When we wrote AMD's Kabini: Jaguar And GCN Come Together In A 15 W APU almost one year ago, AMD told me that it hoped that its Kabini and Temash APUs would bridge the gap between low-power ARM-based tablets and higher-performance notebooks. While the SoCs arguably achieved their goal, the most popular x86 tablets (Lenovo's ThinkPad, and Dell's Latitude and Venue Pro) employ Intel's Atom platform. The Temash APU never delivered a sub-4 W sweet spot the way Bay Trail did, and the Atom Z3770/D and Z3740/D became the weapon of choice for Windows-based tablets for good reason.
That's not to say AMD's Jaguar architecture isn't perfectly capable (both Microsoft and Sony would argue it is). And at higher thermal ceilings, I'd even suggest that the desktop-oriented version of Kabini is superior to the Bay Trail-D design. But to be truly competitive in the mobile space, AMD needs to do more with less power.
And that was the company's mantra as it created the Mullins and Beema APUs, both low-power SoCs destined to replace the Temash and Kabini solutions.

In presenting its two newest processors, AMD makes some bold claims. For instance, the company says that Mullins boasts two times the graphics performance per watt, and twice the system productivity per watt, compared to Temash. Beema is purported to offer a greater-than 10% graphics performance improvement over Kabini at a TDP that's 40% lower. And compared to its competition, AMD says that Beema serves up better graphics performance than both Bay Trail-T and Haswell-Y. The recurring theme is significantly lower power consumption and more speed. Exactly what kind of magic is involved in bringing these APUs to life?

Perhaps it makes the most sense to talk about what doesn't change. Beema and Mullins are manufactured on a 28 nm node, just like Kabini and Temash. As for the underlying architecture, it turns out that Puma+ offers the same IPC as Jaguar, the design that precedes it. Despite the nomenclature change, cores, caches, and schedulers remain the same. The graphics complex is also similar to the previous generation; the newest APUs similarly sport 128 GCN-based shaders.
Put simply, if you run the Beema/Mullins chips at the same frequencies as Kabini/Temash, you get identical performance.

Of course, that means speed increases must come from higher clock rates. How can that be, when we're talking about lower power and the same manufacturing process, all in the same breath? Fortunately, AMD has a lot to talk about on the subject.
The bummer is that it won't tell us where its newest parts are being etched. What we can relay are boasts of impressive process improvements yielding up to 38% lower leakage from the graphics transistors and 19% less leakage from the CPU cores. Company representatives also cite significant power savings attributable to I/O enhancements like an optimized DDR3L-1333 interface, which is responsible for a 500 mW reduction in draw. There's also a 200 mW savings that comes from a more efficient display engine.

Additionally, system-aware power management reportedly enables up to 50% more frequency at nearly half the TDP of AMD's Temash and Kabini APUs. Indeed, the top-of-the-line Mullins A10 Micro-6700T has a maximum 2.2 GHz clock rate and a 4.5 W TDP. Compare that to the fastest Temash-based A6-1450, which capped out at 1.4 GHz with an 8 W TDP. Of course, real-world frequencies are going to depend on the workload you're running. But AMD's saying it's better able to balance between high clock rates in single-threaded apps and lower frequencies in more parallelized tasks. The 15 W Beema-based A6-6310 tops out at 2.4 GHz, compared to the 25 W Kabini A6-5200 at 2 GHz.
As far as graphics go, the highest-clocked Beema and Mullins APUs run as fast as 800 and 500 MHz, respectively. Kabini and Temash topped out at 600 and 400 MHz.
If all of this sounds too good to be true, note that AMD is listing the highest possible clocks for Mullins and Beema, not their base frequencies. That'd be like Intel rating its processors at their peak Turbo Boost settings, or Nvidia marketing graphics cards at their typical GPU Boost figures. This isn't a new behavior from AMD though, which already takes the same approach with some of its newer graphics and general-purpose processing products.
The company claims that intelligent power control avoids waste by boosting only the applications that benefit from it, tying in with thermal management. In the case of the platform AMD gave us to play with, we saw the A10 Micro-6700T bounce between 1 GHz and a maximum 2.2 GHz clock rate. I fired up a single thread of Prime95 and recorded a 2.2 GHz frequency. Repeating the experiment with two, three, and four threads, we came up with 1.6, 1.4, and 1.2 GHz ceilings. As the SoC heated up, even those settings slid, though.

