|Introduction and Specifications|
AMD launched the first volley of their Fusion technology salvo late last year and since then thin-and-light notebook offerings from various major OEMs have come to market. AMD's Brazos platform and specifically the Zacate E-350 processor were obvious successes and the company was able to pull down many strong design wins with manufacturers like Lenovo and HP leading the charge, marketing the product's intrinsic benefits. AMD's E series Fusion processors offer Atom-like power consumption with more robust CPU and GPU performance, at netbook-like prices. To say the product was a success would be a total understatement. In fact, the company claims to have "sold out" of low cost Fusion processors and shipped over 5 million units. No matter how you do the math, it's huge.
Today, AMD is lifting the veil on their next generation mainstream mobile processor, code named Llano. This latest volley by AMD is aimed at hitting Intel right in the mid-section, where the bulk of multimedia capable notebooks are sold. These are higher-end machines, where the likes of Intel's new Sandy Bridge-based Core i5 and Core i7 dual and quad-core processors live.
The platform we'll give you a look at today is AMD's A Series; specifically we have an A8 APU-based machine that we'll break down for your edification and spin up on the test bench so you can see its performance profile. Let's get a look at some specifics of the A8 Llano APU or "Application Processing Unit," as AMD calls it, along with its accompanying FCH or Fusion Controller Hub.
AMD Llano A-Series A-3500M APU Pin Layout
If you look closely at the spec list above (and didn't allow your eyes to glaze over), you'll note that the A-Series platform has virtually all the latest technologies built in. There is support for HDMI, DisplayPort, PCI Express Gen 1 and 2, 6Gbps SATA and native USB 3.0. In terms of all of the major notebook check-box items, AMD's A-Series has it all.
|The Llano Fusion Microarchitecture|
AMD is keeping some of the intricate details of their latest Fusion APU architecture close to the vest, but below is a relatively detailed block diagram of what you get with the A8 processor that we'll have benchmarks numbers for you on, later in this article.
The A8 consists of four AMD Phenom-class cores with 128-bit FPUs. Each core has its own 1MB L2 cache and the chip now supports low power C6 idle states, as well as CC6 power gating, allowing individual cores to power down completely. More on this later. Also on board is a DX11 capable GPU; you can see how large the graphics array is, as well as its companion Display output and UVD3 (Unified Video Decoder) blocks. Finally, on chip also resides a dual channel DDR3 memory controller supporting speeds up to 1600MHz (though our system came configured at 1333MHz), as well as a X16 PCI Express link and four X1 PCI Express links.
Here AMD illustrates the various interfaces employed between each sub-block of the A-Series APU architecture. Since we're already familiar with the DDR memory and CPU bus interfaces employed in the CPU block, of particular interest are the two new interfaces, known as the "Radeon Memory Bus" and "Fusion Compute Link." The FCL is a bit misleading in its terminology, however. With the word "link" in there, you might confuse it for a serial connection (commonly referred to as links) but it is in fact a parallel bus interface, believe it or not. Here's what AMD had to tell us about it: "The Fusion Controller link is similar to the HT interface used to communicate to our Integrated Graphics Chips, but with the Physical layer stripped off. As such it uses an HT like protocol, but is implemented as a wide parallel bus interface." Interesting. It can't really be a "bus" though can it? When we think of bus architectures we think of things like arbitration and latency but that would be on a "shared" bus architecture. This is a dedicated connection. Regardless, this is all the information AMD was willing to share with us at the moment on FCL.
The Radeon Memory bus is a bit easier to wrap your head around. Here's what AMD had to offer on it: "The Radeon Memory bus aggregates requests from all the Graphics and Multimedia clients and presents a well-ordered (for DRAM efficiency) stream of requests to the back end of the NorthBridge. These requests bypass any of the coherency mechanisms associated with the front-end of the Northbridge. The Radeon Memory bus is capable of making all the system memory bandwidth available to the Radeon compute Units." Ahh, an intelligent DMA Engine and interface that essentially circumvents the Northbridge memory manager. Got it.
In addition to improved IPC (instructions per clock) throughput for each of the four CPU cores, AMD has enabled Llano with their Turbo Core technology, allowing the processor to scale up frequency on a per-core basis depending on workload, or ramp down and completely power off cores that are not in use. Again, this can happen on a per-core basis at the CPU level and gating of the GPU and UVD blocks is available as well. AMD even developed a method of optimizing display power by balancing backlight and pixel brightness to provide an optimal image while reducing power consumption. In the bottom right image above here, you can see the results of the Graphics and UVD blocks in a gated state. In this state, the only power consumption is due to intrinsic leakage current in the circuit areas but all clocks are off. In the thermal images above, blue areas represent how cool the powered-down blocks are, reaching ambient air temp.
