AMD Fusion APU Llano Bulldozes Sandy Bridge in Multi-Tasking Demonstration

i want.. i want.. god damnit AMD release the Llano already!!!
 
If this is real... the "A8-3510MX" is one hell of a mobile apu. (Please AMD, for God's sake, change that name! Phenom !!! (3) has a nice ring to it...)

In the specs there was a "Radeon 6620M Graphics", knowing that every HD6-series part below the 6950 is VLIW-5 or "Evergreen"-based, this specific chip has graphics capabilities between Redwood/Cedar and Juniper. If Llano only has 480sp as rumored, it is pretty impressive.

I wonder if there will be desktop versions of Llano, would make for a nice low-cost, all-around system.
 
And of course when we have both HW in the open, nobody would be able to replicate the performance. then somebody will discover that AMD used a pre-release Sandy Bridge mobo, with buggy drivers, that it did not have the proper video driver loaded, that it had 5 HDDs instead of one and so on.
Kinda of like Nvidia benchmarked its soon-to-be-released Tegra vs. Core CPU from 2006 in notebooks and used GCC 4.4 for Tegra and 3.5 for Core and claiming a resounding win. But when people used also GCC 4.4 on the Core platform it outscored the Tegra by 30%...

http://www.softpedia.com/reviews/wi...-Intel-s-Core-2-Duo-T7200-Review-185406.shtml
 
If this is real... the "A8-3510MX" is one hell of a mobile apu. (Please AMD, for God's sake, change that name! Phenom !!! (3) has a nice ring to it...)

In the specs there was a "Radeon 6620M Graphics", knowing that every HD6-series part below the 6950 is VLIW-5 or "Evergreen"-based, this specific chip has graphics capabilities between Redwood/Cedar and Juniper. If Llano only has 480sp as rumored, it is pretty impressive.

I wonder if there will be desktop versions of Llano, would make for a nice low-cost, all-around system.

no these should not be called phenom III's. these are the low power APU processors and should not in any shape or form carry the phenom III name. while i wish they would of used the phenom III name with bulldozer i kind of understand why they went away from it and i really dont mind since i'm sick of seeing people calling the phenom II's PII's. the abbreviation was already taken by the pentium II, use some friggin creativity when you come up with an abbreviation or just call it a phenom II. though with the new naming i really hope they use the FX part to signify a black edition/unlocked processor like the original FX processors instead of just calling them all phenom FX and then using the black edition add-on for the unlocked versions. that is if they have any unlocked versions of the bulldozer.

my guess is the 6620 is just a higher clocked version of the 6300 with maybe a few more shaders.

yes they are using the VLIW-5 shader technology in these and not the VLIW-4 that are used on the 6950/70/90.

and yes there will be a desktop version using m-itx motherboards. asus, gigabyte, sapphire already have versions with the e-350 brazo versions.

And of course when we have both HW in the open, nobody would be able to replicate the performance. then somebody will discover that AMD used a pre-release Sandy Bridge mobo, with buggy drivers, that it did not have the proper video driver loaded, that it had 5 HDDs instead of one and so on.
Kinda of like Nvidia benchmarked its soon-to-be-released Tegra vs. Core CPU from 2006 in notebooks and used GCC 4.4 for Tegra and 3.5 for Core and claiming a resounding win. But when people used also GCC 4.4 on the Core platform it outscored the Tegra by 30%...

http://www.softpedia.com/reviews/wi...-Intel-s-Core-2-Duo-T7200-Review-185406.shtml

its a manufacture demo, i'm sure they are canned benchmarks. we will probably see the same thing from intel with the sandy bridge against the Llano saying sandy bridge is better and trying to make AMD look like they cheated. its nothing new, been like this since the beginning with Intel and AMD and well any other hardware/software manufacture.

i didn't go into seeing the demo to see how much better it was then the sandy bridge processor, i just wanted to see how much multi-tasking they actually did and what they used. but the big thing for me was power usage and quite frankly it looks pretty damn good.
 
Last edited:
Ok but this was a demo comparison of laptop processors no? I know (or think I know) that the Llano chip spans laptops and desktops - meaning it will be used for both - but my question is: Is this performance indicative of the what the desktop chips will produce also or will there be more powerful, higher freq Llano chips that will be able to compete with the 2600K and such?

