Open vs. Closed Driver Benchmarks Of AMD's A10-7850K Kaveri On Ubuntu 16.04

[phoronix]

Phoronix: Open vs. Closed Driver Benchmarks Of AMD's A10-7850K Kaveri On Ubuntu 16.04

A number of Phoronix readers have been requesting some fresh AMD Kaveri Linux graphics driver benchmarks, so here you go. For your viewing pleasure today is an AMD open vs. closed-source driver comparison on Ubuntu 16.04 plus some extra runs featuring upgrades to the Linux kernel and Mesa as well as manually enabling DRI3 support.

http://www.phoronix.com/vr.php?view=22855

[haplo602]

APU benchmark without ANY mention of memory sticks used. APUs are notoriously memory speed dependent for graphics workloads. Thanks for adding crucial information to the article.

Otherwise very nice tests ...

[duby229]

Yeah, that is very true. But as long as Catalyst was benched on the same hardware at the same settings, then the important part is the difference in performance. It isn't possible to exactly replicate the way micheal likes it.... But he was always wrong on that point anyways.... The point in benchmarking was always to measure the performance of the hardware and software you use in the configurations that you use. Which necessarily means replication isn't usually possible. Because hardware is usually different between machines, settings are usually different between machines.

EDIT: In other words, the point in benchmarking is usually to -find- the best possible settings. Measuring the difference -between- settings. Sometimes he does this as a side affect, but rarely does he make it the point.

[Michael]

APU benchmark without ANY mention of memory sticks used. APUs are notoriously memory speed dependent for graphics workloads. Thanks for adding crucial information to the article.

Otherwise very nice tests ...

It was DDR3-2133 4GB x 2. It would be mentioned in the article, but Linux still doesn't sadly allow way to expose the data nicely unless you're running as root. So only when benchmarking as root does the system table show the frequency / topology / model numbers of memory.

[Kano]

All games tested I never play. Most likely I would compare DiRT Showdown and L4D2 or TF2 (or CS:GO if I would have got it).

[haplo602]

Yeah, that is very true. But as long as Catalyst was benched on the same hardware at the same settings, then the important part is the difference in performance. It isn't possible to exactly replicate the way micheal likes it.... But he was always wrong on that point anyways.... The point in benchmarking was always to measure the performance of the hardware and software you use in the configurations that you use. Which necessarily means replication isn't usually possible. Because hardware is usually different between machines, settings are usually different between machines.

EDIT: In other words, the point in benchmarking is usually to -find- the best possible settings. Measuring the difference -between- settings. Sometimes he does this as a side affect, but rarely does he make it the point.

I got that. I just wanted to know if there's room to grow (2400MHz memory in this case) or Michael already used the best possible. So far the gap to Catalyst is negligible in my view.

[tomtomme]

Phoronix: Open vs. Closed Driver Benchmarks Of AMD's A10-7850K Kaveri On Ubuntu 16.04

A number of Phoronix readers have been requesting some fresh AMD Kaveri Linux graphics driver benchmarks, so here you go. For your viewing pleasure today is an AMD open vs. closed-source driver comparison on Ubuntu 16.04 plus some extra runs featuring upgrades to the Linux kernel and Mesa as well as manually enabling DRI3 support.

http://www.phoronix.com/vr.php?view=22855

whats the openbenchmarking url, so that I can compare to the intel results?
I searched for "kaveri 7850 ubuntu 16.04" on openbenchmarking but did not find it.

[jakubo]

so apart from the usual nagging... what do these graphs tell us? is it cpu or RAM bandwidth bottlenecked? is there a copy less in fglrx? apparently they have some kind of "optimisation" for higher resolutions. how would the relationship shift when CPU freq was increased or held down for all tests? different ram frequencies applied? is the si-scheduler on? is there still an issue with CPU not using highest frequencies due to load no being high enough?
ok, just kidding.
looks quite nice as far as i can tell.

[duby229]

I got that. I just wanted to know if there's room to grow (2400MHz memory in this case) or Michael already used the best possible. So far the gap to Catalyst is negligible in my view.

I agree. It's all that was ever promised. And in a lot of cases the OSS drivers are performing much better than promised.

I still suspect that Catalyst is using some hardware somewhere to schedule loads where the OSS drivers are relying on the compiler to order instructions.

[haagch]

whats the openbenchmarking url, so that I can compare to the intel results?
I searched for "kaveri 7850 ubuntu 16.04" on openbenchmarking but did not find it.

Look at the image with the specs: its url contains 1602217-GA-AMDKAVERI64. That's the id of the test.
Then you can go to http://openbenchmarking.org/result/1...GA-AMDKAVERI64