Finally, a quick look at the full system implementation here, shows a base two-chip solution that employs AMD's new Fusion Controller Hub Southbridge. Specifically, the A70M version of the FCH is what enables USB 3.0 connectivity, though both the A60M and A70M support 6Gbps 3G SATA connectivity as well. As you can see, Llano itself supports 2 DDR3 DIMM sockets, as well as DP and HDMI outputs and a X16 PEG (PCI Express Graphics) link to an optional discrete GPU. This is how our system was setup, with an additional Radeon HD 6630M series chip on board. Incidentally, AMD enables CrossFire rendering with Llano's IGP, for discrete GPUs in tandem, allowing for higher performance.
|Design and Implementation|
AMD sent over a 14-inch Compal notebook that housed their new platform. To look at the system isn't exactly an awe-inspiring experience but the system overall is relatively indicative of the class of machine that the Llano A8 APU will be powering. For Llano, a range of machines are possible from thin and lights to larger, more full-featured multimedia platforms. This smaller Compal unit we've been testing sort of a strikes a balance.
Based solely on AMD's "Sabine" platform, employing the Llano APU and A70M Fusion Controller Hub chip, our system offered an optical drive along with a pair of USB 2.0 ports, one USB 3.0 port, HDMI and VGA output, integrated 801.11n WiFi, and Gigabit wired Ethernet. Also on board was a flash memory card reader and multi-touch track pad with two button mouse. Again, this is standard fare pretty much for mainstream notebooks these days. However, no third-party chips are required to enable USB 3.0, so there's a cost advantage associated with the Sabine platform in that respect, currently. We should also note that even under full load under various mutlimedia and gaming test conditions, in our benchmark runs, the system remained very quiet and didn't emit excessive heat beyond what we would consider to be reasonable levels.Switchable Graphics and CrossFire Empowered:
In support of the new graphics capabilities of the platform, AMD has developed switchable graphics technology, similar to that of NVIDIA's Optimus solutions. You can configure a Llano-equipped notebook to take advantage of a discrete AMD Radeon GPU (in our case a Radeon 6700M series chip) for graphics-intensive workloads, while falling back to Llano's IGP for power saving modes. All of this can happen on the fly and also be configured at the application level, to enable one GPU or the other, or both in CrossFire rendering mode.
Enabling CrossFire rendering between the in-system discrete GPU and the IGP is implemented with a simple check box item in AMD's Vision Control Center driver control panel, much like the way it's done on the desktop. Switchable Graphics modes, directing workloads to the IGP only or discrete GPU, can be activated via workload or power supply condition by checking a simple radio button. Configuring your graphics engines based on power source conditions is pretty self-explanatory. Configuring AMD switchable graphics at the application level requires the user to assign a "high performance" or "power saving" state for each application on the system. However, AMD informed us that retail OEM systems will have a simpler UI that will give the user an option of running strictly on the IGP in Llano for power saving modes, or drive a dual graphics CrossFire mode under graphics-intensive workloads.
|Vital Signs and A8-3500M APU Details|
Our test system came configured with Windows 7 Ultimate X64 and device manager gives us a view of what's under the hood. The AMD A8 Series Llano APU offers four full dedicated CPU cores to the OS, without the need for HyperThreading, though as you'll see later in this piece, with power consumption that looks more like a dual core chip.
In the Display Adapters section you can see we have two GPUs present. The Radeon HD 6620G is Llano's IGP and the Radeon 6600/6700 Series listing is the discrete graphics core in this notebook. Let's get a good look at the brains of the operation here...
The Sabine platform and Llano A8 implementation we were sent for testing, has a fairly simple layout and an obviously low cost bill of materials. Llano's thermal solution, which incorporates a discrete GPU in this particular build, as you can see, is rather svelte. There is a small network of cold plates and heatpipes that connect to a radiator that is cooled by a single turbine exhaust fan, and that's it. On the left, you can see the 722 pin socket for Llano and in the middle is the tiny AMD FCH (Fusion Controller Hub) Southbridge IO chip. On the right is the discrete Radeon 6700M graphics chip and you can see a few GDDR memory chips around it as well. Beyond its two DIMM slots, plunk down some passive and active power components a clock generator and PLL or two, plug in a wireless radio card and there's really not much more to it. Overall, the circuit design is rather elegant and you can imagine how much more simplified it gets with just Llano in there and no discrete GPU.
