Why should we care about 3DMark?

DukenukemX

DukenukemX

[H]F Junkie
Joined
Jan 30, 2005
Messages
11,751
Seriously. Seems that Beyond3D got a chance to do an interview. Giving us insight on the next 3Dmark.

Am I the only person who skips 3DMark benchmarks I see on websites? As a gamer it's useless. No realistic real world performance. Besides the fact that no game engine performs the same but what about other things in games like sound?

Funny thing about gamers is that we buy an expensive Video card an maybe get 10 to 15 more FPS if we're lucky but by having a nice sound card like the Audigy 2 we also get another 10 to 15 FPS depending on what we had before.

Why do we have 3DMark when we have lots of games that'll give us more info in the performance of a system?
 
If you understand how 3dmark works, you can get some useful information from the results...the problem is that almost everyone takes a 3dmark score as the end-all be-all of a video card performance...which is increasingly not the case, because it's impossible for futuremark to accurately benchmark a cards ability to play games without using a real game engine...

Personally I think 3dmark is a quick and dirty way to classify cards...but to truly gauge a card you have to look at the way it runs actual games...
 
Well for one, why not? Its bragging rights... I myself would not care if they did away with it but thats not going to happen.

People just like to get an idea of how good their card is.

Like the person above said, its a way of classifying cards as well.
 
It's a standard tool. Not everyone owns every popular game used in the benchmark sites. Even the framerate from an actual game isn't real world performance.
 
^eMpTy^ said:
If you understand how 3dmark works, you can get some useful information from the results...the problem is that almost everyone takes a 3dmark score as the end-all be-all of a video card performance...which is increasingly not the case, because it's impossible for futuremark to accurately benchmark a cards ability to play games without using a real game engine...

Personally I think 3dmark is a quick and dirty way to classify cards...but to truly gauge a card you have to look at the way it runs actual games...
Click to expand...

do u have a link so i can read up on this plz?
 
^eMpTy^ said:
Personally I think 3dmark is a quick and dirty way to classify cards...but to truly gauge a card you have to look at the way it runs actual games...
Click to expand...

That's about the extent of it.

If you have one card that gets 9000 in 3dMark, and another that gets 3000....well, you know that the first is better than the second. A LOT better, in fact.

If you have one card that gets 9150 and another that gets 9035...uhhh....doesn't tell you much about the cards.

And it certainly doesn't say how well any card plays any game.

It just serves as a quick guesstimate to get a feel for overall performance levels.

And it does just fine at that.

Plus, hey, it looks pretty damn cool. For almost a whole year (until Q4 and FEAR hit), 3dMark05 was just about the best looking thing you could run on your computer to show it off.
 
Benchmarks is also usefull if u tweak setting on ur pc, specially memory timing. What might work well on a pc might not work well on another, it also helps u compare to ppl with similar specs, a 6800gt ot a x850xt will score lower on a lower end gpu but some chips are more independent from processor. is also good 4 ppl with dual core processors since the require some manual adjustments.
 
It's very useful if you don't track average framerates in game. You want to know if your video card is behaving normally, just check someone's 3dmark with a similar setup.
 
It's also a quick way to confirm that your rig is working properly. One of the first things I did was run 3dmark when I first built my computer... then never touched it again. It was to just confirm that everything was A OK... since 3dmark is standardized. And also when I OC, it gives you a guestimate of how much you gained from the OC.
 
You want to know if your video card is behaving normally, just check someone's 3dmark with a similar setup.
Click to expand...


Yup, that's pretty much all it is good for now. The program tends to get less accurate the older it is (not exactly surprising).
 
I think it's mainly just for bragging rights. People that spend $400+ on gpus should be proud of their 9,000+ points and so on. And...it gives you a general idea of how your pc is performing.
 
I second,... or third the whole quick and dirty comment. Its kinda like cetaris parabis, all other things held equal, 3dmark scores give rough estimates of the quality of systems.

Assuming that it isn't optomized for anything particular card, then it gives a good idea.
 
3DMark does an excellent job of detecting driver issues and the like.
 
