News for hackers. Stuff that matters.
Re:West Virginia Engineer Proves to Be a David to VW's Goliath by Gravityloss on Wednesday September 23, 2015 @10:58AM (#10265389)
Graphics drivers can detect when a speed test is being run, and degrade the output quality to get a higher frame rate.

I don't think anybody called that a mistake, it's just straight cheating.

If anything, they've probably seen that and thought it was a great idea!

http://techreport.com/review/3089/how-ati-drivers-optimize-q...

Re:West Virginia Engineer Proves to Be a David to VW's Goliath by gambiting on Wednesday September 23, 2015 @12:24PM (#10265997)
As someone who works in games industry - we have a line to AMD and NVIDIA open all the time. They change their drivers to work with our game just as much as we change our games to work with their drivers. Each driver has thousands of profiles for different games, and that has nothing to do with cheating - it's because no one writes perfect, 100% standard-compliant shaders or other gpu related code. The article you linked is ancient, but this still happens - drivers know that in your specific game there's a problem with the textures so they will automatically reduce the fidelity of texture shaders, or do some other changes. If you changed the name of our game internally, the drivers would no longer pick it up and you would lose 10-20 fps depending on the situation. That's not "cheating".
Re:West Virginia Engineer Proves to Be a David to VW's Goliath by MengerSponge on Wednesday September 23, 2015 @12:51PM (#10266201)
Optimizing the device's profile for a known input is generally celebrated. Deliberately misrepresenting the device's capabilities to deceive a test is what folks are generally upset about. It's a subtle distinction, but I'm happy to use Justice Stewart's "I know it when I see it" criterion.

http://www.tomshardware.com/reviews/live-memory-test,1241-10...

http://www.anandtech.com/show/7384/state-of-cheating-in-andr...

Re:West Virginia Engineer Proves to Be a David to VW's Goliath by ryandrake on Wednesday September 23, 2015 @01:43PM (#10266586)
To me, the line between cheating and ensuring good compatibility is drawn when you deceptively "optimize" for particular benchmark measurements at the expense of other qualitative un-measured or difficult-to-measure qualities. This undermines the purpose of the benchmark.

There's a huge difference between "Game XYZ uses these inputs and shaders, let's make sure our driver works as well as possible when we see this pattern" and "When we detect the user is running Benchmark ABC, make everything look like shit so we come out with a high benchmark score that we can use in our marketing materials!"

I'll admit, however, the line gets fuzzy when the game itself is used as a benchmark.

Re:West Virginia Engineer Proves to Be a David to VW's Goliath by Gravityloss on Wednesday September 23, 2015 @05:57PM (#10268533)
Come to think of it, with computer proliferation this will permeate everything, and already has.

Trying to buy airline or cruise tickets? Some kind of shenanigans are bound to happen. The system is trying to infer if you might be willing to pay a lot.

Though, in the old times this haggling was done by real people.