothing in PC scene (not that I pretend to know all ofcourse 🙂 feel free to enlighten me )

      • Pwnstar
      • 5 years ago

      Civilization is notorious for being CPU heavy.

      But it’s not the only game to be so. You must only be comparing games you’ve played.

      • Firestarter
      • 5 years ago

      Planetside 2 is also CPU bottlenecked for anyone with a decent GPU. Mantle or DX12 would help immensely as the biggest problem is its single DX9 rendering thread

        • LostCat
        • 5 years ago

        Planetside 2 is scheduled to get DX11 finally, at least.

        At that point I plan to finally play it.

          • Firestarter
          • 5 years ago

          Planetside 2 was/is scheduled to get a lot of things, that doesn’t mean it will happen

            • LostCat
            • 5 years ago

            No, but I’m assuming it will. Sony and PC gaming have a poor relationship at best, but they keep trying.

    • anotherengineer
    • 5 years ago

    What about counter strike??

    I want 1200 fps!!!!

    • chuckula
    • 5 years ago

    As for the game itself…
    1. Linux support… w00t.
    2. There better be terraforming support!

      • willmore
      • 5 years ago

      I just fired up the HTPC and updatred the OS and Steam. Lots more games there, now. Woo, hoo!! BL2, L4D2, Civ5, Oil Rush, most of HL2(?) And my seven year old son’s favorite “Goat Simulator”. Seems there is something so funny about that game that it renders him into a giggling little ball rolling about the floor. In additional news, the cat hates that game.

    • chuckula
    • 5 years ago

    [quote<]The folks at AMD have been working with its developer, Firaxis, to optimize the game for Radeon graphics cards. Most notably, Firaxis and AMD have ported the game to work with AMD"s lightweight Mantle graphics API.[/quote<] That's wonderful and all but... [url<]http://anandtech.com/show/8640/benchmarked-civilization-beyond-earth[/url<] [quote<]As for Mantle, it remains an interesting option if you have an AMD card, but so far we haven't really seen a huge benefit. [/quote<]

      • Alexko
      • 5 years ago

      AMD showed a benchmark at 4K with MSAA 8X and an 8GB 290X; Anandtech benched at 4X with a 4GB 290X.

      Since the GTX 980 has a 256-bit bus, you’d expect it to take a relatively larger hit than the 290X, especially if the latter is in an 8GB configuration and doesn’t have to worry about running out of VRAM. So AMD’s benchmark looks plausible to me.

      Of course, AMD’s settings were cherry-picked, but in Anandtech’s test the 290X still performs quite well.

      • Pwnstar
      • 5 years ago

      The benefit is the AMD card is $150 cheaper for the same performance in this game.

      • jensend
      • 5 years ago

      Yeah, the minimum frame rate increased by only 53%. Totally pathetic and never worth using. We all know the average frame rate is the only thing that matters.

        • soryuuha
        • 5 years ago

        “We all know the average frame rate is the only thing that matters.”

        not sure if trolling or..

          • jensend
          • 5 years ago

          Not sure if totally oblivious to over-the-top sarcasm or. . .

          The point was, though the reviewer was unimpressed because they’re focused on average FPS, chuckie’s link simply disproved the point chuckie was trying to make.

        • Klimax
        • 5 years ago

        Interestingly, only 980/780 and 290x(Mantle/DX), no other card saw such improvement. Which is quite an anomaly.

        Can’t say what is exactly going there, but my initial guess based on rest of results is that ROP/bandwidth is key. (In Maxwell case it is architectural upgrade, in Radeon case it would be engine optimization for particular card) Question is, why no other card got there.

          • LostCat
          • 5 years ago

          They didn’t test any other GCN 1.1 parts, where Mantle has historically done better.

            • Klimax
            • 5 years ago

            Well, 285 could be interesting case. Would it work propel or would it be same as with BF4 or Thief… (New compression, different trade offs between bandwidth/shaders/ROPs/…)

            It more or less requires separate optimization path then rest of line up.

            • kalelovil
            • 5 years ago

            AMD’s provided Mantle benchmarks show the R9 285 performing above the R9 280, so presumably the drivers/Civ: BE have made accommodations for its architecture.
            [url<]http://community.amd.com/community/amd-blogs/amd-gaming/blog/2014/10/23/amd-reigns-supreme-in-sid-meiers-civilization-beyond-earth[/url<]

            • Klimax
            • 5 years ago

            Small problem is, drivers can’t do a thing when Mantle is used. (whole “point” of that exercise) It is all up to developers themselves.

          • jensend
          • 5 years ago

          Yeah. The much more informative [url=http://www.pcper.com/reviews/Graphics-Cards/Civilization-Beyond-Earth-Performance-Maxwell-vs-Hawaii-DX11-vs-Mantle/2560×1<]PCPer frame time graphs[/url<] also showed fabulous improvements from Mantle for the 290X, so this isn't just an anomaly with anandtech, but it'd be interesting to know more about why this card and these settings show such an improvement while the improvement with other cards and settings is much more modest. One thing at issue is the fact that both of these reviews have used ultra-high-end CPUs- a 4770K overclocked to 4.1GHz for anandtech and the 3960XXXTREME!!11 at PCPer. This is in line with old reviewer traditions of trying to isolate the video card variable but doesn't reflect the reality faced by many mainstream gamers in a CPU-intensive title like this. Since Mantle and DX12 are largely supposed to reduce CPU overhead I have to wonder whether the benefit would be higher for the 280 and 280X when used with an i3-4330 or i5-4460 or such.

        • chuckula
        • 5 years ago

        As Damage just noted:
        [quote<]Yeah, after I wrote what I did yesterday, I decided to post an update with performance results from Mantle with CrossFire. Unfortunately, despite everything that AMD said, I am seeing zero impact from enabling a second GPU. The frame times are identical: [url<]https://dl.dropboxusercontent.com/u/68992780/290x-mantle.gif[/url<] That ain't right. I am looking into it, but it looks like Mantle w/SFR in Civ:BE is just broken right now.[/quote<] As for any "official" numbers from AMD/Nvidia/etc.: Trust but Verify. (and TR does the Verify)

          • jensend
          • 5 years ago

          What I posted has nothing to do with any “official” numbers and everything to do with the information readily available at the link [i<]you[/i<] posted (as well as the confirming results of the PCPer review). Yes, the CF stuff is apparently not working well. But the Jarred Walton comment you focused on is totally at odds with his own data as well as PCPer's frame time graphs. With the 290X they absolutely did "really see a huge benefit" from Mantle. The improvements with the 280 and 280x also look impressive, though a frame time graph or quantile plot or such would be much more helpful in assessing that than anandtech's minimum frame bar charts.

Pin It on Pinterest

Share This