I might not be making any sense but I guess what I am saying is that, yeah, this Llano looks great compared to Intel's mobile chipset and we should be glad about that but what about vs. the desktop versions?
 
Brazo is AMD's ultra low power netbook/HTPC APU processor. Llano is AMD's low power laptop/notebook/desktop APU processor. bulldozer is AMD's high end performance processor.

Llano and brazo are not meant to compete with any of intels desktop processors. they are meant to compete against Intel's atom(AMD's brazo) and Intel's mobile sandy bridge processors(AMD's Llano).

yes there will be desktop versions of the Llano and yes there are desktop versions of Brazo but those are a secondary part created by motherboard manufactures, those are not the primary purpose of the processors.
 
It wouldnt surprise me at all to find out thats real.. Also if you were to change the order the Intel should stomp the AMD in most tests there, but amd put intel's weakest link 1st to slow down everything else.
 
Showing off the graphics portion of the APU kinda irks me because it doesn't show any CPU-to-CPU comparisons, but they were putting a decent amount of stress on both there it appears. I'd really like to see the CPU load, and I'd also like to see Win7 get an update that will show the GPU load on these new hybrid chips in the task manager. Either way, AMD's APU is faster than Intel's APU at certain things. We'll still need to wait a bit longer before we can put them into a more fair arena.
 
If this is all fair and good, then this is an impresive showing by AMD. Since this is a quad core, I am assuming that is is using 2 Bulldozer modules with 4 integer cores and 2 FPU's.

However how hard is demo stressing the graphics portion of the APU versus the CPU portion? I wonder about this because we know that AMD's APU's have done better jobs handling graphics.
 
If this is all fair and good, then this is an impresive showing by AMD. Since this is a quad core, I am assuming that is is using 2 Bulldozer modules with 4 integer cores and 2 FPU's.

However how hard is demo stressing the graphics portion of the APU versus the CPU portion? I wonder about this because we know that AMD's APU's have done better jobs handling graphics.

Nope... Llano is based on the 'Stars' cores, aka Deneb, Thuban, Propus... K10.5 arch, not Bulldozer.
Posted via Mobile Device
 
If this is all fair and good, then this is an impresive showing by AMD. Since this is a quad core, I am assuming that is is using 2 Bulldozer modules with 4 integer cores and 2 FPU's.

However how hard is demo stressing the graphics portion of the APU versus the CPU portion? I wonder about this because we know that AMD's APU's have done better jobs handling graphics.

these are NOT bulldozer modules. This is Llano, not bulldozer- bulldozer is the high-end desktop part, not the laptop/low powered desktop.
 
Nope... Llano is based on the 'Stars' cores, aka Deneb, Thuban, Propus... K10.5 arch, not Bulldozer.
Posted via Mobile Device

these are NOT bulldozer modules. This is Llano, not bulldozer- bulldozer is the high-end desktop part, not the laptop/low powered desktop.

Ahh, thanks for clearing that up guys. If it is based on the current Stars, then CPU performance probably isn't going to be all that great.
 
And of course when we have both HW in the open, nobody would be able to replicate the performance. then somebody will discover that AMD used a pre-release Sandy Bridge mobo, with buggy drivers, that it did not have the proper video driver loaded, that it had 5 HDDs instead of one and so on.
They list the mobo/drivers at the end of the video, don't seem to be playing games there. Given how Llano's GPU is supposed to have up to 400 SP's its not unreasonable to expect it to whomp pretty much any SB in graphics heavy work loads.
 
but my question is: Is this performance indicative of the what the desktop chips will produce also or will there be more powerful, higher freq Llano chips that will be able to compete with the 2600K and such?
Mobile chips are just binned versions of desktop chips, selected for power usage over clock speed. So on a per clock basis the desktop and mobile versions of Llano will perform the same but the desktop versions will likely be the fastest over all since they'll have a higher thermal capacity to work with which means higher clocks is a given.

