Excavator core improve efficiency
During the International Solid-State Circuits Conference (ISSCC) 2015, AMD has revealed more details regarding its upcoming next-gen 28nm Carrizo APU, which combines Excavator CPU cores with GCN graphics cores and Fusion Control Hub (chipset) on the same chip as well as offer much better efficiency.
Although still based on a 28nm manufacturing process, the new Carrizo APU will offer significant performance and efficiency leap as well as offer quite a few new features.
The biggest novelty is the Excavator CPU module that Carrizo will be based on. Although based on the same 28nm manufacturing process, the Excavator CPU model, which features two x86-64 CPU cores, is 23 percent smaller in area when compared to Steamroller, mostly to a new high-density library design which reduced die-area of the module.
AMD also made the floating point scheduler 38 percent smaller, fused multiply-accumulate (FMAC) units are 35 percent smaller and instruction-cache controller is 35 percent smaller. This allowed AMD to have much lower power consumption and less loss due to leakage, all without redesigning the manufacturing process.
Although being smaller, Carrizo still packs 3.1 billion transistors which is about 29 percent increase compared to Kaveri. According to numbers from AMD, all these design optimizations, including a new adaptive-voltage algorithm, allowed AMD to have a 10 percent CPU clock bump over Kaveri at the same power, with CPU using 19 percent and GPU using 10 percent less power compared to Kaveri APU design.
As far as the GPU part goes, Carrizo will feature a Graphics Core Next 1.3 GPU with 8 Compute Units (CUs), which adds up to 512 Stream Processors (SPs). It will feature support for Mantle and DirectX 12 API support and H.265 hardware-acceleration.
While it did not talk about specific performance or chips that will show up AMD has already announced Carrizo APUs back in November and we know these will be available in two versions, Carrizo and Carrizo-L for notebooks. The Carrizo APUs are expected to launch sometime in Q2 2015 and all those promises of performance improvements with high power efficiency certainly sound impressive.
Source: AMD.com.
Although still based on a 28nm manufacturing process, the new Carrizo APU will offer significant performance and efficiency leap as well as offer quite a few new features.
The biggest novelty is the Excavator CPU module that Carrizo will be based on. Although based on the same 28nm manufacturing process, the Excavator CPU model, which features two x86-64 CPU cores, is 23 percent smaller in area when compared to Steamroller, mostly to a new high-density library design which reduced die-area of the module.
AMD also made the floating point scheduler 38 percent smaller, fused multiply-accumulate (FMAC) units are 35 percent smaller and instruction-cache controller is 35 percent smaller. This allowed AMD to have much lower power consumption and less loss due to leakage, all without redesigning the manufacturing process.
Although being smaller, Carrizo still packs 3.1 billion transistors which is about 29 percent increase compared to Kaveri. According to numbers from AMD, all these design optimizations, including a new adaptive-voltage algorithm, allowed AMD to have a 10 percent CPU clock bump over Kaveri at the same power, with CPU using 19 percent and GPU using 10 percent less power compared to Kaveri APU design.
As far as the GPU part goes, Carrizo will feature a Graphics Core Next 1.3 GPU with 8 Compute Units (CUs), which adds up to 512 Stream Processors (SPs). It will feature support for Mantle and DirectX 12 API support and H.265 hardware-acceleration.
While it did not talk about specific performance or chips that will show up AMD has already announced Carrizo APUs back in November and we know these will be available in two versions, Carrizo and Carrizo-L for notebooks. The Carrizo APUs are expected to launch sometime in Q2 2015 and all those promises of performance improvements with high power efficiency certainly sound impressive.
Source: AMD.com.
News by Luca Rocchi and Marc Büchel - German Translation by Paul Görnhardt - Italian Translation by Francesco Daghini
Previous article - Next article
comments powered by Disqus
comments powered by Disqus