AMD Llano "A8 3500M" (Fusion) review's

[jimbo75]

Llano has better performance per watt than SB in gaming, but in other things it doesn't

Performance per watt doesn't look better than i5 2520m in non-gaming applications
http://www.anandtech.com/show/4444/a...apu-a8-3500m/7

Actually it blows it away on cpu power draw as well but the A8-3500M review sample it is that turbo isn't working so its stuck on 4 core speed and can't get back into idle as fast on light loads.



Even without that it's still practically identical to the 2520M in power draw. Expect this to improve even more with the working bios.

Also note that according to hexus...

http://www.hexus.net/content/item.php?item=30828&page=5

we're doing the Llano laptop a disservice here because it can comfortably run the 720p looping clip on the power-saving state. Doing so increases time to hibernation to 307 minutes, or over five hours.

Next attempt?

[naokaji]

There are some tests on the Anandtech review that weren't very good: scaling was horrible (negative on some titles) and compatibility was awful (DX10/11 only, and with video corruption on some titles), but hopefully it will be at least partially sorted out when retail products start to show up.

Seems really odd AMD would allow previews of such a unfinished product.

90 GFlops @ 18w is pretty kick ass for OpenCL distributed rendering, btw...

How many decades to go until the majority of software is capable of exploiting the power on offer?

[Boissez]

... it needs fast RAM to shine: (from Anandtech)



Hopefuly OEMs won't cheap out too much with 1066/1333 DDR3

[undone]

... it needs fast RAM to shine: (from Anandtech)



Hopefuly OEMs won't cheap out too much with 1066/1333 DDR3

Seems 1600 is a breakthrough point, better than 1333 about 10-15% When comes to 1866 the gap become smaller. Looks like Llano doesn't need 2000mhz memory.

[dartaz]

Even without that it's still practically identical to the 2520M in power draw. Expect this to improve even more with the working bios.

i5 2520 and A8- 3500m are not comparable in CPU performance. i5 2520 is much faster.

For example, i5 consumed less power in H.264 test and it performed better in that test according to anandtech, so who have better performance per watt ?

[Motiv]

Seems 1600 is a breakthrough point, better than 1333 about 10-15% When comes to 1866 the gap become smaller. Looks like Llano doesn't need 2000mhz memory.

looks like any reviewer who uses 1333 in a llano review, will be ridiculed from this point onwards.

Quite impressive increases in speed going up to 1600, and cheap enough to be cost effective.

[Greg83]

For example, i5 consumed less power in H.264 test and it performed better in that test according to anandtech, so who have better performance per watt ?

it could be that i5, with that gpu had gpu acceleration of h.264 enabled where the Llano could be having a driverbug or it hasn't been configured properly to use the gpu acceleration of the h.264

I saw some slides that showed a UVD chip in there.

So that could explain why. Using the cpu for h.264 instead of the gpu part.

[kl0012]

I don't understand why people complain about the "not working BIOS" and lack of turbo. After all it was AMD which sends the notebook for testing. It seems like AMD thinks that the platform is mature enough.
BTW, the test platform has 14" screen while all Intel platforms have larger screens. So the comparisons are not completely equal.

[Boissez]

I don't understand why people complain about the "not working BIOS" and lack of turbo. After all it was AMD which sends the notebook for testing. It seems like AMD thinks that the platform is mature enough.
BTW, the test platform has 14" screen while all Intel platforms have larger screens. So the comparisons are not completely equal.

From the Anandtech (p)review:
"It’s also early hardware, as we haven’t received anything from the usual suspects, but performance and battery life should be representative of what we’ll see in shipping hardware. There’s still room for BIOS, firmware, and driver optimizations, so if anything we’d expect some scores to even improve from what we’re reporting, but for now we can get a starting point for what to expect from shipping Llano laptops and notebooks"
It's not final hardware and there's obviously room for improvement as both turbo and ACF seems somewhat broken so far.

[marten_larsson]

While you are correct that AMD seems to think it's mature enough doesn't mean it can't improve with retail samples. I mean, if this equals "as bad as it gets" then it's looking really good since it's probably not going to get worse with working turbo and better power saving features...

[FlanK3r]

awesome performance!

[PatRaceTin]

i want to see some desktop FM1 overclock

[undone]

looks like any reviewer who uses 1333 in a llano review, will be ridiculed from this point onwards.

Quite impressive increases in speed going up to 1600, and cheap enough to be cost effective.

Not to mention even when you choose dedicated GPU for gaming, the DRAM freq still slightly affect gaming perf. If take this into account you'll find the bandwidth isn't problem for APUs.
Sure if you have faster DRAM the gaming will be faster cuz you have both fast system memory & graphic memory at the same time, especially when you overclock it.

[Lightman]

I don't understand why people complain about the "not working BIOS" and lack of turbo. After all it was AMD which sends the notebook for testing. It seems like AMD thinks that the platform is mature enough.
BTW, the test platform has 14" screen while all Intel platforms have larger screens. So the comparisons are not completely equal.

