AMD A8-3500M Llano Linux Benchmarks

[phoronix]

Phoronix: AMD A8-3500M Llano Linux Benchmarks

This morning there will be an introduction by Advanced Micro Devices of Llano, their next-generation Fusion APU. In this article are some launch-day Linux benchmarks of this newest 32nm accelerated processing unit using Ubuntu 11.04 with the A8-3500M part. This article is in continuation of our Radeon HD 6620G Linux benchmarks that were published at midnight.

http://www.phoronix.com/vr.php?view=16145

[werfu]

I guess these new processors are a lot more performant, but I think you missed the spot by comparing them with the E-350 and the Atom. The A8-3500M isn't targeted at the same market and I'm pretty sure this will be seen on power consumption and pricing.

About power consumption monitoring, the Phoronix Test Suit could monitor power drawn from the batteries by watching /proc/acpi/battery on systems that support it.

[Michael]

About power consumption monitoring, the Phoronix Test Suit could monitor power drawn from the batteries by watching /proc/acpi/battery on systems that support it.

The Phoronix Test Suite already has full support for monitoring power consumption, among other sensors, but still can't handle it remotely when I can't yank the power cord across the Internet for the system I had access to and there wasn't any IPMI power switch or other means...

[werfu]

The Phoronix Test Suite already has full support for monitoring power consumption, among other sensors, but still can't handle it remotely when I can't yank the power cord across the Internet for the system I had access to and there wasn't any IPMI power switch or other means...

LOL, totaly true! Some system can stop the battery from charging even plugged, maybe it could be a possibility.

BTW, glad to see you're so on the edge with this product launch. KUTGW!

[ssam]

A8-3500M [1]
2011, quadcore, 35W TDP

atom 330 [2]
2008, dualcore, 8W TDP

maybe an i3 or i5 would be a more fair comparison. performance per watt is pretty important. (unless intel and AMD have vastly different ways of measuring TDP).

[1] http://en.wikipedia.org/wiki/AMD_Fusion
[2] http://en.wikipedia.org/wiki/List_of...icroprocessors

[Michael]

maybe an i3 or i5 would be a more fair comparison. performance per watt is pretty important. (unless intel and AMD have vastly different ways of measuring TDP).

[1] http://en.wikipedia.org/wiki/AMD_Fusion
[2] http://en.wikipedia.org/wiki/List_of...icroprocessors

Once AMD sends out some hardware I'll worry more about a proper comparison.

[Cyborg16]

Doesn't look that amazing in terms of CPU performance: http://www.tomshardware.com/reviews/...u,2959-16.html

Shocking that AMD let you benchmark with the restriction of using only netbook processors as comparators though!

[PsynoKhi0]

Doesn't look that amazing in terms of CPU performance: http://www.tomshardware.com/reviews/...u,2959-16.html

Not surprising since:
- The first Llano gen is based on the Athlon II/Phenom II architecture
- More focus has been put on balancing cpu/gpu performance than raw x86 punch
- Llano is a massive improvement over previous mid-range mobile platforms from AMD in terms of battery consumption, so I guess a fair bit of engineering went into this as well

Shocking that AMD let you benchmark with the restriction of using only netbook processors as comparators though!

As for Llano, hoping to see more about price points in retail soon. (And yes, a comparison with i5/i7 would have been more adequate, though it seems that'll have to wait)

[smitty3268]

From other reviews (non-linux):

CPU gets crushed by sandybridge.
GPU is typically 50-100% faster than sandybridge. Although heavily CPU dependent games (like Starcraft 2) can bring performance down to on par.
Price is cheaper than sandybridge.
Power usage is on par with sandybridge.

[curaga]

Anand's desktop preview showed the cpu to be ~ equivalent to an Athlon II x4 645.

Now it only depends on price


Oh, and one of the AMD slides showed driver-independent GPU power management. Finally proper idle gpu wattage on open drivers w/o forced low clocks?