Benchmarks Of AMD's Newest Gallium3D Driver

[pingufunkybeat]

It makes sense, since at least one result was higher than 60fps.

Still, I think that the more recent driver versions with all the latest patches should perform considerably better.

[elanthis]

Its about 100 developers working full time.

Dave.

A slightly more useful answer might include what the specific bottlenecks are that you'd put the first 5 of those 100 developers on. Pretend I was a GPU driver development rockstar, I was hired by AMD full-time, you were my boss, and you could tell me to work on anything you wanted in any particular order: what would you want me to do first?

[darkbasic]

I've talked it over with Alex before and he said vblank_mode 0 should work fine.

I tried and it doesn't. My HD3870 runs Openarena at 100fps @2560x1600 w/ Very High Quality (with mesa classic, gallium should be even better). How can an HD4870 do only 47fps with a lower resolution? .-.

[Beiruty]

It seems the Gallium3d is starving the HW. 80% of the time HD is just idling?
Any one is monitoring the GPU command queue?

[HokTar]

A slightly more useful answer might include what the specific bottlenecks are that you'd put the first 5 of those 100 developers on. Pretend I was a GPU driver development rockstar, I was hired by AMD full-time, you were my boss, and you could tell me to work on anything you wanted in any particular order: what would you want me to do first?

Maybe you should read Jerome's letter and the rest in that thread?

[smitty3268]

I tried and it doesn't. My HD3870 runs Openarena at 100fps @2560x1600 w/ Very High Quality (with mesa classic, gallium should be even better). How can an HD4870 do only 47fps with a lower resolution? .-.

The r600g driver is mostly CPU limited right now, i think. Maybe your cpu is just faster?

tested on a system with an Intel Core i5 750 quad-core CPU clocked at 2.67GHz,

[elanthis]

Maybe you should read Jerome's letter and the rest in that thread?

Indeed, that's a good read, and excellent work on the part of jglisse. Thanks.

[not.sure]

Maybe you should read Jerome's letter and the rest in that thread?

Interesting link, thanks.

[bridgman]

A slightly more useful answer might include what the specific bottlenecks are that you'd put the first 5 of those 100 developers on. Pretend I was a GPU driver development rockstar, I was hired by AMD full-time, you were my boss, and you could tell me to work on anything you wanted in any particular order: what would you want me to do first?

The first task would be to find the bottlenecks. With a graphics driver that is usually a pretty significant task on its own. The graphics driver and the GPU are largely decoupled by a large command buffer and periodic synchronization events so your normal "run the app and see where the time goes" approach doesn't work so well.

It's more like "run the app, see where the time goes, develop a theory, talk to a few people, if they don't totally demolish your theory then write a bunch of code, see if it makes that app go faster, be happy because it does, test your change on other apps, get PO'ed because your change makes the other apps run slower, repeat".

Finding bottlenecks would be a good job for the first 5 developers.

[bridgman]

BTW one of the differences between the proprietary drivers and the open source drivers is that the proprietary drivers are multithreaded (using a worker thread to do most of the processing) while the open source drivers do driver processing inline with application calls.

There are probably some simpler changes that can and should be made first, but as others have said finding the right spots to change is the hardest part of the job.

The driver probably is starving the hardware at this point (which would minimize the difference between fast and slow GPUs), but even that has not yet been confirmed. What we have seen is that overall performance is more a function of CPU speed than GPU power right now.