From Wikipedia, the free encyclopedia
Jump to: navigation, search
WikiProject Computing / Software (Rated C-class, Mid-importance)
WikiProject icon This article is within the scope of WikiProject Computing, a collaborative effort to improve the coverage of computers, computing, and information technology on Wikipedia. If you would like to participate, please visit the project page, where you can join the discussion and see a list of open tasks.
C-Class article C  This article has been rated as C-Class on the project's quality scale.
 Mid  This article has been rated as Mid-importance on the project's importance scale.
Taskforce icon
This article is supported by WikiProject Software (marked as Mid-importance).


I am editing this article as a Futuremark employee. I am happy to discuss any issues, questions, changes or suggestions for the article with admins and other users. Tripleblade (talk) 14:11, 24 March 2011 (UTC)

Awesome. I just read this article and it mentions Android OS as well as iOS. So how come this article does not mention OpenGL ES? Are you paid by Microsoft to advertise their stuff or are you just not interested in documenting better what 3DMark supports and what not? User:ScotXWt@lk 21:24, 4 June 2014 (UTC)

Just video cards?[edit]

The article suggests that 3dmark is used solely to evaluate video cards. In fact, the 3dmark score you get is an overall value of the whole system - when a new 3dmark comes out (eg 06) the video card tends to be the bottleneck, and therefore the dominant factor, but the idea is always to get a score that represents the system as a whole. Anyone care to correct this or prove me wrong? 16:05, 21 July 2006 (UTC)

At least one of the 3D marks included a CPU benchmark (i.e. a software-rendering graphics demo), which supports that hypothesis -- (talk) 08:44, 22 November 2008 (UTC)
Nope. Another product - PCMark - is used for fuller evaluation of the complete PC performance. Also, disk storage seems to be the bottleneck these days (2014 onward). Naki (talk) 18:18, 4 June 2016 (UTC)

Claims of boosterism by card manufacturers[edit]

Some makers of 3D video cards also shiped their cards with custor drivers that detected if a test was in progress and ignored some operations to make the test report higher numbers. This ofcourse led to larger drivers and that the cards were actually slower for most usage.

  • I've moved the paragraph above here until it can be verified and a source cited. As of now, it's too vague. Specifically, which video card manufacturers and models attempted to inflate benchmark stats? Which tests were being manipulated? If possible, always cite an external source when making a controversial statement. —RaD Man (talk) 17:21, 8 Apr 2005 (UTC)

Google is your friend:


There is also an article called 3dmark. // Liftarn

There was also one called 3dMark, which now links here

old versions[edit]

do older versions still have a use, or should you ONLY use the new ones? The old ones now redundant? I don't know JayKeaton 20:01, 13 September 2006 (UTC)

You can use older versions to test older hardware, i.e. older video cards. Naki (talk) 13:29, 21 May 2016 (UTC)


Should I make things like 3DMark06 redirect to this article? If someone wonders what 3DMark05 is and they type it in, they might not think to type in just 3DMark. Geekosaurus 22:04, 19 September 2006 (UTC)

It makes sense, and there is no harm in doing so either. ~~ Peteb16 22:11, 19 September 2006 (UTC)

3DMark Performance Evaluation/Old Versions[edit]

Two previous topics mentioned here are 3DMark performance and using old versions. First, 3DMark does evaluate more than just the video card performance. At one point during the benchmarks, 3DMark synthesizes 3D video performance by using the CPU to perform the 3D tasks. When this test is performed, there is no variation in the frame rate value. Adam Kemp 01:10, 10 January 2007 (UTC)

In 3DMark03 and 3DMark05, the CPU tests perform the vertex shading on the CPU; the rest of the 3D rendering is performed by the graphics card. In 3DMark06, all of the rendering is done by the graphics card. No version of 3DMark does all of the graphics work on the CPU. (Neeyik 23:41, 22 March 2007 (UTC))

Second, there was a question as to whether older versions of 3DMark should be used for video testing. The answer is yes, as each version is tailored to evaluate a certain version of DirectX. It is a good idea to match the version of 3DMark with the approximate age or generation of the video GPU. After all, many gamers prefer using many different benchmarking programs to give them an idea of the performance of their video card. Adam Kemp 01:10, 10 January 2007 (UTC)

