Rage3D new performance analysis method

[caveman-jim]

Used in the HIS Radeon 7970 review, take a look and let us know what you think.

http://www.rage3d.com/board/showthread.php?t=33997763

[KAC]

I still prefer the old min, max and avg methodology. Gives me all I need to know. Frametimes is not something I give a shite about and perceived smoothness etc is something that only matters in the most stressful of games. Adding it in is fine but removing the traditional min, max and avg will not be a good idea imo (I haven't read your complete article).

However, this brings me to another point that I have made at least once in the past. Unless you pit the competitor products against AMD, the reviews will always be not useful to me since that is the only thing I am concerned with when buying a card. Your reviews are good read and have usually some nice insights from a single vendor perspective but they become ultimately meaningless for any sort of card buying decision and remain in the mental masturbation category of my brain never moving into the action category.

Either way, good luck with your reviews and hope some people who have more attention span find it useful.

[caveman-jim]

Thank you for the feedback and those are valid points.

I am trying to get samples from NVIDIA and their partners, hopefully with the new generation due this year Rage3D can be on the sample lis, especially as we're now being more analytical of the cards more than just plain FPS. Short of buying them myself I don't have access, which I can't currently stretch to. Maybe I should start a 'fund me' campaign? I don't know how that would be received.

Minimum and Maximum FPS are still present, as well as a new perceived average which is more accurate than the simple average that FRAPS pumps out, IMO. Frametimes gives you a glimpse into what's not being told in the minimum and maximum FPS rates, because those are averages which can hide the stutters that people complain about. People differ in their sensitivity to them which is why I've blended lots of results.

As a single card look, it shows the effects of things like frame limiting and vsync as well as resolution, multi-GPU, which are interesting to people.

[Lazy8s]

Imo this is the only way cards should have ever been benchmarked.

Min can be deceptive. My Metro 2033 bench always gave me about an 18 fps minimum (oops, meant average), but that was due one single instance where (for whatever reason) framerate would plummet to about 10 fps, then immediately recover to the average 22-28 for the rest of the test. It was enough to skew the results.

Anything over 40 fps looks smooth *to me* so advertising a card that can hit a max of 150 fps really doesn't tell me what I would like to know.

Does the bench/game run smoothly to the human eye in spite of what may look like low minimums?

If yes, go to next res/batch of settings etc and try again.

The min/max and average are good and needed info, but it just doesn't tell the whole story imo.

Just my $.02 fwiw

[SIrPauly]

Thanks, nice to read reviews that try to be beyond frame-rate!

[Faceless Rebel]

I think frame time is very important, especially in multi-GPU configurations. In single-GPU configurations it's important too, if you've ever tried to play any game on Intel graphics you know this because if the incredibly inconsistent frame timing on Intel's awful drivers. But it's crucial to a good experience in multi-GPU configurations where microstuttering can make or break the gameplay experience.

[icecold1983]

well done jim, hopefully you get some samples to compare cards to their competition

[acroig]

Originally Posted by icecold1983

well done jim, hopefully you get some samples to compare cards to their competition

Agreed.

[NIGELG]

[noko]

Originally Posted by caveman-jim

Thank you for the feedback and those are valid points.

I am trying to get samples from NVIDIA and their partners, hopefully with the new generation due this year Rage3D can be on the sample lis, especially as we're now being more analytical of the cards more than just plain FPS. Short of buying them myself I don't have access, which I can't currently stretch to. Maybe I should start a 'fund me' campaign? I don't know how that would be received.

Minimum and Maximum FPS are still present, as well as a new perceived average which is more accurate than the simple average that FRAPS pumps out, IMO. Frametimes gives you a glimpse into what's not being told in the minimum and maximum FPS rates, because those are averages which can hide the stutters that people complain about. People differ in their sensitivity to them which is why I've blended lots of results.

As a single card look, it shows the effects of things like frame limiting and vsync as well as resolution, multi-GPU, which are interesting to people.

Might go over well if the tested card becomes a Rage3D member lottery card, or Rage3D member auction card which the money can go back into buying another card etc. The sucky cards that no one wants can be videoed in different ways to be recycled back into nature; from gasoline, fire crackers, garbage disposals to chainsaws, barbacues, bounce test, bridge toss, baseball bat etc. Probably will get a lot more hits once the news gets around Rage3D is about to take some Rage on a card

If all the cards are bought by the community, it would be harder to say the report was jadded by free gifts.

[demo]

Thanks for adding another useful metric to your reviews, R3D and Jimbo. Imo min/avg/max frame rates are still essential, but this is a great addition and helps paint a much clearer picture overall.

[caveman-jim]

Thanks for the feedback and suggestions guys, appreciated. Are there graphs or data points on there that don't need to be there or aren't as helpful?