Latest rumours on Intel Haswell architecture
Latest rumours suggest that Intel will launch its next generation Haswell processor during the month of June, 2013.
It is well known that Ivy Bridge was a slight refinement of previous Sandy Bridge and has improvements in terms of power consumption, manufacturing process and performance per cycle.
So Intel is aiming to improve both performances per cycle and energy consumption further, with new Haswell architecture.
HWBOT has come out with some expected details of Intel’s fourth generation microprocessor. It will be built using new micro-architecture called Haswell.
Intel promises to deliver higher performance through Haswell architecture by making it to simultaneously execute up to eight instructions per cycle. It is notable that Ivy Bridge and earlier version can execute only up to 8 instructions per cycle.
It is also believed that the new architecture will offer rich set of instructions like AVX 2.0, FMA3, and TSX and so on.
Sub-redesigned Slices in Intel Haswell
Intel has planned to use fourth generation modular architecture code named Denlow in Haswell.
The modular architecture of Haswell will be based on various modules called Sub-slices, where each Sub-slice will be composed of multiple computational units. In previous architectures a Sub-Slice consisted of six to 8 EUs, but in Haswell each Sub-Slice will have 10 EUs.
Intel has added the ability to use multiple Slices in Denlow, so it will have following configurations:
- Denlow GT3 with two slices (40 EUs).
- Denlow GT2 with one Slice (20 EUs).
- Denlow GT1 with one Slice, but with a Sub-Slice disabled (10 Eus).
Video Quality Enhancement in Haswell
Haswell will have a new video unit in Denlow, which will act as hardware-accelerated video unit.
It eliminates the need of EUs for video decoding calculations and frees it for performing other actions like gaming graphics calculations.
With new Video Quality Enhancement unit, Intel promises to improve the quality of the videos by introducing new features like video stabilization.
Haswell vs Ivy Bridge in terms of IGP
Compared to IGP of Ivy Bridge codenamed Carlow, the new Denlow IGP of Haswell is supposed to have improvements like
- It will have new graphics engine that supports DirectX API 11.1, OpenGL 4.0 and OpenCL 1.2.
- It will have Mayor independence CPU internal bus
- It will have improved texture units and 4 times faster sampler
Benchmark result of Intel Haswell microarchitecture
Russian OCLab.ru website managed to get the first testing results from the upcoming Intel Haswell architecture.
They put the CPU through four popular benchmarks for overclockers: SuperPI 1M, SuperPI 32M, PiFast, Wprime 32M and Wprime 1024M. Haswell B0 stepping CPU clocked at 2.8GHz was compared to an Ivy Bridge clocked at 2.8GHz.
Note: The results are in seconds and lower is better
Super PI 32M
Haswell: 11 minutes 27.505 seconds
Ivy Bridge: 11 minutes 49.094 seconds
Haswell: 24,01 seconds
Ivy Bridge: 25,5 seconds
Haswell: 12.8x seconds
Ivy Bridge: 13.97 seconds
Haswell: 7 minutes 11.181 seconds
Ivy Bridge: 7 minutes 11.8xx seconds
Super PI 1M
Haswell: 14,1 seconds
Ivy Bridge: 13,1 seconds
It is speculated that Intel Haswell will yield 10% more performance per cycle than Ivy Bridge architecture.
In the preliminary test, Haswell-DT CPU shows 0.14 to 9.14% higher yield than Ivy Bridge-DT.
But it is notable that performance is strangely 7.09% lower in the SuperPi 1M test, which may be due to the premature state of the microprocessor tested. So it is expected that the final version of chip will be more consistent in all applications.