More Linux Benchmarks Of The AMD A8-3500M Fusion APU

[not.sure]

Isn't the AMD A8 the high end series of AMD APUs? And it has roughly the same CPU performance as the i3, which is Intel's entry level series?

I think it's more mid-range. Only later this year the actual new stuff bulldozer/trinity or something is positioned against i5/i7. Let's hope those parts are really good.

[FunkyRider]

4 core llano loses to 2 core sandy bridge in CPU benchmarks, what can I say, shame on you AMD! It's like Pentium D vs Athlon 64 single core, but in opposite sides

[blackshard]

4 core llano loses to 2 core sandy bridge in CPU benchmarks, what can I say, shame on you AMD! It's like Pentium D vs Athlon 64 single core, but in opposite sides

It means nothing. Hyperthreading in modern intel processors is very different than hyperthreading in old intel processors. We can roughlt say that Intel cores are more complex (in terms of execution units) but there are less, AMD cores are less complex, but there are more. I like much more the Intel approach because it will give more speed in single core applications.

Anyway you should compare performance per watt ratio and not performance per core ratio.

It would be really nice to see how graph changes in case of a different compiler.

[Fenrin]

I'll wait on Trinity either for my computer upgrade or for a new notebook. Vice President Rick Bergmann told that Trinity will be about 50% faster. It's not unlikely that Trinity will be released in Q1 2012, if we consider that AMD has already first samples.

My Phenom X2 II + Radeon HD 3200 works still ok for me. I can wait

[Jonno]

Isn't the AMD A8 the high end series of AMD APUs?

Yes, A8 is the high en series, but this is the "Mobile" model (A8-3xxxM = 35W) in that series. The "Mobile Performance" models (A8-3xxxMX = 45W) will be 20/26% faster (1.8/1.9GHz vs 1.5 GHz) and the desktop (A8-3xxx = 65W) and performance (A8-3xxxP = 100W) models will be even faster, rumours say up to twice as fast (3.0GHz vs 1.5GHz).

And while comparing a 35W Intel Core i3 with a 45W or 65W AMD A8 might seem unfair, remember that the Core i3 needs to be combined with a 35W NVidia GPU to get acceptable graphics performance, making it 70W total...

[psycho_driver]

The llano is basically an athlon II paired with a decent igpu. The Sandy Bridge integrated graphics were already fairly impressive for being on-die, and the anandtech preview of the desktop llano part showed that we can expect a bit more than double the performance from llano's gpu. Virtually everything they tested was playable.

Of course the really big deal will be that WoW will be very playable without a discrete GPU. This could win AMD a lot of battles with PC manufacturers. The biggest factor in llano's success will be how its priced. The speculative prices I've seen so far seem to be a bit on the high side if AMD is serious about taking back some market share.

The rumor that next year AMD will release a new Fusion processor based on the Bulldozer core with an igpu based on 6900 architecture is pretty interesting. Even if Bulldozer ends up being subpar compared to Sandy Bridge, with that sort of integrated gpu grunt it should still pose a pretty serious threat to Intel. I do not believe they can hope to catch up in that department between now and then. That probably has nVidia a little worried as well.

[Kivada]

As usual...

Surprisingly, Llano has an advantage when it comes to Nexuiz where the Radeon HD 6620G with the proprietary Catalyst driver is able to outperform the GeForce GT 425M also on its binary driver.

No... Theres actually no surprise at all in the Open Arena and World Of Padman results since they are so low quality they may as well be GLX Gears as far as the GPU is concerned. It's very apparent that they are CPU dependent at this point, the entire reason they are faster on the i3 is because of it's faster integer performance over the AMD K10.5 architecture. AMD has even stated that they don't care much about single threaded integer performance anymore, since its already plenty fast and that the real future is in multithreaded tasks.

Also, unless theres a huge difference between Win7 and Linux drivers, something is up in your multithreaded tests, since the A8 always passed the i3 in things like Cinibench, physics and 7zip in Win7.

[renkin]

Very nice

http://www.tomshardware.com/reviews/...u,2959-22.html

Very... very nice.

[chithanh]

Why were Catalyst 11.6 and Nvidia driver 275.09.07 not used for this review?

Also, Anandtech was heavily criticized for doing everything to make Llano look bad in their review. For example, they initially used DDR3-1333 memory and withheld DDR3-1866 results, until they were called out on it by their readers. Then they used low resolutions in some benchmarks (while even the mobile Llano is capable of running modern games at 1366x768).

[smitty3268]

No... Theres actually no surprise at all in the Open Arena and World Of Padman results since they are so low quality they may as well be GLX Gears as far as the GPU is concerned. It's very apparent that they are CPU dependent at this point, the entire reason they are faster on the i3 is because of it's faster integer performance over the AMD K10.5 architecture.

The Quake 3 engine has been known to be memory bandwidth limited for about 10 years now. Reviews of Llano show that to be a particular weakness, with faster RAM giving 50% speedups in benchmarks. So that's basically the worst case scenario for this hardware, and not representative of actual performance (which is why i always say that benchmarks that go over 100fps are of limited value - you start hitting performance bottlenecks that may never actually be hit in anything that matters - no one can see the difference between 300fps and 400fps).

Nexuiz is actually using the shader hardware, which explains why the results there are completely different.