AMD Llano Unveiled - A Notebook Platform Preview

The new AMD A-Series APU (Llano) is upon us and we have a first hand preview of their new Sabine notebook platform. Would AMD's latest Fusion processor be a Sandy Bridge killer?

Llano - The Latest AMD Fusion A-Series APU

Building on the which saw tremendous sales in the first quarter of this year, AMD is now making available the next iteration of their Fusion APUs known as . The new APUs will be known as the A-Series APU and it will be offered in various flavors covering systems with price points from US$499 and beyond. The new A-Series APUs will go head-to-head with and will attempt to compete hard with Intel in the budget and mid-range segments (Core i3 and i5).

The new A-Series APU will feature up to four x86 CPU cores and will be introduced in the mainstream notebook platform (check the link for a comparison of Brazos vs. Sabine). The CPU cores will feature AMD's own Turbo Core technology which dynamically increases the core frequency for added performance when applications call for it. Besides that, the APU also features an array of Radeon cores which gives it discrete-class DirectX 11 graphics performance. This means that without adding an additional graphics card, the APU will be able to deliver smooth 3D graphics using the latest DirectX 11 API. On the other hand, Intel's Sandy Bridge, which also similarly supports rendering of 3D graphics, is only capable of supporting the older DirectX 10.1 API and in terms of graphics quality, it won't be as realistic as what AMD's A-Series can deliver with DirectX 11.

The new APU also comes with a third generation Unified Video Decoder (UVD), dual-channel DDR3 memory controller, two dedicated digital display interfaces and PCIe Gen2 interfaces. It can be paired with the latest AMD Fusion Controller Hub (A70M) which lends support to four USB 3.0 ports, 10x USB 2.0 ports, SATA 6G, PCI and VGA DAC. Let's take a look at the full series of A-Series processors from AMD coming our way:-

Model
Radeon Brand
TDP
CPU Cores
CPUClock (turbo / idle)
L2 Cache
Radeon Cores
GPUClock
Max DDR3
A8-3530MX
HD 6620G
45W
4
2.6GHz / 1.9GHz
4MB
400
444MHz

DDR3-1600 DDR3L-1333

A8-3510MX
HD6620G
45W
4
2.5GHz / 1.8GHz
4MB
400
444MHz

DDR3-1600 DDR3L-1333

A8-3500M
HD 6620G
35W
4
2.4GHz / 1.5GHz
4MB
400
444MHz
DDR3-1333 DDR3L-1333
A6-3410MX
HD 6520G
45W
4
2.3GHz / 1.6GHz
4MB
320
400MHz
DDR3-1600 DDR3L-1333
A6-3400M
HD 6520G
35W
4
2.3GHz / 1.4GHz
4MB
320
400MHz
DDR3-1333 DDR3L-1333
A4-3310MX
HD 6480G
45W
2
2.5GHz / 2.1GHz
2MB
240
444MHz

DDR3-1333 DDR3L-1333

A4-3300M
HD 6480G
35W
2
2.5GHz / 1.9GHz
2MB
240
444MHz
DDR3-1333 DDR3L-1333

As you can see in the specifications above, the A-Series will come in three versions, namely A8, A6 and A4. The A4 series of processors will go head on to compete with Intel's Core i3 while A6 will compete with lower-end Core i5 and higher speed Core i3. The A8 series of processors are slated as the highest performance parts you'll find in the Llano line-up and they are positioned to go against Intel Core i7 and higher-end Core i5 processors.

The CPU clock speed indicated in the table above shows its speed when Turbo Core is active while the lower clock number is the lower threshold value of the clock speed at low/idle workloads.

For now, AMD is releasing processors for the notebook market and you can see in the table above that these have rather low TDPs at 35W and 45W. You can see that all the APU models denoted with an "X" correspond to the 45W TDP threshold and signal that these are the more 'extrme' versions within the series. Future Llano APUs for desktops are expected to carry an even higher TDP rating of 65W and 100W and we expect these to have high CPU and GPU clocks.

On the graphics portion of the APU, you'll find that AMD has rated its Radeon GPUs based on the number of Radeon cores allocated to the GPU, as well as its GPU clock speed. A notebook with either one of these APUs should give rather decent entry-level graphics performance if we are to compare the Radeon model number with its discrete counterpart. In fact, AMD claims that the A4-3300M part is more superior than the discrete version of the AMD Radeon HD 6470M commonly used in mobile platforms.

Since the graphics architecture employed within the APU is scalable, we can expect to see future APUs to feature even faster GPUs. Even with the current design, AMD indicated that its GPUs are targeted to reach clock speeds of up to 600MHz, possibly meant for higher TDP parts used in its desktop platform.

