The A10-7870K APU is built on the Kaveri architecture and refreshes the A10-7850K, mostly by way of small frequency boosts to the CPU and GPU. This AMD A10-7870K review compares the APU against budget CPU + GPU purchases – like the G3258 – and Intel's own HD 4000 series IGPs.

AMD's A10-7870K is priced at $150 at most retailers. Note that, at the time of this review, Amazon is selling the A10-7870K for $122 as a “daily deal.” Pricing will return to $150 once the day has concluded.

We understand that AMD's market positioning and target may differ in Asiatic countries. This review strictly looks at the product from a US / Western perspective, where the price structure of competing and ancillary products is more familiar.

AMD A10-7870K Specs

Changes from the A10-7850K

The A10-7870K reflects similar feelings to the R300 refresh. It's not a new architecture, but there are some small changes instituted to the existing platform. As with the R300 vs. R200 references, we would consider the 7870K a “refresh” to the 7850K; that is, it's not a true “rebadge” – the selling of an existing product under a new name – but isn't entirely new, either. There is no inherent flaw with refreshing a product as long as the new price and feature updates reflect progress. Most companies, even non-electrical manufacturers (like those who make cases and coolers), will refresh products between cycles.

AMD's 7870K has increased the CPU frequency from 4GHz / 3.7GHz (turbo / standard) on the 7850K to 4.1GHz / 3.9GHz (turbo / standard) on the 7870K; that's a 0.2GHz pre-overclock. The GPU frequency is slightly more noteworthy, increasing from 720MHz to 866MHz. The rest is the same – Dx12, HSA, and FreeSync are all supported on both the 7850K and 7870K. The 7870K still hosts 4 CPU and 8 GPU cores.

AMD's Kaveri platform hosts Steamroller, for those who recall the now-aging Bulldozer / Piledriver / Steamroller nomenclature. The architecture remains heavy on COMPUTE and arithmetic (integer) processing. Steamroller drove IPC performance more heavily than its power-focused predecessor, Piledriver. Kaveri hosts 2.41B transistors on a 245mm^2 physical die package, using 28nm manufacturing process.

Kaveri's programming model allows developers to tap into its heterogeneous system architecture (HSA) for more equal loading of the coupled CPU and GPU components. The CPU and GPU are able to jointly access most sets of data in the same memory locations, theoretically allowing for co-processing in a fashion that accelerates applications. This is something that must be optimized for in the codebase and is not necessarily a “free” performance improvement.

Test Methodology

We tested using our updated 2015 GPU test bench, detailed in the table below. Our thanks to supporting hardware vendors for supplying some of the test components.

The latest AMD Catalyst drivers (15.7.1) were used for testing. NVidia's 353.62 drivers were used for testing. Game settings were manually controlled for the DUT.

Testing was split between CPU bottleneck analysis and IGP benchmarks. To test at what point the CPU bottlenecks a high-end GPU, we equipped all platforms with a GTX 980 Ti and performed tests using the following settings:

• Metro: Last Light - Very High w/ high tessellation.
• GRID: Autosport - Ultra, 4xMSAA.
• Shadow of Mordor - Ultra preset.
• GTA V - Very High / High.
• GTA V - Very High / High with Advanced Graphics (to find bottlenecks).
• FireStrike (normal).

IGP testing was performed with a lower-end set of games due to limitations on the on-die GPU:

• CSGO - High / Very High with AA disabled.
• GRID: Autosport - Medium.
• DOTA2 - Medium on the sliders, other settings as below.
• GTA V - Lowest (normal).
• FireStrike (normal) for physics.

All tests had the resolution set to 1080p.

Each game was tested for 30 seconds in an identical scenario, then repeated three times for parity.

Average FPS, 1% low, and 0.1% low times are measured. We do not measure maximum or minimum FPS results as we consider these numbers to be pure outliers. Instead, we take an average of the lowest 1% of results (1% low) to show real-world, noticeable dips; we then take an average of the lowest 0.1% of results for severe spikes. Anti-Aliasing was disabled in all tests except GRID: Autosport, which looks significantly better with its default 4xMSAA. HairWorks was disabled where prevalent. Manufacturer-specific technologies were used when present (CHS, PCSS).

Here is the Intel test bench:

And the AMD bench:

Continue to page 2 for the benchmark results!