Poll: Do you consider 3DMark a valid benchmark?

[Deeko]

Ok, I'm getting tired of complaining in every 3DMark thread, and there are soooo many. Here's the deal: Do you consider 3DMark a valid benchmark?

 

[Methodical]

yes, so long as you dont use it to compare computers with nvidia graphics cards in them, it is valid. you cant compare graphics cards with the benchmark because nvidia tweaks drivers especially for 3dmark, i hear

 

[urbantechie]

I don't use benchmarks.

 

[xtreme2k]

I wont say it isnt 'valid' but then i must say it doesnt reflect real gaming performance.

THe V5 5500 is slower than even a G256/SDR, i seriously doubt that is the case in real life

3dmark2000 stressed too much on the TnL potion of the game.

i mean if you are bored and want to see some graphics on your screen, then feel free to run 3dmark2000
but just dont take those numbers seriously

 

[Dug]

3dmark only collects certain parts of information from the test to caculate a score. For instance, it doesn't use the fillrate of a card as part of the score even though it tests it.

This is what I heard, but I'm not sure. Can anyone confirm this?

 

[xtreme2k]

yep i think they only use the
Game1
Game2
and High poly count
to determine the score

 

[Wingznut]

I think it's fine for comparing "tweaks" (setting changes, driver updates, etc) in your current system.

But I think it's a steaming load of crap, for trying to compare different pieces of hardware.

 

[TGCid]

No benchmark is absolutely perfect. It is good for measuring the overall performance if the influencing variables are kept in check. Even so it is good when using in combination with other benchmark but by itself doesn't mean anything.

 

[compuwiz1]

<< It sucks as a benchmark >>



Oh....whatamatter Paulson; it wouldn't run on your Compuke's integrated video system?

 

[optoman]

I don't use it to benchmark my comp but I do use it to test stability.

 

[xtreme2k]

well it is good if you test within changes in your system

let say you tweak the BIOS's setting (cas lantency/ precharge/ cache lantency etc)
then it is a great program to measure as it gives consistant results

also
let say you overclock your CPU/3dcard

3dmark is also a good benchmark to test how you are performing after the tweaks


however, it areent the best to test in between CPU or 3dcards

 

[xMikey]

I class it as a valid benchmark program.





Mikey out!!

 

[Weyoun]

I dont use it to compare systems and video cards, but like xtreme2k said, it's good for testing new modifications on your system. Comparing a 3Dmark with both 3dfx and nVidia cards is just ludicrous and their marking methodology is very absurd. Just averaging framerates isn't what I'd call a conclusive benchmark in any light. It would be better to integrate some image quality tests, perhaps put this framerate fluctuation fiasco to death and integrate these into the mark given at the end. It has a lot of room to improve in as a synthetic benchmark and should give an overall indicator as to how good your gaming experience is...

 

[peemo]

Pretty lousy. It doesn't indicate whether you will have smooth frame rates in real games and does nothing to compare the quality of the signal to your monitor.

I think it's responsible for countless wasted hours people spend tweaking or even changing hardware because they think something is wrong with a system when it is really just buggy software. 3DMark also encourages unnecessary overclocking which shortens the lives of CPUs and video cards for no good reason. I bet many people who overclock can't tell the difference in normal system use and so resort to &quot;benchmarks&quot; like 3DMark 2000 or the new VBr8R to convince themselves that the money they spent on that peltier and waterblock was worthwhile.

 

[Deeko]

I added &quot;Not a good benchmark but good for a stability test&quot;. That's what I would vote for.

 

[AMB]

I voted for

<< Not a good benchmark, but good for a stability test >>



It does, to some extent show the difference between computers, but just not well enough

 

[BenSkywalker]

I consider it probably the best gaming benchmark out, but the 3DMark score is USELESS.

That is the frustrating part for me, the bench is very detailed and breaks down nearly every aspect of video card performance, but people focus on the useless 3DMark score.

Fillrate test for both sigle and multitexture,

High texture load performance,

T&amp;L performance,

CPU performance,

Full D3D feature test,

And a complete readout out of every feature, game related or not, that you board supports under DX. I think for 3DMark2K1 they should drop the 3DMark score and make people look at the numbers the test is generating. The 3DMark score itself is a great indicator of how the &quot;MaxPayne&quot;(If I recall properly) engine should run, much as UT is a good indicator of how UT engined games will run, but not overall DX performance.

3DMark2K is, IMHO, the best DX bench that is out, it just horribly misused by too many people.

 

[Rigoletto]

Don't forget that there is an image quality benchmark in 3Dmark2000, just not the freeware version. Magazines regularly use them.
Whether NVidia optimise for 3D mark or not, I doubt very much the Voodoos are any longer in the stratosphere of speed. Whatever system you're using to mark them.

 

[VladTrishkin]

3d mark is only &quot;valid&quot; under win 9x... But Q3 will give you the right scores...

 

[MaJik]

Well, to start ill reply to some of the things you guys said:


<< THe V5 5500 is slower than even a G256/SDR, i seriously doubt that is the case in real life >>


Anandtech's Tests
As you can see, the G256/SDR is in most cases faster then a V5 5500



<< I consider it probably the best gaming benchmark out, but the 3DMark score is USELESS. >>


The Score is just a general score figure, what u want one number to tell you everything?

As for the rest of you


<< It sucks as a benchmark >>




<< But I think it's a steaming load of crap, for trying to compare different pieces of hardware. >>




<< Pretty lousy. It doesn't indicate whether you will have smooth frame rates in real games >>



Well, 3DMark is used to messure the 3D performance of your system, why would you want to do that? to give you a idea of how great it will perform when using it? right. Well if you compair 2 cards, say Voodoo5 5500 and a Geforce SDR. Try them in Quake3Arena, they may be close to the same, then try them in UnrealTournament, one may be much better then the other in that! Try a few more games, you may notice that, the cards don't exactly show the same results with every game, one may always be faster, but by how much? if the Geforce is 5% faster in Q3 and 35% faster in UT don't you think that makes a difference?

Well my point is, 3DMark does its job, it messures the proformance of the setup in 3Dmark nothing else. Everything is different, there is no way it can tell you how well its going to perform in every application you may use it in.

 

[AtomClock]

I use Wintune to check my system after a tweak. It seems to be a better test for overall performance. I use 3DMark just to test stability of my drivers. I also use Sandra to test individual subsystems.

$0.02

 

[Dravic]

I simply use it to see how my system scores to &quot;similar&quot; setups. How it runs after tweaks, and stability

 

[kami]

Yeah, I'd only good for stability tests.

 

[AmdEmAll]

3DMark IS SUCK


Graphics look nice though

 

[sandorski]

I'm with Wing on this one, it's good to test system tweaks, maybe. I don't consider it much of a stability test, it's seems buggy to me. Sometimes it works, sometimes it only works after a re-install, as far as I know, no ones ever proven 3dmark as a stable piece of software.

Games are really the only usefull bench for game performance, IMO.