• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

Is 3Dmark synthetic valid?

Do you think synthetics are valid like 3Dmark05?

  • Yes

    Votes: 41 20.6%

  • No

    Votes: 66 33.2%

  • stability only

    Votes: 92 46.2%
  • Total voters
    199
  • Poll closed .
Brent_Justice said:
Valid for what?

I think that needs to be expanded on before one can answer the question...
Click to expand...

Don't bother Brent. mal usually thinks that the only benchmark that matters is the one he can win or at least compete in. Which sorta begs the question of why is he pushing 3D mark? Dunno myself, he's been getting his ass kicked in benchmarks for the last 5 years to people that have supposedly inferior hardware (usually anything that has something to do with AMD).

I could post up some linkage or mals greatest quotes but I'll let his posts on [H] forum speak for themselves. ;)
 
maleficarus said:
After starting a thread I thought this poll would be a good one to start
Click to expand...
I think that maybe the score they give is not valid but how you see your computer performing in them is valid. If I see that my computer is stuttering and choking as it trys to get past the stressful benchmarks of 3DMark03 then I know that its not going to perform in games as well as a computer that goes through 3DMark03 with flying colors.

Also, they are good to measure if your upgrade made a difference. For example, say you had a GeForce MX and your avg score in some test was 30 FPS. Then you got a 6800 GT (keeping the same drivers/ and not allowing any other big variables) and your avg FPS is 60 FPS or so. Nice way to give you an idea if your upgrade/ mod made any sort of impact.
 
maleficarus said:
After starting a thread I thought this poll would be a good one to start
Click to expand...

Why?

Whether you are looking at synthetic benchmarks or game scores, you have always been scoring at the bottom of the charts due to poor hardware choices...

So I don't see why you would care about either.
 
Theres a discussion going on in the forums on the HOCP bias or not as well as about synthetic and FRAPS benchies. Read that (link on HOCP main page) b4 U question or answer this thread it will enlighten the meaning of this Poll somewhat.
 
Synthetics used to be great for performance from one driver to the next, but after the cheating fiasco, I don't trust them for that anymore (which means I have to use timedemos instead - more accurate, but takes longer).

Synthetics are also great for a sneak peek at where we'll be at in 6 months engine-wise. They're great to look at, and make use of tons of advanced features.
 
Personally I don't trust them to compare one product verses another, but to compare what I do on my machine I like to use them to see the performance increase. Example: unlocking my 6800 OC to 16x6 v. 15x5 I used 3DMark to measure my performance increase.

 
KevinO said:
Personally I don't trust them to compare one product verses another, but to compare what I do on my machine I like to use them to see the performance increase. Example: unlocking my 6800 OC to 16x6 v. 15x5 I used 3DMark to measure my performance increase.

Click to expand...
Exactly... Too much variables between different systems make benchmarks invalid. It is only good to check for performance increase or for stability..,. Myself, i would trust real gaming performance only for the most accurate feeling.
 
3D Mark tells you one thing and one thing only - how well your computer runs that particular version of 3D Mark. Since it's not a game and I really could care less what I get for a score, it's worthless to me. I guess it's OK for testing stability.
 
I use 3dmark for stability testing, and when I want to sit back and enjoy a 3D demo.

I'm a big demo scene fan, so I like to sit back and watch 3D demos.
 
For me, these kinds of 3D benchmark apps are good for stability testing on new PC builds. I don't draw any conclusions however from a "score" that really means nothing. If I pass all of the tests after multiple loops and all is well, I move on from there with more testing and finally onto games. Because that's what I am, a gamer, not a benchmarker.

Although I understand that people are into that, kind of like drag racing, so that's cool if it's your thing.
 
people are just pissed off with 3dmark05 because all the money they dumped into systems to get 25k+ in 3dmark01 and 13k+ in 3dmark03 are only getting around 5000 in 3dmark05. as cards get faster scores will go up and sooner or later we will have a single card that can beat out 2 6800u in sli. and then when 3dmark06 or 07 comes out we'll be back where we started with low scores.
 
A benchmark is just that, what happens on the test bench. The numbers derived are pretty much the equivilent of the City and Highway MPG ratings you see on cars on the dealer showroom floor. No one expects to get those ratings when driving in the real world. Conditions vary.

The place where synthetic benchmarks work is in describing the relative power of a broad range of cards. It is useful and valid to be able to quantify the difference say of a 9600XT and a 6800GT. This can be useful for someone considering an upgrade and wants a really quick and dirty answer to see if it's even worth considering. If you take them as in the mileage rating analogy, synthetic benchmarks have their place.

However, many people buy their next videocard with a specific game in mind, and are trying to decide among three or four different cards in the same or adjacent price points. For that scenario the answer is obvious that the synthetic is not the tool to use.
 
As a Grad Student in Software Engineering I have heard lots of PhDs talk about benchmarks. They are synthetic, and nothing more. They are an attempt to reproduce actual usage and measure them. BUT in reality, systems are so complicated and random that it is literally impossible to capture any realistic set of features or measurements, unless as a tester, you decide to limit yourself to a very small set of tests.

For example, does overclocking my video card increase performance? This is a good test for a benchmark, since it involves very few variables.

However, using a benchmark to choose a new video card is a poor idea because a benchmark cannot accuratly simulate REAL world performance. The only benchmark that truly could, would be a collection of all the games that everyone every played, and that would be infinitly large.

