It's just idiotic to test MSAA when the non-MSAA test is running at 7fps
Seriously, michael, I don't know what you were thinking.
Thanks for the tests with some usable info, Marek.
Michael did run the tests with FullHD screen (1920x1080), maybe that's it?
Originally Posted by dungeon
Nah, when benching a GPU you want to run the games at as high a resolution and as high a graphical settings as the game offers so you can fully beat on the GPU and drivers as hard as possible.
Originally Posted by tomato
If this where a CPU test you'd do the opposite, you run it at minimum graphical settings at 640x480 or 800x600, whatever the bare minimum the game will allow you to run it at to see how fast the CPU can run the game engine without the GPU being the bottleneck.
The GPU test is closer to how most people actually play their games though. I.E. as high a detail settings as possible while not falling below the 30FPS threshold where the game will get choppy and laggy.
That's not true at all.. When you're talking 1080p + MSAA on an old card, the only thing you're testing is the memory bandwidth.. The CPU, GPU and GPU drivers are sitting there practically idle waiting for data to flow between the GPU and GPU memory..
Originally Posted by Kivada
In order for anti-aliasing to work, geometry gets scaled up to double their resolution in GPU memory. An old card, running at 1080p with anti-aliasing is absurd.. It was never done then, and no gamer is going to do it now on an old card because the framerates produced are too low to be playable and it has nothing to do with the GPU chip or the driver.
IMO, the benchmarks don't seem very useful at all.. He should be comparing low resolution + AA vs. high resolution without AA.. If high resolution looks better and is a lot faster, then you can conclude there is still some optimization work to be done in the AA.. AA has always been used by gamers as a way to get around needing to have a higher resolution display.
The only thing the benchmark shows, in my opinion, is that old cards were badly memory bandwidth bottlenecked.. Which gamers already know. When those cards were out, nobody had 1080p screens and if they did, they didn't need any AA.
Last edited by Sidicas; 01-16-2013 at 06:51 AM.
The issue is that (a) the same card produced much higher frame rates with the open source driver 18 months ago -- high enough that MSAA would probably have been useful even at 1920x1080 and (b) users are reporting higher frame rates even today which suggests it's not as simple as an across-the-board regression.
Could be that the older frame rates were wrong, or the new ones, or there's a partial regression, or some setting/configuration-dependent thing.
My suggestion is that developers issue a public recommendation/profile for Michael how to test properly and Michael reads own forums
Result : everyone happy
AMD should build a statue for Marek so everyone(incl. AMD) can ritually sacrifice him some "beers"
In fairness, we're not aware of anything wrong with the tests. The results just seem odd and we don't know why.
Agree on the Marek statue. Questions is where it should be built so that everyone could get to it. Maybe we could put a little picture of Marek in the lower right corner of the screen when running open drivers, just like the <testing> icon fglrx displays on untested hardware
Oh shit!! Now we even have Bridgman bowing to the mighty Marek!
Seriously You guys, both in AMD and independant developers, have been doing a fantastic job. I have a 6850 that I could only be slightly happier with. Keep up the good work guys.
Power management and video decode is really all thats left. Well at least for my needs anyway.
He, he, it is time for Michael to do some bisecting and not benchamarking