Before we dive into the nitty-gritty numbers, here's what Windows 7 X64 thinks of the AMD A8-3500M APU, along with its graphics, memory and storage subsystems. Surprisingly the lowest score here was the graphics subsystem but again, we're looking at the IGP itself, which is significantly better than we've seen in competitive Intel HD Graphics scores with the latest Sandy Bridge dual-core processors and almost on par with Sandy Bridge quad-core chips. All in all, things are looking not too shabby here for Llano.
|Test Systems and PCMark Vantage|
|For this article, we were presented with a bit of a challenge with respect to providing relevant and reasonably balanced reference numbers for benchmark comparisons. The Compal Llano mobile reference platform is more of a mainstream configuration, rather than a high-performance or gaming-targeted model. As such, for productivity, content creation and multimedia testing, we wanted to compare it to other machines in its class, based on Intel's current architectures. These test machines are listed in detail below.
Primary Test System
First, we our test systems through Futuremark’s latest system performance metric, PCMark Vantage. PCMark Vantage runs through a host of different usage scenarios to simulate different types of workloads including HD TV and movie playback and video manipulation, gaming, image editing and manipulation, music compression, communications, and productivity. Most of the tests are multi-threaded as well, so the tests can exploit the additional resources offered by multi-core CPUs.
For general purpose compute tasks and lighter duty everyday work loads, Llano shows itself mostly competitive versus Intel's Core i5 dual cores with HyperThreading, but no match for the quad-core Core i7 2820QM. An area where Llano trails significantly is the Communications test, where Intel's latest security instruction set support and the on-board IPSec processing engine in Sandy Bridge, take this test up a notch or three.
|Futuremark PCMark 7|
|Futuremark's PCMark 7 is the latest version of the PCMark suite, recently released this spring. It has updated application performance measurements targeted for a Windows 7 environment. Here's what Futuremark says is incorporated in the base PCMark suite and the Entertainment suite, the two modules we have benchmark scores for you here.
The PCMark test is a collection of workloads that measure system performance during typical desktop usage. This is the most important test since it returns the official PCMark score for the system
Here AMD's new Llano A8 Fusion processor fairs significantly better, coming within striking distance of Intel's Core i5-2520M mobile processor, within about a 10% margin or so. Also, note that in this test, as was evident in our PCMark Vantage test from the previous page, the Llano / discrete CrossFire configuration doesn't offer much performance improvement. Obviously, these tests are not heavily weighted with 3D graphics workloads.
|HD Media Playback and Encoding|
|On the video decode side of the equation, we selected an HD movie trailer clip to look at from a resource utilization standpoint. With a powerful chip like the AMD A8-3500M, we didn't expect HD video playback to be an issue but regardless, we looked at CPU utilization playing back this 1080p H.264 QuickTime clip. We fired up the I am Legend trailer .mov file with Windows Media Player, while taking note of thread activity in Windows Task Manager Performance Monitor.
1080p H.264 QuickTime Movie Trailer - I Am Legend - 13% CPU utilization and no sweat
1080P Flash Video, Youtube: Avatar Movie Trailer
We also streamed a 1080p movie trailer on Youtube in full screen mode, a notoriously taxing proposition for ultralight platforms, netbooks and the like. In the H.264 Quicktime clip, performance monitor showed activity only across two threads and oscillating around 5% CPU utilization. In the flash video test, the Avatar trailer only consumed about 15 - 17% but did so across 4 active threads. Regardless, Llano didn't even break a sweat obviously in either of these two HD video playback scenarios.
Cyberlink's Media Show Espresso is a video conversion tool that imports various video media files types and converts them to other standard video formats for publication, distribution and streaming. In this test, we take a 224MB high definition 1080p AVCHD video clip and compress and convert it to a 720p H.264-encoded .mov file. Times are measured in minutes:seconds with lower times representing faster throughput in the video conversion process.
The version of the application we used actually does support video transcode processing on AMD GPUs, so what you're looking at here is what Llano's IGP can do, rather than a software conversion on the CPU. Clearly Intel's QuickSync engine outclasses all other test systems but AMD's Llano processor holds up relatively well versus the higher-end, higher power consuming Core i7 quad core with 8 threads processing the workload. Llano is also able to outpace a similarly clocked, previous generation Intel dual core (Arrandale) that doesn't have QuickSync technology at its disposal.
|LAME MT Audio Encoding and Cinebench R11.5|
In our custom LAME MT MP3 encoding test, we convert a large WAV file to the MP3 format, which is a popular scenario that many end users work with on a day-to-day basis to provide portability and storage of their digital audio content. LAME is an open-source mid to high bit-rate and VBR (variable bit rate) MP3 audio encoder that is used widely around the world in a multitude of third party applications.
