View Comments
3DMark11_intro2.jpgImageFuturemark proudly continues their tradition of presenting the new version of the most popular 3D benchmark program in the world every two years (give or take a couple of months). This developer first reached global popularity while under the name MadOnion, when they premiered 3DMark 2000, soon followed by 3DMark 2001 (SE), only to revert back to their original name, “Futuremark”, in 2002, taken way back in 1997. Besides this benchmark family, Futuremark is also famed for their PCMark, a complete system benchmark suite, but also 3DMark Mobile, a program for mobile devices. This time, we're reviewing the latest 3DMark 11, which implements all new technologies that DirectX 11 has to offer, and we have to admit that it looks impressive.

3DMark 11 is based on DirectX 11, and according to Futuremark, uses most of the advanced techniques offered by that technology in order to simulate the hardware requirements of upcoming games and also conserve usability through years to come. It consists of six tests, four of which are exclusively graphics tests. The fifth is the “Physics test” which tests the CPU's capability of performing physics calculations, while the last is the “Combined test”, which simulates a situation in which each technique previously used is utilised to a certain extent. If you're familiar with 3DMark Vantage, you'll recall that it contained presets, i.e. pre-defined settings in order to cover the usual situations, but also create somewhat of a standard that would ease comparison with other users and their PCs. The new 3DMark uses the same system, except that presets are now separated according to purpose.

The first one is Entry, with minimum settings and a resolution of 1024x600, allegedly targeting netbook devices. We find this rather comical, since we can't imagine a netbook that would yield a usable 3DMark score even in two years' time. We do believe in Fusion and similar concepts, but have in mind that the new 3DMark really is an extraordinarily taxing benchmark. The second preset is 1280x720, aka 720p, with a medium detail level, covering the mainstream portion of the market, while the Extreme preset is exactly as it states, extreme, with maximum settings together with a Full HD resolution (1920x1080). In this mode, even GeForce GTX580, the strongest single-GPU card, has an incredibly hard time, making tests look more like a slideshow than a fluent animation.

Preset_conf_t.png

In order to achieve a usable 3DMark 11 score, you have to go through six tests.

Graphics test 1 and 2 are scenes displaying submarines with the Futuremark logo, located at the bottom of the ocean and patrolling around an ancient wreckage. These two tests contain a sizeable number of lights (shadow casting and non-shadow casting), volumetric lighting which simulates light dissipation through murky water with added “noise” (the bottom of the ocean isn't the clearest place on Earth, is it?). The graphics card is primarily processing a large number of lights and shadows in these tests. It's worth mentioning that the first test has no tessellation at all, while some is present in the second one.

sshot_1_t.jpg sshot_2_t.jpg
sshot_3_t.jpg sshot_4_t.jpg

The other two tests, Graphics test 3 and 4, display a jungle scene, with a few objects and thick vegetation. This scene uses geometry tessellation, with a single directional light source in the form of the Sun. Having in mind the thick vegetation, there is a large number of shadows, coming from the sunlight. Besides this light source, there is a large number of non-shadow casting light sources. DOF (Depth of Field), which blurs objects outside of focus, has been added to post-processing and fairly easy to notice in this scene. Graphics test 4 is basically the same with increased complexity, and consequently hardware requirements. The number of shadow-casting lights is larger (only one in the previous test), as well as the amount of tessellated geometry (pay attention to the stone dragons).

sshot_5_t.jpg sshot_6_t.jpg
sshot_7_t.jpg sshot_8_t.jpg

The Physics test was first seen in 3DMark Vantage, used to put strain on the CPU through simulating physical effects of firm objects colliding with a large number of objects. The rendering itself has been simplified to a great extent, in a fixed resolution for any preset, in order to minimise the impact of the graphics card. During the scene, firm objects collide with each other, which makes the CPU calculate the effects of each collision and the forces involved. This test uses all CPU abilities by detecting the number of cores and threads available, afterwards splitting operations accordingly in order to achieve optimal execution. In order to keep the conditions identical for any CPU, after the number of threads and cores has been detected, the simulated world is split into an appropriate number of isolated regions that don't interfere with each other. 3DMark uses the Bullet Open Source  library for C++.

The last of the “required” tests is the Combined test, which combines (duh) firm and soft objects in an interaction, which practically puts the strain on the entire system, since, unlike the Physics test, this test puts considerable load on both the CPU and the graphics card.

3DMark yields a score in the form of a single number, preceded by a letter of prefix signifying which preset was used (E, P, X), and is a result of a complex calculation. Three results are calculated first, Graphics, Physics and Combined, with a different formula for each.

The graphics score is the most complex one, since it incorporates results from all four graphics tests into a single one, using the following formula:

Graphics_score.png

In this formula, S(graphics) is the graphics score, whereas C(graphics) is the constant varying depending on the preset and serving the purpose of scaling the result. F(gtx) is the average frame rate from each of the graphics tests.

For the other two tests, the calculations are much simpler – a simple multiplication of the C(physics/combined) constant with the average yielded framerate.

The result of the Physics test is calculated using the following formula:

Physics_score.png

The situation is almost identical in the Combined test scoring:

Combined_score.png

These three results are all required to get a total 3DMark 11 score, with the following formula:

3DMark_score.png

In this formula, W(x) are constants, the values of which depend on the chosen preset.

Constants.png

As was the case with previous versions, 3DMark 11 is already considered standard. Unlike Vantage, which was a bit of a disappointment visually, despite using advanced graphics technologies for its time, the new 3DMark is going back to its roots and is a truly stunning visual experience. This is why its Demo mode, serving only the purpose of fun, will continue to be used for a long time. Certainly, 3DMark has useful tools for checking the quality of 3D display, albeit reserved only for the most expensive version. 3DMark 11 will be used for InsideHW's test battery in the future as well, of course, whenever there's a new graphics card to review, so you'll be seeing plenty of models handling it. Besides, the online community established thus far will give you plenty of users to compare and compete against. Just make sure to restrain your ego first, since you're bound to get what may seem like a pathetic score, no matter which graphics card you're using.

Three Versions

3DMark 11 is available in three versions: Basic, Advanced and Professional. The first of these is the free version, available for download from Futuremark's website owing to the sponsorship of MSI and Antec. The other two versions are commercial, but also far richer in options.