3DMark Vantage: The worst 3DMark Ever?

Before any of you go around and brag about 3DMark, I'd like you guys to read this first.

Futuremark, the creator of the popular 3DMark series of benchmark has made very idiotic mistakes on their latest benchmark, the 3DMark Vantage.

The true last good 3DMark ended with 3DMark05. 3DMark06 isn't as bad as Vantage, but it is still a horrible benchmark. Take a look at why.

http://www.tomshardware.com/reviews/amd-785g-chipset,2381-8.html

3DMark Vantage
9300/E7200: 3682
790GX/PII X2 520: 2931

Left 4 Dead: 790GX outperforms 9300 in lower settings and is equal at higher settings
Crysis: Equal
HAWX: 790GX is faster
WIC: Equal

The problem with Vantage is that it takes CPU scores are taken way too seriously in the final score. It seems fine and dandy until we realize that each game component the benchmark runs on is also affected by the CPU. So the faster CPU will result in increasing the individual score and AGAIN on the final score due to the CPU portion of the benchmark becoming better. Its akin to someone taking fps scores from a game, and scores from SuperPI and using some arbitrary equations to make a final score.

But we want a GPU benchmark, not a combined CPU/GPU benchmark. It's ironic they call Vantage a "3DMark" as it doesn't really benchmark the 3D component and betrays the fine tradition of making a worthwhile benchmark like they used to.

Again, I repeat this wasn't always true. 3DMark05 actually reflects real world scenarios pretty well, due to the fact only the GPU portion of the benchmark affects the final score.

This problem gets especially worse as the GPU gets slower, as is true with mobile and integrated solutions.

 

http://i35.tinypic.com/2a98g3b.jpg

here's my highest

 

3dmark vantage is not really designed to test such low end hardware, that's more the job of 3dmark 06. Vantage is calibrated to test high end desktops more than low end integrated chips.

Please note however that vantage will give individual scores to 3d and CPU and in such cases the 3d score is quite CPU independant and would work better in this case.

 

And 3dmark06 isn't CPU dependent? Why do i get a 6-7% boost by OCing my CPU 20%?

Actually, the GPU score in Vantage is one of the best CPU independent benchmarks available at the moment.

 

.... vantage is the best product out for benching GPU's.

turn physix off if you have nvidia.

Now 3dmark is WAYYY to dependent on CPU, when my M17 can score 15k on it with a t9400.... something is really wrong. Your rant is flawed in many ways, and i assure we are very capable people here and know what we are doing. We use vantage for a reason : its reliable, and gives a true answer

 

here it is

 

4,000 point difference just because your CPU is .7 ghz faster? Now which one is worse?

 

I don't know why people complain about synthetic benchmarks. It's just a benchmark, that's it. It's never going to be perfectly reflective of real world gaming performance, especially considering how much games vary (CPU or GPU dependent) based on how the developers code their product.

It does give everyone a platform to compare and test systems and push them to the limit. That's all it's designed for...

 

+1 to that, everyone seems to love just comparing marks...

 

Of course it does. But you gotta realize the "score" itself is there, which COMBINES the CPU AND the GPU portion. Not many sites show the result for all 3 results.

You are SUPPOSED to get some improvement when having a faster CPU even in GPU benchmarks. But the 3DMark problem is that the "CPU" portion of the benchmark ADDS to the final score in addition to the GPU component. So the CPU differences are amplified! In reality its more accurate to just not include the CPU portion.

Yes. I'm pretty sure that the GPU score component is excellent. It's when it comes to the FINAL score it becomes bad. Except in most product comparisons reviews it shows the final score, not the individual component score. It is very annoying to hunt down the sites that do more than just show the final combined score to see how the products really perform!

3DMark06 is bad, but not as bad as Vantage.

With 05, it does not calculate CPU scores into final benchmark scores. With 05, if you compared two video cards on multiple 3D apps and then looked at 3DMark score, the average over multiple apps roughly equaled the 3DMark score difference.

(I'm pretty sure this isn't what people want to read. Yes, people like their "marks". But it should be a wake-up call to Futuremark that markets themselves as a GPU benchmark, yet it doesn't do that)

 

I just use vantage and 06 to make sure my laptop or desktop is performing like its counterparts. If I want to know how it performs in a game, well, I go ahead and run that game.

Seems like a no brainer to me.

 

On these forums, all we do with Vantage is run is the GPU portion anyway.

 

I still haven't figured out why my Sager (yes, that beast in my signature) only scores ~4500. Vantage runs really slow the whole way through, and at the end, it ranks my computer below the cutoff for "really old computers".

What the hell!? I've got the best single mobile GPU available, and a core i7! Really old, my arse!

Still, performs admirably in most games.

 

I'll run it on my G72 and see what I get.

 

Look at the GPU scores too. At 2.0ghz, your CPU is bottlenecking your card

 

i'l run 1.5ghz and 1.0ghz for kicks later.

 

I also prefer 3D mark as well. IM so use to measuring it through there that is hasn't let me down.

 

533 is the lowest fsb my mother board will give me

 

Interestingly, HDR/SM3.0 score is the same for 1.6ghz and 2.5ghz.

 

that is weird but i am using a different driver so may explain it. everything else is the same....


that is very weird.

 

A few comments...

- All benchmarks are unreliable, and this should be expected to a certain extent.

- LOL at dondadah's results.

- Even if the total score is skewed, just don't worry about your total score and look at the other scores, or set it to not even run the CPU specific test.

- It would be nice though to have a GPU test that is largely CPU independent though, but it would also be hard to implement well.

 

First of all, if you average out the scores in multiple games, it will roughly turn out to be the score in 3DMark. The problem is that averages don't really show anything more than being just a comparative guide. Enough for comparing two video cards and say which is better in general though.

Not all review sites do the CPU/GPU specific tests, which is what irks me. Individual people may do it, that's good.

We don't need a GPU test that's CPU independent(because its unrealistic as CPU DOES affect performance), we need a GPU test that runs 3D code that's based on real world games so its relevant to us. IMO that used to be 3DMark05.

 

No, the actual issue is that 3Dmark, and vantage use a screen resolution that is WAY too low. The resolution used barely stresses even midrange GPU's, and becomes CPU dependant. The benchmark will become several times more accurate when the resolution is raised, the higher the res, the more accurate it becomes. I think the new standard should be 1680x1050 , no one uses 4/3 anymore, and 1680x1050 is now the most common, and is compatible with higher res monitors. Do this and the benchmark will be FAR more accurate.

 

Vantage's performance test (default one) uses 1680x1050 IIRC

 

I did a x run which is 1920*1200 and overclocked I got 59?? On the gpu and the CPU was the same as p run

 

I said 3dmark 06 is for low end intigrated chips, a 4850 is more what vantage was designed for. Plus yes your 2ghz CPU is a bottleneck and look, the benchmark shows that....

 

I know that. I was just doing runs for kicks. The post above you was my vantage run doing extreme which maxes out the gpu at 1920*1200.

 

Good information there, dondadah (although your ability to overclock your T9500 makes me envious).

 

A unlocked BIOS is a beautiful thing ...(***cough Clevo **)

 

lol. You want to buy my t9300 with it's overclocking capability?

 

Tempting, but no. I'd rather get an SSD for that kinda cash.