ARM’s Cortex A76 CPU Will Clock At 3.3GHz With A 35% Power Boost And 33% IPC Gain As Company Directly Targets Intel
This year, it's all about ARM powering Windows 10 PCs, as several manufacturers have started to move in the direction. As ARM's Cortex cores start to increase in power and decrease in size, their capabilities are starting to prove more than adequate to perform everyday computing for the majority of users.
With this trend starting to show signs of early maturity, ARM has now launched its latest Cortex A76 cores. These represent a completely new microarchitecture over their predecessors, as ARM makes some massive upgrades this year. In fact, the company describes the processor as capable of meeting desktop requirements with the power efficiency of mobile chips.
ARM Launches The Cortex A76 With 35% Performance Boost, 40% Power Efficiency, And 4X Machine Learning Improvement; Brand New Microarchitecture Allows For Smoother Performance Over Predecessors
The Cortex A76 launch from ARM opens up a lot of dimensions for the company's designs. However, before we get into these, it's relevant to take a look at what's changed on the core over the year. In terms of basic numbers, the ARM Cortes A76 will offer a 35% performance boost and 40% power efficiency over a Cortex A75 clocked at 2.8GHz. The CPU will also reach up to 3.3GHz in frequency.
ARM has also increased L3 cache on the A76, by 2MB. The new L3 for the core sits at 4MB and is expected to offer at least twice the performance of the A73. As the A76 is expected to utilize 7nm and clock above 3.3GHz, both of these will contribute significantly to its performance gains. In addition, as mentioned above, the A76 also introduces a new microarchitecture - with this also contributing towards the aforementioned gains.
The Cortex A76 features an all new branch prediction unit. The branch unit now has twice the IPC of the fetch unit. This gives it a significant performance edge and allows the processor and the fetch to continue operations in case of a mispredict. The A76's CPU is also able to reduce cycles on instruction latency over the A75 and stay at par with Samsung's Mongoose M1.
ARM Aims Directly At Intel With The Cortex A76; Mike Filippo, Lead Processor Architect Outlines Intel As His Company's Biggest Rival
One big avenue that the Cortex A76 will target (or satiate), is performance requirements for Windows 10 PCs. ARM's head of Intellectual property Rene Haas outlined the A76 2x performance gain over the A73 used on current Windows 10 Always Connected PCs. One major drawback of these devices is their power - a fact that has also led to friction between ARM and Microsoft as the latter hopes to take on Google's Chromebooks.
ARM's lead processor architect Miki Filippo highlighted the A76's IPC gains over the A75 and A73. The A76 has a decode width of 4 IPC, a gain of 33% over its predecessor. Additionally, Filippo also claims that the 35% performance boost advertised by ARM represents the minimum gain for the processor. The A76 is capable of a staggering 90% performance gain for software which requires fast memory access. Of course, since there's still a good 10 months left before we see the A76 in the flesh, don't take these claims as an absolute fact.
Alongside the Cortex A76, ARM also announced the Mali G76 GPU and V76 video processor. The GPU is able to offer a 30% speed increase over its predecessor while the V76 is able to support 8K displays. ARM's strict focus towards comparing the A76 with the A73 demonstrates the company's eagerness to establish a strong presence in the notebook market. The sharks smell blood, Intel.
ARM's Big Promises With The Cortex A76 Hope To Directly Take On Intel; Will The Company Succeed In This Goal?
Mike Filippo also expects that the Cortex A76 will match Intel Core i5-7300 in performance terms and can also match the higher end Core i7 if equipped with more cache. Intel's inability to move towards 10nm and its insistence to continue refining the 14nm node are well known. Right now, it's the Apple/ARM camp that continue to make gains and while TSMC or Samsung's 10nm node is not similar to Intel's, it can achieve parity over the next couple of years.
Even as Intel's (and x86) lead is solid when we talk performance, Santa Clara's processors are notorious power hogs. It's this segment which holds the greatest potential of entry by ARM into the mainstream computing space. While ARM claims to have demonstrated a laser-like focus on microarchitectural improvements and detail on the A76, it still shouldn't be able to match Intel in terms of performance at least.
Even if it does, backward compatibility for the x86 ISA will have to be emulated and you can imagine the performance hit there. Additionally, Intel has decades of semiconductor experience and while x86's drawbacks or ISA impact on performance are debatable, a large chunk of the current freeze is rightly attributable to physics. While ARM can tinker around microarchitecture all it wants, beating physics will prove to be truly miraculous and therefore, completely, if not highly improbable.
Thoughts? Let us know what you think in the comments section below and stay tuned. We'll keep you updated on the latest.