AMD also spoke to us at length about Skin Temperature Aware Power Management, or STAPM. The thermal limit of a tablet is often constrained by the temperature of its chassis, rather than the SoC's ceiling, since a piece of silicon withstands higher heat levels than your lap. Most devices are bound by the highest clock rate a processor can sustain without pushing the skin temperature beyond the user's sensitivity limit. Using STAPM, an APU ramps up aggressively until its host device's enclosure reaches a defined maximum, allowing higher performance for brief periods. Since many mobile applications involve holding onto a tablet for short durations, it's easier to get a speedier experience this way.

Finally, memory support evolves, allowing the top-end Beema APU to handle DDR3L-1866. Previously, the top-end mobile Kabini APU peaked at DDR3L-1600.
But before we tackle the performance implications of AMD's adjustments, let's take a closer look at the new on-die Platform Security Processor and the specific models planned for introduction.

AMD's updated low-power APUs are the first (and so far, only) x86 processors with an integrated ARM core. Dubbed the Platform Security Processor (PSP), this core is actually a 32-bit ARM Cortex-A5 with its own ROM and SRAM. The PSP is designed to provide a secure processing path, a trusted execution environment (TEE), a Trusted Platform Module (TPM), and a crypto co-processor capable of operating within ARM's TrustZone blanket of hardware-enabled security services.

At least on paper, the Platform Security Processor is an interesting IP add-on that we'd imagine stems from a desire to make mobile devices in the workplace easier to manage and control, since these aren't capabilities most home users want or need. AMD appears to be courting a segment currently under-serviced by existing ARM-based SoCs and Intel's own Bay Trail devices. However, it remains to be seen whether ISVs develop around the on-die hardware to expose its capabilities in software.
Models And Market Positioning
By now it's pretty clear what we're dealing with at the heart of AMD's newest mobile APUs. The following chart spells out how AMD is segmenting its Mullins-based models:
| Radeon Brand | SDP | TDP | CPU Cores | Max CPU Clock | L2 Cache | Shaders | Max GPU Clock | Memory | |
|---|---|---|---|---|---|---|---|---|---|
| A10 Micro-6700T | R6 | 2.8 W | 4.5 W | 4 | 2.2 GHz | 2 MB | 128 | 500 MHz | DDR3L-1333 |
| A4 Micro-6400T | R3 | 2.8 W | 4.5 W | 4 | 1.6 GHz | 2 MB | 128 | 350 MHz | DDR3L-1333 |
| E1 Micro-6200T | R2 | 2.8 W | 3.95 W | 2 | 1.4 GHz | 1 MB | 128 | 300 MHz | DDR3L-1066 |
The TDP range is tight, hanging between 3.95 and 4.5 W. Further, AMD attaches an SDP rating of 2.8 W up and down the line-up, conveying the power use expected during normal activity. Both top-end versions are quad-core configurations, receiving A10 and A4 branding, but differentiated by the addition of Micro as a prefix to their model number. The lone dual-core SoC bears the E1 designation.
AMD takes a different tact with its graphics marketing, using R2, R3, and R6 to indicate performance at multiple clock rates, since all three options sport 128 shaders.

Clearly, AMD hopes that Mullins will compete against Bay Trail-T and the more powerful Haswell-Y processors. And although there aren't any products with AMD's hardware inside yet, we're going to be benchmarking the potential of A10 Micro-6700T in a passively-cooled reference-class tablet.
| Radeon Brand | TDP | CPU Cores | Max CPU Clock | L2 Cache | Shaders | Max GPU Clock | Memory | |
|---|---|---|---|---|---|---|---|---|
| A6-6310 | R4 | 15 W | 4 | 2.4 GHz | 2 MB | 128 | 800 MHz | DDR3L-1866 |
| A4-6210 | R3 | 15 W | 4 | 1.8 GHz | 2 MB | 128 | 600 MHz | DDR3L-1600 |
| E2-6110 | R2 | 15 W | 4 | 1.5 GHz | 2 MB | 128 | 500 MHz | DDR3L-1600 |
| E1-6010 | R2 | 10 W | 2 | 1.35 GHz | 1 MB | 128 | 350 MHz | DDR3L-1333 |
The Beema-based parts have TDPs between 10 and 15 W. There aren't any SDP ratings; that's because SDP is reserved for devices designed with a touch-first usage model in mind, and those aren't what Beema will drop into.
Again, most of the SoCs are quad-core configurations; just one, the E1-6010, is a dual-core SoC. As with the Mullins-based parts, branding of the Radeon engine depends on clock rate, since every model includes 128 shader cores.

