Why is everyone picking on 3DMark06?

[SpeedyVV]

Ok, this is a real question, because I really would like to know the usefullness of 3DMark06. Please no reason to get upset.

1 - While i understand that for us gamers, game benchmarks are the real measure of performance, is there no value in a synthetic benchmarks that is pretty much the de-facto standard, allowing us to compare different systems?

2 - And does the 3 sub benchmarks not give a good relative indication of performance of both GPU and CPU?

3 - If my 3DMark06 score goes up, is this not an indication that the performance of my system has improved (regardless of impact on games)?

4 - Is it possible for 3DMark06 score to go up, and game performance to go down?

 

[RangerSVT]

I still value 3dMark results for overall system performance.

I'm pretty sure that if 3dMark is showing a score of 4K below average for that particular setup, most games will reflect what 3dMark is reporting.

People like reviews of actual games, to determine actual performance numbers from a particular setup with particular games, makes sense, considering it's the most detailed analysis. Compared to 3dMark which is an overall analysis with not much detail to go into play, considering some 3dMark versions judge the score by GPU rather than CPU, and some judge the score by CPU rather than GPU, while the latest I believe judges on GPU.

 

[The Null]

1) while it is a 'standard' 3dmark06 has its own little qwirks, 2 complete different cards may get the same performance in 3dmark06, but in game, the performance difference is quite noticeable. eg the 8800GTS/GTX and the 2900XT

2) u mean the CPU tests right?

3) your score in 3dmark can go up, but performance in game can go down, upon saying this, rest assured that most of the time if 3dmark goes up, your games will most likely go up as well, depends on the game.

 

[SpeedyVV]

1) while it is a 'standard' 3dmark06 has its own little qwirks, 2 complete different cards may get the same performance in 3dmark06, but in game, the performance difference is quite noticeable. eg the 8800GTS/GTX and the 2900XT

Ok, this then would make sense why people pick on 3DMark06. I did not know this was possible. ie that a lower performing GPU would score the same as a higher performing one.

2) u mean the CPU tests right?

I thought 3DMark06 had 2 GPU sub scores and 1 CPU sub score?

3) your score in 3dmark can go up, but performance in game can go down, upon saying this, rest assured that most of the time if 3dmark goes up, your games will most likely go up as well, depends on the game.

This is the part that I dont understand. The only way I know of (ok, so I am a n00b), to improve 3Dmark06 is to go higher on OC of GPU or CPU (yes, mem and HDD), so why would a game performance ever go lower if I up my OC?

 

[NoxTek]

I'm going to cut through the inevitable flood of bullshit answers you are going to get, and give you a straight one:

The reason almost everyone here poo poos on 3DMark or anyone who mentions it is that a few years back Kyle suddenly decided to stop using Futuremark (Formerly MadOnion... err... and FutureMark before that) software in any of the benchmarks done on [H]ardOCP. Read into it whatever you want.... Thereupon Kyle and the rest of the [H]ardOCP staffers proclaimed that not only were Synthetic benchmarks "BAD" but anyone who so much as hinted at using them in the forums were to be hunted down, lynched, and boiled alive in their own bile. (OK, I made the lynching part up... but it sure seems that way). Thus the [H]ardOCP "Apples to Apples" method of benching was born and since then it has been used religiously in all reviews you see done on the main site.

Despite the fact that [H]ardOCP has abandoned 3D synthetic benchmarks many of us still see value in them and use them TOGETHER with real world gaming to gauge the performance of our rigs.

 

[Frosteh]

3DMark is just not a good example of typical games on the market today, it's scores don't accurately reflect video cards capabilities in real world gaming, partly due to it's own shortcomings, but also partly due to the fact that AMD and Nvidia both optimise driver code specifically for synthetic benchmarks.

It's only real use I've found, besides being a pretty tech demo, is to gauge FPS return on overclocking the same piece of hardware, and looping to check for artifacts (stress test)

 

[stoney_titan]

I wouldn't build a new system without it. Call me whatever you'd like but if my sig rig scored an 8500 i know somethings wrong. I usually test it to make sure my gpu drivers are in check and that everything is mostly stable running 3d intensive apps. I have a dream. That one day futuremark uses modern day games as benchies to give us a score that represents more than just a number. I have a dream. *sorry MLK*

 

[SpeedyVV]

So, if I understand it correctly, it is fine to use it to understand the relative performance of one's system, in various stages of OCing and to compare it to other "similar" systems that use "similar" hardware.

But it is bad to use, to compare againts other disimilar systems, and also as predictor of gaming performance or FPS.

Thanks for all the replies.

 

[Frosteh]

So, if I understand it correctly, it is fine to use it to understand the relative performance of one's system, in various stages of OCing and to compare it to other "similar" systems that use "similar" hardware.

But it is bad to use, to compare againts other disimilar systems, and also as predictor of gaming performance or FPS.

Thanks for all the replies.

Pretty much dead on.