Specifically, 3DMark2001 is useful for evaluating DX7 and early DX8 GPUs; 3DMark03 for DX8, DX8.1 and early DX9 GPUs. 3DMark05 is for DX9 (SM2.0 only), whereas 3DMark06 is DX9 (SM2.0 and SM3.0) (Neeyik 23:41, 22 March 2007 (UTC))
What about DirectX 10, 11, and 12?? Naki (talk) 21:23, 30 June 2016 (UTC)
To partially answer to myself - in July 2016, more exactly July 14th 2016, Time Spy test was added, as an extra test/optional DLC. It tests DirectX 12, and as such requires at least Windows 10. Naki (talk) 21:02, 17 July 2016 (UTC)


No Vantage criticism over the fact that the PhysX test is not done in a fair manor and boosts the scores unfairly?
Em27 (talk) 16:30, 23 September 2009 (UTC)

Certainly not without a very good reference (i.e. not just forum chatter, someone of authority who has demonstrated it and published the results). If you can find such a reference, by all means add it in. Halsteadk (talk) 19:21, 23 September 2009 (UTC)
Well, there is this article, Is Nvidia cheating on Futuremark's 3DMark Vantage?. The thing is that it's not showing how the GPU does two separate tests, but how the CPU test is "helped" by the GPU through PhysX acceleration. Related to this there is the official word of Futuremark on the Approval Condition(s) of drivers and a list of nvidia drivers that enable the uses to cheat, but this isn't really a criticism, since they took measures to prevent users from getting their sores certified, i.e manually turning off physx. I don't know how to classify this since they took measures, if it's worth anything. I'll still look for the actual "cheat"...
Em27 (talk) 02:23, 9 March 2010 (UTC)


It would be nice to have a section with a scale of scores achieved with some well known, reference hardware. this video shows 3DMark05 reaching a score of 45,474 in some very extreme conditions, so I am sure it's impressive ;) however there is no comparison points given in the wikipedia article. AugustinMa (talk) 01:58, 22 October 2009 (UTC)

Wouldn't that be to much of a hassle? HW changes every 6 months so scores posted here would have ti be constantly updated. It's easier to post to the official halls pages where certified scores are actually used.
Em27 (talk) 02:25, 9 March 2010 (UTC)

More on Scores[edit]

The article does not seem to provide the most basic fact about scores - is bigger better or worse.

Bigger score means better. Naki (talk) 15:11, 9 December 2010 (UTC)
Added this to the introduction section alongside a note that scores cannot be compared across versions. Tripleblade (talk) 09:55, 23 March 2011 (UTC)

3DMark 11 Compatibility[edit]

"3DMark 11 will install on a system with a DirectX10.1 (or lower) card, but will not run when you try it (shows an error message that DirectX 11 compatible hardware is required)." Is this statement really necessary? The minimum specs listed on Futuremark website state "3DMark 11 requires DirectX 11, a DirectX 11 compatible video card, and Windows Vista or Windows 7." Is it likely that a reader would expect it to work outside those requirements, i.e. on a system with DirectX10 or lower?Tripleblade (talk) 13:59, 24 March 2011 (UTC)

Sorry, it was me being too much detailed here. :) Not necessary, of course. Naki (talk) 10:29, 8 August 2016 (UTC)

External links modified[edit]

Hello fellow Wikipedians,

I have just added archive links to one external link on 3DMark. Please take a moment to review my edit. If necessary, add {{cbignore}} after the link to keep me from modifying it. Alternatively, you can add {{nobots|deny=InternetArchiveBot}} to keep me off the page altogether. I made the following changes:

When you have finished reviewing my changes, please set the checked parameter below to true to let others know.

YesY Archived sources have been checked to be working

Cheers. —cyberbot IITalk to my owner:Online 13:36, 17 October 2015 (UTC)

File Sizes for each Major Revision[edit]

Filesizes should be included. The compressed file size for the 2013 version with fire strike is 1.5 gb. The 2016 version appears to be around 5 gb (verification needed)--Lmstearn (talk) 14:03, 30 July 2016 (UTC)

Yes, good idea - please add this back. It used to be there before, seems it was removed at some point. Naki (talk) 10:12, 8 August 2016 (UTC)