Analysis of the AMD Kaveri Notebook Platform

Till Schönborn (translated by Martina Osztovits), 06/05/2014

Kaveri goes mobile. The wait is over. The frugal notebook models follow the desktop models, which have already been available for a while. A total of nine APUs should compete with Intel's Haswell platform and especially score points with their superior graphics performance. We analyzed the potential of the new chips.

For the original German article, see here.

AMD already introduced the new Kaveri architecture less than four months ago. However, the model range only consisted of a few desktop models. Finally, the previous Richland platform, based on the above two-year-old Trinity design, was also replaced in the notebook range. Therefore, it was high time to launch a completely revamped successor.

Lithography & Design

Before we take a close look at the design of processor and graphics unit, we will give a general overview of the chip. A first and decisive innovation is already obvious here: While Trinity and Richland were still produced in 32-nanometer lithography, Kaveri features the finer 28-nanometer process. Apart from the fact that so-called half node shrink is used (CPUs are usually always produced in full node processes), its technological focus is surprising: While the old 32-nanometer SOI process (High-k + Metal-Gate) still mainly focused on CPUs with high clock rate, packing density got more important in the new 28-nanometer lithography. This is especially important for the complex GPU part.

That is why the die has suddenly space for more than 2.4 billion transistors instead of only about 1.3 billion in Trinity/Richland, although the size of 245 mm² remained virtually the same. Nevertheless, it still only features two CPU modules or four integer cores and a dual-channel DDR controller, which (depending on the model) also supports faster modules up the DDR3-2133. So, the GPU benefits primarily from the higher number of transistors - we will come back to this soon. Other enhancements include support of PCIe 3.0 and TrueAudio technology.

  • 28 nm SHP process (Globalfoundries)
  • 2.41 billion transistors
  • 245 mm² die size
  • 3rd generation Bulldozer cores ("Steamroller")
  • 2 modules (4 integer cores)
  • 4 MB L2 cache
  • DirectX 11.2 GPU with 512 Shader ALUs (GCN 1.1 architecture)
  • UVD 4.2, VCE 2.0
  • Memory controller up to DDR3-2133
  • PCI Express 3.0
  • TDP of 17, 19 or 35 Watts

Model Overview

Precise specifications of all Kaveri APUs have already been circulating on the Internet for a while and are now officially confirmed with a single exception: The previously announced A6-7000 dual-core APU is no longer listed in the data sheets. This reduced the line-up to nine models, six of them in the consumer range. (Nearly) all models feature two modules or four CPU cores and an integrated graphics unit. Apart from the usual distinguishing features, clock rate and GPU module, the models also belong to different TDP classes: SV models (standard voltage) belong to the 35-Watt category and ULV models (ultra low voltage) to the 19-Watt category. The more frugal models also lack support of bigger DDR3-1600 modules and the faster PCIe 3.0 standard.

The Pro series in the business range is new. However, its specs are basically identical with the consumer chips of the 19-Watt category. But, AMD guarantees an especially high stability and durability (in terms of availability of the APUs for several years), which the probably cheaper consumers do not offer.

Name Modules / Cores CPU clock L2 cache Graphics unit Memory PCIe TDP
A-Series SV              
FX-7600P 2/4 2.7 - 3.6 GHz 4 MB Radeon R7
(512 ALUs/686 MHz)
DDR3-2133 1x 16x-PCIe 3.0 35 W
A10-7400P 2/4 2.5 - 3.4 GHz 4 MB Radeon R6
(384 ALUs/654 MHz)
DDR3-1866 1x 16x-PCIe 3.0 35 W
A8-7200P 2/4 2.4 - 3.3 GHz 4 MB Radeon R5
(256 ALUs/626 MHz)
DDR3-1866 1x 16x-PCIe 3.0 35 W
A-Series ULV              
FX-7500 2/4 2.1 - 3.3 GHz 4 MB Radeon R7
(384 ALUs/553 MHz)
DDR3-1600 1x 8x-PCIe 2.0 19 W
A10-7300 2/4 1.9 - 3.2 GHz 4 MB Radeon R6
(384 ALUs/533 MHz)
DDR3-1600 1x 8x-PCIe 2.0 19 W
A8-7100 2/4 1.8 - 3.0 GHz 4 MB Radeon R5
(256 ALUs/514 MHz)
DDR3-1600 1x 8x-PCIe 2.0 19 W
A-Series Commercial              
A10 Pro-7350B 2/4 2.1 - 3.3 GHz 4 MB Radeon R6
(384 ALUs/553 MHz)
DDR3-1600 1x 8x-PCIe 2.0 19 W
A8 Pro-7150B 2/4 1.9 - 3.2 GHz 4 MB Radeon R5
(384 ALUs/533 MHz)
DDR3-1600 1x 8x-PCIe 2.0 19 W
A6 Pro-7050B 1/2 2.2 - 3.0 GHz 1 MB Radeon R4
(192 ALUs/533 MHz)
DDR3-1600 1x 8x-PCIe 2.0 17 W