 

[Mac[X-D]]

this is my biggest problem with 3dmark, when new drivers come out for either nv or ati, everyone installs them and the for the most part the first thing they post about is, my 3dmark went up some percentage or went down down some percentage or everything stayed the same. and i always sit back and think, well ok, now what? you got a 100 more or less points in 3dmark, do you play 3dmark? do you notice it when you are online gaming? no because it doesn't mean anything. what do your games tell you, how does your shit really run. did new drivers hurt or help the games you play. thats the shit that matters, thats what i want to know from other people when new hardware or drivers come out. i could give a fuck about gaining anything in 3dmark.

plus there are so many tweaks you can do to get every little once of performance out of your system to get the highest 3dmark score possible and its useless, and i have been using 3dmark to test hardware since it came out and i have always felt this way, we see it here everyday post after post about "why is my 3dmark score low" or "does my 3dmark score seem right" and yet a lot of the times the person never says my games play shitty or great, their overall system seems to be running great but yet they are worried about their 3dmark score. and the same goes when over clocking, at stock they got this much and now they get this much, well ok, i think logic tells us that there will be a increase from stock to a overclocked system. but they are never happy, its always "at stock i was getting this at my overclcoked speeds i am getting this shouldn't i be getting more" well who really knows. maybe other peoples shit just runs better, or they have tweaked better, who knows, once again its why its a useless tool, plus and one of the biggest things that bugs me about futuremark, is since its a synth bench why no set of base scores? a chart to compare to? could it be because with all of the different hardware configs possible, it would be near impossible to put a baseline together? maybe.

But since every system here at the [H] is different even the ones that are they same are different, there are just to many uncontrollable variables when testing 3dmark,

i think a lot of this has really come to pass lately with a lot of the new games that are out, betas, demos, or full retail. how many people here have been discussing game performance of a particular game and peoples very similar if not exact systems find that a game is completely playable for them at one level of settings and the next person says the exact opposite and that level is unplayable on my system. yet they have very comparable hardware or identical hardware in some cases. but yet they could get the exact same 3dmark score and yet their gaming experiences are completely different.

anyways this post has become much longer then i expected sorry for my ramblings.

 

[sabrewolf732]

I'm going to cut through the inevitable flood of bullshit answers you are going to get, and give you a straight one:

The reason almost everyone here poo poos on 3DMark or anyone who mentions it is that a few years back Kyle suddenly decided to stop using Futuremark (Formerly MadOnion... err... and FutureMark before that) software in any of the benchmarks done on [H]ardOCP. Read into it whatever you want.... Thereupon Kyle and the rest of the [H]ardOCP staffers proclaimed that not only were Synthetic benchmarks "BAD" but anyone who so much as hinted at using them in the forums were to be hunted down, lynched, and boiled alive in their own bile. (OK, I made the lynching part up... but it sure seems that way). Thus the [H]ardOCP "Apples to Apples" method of benching was born and since then it has been used religiously in all reviews you see done on the main site.

Despite the fact that [H]ardOCP has abandoned 3D synthetic benchmarks many of us still see value in them and use them TOGETHER with real world gaming to gauge the performance of our rigs.

qft

 

[Kowan]

3DMark is just not a good example of typical games on the market today, it's scores don't accurately reflect video cards capabilities in real world gaming, partly due to it's own shortcomings, but also partly due to the fact that AMD and Nvidia both optimise driver code specifically for synthetic benchmarks.

It's only real use I've found, besides being a pretty tech demo, is to gauge FPS return on overclocking the same piece of hardware, and looping to check for artifacts (stress test)

Other than checking to see if your system is way out of wack compared to similar setups it's something to run when bored.

 

[jmackay]

3dmark is great for comparing your system to other people using the exact same setup. However trying to compare to other hardware is where it fails, and is also what most people use it for.

Eg: Dual vs Quad, 8800 vs 2900

A C2D @ say 3ghz with an OCed 8800gtx will score much lower in 3dmark than a Quad @ 3ghz with a 2900xt yet will do better in games. Now the quad can do non-gaming stuff better, but most people that use 3dmark do so for gaming comparisons not general computer usage stuff.

So while the world record is (I think still, don't care anymore tbh ) held by a quad / CF 2900xt, a Quad / SLI Ultra would beat the snot out of it in games.

 

[Arcygenical]

The reason almost everyone here poo poos on 3DMark or anyone who mentions it is that a few years back Kyle suddenly decided to stop using Futuremark (Formerly MadOnion... err... and FutureMark before that) software in any of the benchmarks done on [H]ardOCP. Read into it whatever you want.... Thereupon Kyle and the rest of the [H]ardOCP staffers proclaimed that not only were Synthetic benchmarks "BAD" but anyone who so much as hinted at using them in the forums were to be hunted down, lynched, and boiled alive in their own bile. (OK, I made the lynching part up... but it sure seems that way). Thus the [H]ardOCP "Apples to Apples" method of benching was born and since then it has been used religiously in all reviews you see done on the main site.

Oh CMON... I can slap in a Qx6600, OC it to 3.6 ghz and score an extra 2000pts in 3Dmark06... Yet see a 1fps increase in games?

The fact that the HD2900 owns the GTX's in 3Dmark, but often loses in-game?

It's not a reliable benchmark. It's great for judging overall system performance, but it should not be the focus of a review like most other sites use it as.

 

[SpeedyVV]

Thanks folks, this has been most informative. I am sure some of you old timers already take this as fact, but for a n00b this is very interesting, and better than, because Kyle said so.

Even though when Kyle says so, I tend to believe is true ;-) It is just good to understand the logic.

I am now going to start another thread to ask, what games would make good benchmark indicators.

Of course, the games i play would be the best ;-) but what I am looking for is something like WiC is good for CPU, FEAR for GPU, x for mem\HDD, etc,.

btw, just examples, I have no idea which ones are good for what.

 

[AdmiralFlameberg]

I dont see 3dmark as a good tool to see how good your comptuer is. I never did. I just saw it as "who can get the most highest score" I remember people use to use it to see what games they could play and what not.

I usually test the games to see if i can play it I didnt use 3dmark to see. And I always told people back in the day to play the game to see if you can play it not look at 3dmark.

Now adays I just skip the 3dmark tests on video card reviews and go right to the games since those numbers from 3dmark doesnt help me. And are just numbers to me.