AMD plans to pit Beema against Bay Trail-M and Haswell-U in low-power laptops. True performance comparisons will need to wait until the company is able to secure design wins. But nothing about the specs suggest the theoretical match-up is unbalanced.
What we do have to test is an A10 Micro-6700T-equipped reference tablet that AMD calls its Discovery tablet, which we'll look at on the next page.
While AMD doesn't provide a true reference platform for its new APUs, the company did have its engineers put together a functional test bed for demonstration purposes. Dubbed the Discovery project, we got to try out the platform at AMD's campus in Austin, Texas.

Equipped with the 4.5 W A10 Micro-6700T flagship, based on the Mullins APU, AMD's Discovery tablet is equipped with 2 GB of DDR3L-1333 inside a generic 10" display form factor. AMD showed off stylized peripherals alongside the device, such as a game controller attachment and dock, and demonstrated the platform wirelessly broadcasting video to a television.
Although it's interesting, Discovery isn't a shipping product. So, the tablet's most important job is simply showing that Mullins operates in a handheld chassis without thermal issues.

Discovery also gave us a sneak peek at the platform's performance capabilities. Naturally, we prefer to put retail hardware through its paces in our lab. However, AMD only has a handful of these things, and there aren't enough to send out to reviewers.
As a result, we were limited to a few hours of testing on-site at AMD. To the company's credit, we were given completely unrestricted access to the hardware and were allowed to install anything we wanted. Still, though, this is a one-off device benchmarked at AMD. There's no way for us to guarantee its clock rates or the amount of time it was able to sustain them represents anything you'll see in the future.

Typically it takes us days of testing to come up with truly useful data. But form factors are so important to tablet benchmarking that going deep into a demo platform would yield limited information anyway. I'd much rather put a commercially available product through the wringer, and that's what I'll do as soon as it's possible. Until then, the few numbers gleaned from AMD's Discovery will hopefully shed some light on the potential of AMD's newest APUs.
Even though I didn't have much time to test, I prepared a couple of interesting comparisons. For instance, I benchmarked Dell's Venue 8 Pro tablet equipped with an Atom Z3740D, giving us representation from Intel's low-power Bay Trail line-up. To better plot where Mullins falls on the spectrum, I also have the Celeron J1900 (Bay Trail-D) and Athlon 5350 (Kabini) in my charts. Those two processors are rated at higher 10 and 25 W TDPs, respectively.
We begin with graphics testing, emphasizing AMD's GCN architecture. The question is whether AMD's design can scale effectively down to a tablet's restrictive power envelope. Is this Radeon-branded engine as effective down at 4.5 W as it is in more complex GPUs?


In the Ice Storm Unlimited test, AMD's A10 Micro-6700T competes against Intel's 10 W Celeron J1900. The Cloud Gate benchmark actually gives Mullins a notable lead.
We keep in mind that the platform and its clocks are configured by AMD. However, the initial numbers still look promising.
How do the Puma+-based cores hold up under the rigors of PCMark 8's Home sub-test?

The benchmark won't install on Dell's Venue 8 Pro, but we come away with insight from Intel's Celeron J1900 all the same.
In the conventional version of Futuremark's synthetic Home suite, the Bay Trail-D platform inches past AMD's custom-built tablet. But with OpenCL support turned on, Mullins achieves parity with Intel's desktop-oriented board.

Onto SiSoftware's Sandra Arithmetic module, where the A10 APU shows up between Intel's Atom Z3740D and Celeron J1900.

When it comes to GPGPU processing, AMD's Athlon posts the strongest finish, followed by the A10 Micro-6700T in this module's Encoding/Decoding module. It's cryptographic bandwidth and hashing performance trail Intel's Celeron slightly.
We brought games with us that we knew would be lightweight. However, the following two titles turned out to be more demanding than a 4.5 W APU can realistically handle, even with detail settings dialed all the way down. Had I been able to take the platform home with me, I would have revisited my choices. For now, the following results still tell an interesting tale about 3D performance, at least.