CPU Architecture

Like its two predecessors, Kaveri also integrates a maximum of four CPU cores based on the Bulldozer architecture or more specifically the 3rd generation called "Steamroller". Compared to the "Piledriver" cores in Trinity and Richland, the new design should bring up to 20 percent higher performance per clock. AMD promises an average performance gain of about 10 percent. Independent benchmarks of the desktop models could already more or less (closely) confirm this.

But what is the reason for the sudden performance gain? An important one could be that there are now two separate decoders per module instead of a shared one. Although it uses up transistors and increases power consumption, it improves the workload and as a result, the performance increases. You could say, in doing so, AMD has slightly blurred the basic module concept again. In addition, an improved load/store unit, bigger registers, an optimized jump prediction and a bigger L1 instruction cache (96 instead of 64 KB per module) should lead to a performance gain.

Graphics Unit

About half of the chip size of Kaveri is used for the graphics unit. This is a lot of space, which AMD used to increase the shader count from 384 (Trinity/Richland) to 512 ALUs, or eight compute cores. Furthermore, the GPU no longer uses the outdated VLIW design, but the modern GCN 1.1 architecture of current Radeon graphics cards. This improves energy efficiency and hardware workload and the APU can also handle DirectX 11.2 "Tier 2" and AMD's Mantle API.

Early benchmarks run by AMD verify that the new Kaveri APUs have remarkable performance reserves. Already the FX-7500 (Radeon R7) and the A10-7300 (Radeon R6) 19-Watt modules should achieve about 1400 to 1500 points in 3DMark 11, which would even outperform several dedicated 3D accelerators from the entry-level segment. Intel's Haswell ULV chips, which only have a TDP of 15 Watts (including chip set), might not stand a chance against Kaveri's graphics performance. We eagerly await the results of the FX-7600P 35-Watt top model, which features the best GPU, the highest clock rate and fast DDR3-2133 RAM. Our measurements of desktop models with similar specs hint on quite promising performance.

As the implementation of AMD's HSA concept (Heterogeneous System Architecture) advances, CPU and GPU moved closer together in Kaveri. It is the first time they can access the same memory area via hUMA. As a result, performance-wasting copying is not needed. Another feature called hQ (Heterogeneous Queuing) makes communication and task distribution between CPU and GPU more flexible if both are involved in a calculation. However, the new features will first pay off with coming, especially optimized, software.

Verdict

AMD Kaveri
AMD Kaveri

The technical data are known. Now Kaveri has to prove its strengths in practice. It is particularly difficult to estimate the continuously available performance of the ULV models because of their big Turbo range – PCMark and 3DMark benchmarks published so far are only a very rough reference. Nevertheless, Kaveri will likely increase competitive pressure on Intel; most models might be significantly better than their Haswell counterparts in GPU performance. However, this is partly achieved with a slightly higher (platform) TDP. Impacts on battery life still remain to be seen.

The pricing of the coming production models might be an important aspect for success or failure of the new APU generation. In the past, the strong processor graphics of AMD chips was often combined with hardly faster, dedicated GPUs. As a result, costs and emissions increased, but the 3D performance did not to the same extent (micro stuttering and driver problems). High-res displays and fast SSDs should also no longer be reserved for Intel based notebooks and ultrabooks. If manufacturers oblige, ultra thin devices with AMD-APU could become a real alternative.

static version load dynamic

Comment this article:

> Notebook / Laptop Reviews and News > Reviews > Analysis of the AMD Kaveri Notebook Platform
Author: Till Schönborn, 2014-06- 5 (Update: 2014-06- 6)