3DMark and you!

[AdamK47]

I use to be a big fan of 3DMark when the first one came out with 3DMark99 (before that I used Final Reality). I used it every time I would test an overclock or a driver change. Since then it seems like video card companies have taken notice of the popularity of 3DMark and have increased performance through their drivers just to make the scores higher. I've noticed time and time again where a driver update would gain 500 points in 3DMark yet decrease Direct3D and OpenGL gaming performance by 3 to 5 percent, or to some extent break compatibility. It would irritate me when I would go into a newsgroup or discussion board and see these sheep praise their video board maker and their incredible new drivers based solely on the number 3DMark is spitting out at them. These types of people fell for it. 3DMark is no longer an accurate benchmarking tool. It?s become more of a marketing tool. How accurate can benchmarking software be if that software has hardware drivers written to take advantage of it? About a year and 1/2 ago I stopped using 3DMark all together. I'm now seeing people who may have started really getting into computers and video cards for only a year to maybe two years use 3DMark like its a religion. Most of the posts I'm seeing here in the Anandtech video forum concerning 3DMark are by people who may have been members for half a year at most. If you were to frequent other boards, such as Beyond3D or others like it who discuss video performance, rarely is 3DMark is ever mentioned. The bottom line is that you won't find many people who have been dealing with testing a 3D-accelerator performance for many years? use 3DMark as the end all determination. It's relevance, to me at least, is pretty much nil.

I'm done with my ranting on the subject... for now.

 

[JellyBaby]

3DMark is pretty useless. The once exception is using it to check to make sure your rig isn't missing out on performance. That's what I do during a major upgrade or system build -- compare to other machines to make sure I'm where I should be.

 

[kylebisme]

i like it because i can run it once, quickly check the online results database and see if my system is matching in preformace similar systems. however, aside from that the benchmark means nothing, i stand by your arguments compleately.

edit: weird it says 2 minutes between post but it took me all of 30 seconds to type that. anyway what can i say to JellyBaby but great minds think alike

 

[JellyBaby]

anyway what can i say to JellyBaby but great minds think alike

Yeah we rule.

 

[MIGhunter]

I didn't know anything about benchmarks before I built my new computer. I read on here all this stuff about 3dmark so I did one. Then I compared it to only equal CPU speeds to make sure I was where I needed to be. The only reason it is in my signature is so that when I am talking about a problem and someone asks me, I can reference it.

I think that any benchmark is just a comparison.

 

[AnAndAustin]

I think you're being a bit harsh on 3Dmark, the same is largely true for any game chipset developer's think will be used for benchmarking and comparing cards, Quake3 being the prime example. To take any one benchmark (or review FTM) as the sole basis of your buying decision is certainly a bit fool-hardy. But 3Dmark is still one of the best, quickest and easiest ways to check your perf is about what it should be, to see what a diff CPU/gfx card (or both) would have on your over-all perf and to for a quick test when trying to establish a card's o/c'ing limit and the true gains any o/c bring.

Too many people knock 3Dmark, but it is certainly true that too many people also rely far too heavily on it too

 

[Rand]

IMHO 3DMark is decent as a very vague indicator of gaming performance, but it definitely does not corellate perfectly with real world application usage, and both nVidia and ATi have heavily optimized their drivers specifically for 3DMark in a manner quite impossible in typical games.

3DMark's worth as a benchmark has been diminishing ever since release, and at the present time it's of only minimal benefit.
It's vaguely useful for determining if their are any severe performance issues, and perhaps mildly useful as a stress test.
As a benchmark it'll really only give you a hint of real world application performance at this point in time.
I definitely consider it virtually 100% useless when comparing graphics cards of differing generations.

Probably the only very useful aspect left in 3DMark is MadOnion's catalogue of average performance recieved with each processor/graphics card configuration. By comparing your setup to other the catalogue of average scores you can get a very rough idea of how much performance gain you would see by upgrading.
Though even that is of dubious benefit as the individual requirements can vary drastically depending upon the type of game in question.


Personally I put more faith in Quake 3 as an indicator of DirectX gaming performance.
This despite the application that is over 2 years old, and stresses hardware in a manner completely different from modern games, utilizes no modern features and was optimized to counteract bottlenecks that are no longer existent. Is heavily optimized for in graphics card drivers, and is of a genre that I have absolutely no interest in.

3DMark isnt totally useless but it's benefits are minimal.


Go to the MadOnion forums if you want a laugh. The commonly given advice on boosting ones 3DMark includes dropping to point-sampling, setting mipmapping to the lowest settings possible, alter LOD bias to the fastest settings available, overclock everything to the absolute limits regardless of whether you get graphical issues or stability problems... as long as it makes it through 3DMark once!
I've even seen people handing out recommendations to diusable rendering of all textures just to get a higher score!

As I've said once before, I count it as a blessing that none of them have yet figured out that you can disable rendering in 3D and allow 3DMark to run through it's testing suite without rendering anything to the monitor.