WAG on my part but since the Llano CPU core itself is a tweaked version of the current 4 or 2 core Athlon II's I'd expect clock speed to be pretty similar to that for both mobile and desktop versions. So around 1.8-2.8Ghz for the mobile 2 and 4 core parts, with the 2 core parts getting the highest clocks and the 4 core have the lowest clocks due to higher heat. Desktop parts will probably top out at around 3.3-3.2Ghz or so, but the higher power envelope means you can have 4 instead of 2 cores at that speed.

SB will have the same per clock performance advantage over Llano as it does over PhenomII/AthlonII. So SB will still beat it handily in CPU performance, but over all system performance will be better with Llano. If you're looking for a AMD competitor for the i5-2600K you'll have to wait and hope that BD is up for the task.
 
It wouldnt surprise me at all to find out thats real.. Also if you were to change the order the Intel should stomp the AMD in most tests there, but amd put intel's weakest link 1st to slow down everything else.
What? You're not making any sense at all. Running the FF-whatever demo last instead of first wouldn't make it any less jerky on the Intel system at all.

What matters is the work load, not the order that you start the work load in.
 
They list the mobo/drivers at the end of the video, don't seem to be playing games there. Given how Llano's GPU is supposed to have up to 400 SP's its not unreasonable to expect it to whomp pretty much any SB in graphics heavy work loads.

You are right of course.

What puzzled me was how badly SB performed in the multitasking scenarios. Also it was dropping frames in the game, although FPS was well over 25. At least what I could see Llano was getting around 2x as much FPS as SB. So far so good.

But why did SB crawled to an almost full stop in the rendering ? I cannot explain that.
 
You are right of course.

What puzzled me was how badly SB performed in the multitasking scenarios. Also it was dropping frames in the game, although FPS was well over 25. At least what I could see Llano was getting around 2x as much FPS as SB. So far so good.

But why did SB crawled to an almost full stop in the rendering ? I cannot explain that.

my guess is due to resource usage between the igp and cpu it was causing some weird bottleneck. because doesn't the SB use portions of all the cores for the video playback? some one correct me if i'm wrong since i really cant remember. i dont pay attention much to low end intel products
 
CPU, GPU, and media engine share the L3 cache via a ring bus. As far as anyone knows there isn't any other resource sharing going on with SB.
 
A 25W version will either have a very low clocked core and/or low clocked GPU. It'd be better than current AMD or Intel IGP's no doubt but I don't think you'll get the level of performance we saw in the video. It'd be like having a dual core ~1.8Ghz Athlon II + 5470M laptop, which isn't bad at all for that power envelope, particularly if they price it right.

The Llano in the video probably a mid or high end mobile Llano given the watt usage reported (ie. high 40's to low 50's). Or maybe even a low watt desktop version.
 
no doubt, your analysis looks sound but all i am after is, enough power for a genuine productivity/gaming experience, in a package low-power enough to provide six hours battery life in a slimline 12.1" chassis, at a price-point which starts at £500.

amd brazoz cannot provide this as it is too gutless
amd danube cannot provide this as it's too high power
intel SB cannot provide this as its expensive with duff graphics

roll on dual-core llano.
 
WAG on my part but since the Llano CPU core itself is a tweaked version of the current 4 or 2 core Athlon II's I'd expect clock speed to be pretty similar to that for both mobile and desktop versions.

Llano is also on a completely new process. They are probably gunning for low power and won't go after super high clockspeeds but the 32nm process may allow them more TDP headroom should they want to pursue it.

Would love to see some black edition desktop versions released to see just how high the new process can go in the hands of a pro with LHe. At 45nm the Phenom II record was 7GHz. If you could disable the built in GPU I wonder if these could get a tad closer to 10 :D
 
all i am after is, enough power for a genuine productivity/gaming experience, in a package low-power enough to provide six hours battery life in a slimline 12.1" chassis, at a price-point which starts at £500..
6 hours would be pushing it unless the lappy had a 7 or 9 cell high capacity battery, everything else sounds doable with a ULV Llano IMO. For reference the Nile platform (dual core 1.3Ghz NeoII+4250 IGP) used around the same amount of power and got somewhere around 4-5 hours battery time in a net book form factor with a 6 cell battery.
 