In this test, we created our own 223MB WAV file (a hallucinogenic-induced Grateful Dead jam) and converted it to the MP3 format using the multi-thread capable LAME MT application in single and multi-thread modes. Processing times are recorded below, listed in seconds. Shorter times equate to better performance.
Ouch. This test is obviously not AMD's forte. The newly enhanced 32nm quad-core Phenom II engine in Llano is only exercised over two threads in this test and as such, AMD's new chip is competing against the massive memory bandwidth and compute throughput of Intel's Core architecture. In short, even against Intel's previous gen Arrandale mobile architecture, it's no contest.
Cinebench R11.5 is an OpenGL 3D rendering performance test based on Cinema 4D from Maxon. Cinema 4D is a 3D rendering and animation tool suite used by 3D animation houses and producers like Sony Animation and many others. It's very demanding of system processor resources and is an excellent gauge of pure computational throughput. This is a multi-threaded, multi-processor aware benchmark that renders and animates 3D scenes and tracks the length of the entire process. The rate at which each test system was able to render the entire scene is represented in the graph below.
Now here is where things start to get interesting for Llano. Though clearly, in terms of CPU throughput, AMD's new mid-range mobile architecture trails Intel's Sandy Bridge architecture by a wide margin, the inverese can be said for AMD's new Llano processor when it comes to on-die GPU horsepower. Here Cinebench shows Llano trailing the Core i5 dual core by 55% or so and it's over-powered by a factor of well over 2X versus the Core i7 quad-core. However, Llano's OpenGL GPU performance in Cinebench is over twice as fast as the Core i5 and almost twice as fast as the high-end Core i7-2820QM. In fact, Llano was able to out-pace a Core i5 system outfitted with NVIDIA GeForce GT 525M discrete graphics.
|DX9 Synthetic Gaming Test - 3DMark 06|
|FutureMark states that, "3DMark's score is an overall measure of your system’s 3D gaming capabilities, based on comprehensive real-time 3D graphics and processor tests. By comparing your score with those submitted by millions of other gamers you can see how your gaming rig performs, making it easier to choose the most effective upgrades or finding other ways to optimize your system. 3DMark06 has been downloaded more than any other 3D benchmark and the ORB database now contains over 8.5 million 3DMark06 benchmark scores from around the world."
For our first dedicated gaming test, albeit a synthetic benchmark, we got pretty much what we expected from Llano; a strong DX9 GPU performance and reasonably strong CPU performance. We decided to mix in one Core i3-350M (2.26GHz dual-core) score for you here from our internal database. As you can see, 3DMark06 rates the 2.4GHz Llano A8-3500M right about on par with the Core i3, performance wise. In terms of total 3DMark score, the Llano-based systems overtake even the high-end Core i7-2820QM base notebook with integrated graphics, and even comes close to catching the discrete GeForce GT 525M and Core i5 combination.
|Gaming: Integrated APU Graphics - Far Cry 2|
|Next we'll get into some actual game testing, with Far Cry 2. These following gaming benchmark scenarios should theoretically align right within Llano's graphics-intensive wheelhouse.
Stark contrast; these would be the two words we'd use to describe the performance of AMD's Llano-based notebook here versus the Core i5-2520M dual-core and Core i7-2820QM quad-core notebooks. In fact, Llano doubles Intel's performance cleanly and is almost able to catch the discrete GPU-driven Core i5 2520M system. Incidentally, the GeForce GT 525M is a 96 shader core 1.2GHz mobile GPU from NVIDIA and is no slouch.
|Gaming: Integrated APU Graphics - Metro 2033|
|Next, we turned up the graphics workload a notch or two, with Metro 2033, a 3D graphics stress test if we ever saw one.
Here the playing field is a bit more condensed, largely in part due to the game's demanding physics engine. Here the AMD A8-3500M shows an approximate 25% lead over Intel's higher-end quad-core Core i7 chip, but falls about the same diffential short of a similarly equipped dual-core Core i5 processor with a discrete NVIDIA GeForce GT 525M GPU at the helm for the graphics workload.
|Gaming: Integrated APU Graphics - Dirt 3|
|We continue to stress the AMD Llano system and its competitive counterparts here with the latest racing simulation title to hit the market, Dirt 3.