Whatever happened the article kyle was gonna do on 3dmark after one of the creators threatened hardocp with a lawsuit a while back? Far as im aware that article never apeared. Also the article on the evga cooler (ac3 or something) that was meant to show if it perfomed well or not..that never appeared either. :confused:
 
I use it to compare tweaks within' my own computer... and to make sure I'm not 2000+ points less than someone with the same hardware (thus signaling that I suck at building computers..)
 
It's not only a benchmark, I use it to detect artifacts when i'm overclocking.
 
3DMark can also work for stability testing . . . with unstable ram for instance, 3DMark05 would often crash on me during one of the tests . . . sure it's used as bragging rights for many people but that's just the nature of benchmarking . . .

It's also very useful for testing how effective OC is and since it's standardized as some have mentioned a good way to check with others if your card is performing as it should . . .
 
3DMark gives a general idea of how a system peforms as compare to others, but it doesn't shows how that system itself would peform in real world gaming. That said, the score difference here have to be sufficiently large, a few hundered would mean almost nothing.

Yes, i mention system there because its somehow affected by the CPU as well...if one wants to do a comparison between 2 graphic cards, he would at least have to run it on a similer system.
 
I think the main issue with the current 3dmark(05) is the vertex setup bottleneck that it has, that isn't reflected in actual gaming. It's models are so complex, that the higher amount of vertex ops you can do without bottlenecking on ROPS and shader ops means you'll get a higher score than a card that has more shader power, but less vertex power. Thats been proven with the 7800gtx vs x1800xt. Take the scores of 7700(7800gtx) vs 9100(x1800xt), that means that the x1800xt should be 18% faster in non anti-aliased low res gaming. Thats not the case, it might be faster with AA, but thats not whats being tested.

I think they need to come out with somethign that stresses memory and shaders, not ROPS either, fill rate is good enough for todays resolutions. I believe the minimum res should be 1280X1024 w/4xAA and 8X AF, that would give a good accounting of what most gamers presently use. That would stress the shaders adn the memory.


Oh, and they should write (or license) an engine that runs on openGL like the doom3 engine and test for that as well.
 
3dmark comes in handy demonstrating overall average performance of the renderer in a wide variety of games. Futuremark employs several *game engines* demostrating their expectations of what is coming down the pipe for the next several years (and they do an excellent job at it).

3dmark does not cover ai, does not have a serious bench-test for physics and is not suppose to give you real world results of all games. 3dmark is a video card benchmark demonstrating real world performance of your card. In other words, regardless (relatively speaking) of cpu and system memory, an amd 1800 xp + x800 xt will have similar scores to a P4 3.8 + x800 xt. HL2 and Doom3 benchmarks do exactly the same thing (excluding the [H] because they do it right).

3dmark is a video card benchmark, 3dmark is advertised as a video card benchmark, Futuremark claims that 3dmark is a video card benchmark, 3dmark is most likely a video card benchmark. IMO, it does a mighty fine job.
 
But I know 2003 does a CPU test, not sure about the later version.

I mainly use the 3Dmark programs to see what differences in my settings make. As for trying to get higher marks than your friends, well.. it's silly. People go and turn down the settings in their drivers to get the crappiest quality available, for the sake of gaining some points.. the video card doesn't suddenly become "better" just because it scored a few more points in 3Dmark. :p
 
I think 3dmark is overrated and is just a e-tape meassure for your e-penis.
 
i think it's pretty good.

Certainly on your own system it can be used for relative scores. You can't really argue that it is a good way to see if you have gained something by overclocking.

It was released a while ago, I think that towards the end of every 3d mark's lifetime, people start bashing it, but that's expected. You don't use BF 1942 to bench your GTX, right?

They seem to have realized their mistake in making 05 too shader intensive and will be trying very hard to make it align with current game engines. They should definitely at least wait to see how Unreal turns out, because that will be a very trend-setting/popular game engine, and who knows, maybe it will be uber-shader intensive.
 
3dmark is good for competitive benching ( fun if you know how ;) ) and showing off the latest effects your card can produce (usually) before a retail game does so. It's not indicative of real world performance however. :cool:
 
You must log in or register to reply here.
Back
Top