Nvidia’s Titan arrives to take the performance crown – 36 Performance Benchmarks

Performance summary charts & graphs


Here are the summary charts of 30 games and 6 synthetic tests. The highest settings are always chosen and it is DX11 when there is a choice; DX10 is picked above DX9, and the settings are ultra or maxed. Specific settings are listed on the Main Performance charts. The benches are run at 1920×1080 and 2560×1600 with separate charts devoted to overclocking, PhysX, and 5760×1080 Surround, as well as dividing games up into easy-to-read charts by their DX pathway and by resolution.

Now 5760×1080 with the same settings as above

All results, except for Vantage, Firestrike and 3DMark11, show average framerates and higher is always better. In-game settings are fully maxed out and they are identically high or ultra across all platforms. As usual, we begin with the synthetics.

Futuremark & Heaven synthetic tests

3DMark 2013 Firestrike is Futuremark’s latest DX11-only benchmark, 3D Mark 11 is also DX11 and Vantage is DX10. Unfortunately, scores are completely meaningless when they are presented in this way but they do offer supporting data to accompany our game benches. Here are the charts with Vantage, 3D Mark11, and Firestrike’s basic and extreme tests:

Now we look at Unigine and our old standby Heaven 3.0.  It has been recently updated to version 4.0 and there is a new synthetic, Valley 1.0:

Again, synthetic tests are interesting but they are not necessarily indicative of real world gaming performance. In all three cases, the GTX Titan “wins” over everything else except the GTX 690 by a large margin. Next up, let’s look at DX9 games.

DX9 Games

We test the popular Source Engine represented by Left 4 Dead 2 and also a demanding DX9 game, Serious Sam 3, BFE with both at completely maxed out settings. We have also added Borderlands 2 with and without PhysX, and the Witcher with maxed out settings including “ubersampling”.

First up is 2560×1600:

Now we see 1920×1080

Titan is generally faster than the other cards except for the GTX 690.  The HD 6990 does match Titan in a few of the benches.

Let’s check out DX10 games

DX10 Games

We test five DX10 games – Just Cause 3, Far Cry 2, Resident Evil 5, Crysis and World in Conflict, Soviet Assault.

First up at 2560×1600

We noticed texture flashing and artifacting with our HD 7970 in this benchmark especially at 2560×1600.  Now at 1920×1080:

Benching these five DX10 games, the GTX Titan simply stands out from the rest, losing only to the GTX 690.

DX11 Games

Most of our testing emphasizes DX11 games and we bench 20. Since the charts get too long, we break them up into charts of 6 or 7 games each.

First up are the older DX11 games at 2560×1600.  The HD 6990 could not complete a single run of Metro 2o33 but would crash after beginning the run.

Now those same games at 1920×1080:

Now the newer DX11 games at 2560×1600:

Now the same DX11 games at 1920×1080:

There is simply no contest – the GTX 690 is the fastest video card anywhere but the overclocked Titan comes pretty close.

Now let’s check out our newest DX11 games

Here are our newest DX11 games at 2560×1600.  Please note that there is an issue with Titan drivers with the Sniper Elite V2 benchmark.  “Advanced Shadows” had to be disabled in the in-game menu or else the run would lock up.

We ran Sniper Elite V2 with the maximum in-game settings including 4.0X SuperSampling which is incredibly demanding but looks awesome compared to running with it off.  On the other hand, “Use Compute Shaders” appears to implement DOF with a huge performance hit and with not as significant an IQ change except to blur part of the scene – so we test with Compute shaders off versus on to compare performance across all of our test video cards.

Now the same DX11 games at 1920×1080:


Super-Widescreen 5760 x1080, Surround, 3D Vision Surround, and PhysX

Here is the main graph made from the chart at the beginning of this section.  All of the settings are the same maxed out settings that we used for our 1920×1080 and 2560×1600 benches:

We noted a lot of jitter running the GTX 690 on Borderlands 2 with high PhysX at this demanding setting.  The HD 7970 continued to artifact with Just Cause 2, and Titan had rare jitter issues with the Call of Pripyat benchmark.

The GTX 690 is again the overall fastest card but it cannot run Max Payne 3 at all at our ultra settings because the 2GB framebuffer is too small; the same thing happens with the GTX 680.  From looking at all 36 of our benchmarks, the stock GeForce GTX Titan appears to have approximately 80% of the performance of Nvidia’s top card, the dual-GPU GTX 690, and it is almost 45% faster than a GTX 680.

Next up, let’s look at PhysX


We test PhysX in two games.  Borderlands 2 and Batman: Arkham City both make great use of PhyxX and it is a shame to play either game without it. In both cases, turning on PhysX, although affecting the frame rate, it is enough to play the game with fully maxed out details and AA with our GTX Titan and our GTX 690 spanned across three 1920×1080 displays.

Please note that the Borderlands 2 benchmark is generally more intensive using PhysX than playing the game is and that the performance hit is a bit higher in the benchmark than it should be because of the way it is set up.



We overclocked our GTX Titan +130MHz on the core and +550MHz on the memory. This is a good overclock on stock voltage, stock thermal, and stock fan profile, and it falls only 20MHz short of the overclock we got with our GTX 690 of +150/550MHz; or a bit further off of our GTX 680 maximum overclock with and offset of +175/575MHz.

From our charts, we can see that the GTX Titan scales extremely well at 2560×1600 but not as well at 1920×1080. Our CPU at 4.5GHz simply needs to be clocked higher at this resolution.

No matter how you add it up, the GTX Titan is generally faster than any other video card except for the dual-GPU GTX 690. It also overclocks very well with the stock voltage and fan profile.  AMD would have to respin and significantly increase the clocks of the HD 7970 to catch the GTX Titan and we expect that it would use a lot more power and be relatively noisy and difficult to cool. We are eager to see what AMD actually brings if they decide to create their own upcoming dual-GPU flagship card, a true HD 7990.

Let’s head for our conclusion.


Pages: 1 2 3 4 5 6


Founder and Senior Editor of ABT.

11 Responses

  1. Bo_Fox says:

    Wow, so you’re actually able to use the 7970 for these games at 5760, at
    PLAYABLE frame rate, with PhysX @ HIGH, running on the CPU?

    Wow! GTX 680 isn’t any faster than 7970GE by much at all…. whoa! Batman: AC and Borderlands 2????

    So, GTX 680 basically loses its PhysX advantage right there! HMMMMM?????

    You rule! Since no other sites are showing this!

  2. Bo_Fox says:

    Also, your site is basically the only site to still include BOTH GTX 590
    and HD 6990 with the latest drivers, compared against each other more
    than 1.5 year after release!

    Another golden one for ABT!

    89 BENCHMARK RESULTS FOR EACH CARD (no minimum or maximum crap, or
    low-resolution crap, or anything else – just 1080p, 25×16, 3x1080p, and
    PhysX)! (While also wisely selecting the maximum playable settings for
    each game for the article!)

    Isn’t that the new record or what? For any review site out there on the internet!!!!!!!!!!!!!!!!!!!

  3. DOOM$ says:

    so dusty lol

  4. Bo_Fox says:

    Oh, Apoppin explained the PhysX results – that the average does not represent the minimums very well, especially when PhysX comes into play. So, never mind my first comment!

  5. nod says:

    all other sites usually show titan,690,680 and 7970….590 and 6990 maybe older gen…but they are also still high end cards and should be mentioned more in the benchmarks….so a big thank you for a comprehensive comparison

Leave a Reply

Your email address will not be published.

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>