The new Llano chip which is based on their new 32nm process technology. There are a total of 1.45 billion transistors on this 228 square mm silicon die.

The new Llano chip which is based on their new 32nm process technology. There are a total of 1.45 billion transistors on this 228 square mm silicon die.

CrossFireX in the Notebook

The new A-series APUs are designed to enhance integrated graphics performance as well as media related acceleration. As such, a lot of the processor's transistor budget went into building the GPU. According to AMD, as much as 50% of the die space have been allocated to the GPU, leaving the rest for the CPU and other subsystems. But this doesn't mean that its CPU is crippled since their latest 32nm process technology does give them the opportunity to integrate more transistors. However, if you compared the CPU subsystem with its current Phenom range of processors, you'll find that it's missing the large L3 cache which is part of Phenom's balanced cache architecture that helps shorten access time to highly accessed data. Instead, AMD increased the L2 cache size of its individual cores to 1MB each. This may help improve the performance of its individual cores, but it probably would not be comparable to their desktop CPUs. One must understand that the APUs are designed for graphics and media and it may not necessarily excel in general purpose computing workloads.

Functional blocks within the Llano APU die.

Functional blocks within the Llano APU die.

However, when it comes to graphics, the A-series APU does look promising from the start. With a Radeon class GPU built into the same CPU die, one can expect very decent graphics performance right off the APU. This presents users with a very affordable gaming notebook, an area that Intel is currently gaining much traction with their 2nd generation Core processors. Intel's effort to capture the integrated graphics market with Sandy Bridge may change when OEMs start shipping notebook models equipped with AMD's latest A-series APUs.

For notebooks that need the extra bump in graphics performance, AMD's latest A-series APUs will offer manufacturers the option to implement AMD's CrossFireX technology in the notebook. The APU's GPU is built with the ability to pair with another Radeon discrete GPU in a CrossFireX configuration to give greater speed improvements to gaming. Obviously, you'll find most implementations using an asymmetrical setup but CrossFireX will work nonetheless.

The ability to configure multi-GPU graphics is the real deal for AMD's latest A-series APUs. The flexibility to pair any discrete Radeon GPU with the APU is one of the most attractive feature we've come across so far. You can say that AMD has one-up Intel this time around and it would be hard for Intel to catch up unless they have discrete graphics solutions like AMD. Even NVIDIA is being sidelined for good since all AMD Fusion gaming-class notebooks would be using an AMD Radeon discrete GPU over NVIDIA's GeForce - simply because of the added performance from CrossFireX. So if you ever come across an AMD Fusion notebook with an NVIDIA GeForce discrete graphics, you may want to reconsider (unless of course the discrete GeForce is way more powerful and cheaper than the AMD CrossFireX option).

Below, you'll see an example of the various configurations using the A-series APU and the discrete Radeon component:-

Discrete Radeon HD Graphics Unit
AMD A8 series (Radeon 6620G) + Discrete
AMDA6 series (Radeon 6520G) + Discrete
6770M
6775G2
6775G2
6750M
6755G2
6755G2
6730M
6760G2
6740G2
6650M
6740G2
6720G2
6630M
6690G2
6680G2
6490M
6645G2
6545G2
6470M
6640G2
6540G2
6450M
6640G2
6540G2

As you can see in the table above, pairing the APU with a relatively powerful discrete part does not necessarily yield a lot more performance, as far as the branding model number goes.

Switchable Graphics Made Easy

Along with the A-series APUs, AMD will be releasing a new set of driver called the AMD VISION Engine Control Center. In it, you can control the various new features available in the APU and it gives the user greater control of their graphics.

One of the main feature in the driver set is its configurable switchable graphics which lets the user specifically assign if an application should run with power saving graphics or using its discrete graphics for performance. This switching is automatic once the setting has been made when you first run an application. Or if you choose to change it, you could do so later. The good thing about this is the seamless switching which does not require any reboot nor hard switches. You can just pre-set your apps to run in the control panel and it would do it for you without any prompting.

The AMD VISION Engine Control Center lets you configure applications to use either power saving mode (integrated graphics) or high performance mode (CrossFireX graphics).

The AMD VISION Engine Control Center lets you configure applications to use either power saving mode (integrated graphics) or high performance mode (CrossFireX graphics).

The other advantage with AMD's version of the switchable graphics is its very seamless use of the GPU resources for computing purposes. As you know, AMD has been a strong advocate of the OpenCL framework which utilizes the GPU for non-graphical computing. With the seamless integration of the discrete GPU with the APU, a lot of these resources can be utilized with ease. Compared with Intel's latest Sandy Bridge (2nd generation Core), the use of its integrated GPU for such purposes can be limited especially when a discrete part is present. However, that has changed recently with the integration of a third party solution using which enables dynamic switching between the Intel Core and discrete graphics. But the Lucid Virtu solution is only available on the Intel Z68 chipset (and a few limited H67 implementations) for motherboards and we have not yet seen such a solution being implemented in notebooks.