The frame rates are low across the board. But the A10 Micro-6700T beats the Celeron J1900 in Grid 2. I was hoping for more in Dota 2. However, it looks like that title prefers Intel's x86 architecture over AMD's. Still, that's not bad for a 4.5 W APU going up against a 10 W SoC designed to drop onto a desktop-class motherboard.
Unfortunately, I couldn't get Dota 2 onto Dell's Venue 8 Pro for Atom Z3740D numbers. That tablet is limited to 1280x800, and I was benchmarking at 1920x1080.
Don't be discouraged by the low frame rates. I've had surprising success getting some great titles running on tablets with Atom processors inside. Hopefully, Mullins pushes progress in that space even further, given AMD's experience with performance-oriented graphics.
Say what you will about the performance of most tablets running desktop operating systems, but at the very least, they should be capable of smooth Web browsing and video playback. With that in mind, we made it a point to squeeze in a handful of Web-based tests during our limited opportunity to audition AMD's engineering sample hardware.



There's one constant through these tests: Intel's Atom Z3740D trails the pack. Frankly, though, Dell's Venue 8 Pro doesn't feel like a last-place finisher. In fact, it seems fairly well-tuned for this exact usage model.
Even still, the A10 Micro-6700T places first, second, or third, depending on the test you're looking at.
I'm impressed with the potential of AMD's newest low-power APUs, just as I was hopeful that the previous generation would enjoy more success than it apparently did. The Discovery platform I was able to benchmark, featuring an A10 Micro-6700T APU, fares well compared to a number of other compelling SoCs.
Of course, the demo platform I played with was purpose-built to showcase the Mullins APU design, and those numbers only serve as an early indicator of what the hardware can do. We need to wait and see whether the company can attract IHVs. Not only does it face Intel, a powerhouse with advanced manufacturing and product shipping at 22 nm, but also the ARM-based efforts from Qualcomm, Samsung, Nvidia, and Apple.
The good news for AMD is that low-cost tablets with Android on them are prolific. They're everywhere. But they have their limitations, too. Typically, stepping up to a piece of hardware with full Windows 8.1 (not that Windows RT dreck) involves a corresponding price hike. If AMD is able to help enable more budget-conscious mobile devices able to run x86 software competently, it has the opportunity to attract a market full of value-seeking PC users unaccustomed to being told their apps don't work on the road.

It's odd that Microsoft managed to fall so far from grace in a segment that Bill Gates tried to invent at the turn of the millennium. The company completely failed to entrench itself before Apple arrived with its iPad, changing everything. Then, Google took the experience and made it more affordable with Android. I own iPad and Nexus tablets, and both satisfy me as media consumption devices. Neither proves to be especially useful for getting real work done, though. I did have an Iconia W3 with Windows 8. But while it checked the important functionality boxes, Intel's Atom Z2760 (that was Clover Trail, remember) was too slow for me to tolerate.
I now use that Dell Venue 8 Pro you saw me benchmark. Armed with an Atom Z3740D (Bay Trail) and Windows 8.1, it's the first tablet I've owned that truly satisfied me from the four perspectives of battery life, functionality, general performance, and portability. It only really lacks the ability to play demanding PC games, which is what's going to keep PCs around most of our houses for years to come.
But that's one of the reasons I see so much potential in AMD's hardware. This is a company with ATI's DNA. It has what it takes to augment graphics performance in mobile devices. Mullins will never cope with Crysis, but it might be able to handle titles that you've never been able to play on a tablet before. And it gives me hope that we're only a couple of generations away from an x86 tablet with real gaming chops.
Whatever shortcomings kept Temash and Kabini out of more shipping products, Mullins and Beema show significantly more promise, if only because of their lower power envelopes. The inclusion of an integrated ARM-based Platform Security Processor is a dark horse that may yield benefits we currently can't evaluate. Of course, Intel isn't standing still, and its 14 nm Bay Trail replacement, known as Cherry Trail, is well on its way. We'll see if AMD is able to gain real traction from Mullins and Beema in the coming months, and hopefully healthy competition can bolster Windows' success in the mobile space.