Benchmarks Of 45 Linux Systems From Atoms, Athlons & Phenoms To Xeons

[phoronix]

Phoronix: Benchmarks Of 45 Linux Systems From Atoms, Athlons & Phenoms To Xeons

For curiosity sake and as part of a more interesting article coming later this week in celebrating 11 years of Phoronix with its birthday on Friday, here are a range of CPU benchmarks on 45 different Linux systems with quite a range of hardware from low-power Intel Atoms and many AMD APUs to dual socket Opterons and Xeons. There's also a mix from laptops to nettops to desktops and servers.

http://www.phoronix.com/vr.php?view=21732

[coats]

For the computational benchmarks, I'm curious how much an additional "-march=native" (which tells gcc "target this computer") would affect the times. It would seem, for example, that the Haswell times don't benefit from the available AVX2 instructions...

[Michael]

For the computational benchmarks, I'm curious how much an additional "-march=native" (which tells gcc "target this computer") would affect the times. It would seem, for example, that the Haswell times don't benefit from the available AVX2 instructions...

I've done such looks in the past - http://www.phoronix.com/vr.php?view=20169 - http://www.phoronix.com/scan.php?page=search&q=tuning.

Though could easily make another benchmark ticket that sets the CCFLAGS/CXXFLAGS across all the systems and then re-runs the tests... Could if there's enough interest level.

[alpha_one_x86]

And your ARM system??? Compute utilite? Other ARM...?

[Michael]

And your ARM system??? Compute utilite? Other ARM...?

At the moment don't have any ARM systems as part of the daily testing... but you can see some within the Compute Stick tests yesterday.

[peapa]

There is some issue with the i5 2500k results being labelled as Radeon HD 4890.

[Michael]

It's not an issue per se. Normally that system is running graphics tests and prior to running this testing simply forgot to rename the system identifier.

There is some issue with the i5 2500k results being labelled as Radeon HD 4890.

[xeekei]

Odd that FX-8350 is slower than FX-8150 in some tests.

I would also like the march=native tests.

[tomtomme]

Odd that FX-8350 is slower than FX-8150 in some tests.

I would also like the march=native tests.

No it isn?t. the 8350 has a HDD and 3.13 kernel while the 8150 system has a SSD and the 4.1 kernel and other components / software differ too...

[whitecat]

Michael,

It would be interesting in your CPU benchmarks that you also monitor the performance on only 1 thread.
At work, I have seen many times application/server deployment where the application is not optimized to use all the processor's cores but needs performance anyway.

To complete this, it would be interesting to show, for instance, the difference in mono thread between a Xeon E3-12xx v3 which have 8 MB L3 and different Xeon E5-16xx v3 which have 10/15/20 MB L3.