Llano is also on a completely new process. They are probably gunning for low power and won't go after super high clockspeeds but the 32nm process may allow them more TDP headroom should they want to pursue it.
The GPU will eat up all the TDP headroom and maybe even then some since power doesn't scale linearly with clocks. A 3.5Ghz quad + top clocked IGP at top TDP (90W IIRC for desktop Llano) sounds about right.

At 45nm the Phenom II record was 7GHz.
With sub zero cooling though right? I wouldn't be shocked if you could get well north of 4Ghz on air with the Llano CPU if you could disable or heavily down clock the IGP. But 5Ghz, much less 7 or 10Ghz, is probably unreasonable. Maybe BD will get to 5Ghz. Maybe...
 
6 hours would be pushing it unless the lappy had a 7 or 9 cell high capacity battery, everything else sounds doable with a ULV Llano IMO. For reference the Nile platform (dual core 1.3Ghz NeoII+4250 IGP) used around the same amount of power and got somewhere around 4-5 hours battery time in a net book form factor with a 6 cell battery.

i'm willing to sacrifice an hour or so to get some decent performance, but i expect it to be closer to six hours than four given that this laptop could get 3.5:

http://www.anandtech.com/show/4033/acers-aspire-5551g-amd-budget-gaming/6
http://www.anandtech.com/show/3883/toshiba-satellite-a660d-amd-phenom-ii-p920/7
 
i'm willing to sacrifice an hour or so to get some decent performance, but i expect it to be closer to six hours than four given that this laptop could get 3.5:

http://www.anandtech.com/show/4033/acers-aspire-5551g-amd-budget-gaming/6
http://www.anandtech.com/show/3883/toshiba-satellite-a660d-amd-phenom-ii-p920/7
They only get close to 3.5hr when idle. Just using the internet with the screen brightness nearly cut in half reduces the battery time to ~2.7hr for the 5551G and ~3.1hr for the 660D. If you want to watch a movie at 720p with the screen at nearly half brightness then battery time is ~1.8hr for the 5551G and ~2.2hr for the 660D.

The 4-5hr for the Nile platform is when actually doing something, like browsing on the internet + youtube at reduced screen brightness. Anand reviewed a Nile platform net book a while back if you'd like to see for yourself.
 
They only get close to 3.5hr when idle. Just using the internet with the screen brightness nearly cut in half reduces the battery time to ~2.7hr for the 5551G and ~3.1hr for the 660D. If you want to watch a movie at 720p with the screen at nearly half brightness then battery time is ~1.8hr for the 5551G and ~2.2hr for the 660D.

The 4-5hr for the Nile platform is when actually doing something, like browsing on the internet + youtube at reduced screen brightness. Anand reviewed a Nile platform net book a while back if you'd like to see for yourself.

actually the idle figure is 300 minutes, otherwise understood to mean five earth hours.

the 3.5 hour figure i quoted was from the relative battery life graph, which while misinterpreted as basic internet usage in hours, we still find the internet figure is 280 minutes.

it is not unbelievable that a 25W llano APU could get a battery life (in combined usage) of around six hours when we consider that many 18W Brazos systems are getting 6+ hours from their cheap 6-cell batteries.
 
Last edited:
actually the idle figure is 300 minutes, otherwise understood to mean five earth hours.
According to the Anandtech review idle time for the 660D is 229 minutes which is ~3.8hr, for the 5551G its 218 minutes which is ~3.6hr. Only the Dell Studio 14 (i5-430M+5470M) has a 300 minute idle time in the review....

the 3.5 hour figure i quoted was from the relative battery life graph, which while misinterpreted as basic internet usage in hours, we still find the internet figure is 280 minutes.
Yea I know but no one is gonna care about "relative battery life", they're gonna care about actual internet battery life period, which Anand gave in the 2nd graph, so I don't know why you'd bother or care about "relative battery life".

it is not unbelievable that a 25W llano APU could get a battery life (in combined usage) of around six hours when we consider that many 18W Brazos systems are getting 6+ hours from their cheap 6-cell batteries.
Actually it is, and we have a real world example in the Nile platform with only used around 20W for the CPU+GPU but got around 4-5 hours in real world usage. Look you're being way too optimistic here, its best not to get your hopes up, especially when it comes to AMD. I like them but they've been fumbling the ball for the last few years straight on the CPU and process side of things.
 
quite right about the 300m figure, my mistake.

i take your point about optimism, but brazos is very good, and whatever extra overhead llano carries from its higher performance must in some measure be ameliorated by its 32nm manufacture process.

we will of course have to wait and see, but i am optimistic.
 
