nVidia 3DFarq2k3 scandal part II

[TdB]

then what would you do to get a feel of the performance of a card in general, and not just in your friends favourite game?

3dmark tries to benchmark a cards general performance in shaders(among other things), unreal2 benchmarks only tells how good the card is in unreal2...

do you buy a card based on how it performs in specific games, or how you think it will perform in general?
both views are reasonable, but im curious about you opinion and your reasoning.

personally, I try to get a general feeling of the performance, because I don´t yet know what game I will buy in my computer´s future lifetime.

some of my friends builds their computer to fit a very specific game, because they think that game will mimic the general trend in game-performance.

[Technoid]

the idea behind a synteticbenchmark is good, unfortunatley there are just to much "cheating" going around

[Jon P. Inghram]

And 3dmark03 isn't exactly the best benchmark anyway.

[TdB]

I agree, I think they spend to much effort to make 3dmark look pretty, they should rather concentrate in specific benchmarks, That actually tests isolated features of the gpu.
I mean, if they, for example, want to benchmark shaders then make one hell of a shader and throw it at the GPU(and don´t put other stuff in that test, but the shader, no trees, no aliens, no polygons, nothing!), all the pretty stuff is just noise, that makes the benchmark less relevant.
the synthetic tests in 3dmark2001 offered far more information about the GPU, than 3dmark2003 does.

[Jon P. Inghram]

I change my mind... I want all the video card companies to cheat like there's no tommorow! Screw the ignorant morons who only care about how many 3dmarks the video card their equally stupid parents buy 'em!

I'm suffering from Rage3D overdose, too many idiots over there that can't even be called retarded because they don't have an actual biological reason for being so damn dumb.

[DukeP]

Lol.
Ah, not ALL users @ rage3d are that dumb, but yes, on average, they DO give that impression.
~~DukeP~~

[TdB]

I think the general problem with messageboards like that, is the age difference, among other things, or that you don´t know how much they know about the topic in the first place.

when you talk to someone i real life you can see how old your conversational partner is, or how knowledgeable he is and how much indepth he is willing to discuss the topic, and easily fit your conversational style and type of arguments accordingly.

It is very important, that both conversation partners talk in the same "level", when they exchange arguments, or else there will be misunderstandings and(possibly) flaming.

but on a messageboard you can´t see the person you are discussing with, and thus there is a chance that the other person doesn´t have an indepth knowlegde about the topic at hand,(or atleast not as much as you would expect him to).

a typical example of this could be that you(not directed at anyone in particular), come with a highly technical explanation of color-precision in shaders in 3dmark, and he will respond that all that matters to him is the framerates in his favourite games.
the problem here is that you expect the answer to be in your your "level", and to be an argument that has something to do with what you said. he on the other hand doesn´t know what youre talking about, but wants to lower the "level" of the conversation, so that he can voice his own opinion, in terms that makes sense to him.
this could result in frustration on your part, because you spend alot of time and effort in writing your post, and the other guy doesn´t seem to respect that, by responding with an equally well-written post, filled with intelligent counter-arguments. He just informs you his opinion, which may be completely off-topic, because it has nothing to do with what you just posted, but rather about what he thinks the topic is.

everybody is entitled to an opinion, but unfortunately, not everybody stops to investigate, if their opinion has anything to do with what you wrote, or maybe you are to blame for the very same mistake: posting in a different level,(this very post might be prime example of just that, blowing a comment about rage3d members up to a discussion about net etiquette ).

there isn´t really any good solution to this, other than to find a place where everybody is able to discuss at the same "level" as you.

[DedicatedFan3D]

@tdb

I believe that there is really nothing that can accurately be done in current benchmarks to determine

For another, the purpose of 3DMark is to test real world performance, yet it is synthetic. That's an oxymoron I believe.

If you want to test real world performance, then that's what you should test. Day to day usage, quality of drivers, limitations, etc.

Instead all we see is a few benchmarks on widely used and hacked programs. I don't trust review sites. Not because of their credibility, but because of the crediblity of the software and hardware vendors.

[TdB]

fair enough!

but your testing methology, requires that you buy the card first, unaware of its performance, and THEN test it.

that is why I like synthetic benchmarks, if we lived in a perfect world, where there were no cheats, a reviewer could benchmark its specific feature(like how fast it can render polygons, or how many pixel shader instructions it can execute in 1 second).
unfortunately, few hardware reviews provide such benchmarks, they just provide the final 3dmark score.

this will not tell me its performance in games, but it can tell me how it will compare to another card, in a potential future game that is programmed to tax these features in the GPU. I won´t know how fast they are at that game, but I should be able to conclude which gpu that will be the fastest.

the reviews tells me very little about the actual product, they don´t tell about driver quality and robustness, or bugs, they don´t even tell me how I will like the card 6 months later.
They can´t tell me that, and even though that is what I will ultimately want to know, I can accept that they can´t tell me that.
It is, however, entirely possible to deliver synthetic benchmarks, so that I can get a feel for its strengths and weaknesses, and how I think it will perform in the games I might buy in the future.

I agree completely about the credibility problem however, Untill now, I thought I could get an(almost) honest view of the hardware, if I just read enough different reviews of the hardware, to avoid the reviewer´s bias. Now it seems that there can be a huge error margin that should be accounted for, making the benchmarks completely unusable.

Im actually quite exited as to how reviewers will handle this, perhaps they will just ignore the numbers completely, take their time when they review the hardware, and just tell us how it feels in everyday use.
this would be great, because then there would be no need for hardware vendors to cheat in benchmarks, and they would have to spend alot of effort in making their product pleasant to the end-user instead.

[DedicatedFan3D]

I know

I honestly believe that you don't know what you're getting till you buy it.

It's just too bad, but that's the reality of it.
I can't review a tomatoe, or a lemon (pun intended) and then expect you to know what it tastes like.

These video cards perform so differently on so many systems, and all these reviewers benchmark the cards on are the highest performing CPU's that the reviewers can find.

What if say, I wanted to see how it would peform on a system with only 256MB of RAM?

Oh well, shit out of luck, gotta buy it to find that out. Or instead of a P4 3GHz, I want to see how it'll perform on an XP 2200???

Again, shit out of luck. No reviewer to date has reviewed a 9700 or 9800 or any GFFX product on anything less than a 2.5GHz p3 or an AXP 2600 with no less than 512MB RAM.

[bsdgeek]

That's not necessarily true, I remember Tom's did a CPU scaling article a while ago with a 9700Pro.

Yes, Yes, I know it's Tom's but still...