So, I think (and it is an opinion, but one shared by many professionals, I think) that benchmarks have their place, but should be taken with a grain of salt.
 
SquireSCA said:
Why?

Whether you are looking at synthetic benchmarks or game scores, you have always been scoring at the bottom of the charts due to poor hardware choices...

So I don't see why you would care about either.
Click to expand...

My following enters....

Dave Squrriels......


LOL
 
tacomaboy said:
As a Grad Student in Software Engineering I have heard lots of PhDs talk about benchmarks. They are synthetic, and nothing more. They are an attempt to reproduce actual usage and measure them. BUT in reality, systems are so complicated and random that it is literally impossible to capture any realistic set of features or measurements, unless as a tester, you decide to limit yourself to a very small set of tests.
Click to expand...

Yup. If you want to see how something will actually perform, you have to look at its actual performance. Synthetic benchmarks can, at best, be approximations (to varying degrees of accuracy) of real-world performance. The only thing that 3DMark shows with complete accuracy is how well a given system runs 3DMark - everything else is an approximation.

Ever wonder why CPU manufacturers stopped advertising MIPS & FLOPS?



However, using a benchmark to choose a new video card is a poor idea because a benchmark cannot accuratly simulate REAL world performance. The only benchmark that truly could, would be a collection of all the games that everyone every played, and that would be infinitly large.
Click to expand...

Not infinite - just ridiculously large. There are a finite number of hardware combinations on the market and, in the worst case, a finite number of states that hardware can be in (if you wouldn't be satisfied with some particular standardized run of the software). All we really need to do is build a matrix of all possible hardware & figure out how long it takes it to get from one state to the next & then corellate that with the state transitions that appear in a given piece of software. ;)
 
3Dmark synthetics are great for testing the stability of changes and checking out relative differences between changes on MY rig with MY SETTINGS. The optimizations people use to get some of their record runs make a comparison to my rig absolutely worthless to me (i.e. turn off anti-virus, firewall software, etc.).

How can anyone make an arguement for the value of 3Dmark's worth in comparing different machines when the software install, operating systems options, and base computer configuration have such a tremendous impact?
 
However, using that logic you also nullify timedemos and other real-game tests that also were not conducted on your exact system.

The way this argument goes, the only valid thing to do is buy all of the cards you're considering, install each one on your system, and then return all but one. Thank goodness Best Buy doesn't have a restocking fee and 30-days for returns, eh?

Would anyone disagree with the following statement (using a bastardization of the old 80/20 rule):

"Generally speaking, when comparing system A and system B, if system A scores 20% higher on a synthetic benchmark, then system A is likely to exhibit better game performance 80% of the time when compared to system B given that the games in question are typical of games that were published at or around the same time as the synthetic benchmark."

I think all the other arguments are valid but can be reduced to the absurd if we're all willing.
 
rcolbert said:
However, using that logic you also nullify timedemos and other real-game tests that also were not conducted on your exact system.

The way this argument goes, the only valid thing to do is buy all of the cards you're considering, install each one on your system, and then return all but one. Thank goodness Best Buy doesn't have a restocking fee and 30-days for returns, eh?

Would anyone disagree with the following statement (using a bastardization of the old 80/20 rule):

"Generally speaking, when comparing system A and system B, if system A scores 20% higher on a synthetic benchmark, then system A is likely to exhibit better game performance 80% of the time when compared to system B given that the games in question are typical of games that were published at or around the same time as the synthetic benchmark."

I think all the other arguments are valid but can be reduced to the absurd if we're all willing.
Click to expand...


I am not sure I agree with that. The benchmark would have to encompass a lot of different game types to count for 80% of games. For instance, if you used a Doom 3 timedemo as a synthetic benchmark, a 20% improvement would not indicate that your card would have better performance in 80% of other applications.

But, synthetic benchmarks ideally try to simulate as many real world systems as possible, so, to a degree you have a point.

Of course, the best way to decide which card to get is to ask your friends how they like theirs, and play the games you like on other systems with those cards. But, that not always plausible.... so, sometimes you have to deal with thea fact that maybe you should have gotten the GeForce instead of the Radeon, etc etc.

Benchmarks are helpful, but should not be relied upon, I think.
 
3dmark is good for testing a new system.. If your score is similar to those with the same hardware as your own, it's a good indicator that you assembled and set the damn thing up correctly. It's all I use it for.
 
I think my 3DMark score is valid for everything:
- picking up girls - "Hey baby, see how big my 3DMark is? It gets bigger when you overclock my balls."
- impressing my boss - "I got a high 3DMark score, you should give me a raise."
- avoiding traffic tickets - "Please officer, you can see from my high 3DMark score, I run everything faster than all those other cars"

bah, that's all I got for now :rolleyes:
 
how about you just play a freaken game and if runs at your preferential level then you have a good card. F all these stupid benchmarks
 
That's great except for one thing -

How do you play the game without purchasing the hardware first?

That's the whole point of a benchmark - to get an idea about which hardware will do what you want without having to actually have the hardware in hand. It's not reasonable to get 4-5 CPU/Mobos and then slap a dozen different graphics cards in there to figure out which one you really want to work with.
 
Just came across this thread, and if it turns out there is something wrong with the original poster's system, then using the synthetic may in this case have provided some valuable diagnostic information beyond the rather subjective "I think my system should be running a little faster" feeling. Just a thought that synthetics have their place.

http://www.hardforum.com/showthread.php?t=840149
 
You must log in or register to reply here.
Back
Top