Been having some trouble finding answers to my questions, so thought I'd ask you guys.

What is the benefit of the APU? I am wondering if there is any benefit to gamers. I have a mid-range graphics card, how does this work with the APU having a built in GPU? Is it only for onboard graphics, or does it also boost performance in conjunction with the video card? They look pretty sweet, I'm just trying to figure out what the benefits are.
 
If you have an add-in videocard, then the integrated GPU will not have any benefit as they haven't made it so you can combine the APU and card in something like Crossfire. That should be coming up though, hopefully soon. The CPU portion of Llano will still be nice, as it'll be the fastest mobile CPU AMD has to offer so you'll see overall performance gains.
 
If you have an add-in videocard, then the integrated GPU will not have any benefit as they haven't made it so you can combine the APU and card in something like Crossfire. That should be coming up though, hopefully soon. The CPU portion of Llano will still be nice, as it'll be the fastest mobile CPU AMD has to offer so you'll see overall performance gains.

Lucid is getting ready to release a software solution to do just that on the intel side. Who knows when or if it'll come to AMD.

LINK
 
AMD tends to make noise when they know their upcoming product is going to be competetive (think Athlon XP, Athlon 64), and tends to be pretty quiet when it's not (think K6-III, Phenom I). I bet the new crop of processors are going to be fairly impressive.
 
AMD tends to make noise when they know their upcoming product is going to be competetive (think Athlon XP, Athlon 64), and tends to be pretty quiet when it's not (think K6-III, Phenom I). I bet the new crop of processors are going to be fairly impressive.

Actually I perceive it the other way round. Phenom I/Barcelona was a build up hysteria with boxing matches, native quad-core vs. kludge, 40% faster and so on.

Current Llano and BD are fairly quiet, if anything, they tempered ( think of JF fighting rumours all day ) the excitement.

Secondly, both projects had ran into a lot of issues.
-BD version I 45nm was cancelled in 2009, what we have is ver II, slightly altered AFAIK
-Llano was supposed to be done on 45nm ( they taped out some 45nm APUs, did not work ), moved to 32nm, lots and lots of issues with 32nm, got delayed 6 months and now will be shipped in Q3 ( July 4th), again later than what most expected.

The available frequency of 1.8GHz for the showcased parts is a bit underwhelming. I was expecting to see 3GHz+ demos with 4 months before launch. Makes me wonder, if they really solved all the process/design issues.

http://www.xbitlabs.com/news/cpu/di...alizes_Shipment_Dates_for_Next_Gen_Chips.html
 
The available frequency of 1.8GHz for the showcased parts is a bit underwhelming. I was expecting to see 3GHz+ demos with 4 months before launch. Makes me wonder, if they really solved all the process/design issues.

The parts demoed so far are made for the mobile space and they actually do pretty well considering the SB platform uses more power when taxed, at least in the benchmarks that we've seen. Expecting 3GHz out of a mobile quad part seems pretty unrealistic to me. The top bin SB mobile quad tops out at 2.5GHz with a 55W TDP (the SB in the demo has a TDP of 45W @ 2GHz, I'm guessing the ACP of the Llano is also 45W).

If this mobile roadmap is real then I am impressed with what they've managed to do so far inside such a low power envelope. If they matched Intel at 55W ACP I would think AMD could get to at least 2.2 or higher. Unfortunately a part like that isn't on that roadmap, maybe once they get more acquainted with the 32nm process we will see some higher clocks.

On the desktop side, imagine what a 95 or 125W Llano could bring and then that 3GHz number sounds like child's play. OEMs are gonna eat them up, especially if they can crossfire it with a cheap Turks level discrete GPU and offer a competent gaming machine to the masses.
 
Back
Top