[chiphell] kepler rumors suggest 15% better than 580.. price and transistors

[Skurge]

Where's that chart from?

http://www.anandtech.com/show/4239/nvidias-geforce-gtx-590-duking-it-out-for-the-single-card-king/5

EDIT: MrTeal as in MrTeal from GTA:Vice City?

 

[Vesku]

After looking at reviews more closely, I have to say I'm a bit disappointed in Techpowerup for using CPU limited results like the Starcraft II numbers for calculating their overall percentages. That game is clearly CPU bottlenecked, barely any difference from the 480 all the way up to the 590/6990. Including results like that really skewed their overall performance graphs.

Hope they clean that up before NVIDIA flagship launch.

 

[Quantos]

After looking at reviews more closely, I have to say I'm a bit disappointed in Techpowerup for using CPU limited results like the Starcraft II numbers for calculating their overall percentages. That game is clearly CPU bottlenecked, barely any difference from the 480 all the way up to the 590/6990. Including results like that really skewed their overall performance graphs.

Hope they clean that up before NVIDIA flagship launch.

Wow, I just looked at the SCII numbers, that's pretty stupid to be using that! Basically all the cards are the same in that test. What that means, though, is that if we assume that there's a test where the cards are all equal, the overall results should be more extreme.

 

[Arkadrel]

Yeah TechpowerUp does have slightly skewed results for alot of reviews because of stuff like that.

I also think its stupid they test things in 1280x768 resolution and put that into the "all resolutions avg"...
I mean really? who buys a 580 or 7970 and games at 1280x resolution?

 

[blackened23]

Yeah TechpowerUp does have slightly skewed results for alot of reviews because of stuff like that.

I also think its stupid they test things in 1280x768 resolution and put that into the "all resolutions avg"...
I mean really? who buys a 580 or 7970 and games at 1280x resolution?

Someone should go tell techpowerup that nobody cares about 720p benchmarks, and someone tell [H] to stop doing highest playable settings benchmarks.

 

[maddie]

After looking at reviews more closely, I have to say I'm a bit disappointed in Techpowerup for using CPU limited results like the Starcraft II numbers for calculating their overall percentages. That game is clearly CPU bottlenecked, barely any difference from the 480 all the way up to the 590/6990. Including results like that really skewed their overall performance graphs.

Hope they clean that up before NVIDIA flagship launch.

Yeah TechpowerUp does have slightly skewed results for alot of reviews because of stuff like that.

I also think its stupid they test things in 1280x768 resolution and put that into the "all resolutions avg"...
I mean really? who buys a 580 or 7970 and games at 1280x resolution?

Wow, I just looked at the SCII numbers, that's pretty stupid to be using that! Basically all the cards are the same in that test. What that means, though, is that if we assume that there's a test where the cards are all equal, the overall results should be more extreme.

Why are you guys saying those things about TPU reviews. The more games and resolutions the better. More information is NOT a bad thing. There are many people who still game at what would be considered ridiculous, by enthusiasts, resolutions.

In my opinion, they try to give everyone some data.

The averages can be affected by cpu limits in some games, but so what. if the game is played, then why not use it?

All of the data is there, do your own average for the games most important to you.

 

[Quantos]

More information is, but sometimes irrelevant to the product review. Including 1024x768 resolution in the review of a $500 card is useless, because it's a given that it's going to be way too powerful for the resolution anyway.

Also, other factors else than the card's actual performance can influence the results. See the following:

http://tpucdn.com/reviews/AMD/HD_7970/images/bf3_1024_768.gif

There's no much that makes sense in this picture. I'm not exactly sure what's going on, but it's clear that the actual relative performance of those cards isn't well represented there.

Edit: The whole point of performance summary is not to have to make our own conclusions. It should be the relevant performance summary, which it isn't really. Even if you look at a specific resolution only, you've got games as mentioned like SCII that are really irrelevant pieces of information.

 

[maddie]

More information is, but sometimes irrelevant to the product review. Including 1024x768 resolution in the review of a $500 card is useless, because it's a given that it's going to be way too powerful for the resolution anyway.

Also, other factors else than the card's actual performance can influence the results. See the following:

http://tpucdn.com/reviews/AMD/HD_7970/images/bf3_1024_768.gif

There's no much that makes sense in this picture. I'm not exactly sure what's going on, but it's clear that the actual relative performance of those cards isn't well represented there.

Edit: The whole point of performance summary is not to have to make our own conclusions. It should be the relevant performance summary, which it isn't really. Even if you look at a specific resolution only, you've got games as mentioned like SCII that are really irrelevant pieces of information.

It might surprise you to know that many people do not know that fact. Show them in addition to telling them.

 

[Vesku]

Yeah, I should probably send them an email about it because I like how they test a lot of games. It's just including games where all cards max out the fps doesn't convey much in the way of useful evaluation information, either find settings where the load is back on the GPU or exclude the CPU limited games from the GPU comparison chart (and note what you excluded and why).

 

[JAG87]

Ah, totally missed the "minimum" bit. No wonder the numbers seemed so odd. Honest mistake, Jag87?

Yup honest mistake. Sorry everyone!

 

[Quantos]

It might surprise you to know that many people do not know that fact. Show them in addition to telling them.

There's nothing stopping it from being a separate piece of information that doesn't influence the higher resolution's reviews. A simple note that recommends the minimum resolution for each card along with simple explanation explaining why could convey this information without affecting the results of the review.

 

[Skurge]

Yup honest mistake. Sorry everyone!

You're forgiven

 

[SirPauly]

Imho,

Over-all performance metrics are certainly not the end-all-be-all but the one that is very solid is the data from ComputerBase.