One should note that the acceleration of applications utilizing the GPU core's computing resources is very much dependent on whether the application is written to support it. Intel's very efficient transcoding engine and media acceleration is impressive but apps supporting these features are still rather limited. The same goes for OpenCL accelerated apps. However, both AMD and Intel have been addressing the software developer community to ensure apps are coded to take advantage of the GPU.

For now, AMD does have an impressive list of software partners which include big names like Adobe, Microsoft, Corel, Sony, Roxio, ArcSoft, EA, Codemasters, Firaxis, Cyberlink and DivX. Obviously, this is not an exhaustive list but you can see that OpenCL is fast gaining traction in the developer community.

Applications that support OpenCL such as Arcsoft's latest Total MediaTheatre 5 will take advantage of the APU's features for enhanced video acceleration and quality.

Applications that support OpenCL such as Arcsoft's latest Total MediaTheatre 5 will take advantage of the APU's features for enhanced video acceleration and quality.

Vision Computing Enhanced

The new A-series APUs also come with the third generation of AMD's Unified Video Decoder (UVD3) which now supports video acceleration of more formats. UVD3 first debuted in the , which the APU's Radeon graphics cores are based upon. To recap, apart from its support for VC-1, H.264 and MPEG-2 (IDCT profile) decoding since the introduction of UVD2, AMD has now added support for Blu-ray 3D (multi-view CODEC), MPEG-4 part 2 and DviX/Xvid. With the dedicated UVD3 in the chip, video playback quality will be enhaced thanks to dedicated decode and processing/scaling engine. In addition to that, by utilizing the UVD3 portion of the CPU, much of the other subsystems can be powered down (in sleep mode) to save battery power. Besides that, UVD3 also supports faster transcoding of video.

AMD is also introducing a new feature known as AMD SteadyVideo in the A-series APU. Using the AMD APP technology (Accelerated Parallel Processing), this feature utilizes both the integrated GPU and CPU to analyze the playback of internet video to produce shake-free video in real-time. In case you miss our point, the keyword here is 'real-time'. The Steady Video algorithm knows when you're watching videos on the internet (e.g. YouTube) and it post-processes video to remove the camera shake. This means that watching home made movies on YouTube is no longer a Blair With Project affair since videos will be less nauseating. Users can also control how much they want SteadyVideo to compensate via the AMD VISION Engine Control Center.

Controls for enhancing video playback, including AMD's SteadyVideo technology for shake-free internet videos.

Controls for enhancing video playback, including AMD's SteadyVideo technology for shake-free internet videos.

In addition to that, AMD's A-Series APU also enhances the quality of video playback. AMD claims that it scores very well in the HQV Benchmark (which it does anyway with its regular discrete line-up of Radeon graphics) when it comes to decoding, de-interlacing, motion correction, noise reduction, film cadence detection and detail enhancement. Such video enhancements will give video playback a much more pleasant experience especially when you're dealing with lower quality videos or when upscaling DVD movies into 1080p (full HD) displays.

Before we head on to benchmarks in the next page, let's compare the specifications of all the three mobile variants in the table below.

VISIONA8 APU
VISIONA6 APU
VISIONA4 APU
Radeon Cores
400
320
240
SIMDs
5
4
3
Texture Units
20
16
8
Render Back Ends
2
2
1
Z/StencilROPs
32
32
16
Color ROPs
8
8
4
Memory Interface
2x64-bit
2x64-bit
2x64-bit
Memory Data Rate (DDR3)
1.6Gbps
1.6Gbps
1.333Gbps
GPUClock Speed
444MHz
400MHz
444MHz
Peak Compute
355GFLOPS
256GFLOPS
142GFLOPS

 

Some Benchmarks

The Llano prototype notebook which we received for the purpose of this preview is based on the AMD A8-3500M APU clocked at 1.5GHz with Turbo Core speeds at a maximum of 2.4GHz. The quad-core APU comes with an integrated Radeon HD 6620G graphics and a discrete Radeon HD 6630M clocked at 485MHz. Under CrossFireX configuration, both the integrated and discrete part would yield a Radeon HD 6690G2 part. The notebook also comes with 4GB of DDR3 memory and a 250GB 7200RPM Hitachi Travelstar hard drive.