You have a valid point.
We really need some tool to display how Turbo is behaving in Llano. I suspect AMD might have set test platform BIOSes very conservatively to look really good in battery tests. It's their choice. On the other hand Llano might really be TDP limited, but from tests with dGPU (fGPU disabled) it doesn't look like this is the case.
I wish some laptop manufactures will give users BIOS option to either use Turbo to it's fullest potential or set it conservatively to preserve battery.

I hope next month desktop launch will shed more light on this topic as well as some upcoming retail laptop tests.

[Apokalipse]

i5 2520 and A8- 3500m are not comparable in CPU performance. i5 2520 is much faster.

Llano has much better GPU performance.
CPU performance is not *supposed* to match or beat i5.

For example, i5 consumed less power in H.264 test and it performed better in that test according to anandtech, so who have better performance per watt ?

*when testing against the preview system with improper BIOS support (broken turbo and P-states).
Also, that's called cherrypicking your tests.

[Manicdan]

... it needs fast RAM to shine: (from Anandtech)



Hopefuly OEMs won't cheap out too much with 1066/1333 DDR3

wheres the source of this? id like to see the timings used

[Apokalipse]

I don't understand why people complain about the "not working BIOS" and lack of turbo. After all it was AMD which sends the notebook for testing. It seems like AMD thinks that the platform is mature enough.

Because the performance and power consumption with the broken BIOS are not representative of what it will be with a proper BIOS on release.

wheres the source of this? id like to see the timings used

The image is directly linked from anandtech

[Manicdan]

The image is directly linked from anandtech

theres no link, id like to see words that go with the chart, im not arguing if its a valid chart.

[Boissez]

wheres the source of this? id like to see the timings used

It's from Anands tweets. It's CL7 for the 1333 and 1600 runs and CL9 for the 1866.

[Manicdan]

It's from Anands tweets. It's CL7 for the 1333 and 1600 runs and CL9 for the 1866.

thanks, that could explain the massive jump for 1600mhz, while a mediocre jump for 1866.
whats needed next is the MBs in read/write/copy and how much that changes relative to FPS increase.

[Hornet331]

looks like any reviewer who uses 1333 in a llano review, will be ridiculed from this point onwards.

Quite impressive increases in speed going up to 1600, and cheap enough to be cost effective.

Depends... Notebooks still are in the 1066/1333 range and i doubt they are going to be faster this year. Desktop on the other hand, there is no reason to not use ddr-1600. But right now its seems the standard is 1333mhz for the big oems (HP, DELL, etc.) they should press the OEMs to pair it with faster ram.

[[XC] hipno650]

hmm interesting with the RAM bandwidth chart. It's good news for system builders considering 1333 the same price as 1600 in our area of the market. for OEM's however they have ALWAYS cheaper on RAM in the past and I see no reason for them to start changing things up. Especially at this price point. maybe near the end of Lano's life OEM's will start using 1600 in desktops but I can't see that trending into this price range of laptops. also for timings, form what I under stand your average GPU dedicated graphics memory has timings MUCH higher than average desktop RAM so the timings should make virtually no difference in performance. I could be wrong on that however.

[duploxxx]

Depends... Notebooks still are in the 1066/1333 range and i doubt they are going to be faster this year. Desktop on the other hand, there is no reason to not use ddr-1600. But right now its seems the standard is 1333mhz for the big oems (HP, DELL, etc.) they should press the OEMs to pair it with faster ram.

they did stress the OEM to user fast DDR for the higher range A series APU, just like they did stress to always use both memory channels.
INdeed on mobile there will be lots of 1066/1333 but this should be updated to 1333/1600 for liano's sake.

hmm interesting with the RAM bandwidth chart. It's good news for system builders considering 1333 the same price as 1600 in our area of the market. for OEM's however they have ALWAYS cheaper on RAM in the past and I see no reason for them to start changing things up. Especially at this price point. maybe near the end of Lano's life OEM's will start using 1600 in desktops but I can't see that trending into this price range of laptops. also for timings, form what I under stand your average GPU dedicated graphics memory has timings MUCH higher than average desktop RAM so the timings should make virtually no difference in performance. I could be wrong on that however.

while OEM might charge more they pay almost the same price for 1333/1600 give or take 1-2$
indeed the ram will have much higher timings on notebooks so the advantage will be less

[Manicdan]

they are probably opted to use cheaper ram so theres reasons for people to buy nicer stuff. just like entry model cars coming with steel rims and hubcaps, even on a 20k$ car

[xlink]

The new IGP falls somewhere between a mobile Radeon 5570 (If there is such a thing) and a GT540m (Well, that's exaggerating a bit). With nVidia, you get much better drivers, better compatibility with Linux, and a proven product. People are going to be hard-pressed to invest in AMD given their recent track record for not fixing their graphics drivers...

umm...
nVidia's drivers are far more prone to crashing than AMD's
and their GPUs are less energy efficient.
and their driver interface is less polished

it might be a pain to mess with AA settings, but no one does that outside of the enthusiasts. These products aren't aimed at enthusiasts.
speaking of irrelevant things, NO ONE CARES ABOUT LINUX SUPPORT. If you're attempting to game in linux, may the Lord have mercy on your soul - I gave up on that long ago.