Steamroller
Third-generation AMD microarchitecture for APUs
AMD developed Steamroller as the third iteration of its Bulldozer design. Released in early 2014, it powers Kaveri APUs and aims to improve parallelism and efficiency over the previous Piledriver generation.
- Released in January 2014 for Kaveri APUs
- Built on a 28 nm Super High Performance process
- Improved instruction per cycle by up to 30%
- Features independent decoders for each core
- Cancelled server variants due to efficiency issues
What it is
Steamroller is a microarchitecture created by Advanced Micro Devices for its Accelerated Processing Units. It arrived in January 2014 as the third generation of the Bulldozer family. This design focuses on improving parallel processing capabilities while maintaining high clock speeds. It serves as the foundation for the Kaveri line of chips, which combine central processing units with graphics cores. The architecture targets desktop and mobile markets, offering better instruction handling than its predecessor.
How it came about
AMD first announced plans for this third-generation Bulldozer core in 2011. Internal documents initially referred to it as Next Generation Bulldozer before the Steamroller codename leaked later that year. The company intended to launch these processors in 2013 using a 28 nanometre manufacturing process. However, the first products based on this design, known as Kaveri APUs, did not reach the market until January 2014. A refreshed version called Godavari followed between 2015 and 2016.
Inside the chip
The architecture retains the clustered multi-threading module structure from earlier Bulldozer designs. Each module contains two cores that share certain resources but feature independent instruction decoders. Improvements include a wider dispatch per thread, better branch prediction, and larger caches. These changes aim to reduce instruction cache misses and boost instructions per cycle. The design also integrates Graphics Core Next technology to support heterogeneous system architecture features. An embedded ARM co-processor handles security tasks in some models.
How it was made
GlobalFoundries manufactured these chips using a 28 nanometre Super High Performance process. The production technique allowed AMD to balance performance and power consumption. While specific transistor counts are not detailed in the provided records, the shrink in process node helped enable higher clock rates. The chips were packaged for Socket FM2+ on desktops and Socket FP3 for mobile devices. This manufacturing approach supported the integration of both CPU and GPU components on a single die.
Where it went
Steamroller powers the Kaveri A-series APUs for desktop and mainstream computing. Top models like the A10-7850K offer quad-core performance with integrated graphics. Mobile variants appear in laptops, such as the FX-7600P. Some desktop parts, like the FX-770K, disable the graphics unit for OEM use. The architecture supports PCIe 3.0 and DDR3 memory. It also enables features like 4K video output and multi-monitor setups through its integrated graphics capabilities.
Versions and family
The primary consumer variant is the Kaveri APU, available in A10, A8, A6, and A4 series. A refresh known as Godavari introduced enhanced cores and higher frequencies in 2015. AMD planned server versions named Berlin and Seattle but cancelled the x86-based Berlin CPU due to efficiency concerns. The FX series did not receive a direct Steamroller update for the AM3+ socket. Instead, limited OEM-only parts were released with disabled graphics.
Why it matters
This generation delivered a modest nine percent improvement in single-threaded performance over Piledriver. Multi-threaded gains reached eighteen percent. Although energy efficiency remained a challenge compared to competitors, it laid groundwork for the subsequent Excavator architecture. Excavator would later double the performance per watt achieved by Steamroller. The cancellation of server plans highlighted the limitations of the Bulldozer lineage in enterprise environments during this period.
Questions people ask about Steamroller
What is the Steamroller microarchitecture?
Steamroller is the third generation of AMD's Bulldozer family. It was designed for Accelerated Processing Units and launched in early 2014. The architecture focuses on improving parallelism and instruction throughput compared to the previous Piledriver design.
Which processors use the Steamroller core?
The primary chips using this core are the Kaveri A-series APUs. These include desktop models like the A10-7850K and mobile parts. A refreshed version called Godavari also uses enhanced Steamroller cores. Some OEM-only FX parts utilize this architecture with disabled graphics.
How does Steamroller improve on Piledriver?
It introduces independent instruction decoders for each core within a module. The design features larger caches, better branch prediction, and a wider dispatch width. These changes result in fewer cache misses and an estimated thirty percent increase in instructions per cycle over the original Bulldozer.
Good reading and sources
- Checking your browserAnandTech, Article
- AMD to add ARM processors to boost chip securityArs Technica, Article
- Steamroller (microarchitecture) on WikipediaEncyclopedia
- More photos on Wikimedia CommonsPhotos
- Wikidata recordOpen data
The facts on this page come from the Wikipedia article "Steamroller (microarchitecture)", written by its contributors and used under the CC BY-SA 4.0 licence, and from its Wikidata record (CC0). EveryChip rewrote the text in its own words, with AI assistance, and compares every page with the article to make sure no passage is copied. It may be reused under the same CC BY-SA 4.0 licence with a link back.
It may contain mistakes, and figures change over time. Nothing here is investment or buying advice. If something is wrong, tell us. Last refreshed 2026-11-14.
Like Steamroller
All CPUs and microprocessors
🇺🇸 United StatesAMD APU
AMD chips combining CPU and GPU on one dieCPU · x86 · 2011
WorldVIA C7
Low-power x86 processor for compact and mobile devicesCPU · x86 · 2005