The trend continues for the AMD Llano system here with Dirt 3. The A8-3500M puts up a frame rate some 55%+ faster than Intel's fastest quad-core chip with integrated graphics. Though it falls short of our GeForce GT 525M discrete GPU-equipped system, it's impressive to see just how capable Llano is on its own. 1280X720 with 4X AA enabled and completely playable in Dirt 3; one word - impressive.
|Power Consumption and Battery Life Analysis|
Before bringing this article to a close, we'll take a look at power consumption of the AMD A8-3500M-based Compal notebook. Throughout all of our benchmarking and testing, we monitored how much power this new AMD Fusion test system was consuming with a power meter, versus other test systems we used for benchmarks in the previous pages. Our goal was to give you an idea as to how much power each configuration used while idling on the desktop and while under a heavy workload Keep in mind, this is total system power consumption being measured here and not the the CPU or GPU subsystems.
Since OEM retail products will be configured with a varying assortment of battery technologies, we felt it was important to forgo traditional battery drain power consumption time trials and just focus on system power draw. Remember, our test system is a 14-inch notebook, fully decked out with 802.11n WiFi, an optical drive, a 5400 RPM hard drive and 4GB of DDR3-1333 system memory. This is not a stripped down ultra-light notebook we're looking at here but a relatively full-featured multimedia machine, though granted it employs an LCD that is on the smaller side. Regardless, the power meter told the story loud and clear. AMD's Llano-based A8-3500M offers easily the lowest power consumption of any notebook architecture in its class. Operating at a little over 10 Watts at idle is almost ridiculously low power consumption, as is under 43 Watts under full load for a quad-core CPU with a robust graphics engine. Intel's quad-core Core i7 Sandy Bridge processor eats over twice as much power under full load. Looking back at the numbers, speaking strictly from a performance-per-watt point of view with multimedia and gaming usage models, AMD's Llano is no joke.
|Performance Summary and Conclusion|
|Performance Summary: Encapsulating Llano's performance, it's obvious where AMD set their design goals. From a general computing throughput perspective, Llano is about as fast as a similarly clocked dual-core Intel Sandy Bridge processor, depending on the workload and level of multi-threading. Llano's lower per-core IPC is apparent, but under highly threaded workloads, and in a lower power envelope, AMD's new quad-core mobile architecture competes well with Intel's dual-core Sandy Bridge mobile platform, but falls well short of the Core i7 quad-core architecture, though again, power consumption is relative here. From a multimedia and graphics performance standpoint, AMD's A8-3500M Llano processor puts the hurt on Intel's Sandy Bridge architecture, whether you consider dual or quad-core variants. AMD's IP in core graphics technology affords them a sizable advantage.
AMD's Llano - Quad CPUs, 400 Radeon Cores, One Socket
Below we have the rest of AMD's Llano line-up coming to notebook platforms near you in the coming months. We're hearing from AMD that retail product will be ready almost immediately, with over 150 claimed design wins in both notebook and desktop products from major OEMs, through Q2 of this year. In fact, we've heard from HP that AMD's new Fusion line-up will be making appearances in both consumer-targeted HP Pavilion notebooks, as well as HP's enterprise and workstation professional-class Pro Book line-up in the coming weeks. And so, it appears that, similar to AMD's low power Brazos platform and the E Series Zacate Fusion APU, AMD is setup for a strong entrance to market with Llano. Will AMD "sell out" like they did with Brazos? That's the sixty-four thousand dollar question.
There's little question that Intel's stranglehold on the notebook market with Sandy Bridge (even after a bumpy, errata prone ramp-up), is going to be an uphill battle for AMD. Regardless, when savvy consumers see what AMD's A Series APUs are capable of, one can easily speculate that AMD should be poised to take a bite out of Intel's notebook market share. The key is going to be price points. There's no denying that when it comes to volume, the SKUs that Best Buy and other major retailers sell most, are competitively priced systems that can handle it all. At $699 for A8 and $599 suggested targets for A6 platform built machines, AMD may still have to limbo a bit more on price to capture upside market share. That said, those of you in the know reading these pages will agree, in terms of multimedia and graphics performance, Llano is a solid proposition at its intended price points.
Also, though we're not ready to pass full judgment on power consumption because we haven't tested actual retail product as of yet, AMD's Llano processor also appears to be dialed in for power-efficient mobile goodness. In the long run, AMD's A Series Llano APUs should prove themselves a more than worthy alternative to Intel's mobile Core series line-up with integrated Intel HD Graphics. In short, if you care about balanced computing, multimedia and gaming performance, AMD's Llano simply cannot be denied.