 

Futuremark 3DMark06

We tested the AMD A8-3500M APU and compared 3DMark06 results with the Intel Core i7-2630QM (on the HP dv6 notebook) and the Intel Core i5-2520M (found in the Lenovo X1 laptop). The Core i7-2630QM GPU clock is slightly slower than the i5-2520M, which explains the overall 3DMark06 scores recorded between the two CPUs. It can be seen that when it comes to 3D gaming, the AMD A8-3500M APU is indeed more superior whether it's running directly from the integrated APU or under the CrossFireX configuration with the discrete Radeon HD 6630M.

The AMD A8-3500M APU would support benchmarks running in DirectX11 like 3DMark11, but since Intel's Sandy Bridge does not support DirectX11, we decided to skip that since there's really nothing to compare there.

 

Far Cry 2 Game

Results from the Far Cry 2 benchmark puts the AMD A8-3500M APU at rather playable framerates although occasionally one might still run into jerky motion when scenes become too complex for the APU to handle. With the CrossFireX running in high performance mode, performance is somewhat better and more acceptable. But with the Sandy Bridge's built-in graphics, Far Cry 2 is probably not the sort of game you want to run without the help of a discrete component.

PCMark Vantage

Finally, we wrapped up our tests with PCMark Vantage, comparing it with the Core i7-2630QM. Although we know it's not a fair comparison since the A8-3500M is targeted against the Core i5, we wanted to show that the new Llano APU is not too far off in terms of performance. Besides, we wanted to compare quad-core processors, not quad (A8-3500M) against dual core (Mobile Core i5).

As mentioned earlier, the AMD A8-3500M is not too shabby when compared with Intel's Core i7-2630QM in the HP dv6. Productivity and gaming scores were quite on par but Intel took the lead when it comes to TV and Movies. The memory subsystem in AMD's new Llano APUs scored better but it did not manage to boost its overall PCMark score. Still, we don't think it's too much of a disappointment as far as PCMark results go.

 

Battery Life

We also tested the battery life of the notebook by performing our standard DVD playback test and measuring the time it takes to run the battery out. Overall, we could comfortably get close to 4 hours of continuous movie playback from the DVD drive using a standard 5200mAh rated Li-ion battery. The test simply demonstrated the low power consumption of the new APU which makes it quite comparable with some of Intel's 2nd generation laptops.

Conclusion

AMD's latest Fusion A-series processor is finally here and is the long wait worth it? We think it is because it is what processors should deliver, a good balance of both compute and graphics performance. While AMD's approach is skewed towards graphics (in-line with their Vision marketing), Intel's current Sandy Bridge is focused on delivering compelling performance in the compute environment which makes it a powerful general purpose processor. That said, it doesn't mean Sandy Bridge is not delivering great performance in its graphics, but AMD's latest A-series APUs will no doubt take some of the limelight away from Intel's Sandy Bridge.

The single most attractive feature found in the A-series APU has got to be its ability to configure its graphics in CrossFireX mode with another discrete GPU. It was good that AMD is leveraging on ATI's CrossFire technology to make this possible and we believe the integration of such technology is simply to AMD's advantage. With the Llano going into notebooks starting from today, you won't find any other affordable notebooks out there with 'Dual Graphics' except from AMD's A-series APUs. Intel would probably need to come out with another strategy to address this, else AMD would surely dominate the mainstream gaming notebook segment since the message that AMD is sending out to users is going to be pretty hard to resist. NVIDIA, on the other hand, would lose its share in the AMD notebook segment since all APUs would benefit when used along with an AMD Radeon GPU. It would be foolish for any OEMs to pair an NVIDIA GeForce with an AMD Fusion A-series APU.

All in all, the new AMD A-series APU is a serious contender to Intel's Sandy Bridge. It may not be the product that beats Intel's 2nd generation Core processor hands down, but it sure is a serious threat to Intel's dominance in the notebook segment. The AMD Fusion processor has a lot going for it especially when it comes to addressing users' need to consume media on laptops and the new APU has all it takes to deliver just that. Never mind that its integer and floating point units are less powerful, but its strength in graphics more than makes up for it. With OpenCL supported in more and more software, the GPU may just be offloading most computational work from the CPU and it will only continue to grow in AMD's favor.

For now, we can expect up to 150 designs for desktops and notebooks featuring the latest A-series APUs from OEMs in the next quarter of 2011 and beyond. While it's interesting to see what kind of products will be announced in the next couple of weeks, we are more interested to see the price point of these products as it will very much determine its success in the consumer market.

Next up - Llano for desktops. Stay tuned.

Next up - Llano for desktops. Stay tuned.

Our articles may contain affiliate links. If you buy through these links, we may earn a small commission.

Share this article