Duke Nukem Forever: Performance Analysis

Benchmark Results: Highest Quality, FSAA

In the following benchmarks, we see the performance of various cards running under the highest-quality settings, and that includes motion blur and post-processing effects. It also includes anti-aliasing, and since there are two anti-aliasing options, we’ll start with FSAA.

The FSAA mode included in Duke Nukem Forever isn’t based on the typical two- or four-sample MSAA technique we’re used to seeing. It uses four samples from three different buffers: the deferred normal buffer, deferred depth buffer, and output color buffer, all the same size as the target resolution. The normal and depth buffer are used to create an edge detection algorithm, and the result is used as a multiplier to the four samples taken from the output color buffer for the result.

At first glance it sounds like twelve samples, but since eight of those are used for edge detection, the result is probably most comparable to standard 4x MSAA.

With FSAA enabled, the GeForce GT 430 is unplayable, while the Radeon HD 5570 and GeForce GT 240 are only able to deliver a minimum 30 FPS at 1280x1024.

There’s a lot of performance variance across the other cards and resolutions. But the important thing to note is that every option achieves at least 30 FPS minimum. That means that even during the difficult parts, things never get too choppy.

Aside from this, the performance advantage that the GeForce cards showed at medium details is no longer present. AMD's Radeon-based cards even seem to be slightly faster at comparable price points with maximum detail and FSAA enabled.

Now let’s try Nvidia’s new FXAA anti-aliasing method.

Create a new thread in the US Reviews comments forum about this subject
This thread is closed for comments
56 comments
Comment from the forums
    Your comment
  • kcorp2003
    I have yet to purchase this game. waiting on a nice sale to pick it up. I waited for this game for so long and i will play it! but not on my #1 priority list of games to buy right now.
    2
  • wiinippongamer
    I never played the first title, yet I played only the first 2 hours or so then uninstalled it for good, the game is a turd.
    5
  • megamanx00
    Got it on sale at Target for like $20 ($40 but half price if I bought another game at the same time ^_^) so yeah I was happy with what I paid for it. I'd say for $40 it's not bad since it's funny, crude, and the graphics are all right. It's a shame that some of the effects that probably tool alot of time to develop (like the rain in the first level) were used only once and sparingly. I thought the monster truck stuff towards the end was cool, but the RC driving in the beginning was a little odd.
    4
  • lmlim
    yawn...
    0
  • RazberyBandit
    Your benchmarks show some rather obvious CrossFire and SLI issues, yet you chose not to even mention it. Why?
    9
  • cleeve
    RazberyBanditYour benchmarks show some rather obvious CrossFire and SLI issues, yet you chose not to even mention it. Why?


    From the benches it seems kind of self explanatory that SLI and CrossFire aren't working. This is the case with a number of other titles, so it's not really a huge surprise. SLI and CrossFire are inconsistent at the best of times.
    4
  • JMcEntegart
    Don ... causing Duke to hand money to a topless dancer and say “Shake it, baby!”


    This was literally my favourite part of Duke 3D. Oh, to be eleven again.

    As far as Forever is concerned, I'm waiting for an irresistible sale.
    3
  • haplo602
    do we still get the fying eyes from killed enemies ? like when you hit an alien with an RPG, you got quite identifiable fying eyes :)
    0
  • tomc100
    Is this a joke? Might as well do a benchmark test on the Sims.
    1
  • youssef 2010
    Like the gamespot review put it

    "If Duke Nukem Forever weren't called DNFE, nobody would be paying a bit of attention to it; it's boring and ugly and........"

    This reviewer really hates the game. but after watching the review, I can't help but agree with him.

    I mean, If a game ever took 10 years to mature, then it should be something like Crysis, Far Cry, COD, Medal of Honor.

    I enjoyed Manhattan Project but this sequel looks just.....too boring to play.

    Disappointed
    0
  • JOSHSKORN
    It took 13 years to make it...after buying it and playing it, I wish I'd waited another 13 years to purchase it.
    0
  • Anonymous
    Why so much work for a review of a very bad and simple game, that you can run on any low end system ?
    -3
  • professorprofessorson
    Talk about being late to the party. This article must have been a afterthought considering how long the games been on the market now.
    0
  • back_by_demand
    The next Duke should be a prequal.
    0
  • kingius
    Where are the six cores in the CPU charts?

    Toms is biased to Intel yet again...
    -5
  • Zeh
    There really wasn't an improvement from 3 to 4 cores on a Phenom II, it's assumed 6 cores wouldn't help.
    3
  • Parrdacc
    professorprofessorsonTalk about being late to the party. This article must have been a afterthought considering how long the games been on the market now.


    This review comes after the sales for June put DNF as #2 selling game for xbox and ps3 despite the bad reviews. Hmmm? http://www.tomshardware.com/news/NPD-Duke-Nukem-Call-of-duty-Pc-Gaming-LEGO,13089.html
    2
  • Owenator
    I enjoyed it. And had a lot of fun playing it. A refreshing break from the current spate of rehashed FPS titles. DNF doesn't take itself seriously and that makes it all the better IMHO. But I'm and old school gamer (Duke was new when I was in college). Apparently you have to be "old school" to enjoy the new Duke. So be it.
    4
  • cleeve
    kingiusWhere are the six cores in the CPU charts?Toms is biased to Intel yet again...


    Biased? You have no idea what you're talking about, sir.

    If we wanted AMD to look bad, we'd add Phenom II X6 results, because they would be identical to similarly-clocked X4 results. No game I know of makes good use of more than four execution cores...
    6
  • Th-z
    Don, maybe you can add more samples in core scaling test in future review, at least one more speed in the mix, e.g. 3.5 GHz, and add Intel CPU in the test, thanks.
    1