What is Gracemont Processor? (Explained)

What is Gracemont Processor?

The Intel Gracemont processor is the successor to the CPUs based on the Tremont microarchitecture from Intel. Technically, these processors are built on the 10 nm manufacturing process.

The Gracemont microarchitecture is thoughtfully designed and manufactured by Intel to take full advantage of the Intel 7 process, formerly known as 10 nm Enhanced SuperFin (ESF).

KEY TAKEAWAYS

• The Gracemont microarchitecture is designed by Intel on a 10 nm manufacturing process.
• This particular architecture is designed to be used in low-power processors used in SoCs made by Intel.
• These processors typically support a larger Level 1 instruction cache and a DDR5 memory module.
• The performance of the Gracemont processors is much improved over that of its predecessor, Tremont, with a much improved and newer instruction set and extension support.
• The CPUs offer much better I/O performance due to their support for PCI Express 4.0 in place of PCIe 3.0.

Understanding Gracemont Processor

The Systems on a Chip (SoCs) made by Intel typically use low power processors built on specific microarchitectures such as Gracemont.

The microarchitecture of this processor is built on a 10 nm manufacturing process and is the successor to Tremont and the predecessor of Cresmont.

But, just like its predecessor, this architecture is also used in a hybrid design as the low-power cores of the 12^th generation hybrid processors from Intel, codenamed Alder Lake, and the 13^th generation hybrid processors, codenamed Raptor Lake.

The processors based on this specific microarchitecture are designed very carefully with a better front-end and back-end.

As for the front end, the design aspects include the following:

• A bigger Level 1 instruction cache of 64 KB per core, which is up from the earlier 32 KB per core and
• An additional On-Demand Instruction Length Decoder (OD-ILD).

As for the back-end, the design aspects include the following:

• Improved Reorder Buffers or ROBs to 256, which is up from the earlier 208
• A 17-wide issue and
• Four Arithmetic Logic Unit Single Instruction, Multiple Data (ALU SIMD) instead of three.

Built on a smaller 10 nm fabrication process, the Gracemont processors are suitable for use in smaller and ultra low-power devices, such as PCs and microservers.

This is due to several improvements in its design such as:

• A DDR5 memory support in place of DDR4
• A better I/O throughput with Peripheral Component Interconnect Express or PCIe 4.0 support in place of PCIe 3.0
• New instruction set support such as AVX2, AVX-VNNI and others.

General specifications

A lot of improvements have been made in the Gracemont architecture to provide better support to the processors while they are performing. Here are some of the general specifications:

• It was designed and released by Intel on November 4, 2021.
• It refers to the 4^th generation of low-power, Out-of-Order Atom microarchitecture.
• There is an increased instruction issue from four to five per clock as compared to the earlier generations.
• It comes with an increased instruction retire rate from seven to eight per clock.
• The number of execution ports or functional units in it has increased from eight to seventeen.
• It is capable of much better branch prediction.
• The maximum CPU clock rate achieved can range anywhere between 700 MHz and 4 GHz.
• The number of cores can be up to four in each module.
• It supports a System on a Chip architecture, which allows it to be used in low-power and smaller electronic systems.
• The 3D tri-gate transistors offer a much better performance.
• The Thermal Design Power is much improved, being 10 watts for desktop processors and 6 watts for mobile processors.

Integrated graphics processor

Few models of Gracemont processors support an Intel Xe Gen 12.2 integrated graphics processor. This specific GPU supports the following features:

• DirectX 12
• OpenGL 4.6
• OpenGL ES 3.2
• OpenCL 3.0
• Vulkan 1.3

Pipeline

The CPUs designed on the Gracemont architecture support a superscalar pipeline that is capable of the following:

• Out of Order Execution or OoOE
• Speculative execution
• Register renaming

Instruction Set Architecture and extension support

The Gracemont processors support x86-64 Instruction Set Architecture (ISA) and several different extensions of theirs such as:

• MOVBE or Move Data After Swapping Bytes
• MMX or MultiMedia eXtensions
• SSE or Streaming SIMD Extensions, along with all its variants such as SSE2, SSE3, SSSE3, SSE4.1, and SSE4.2
• POPCNT or Population Count
• CLMUL or Carry-less Multiplication
• XSAVE or Save Processor Extended States
• XSAVEOPT or Save Processor Extended States Optimized
• FSGSBASE or FS/GS Base Access Instructions
• PTWRITE to write data to a processor trace packet
• RDPID or Read Processor ID
• SGX or Software Guard Extensions
• UMIP or User Mode Instruction Prevention
• GFNI-SSE or Galois Field New Instructions Streaming SIMD Extensions
• CLWB or Cache Line Write Back
• ENCLV or Enclave instructions
• SHA or Secure Hashing Algorithm
• AVX or Advanced Vector Extensions, along with AVX2
• FMA3 or Fused Multiply Add
• AVX-VNNI or Advanced Vector Extensions Vector Neural Network Instructions
• AES-NI or Advanced Encryption Standard New Instructions
• RDRAND or Read Random
• TXT or Text File extension
• VT-x and VT-d or Virtualization extensions

Memory system

The Gracemont processors come with L1, L2, and L3 cache memory that helps with their performance.

• The Level 1 cache comes with 96 KB per core, out of which 64 KB is for instructions and 32 KB is for data
• The Level 2 cache   measures anywhere between 2 MB and 4 MB in each of the four-core modules
• The Level 3 cache measures up to 3 MB per module

Gracemont vs Golden Cove

• The maximum operating speed of the cores of the Gracemont processors ranges between 700 MHz and 4 GHz. On the other hand, the maximum CPU clock speed in the case of Golden Cove is 1 GHz to 5.5 GHz.
• The Level 1 cache size of the Gracemont processors is usually 96 KB, with 64 KB reserved for instructions and 32 KB for data. On the other hand, the Level 1 cache size of the Golden Cove processors measures 80 KB, with 32 KB reserved for instructions and 48 KB for data.
• The Level 2 cache size of the Gracemont processors ranges between 2 MB and 4 MB per module, but in comparison, the Level 2 cache size of the Golden Cove processors is 1.25 MB and 2 MB for client and server processors, respectively.
• The Gracemont processors are typically manufactured on the 10 nm manufacturing process, but in comparison, the Golden Cove processors are manufactured on the Intel 7 technology node.
• The predecessor of the Gracemont processors is Tremont. On the other hand, the predecessors of the Golden Cove processors are the 10 nm server Sunny Cove, the 14 nm server Skylake, the 10 nm Willow Cove, and the 14 nm desktop Cypress Cove processors.
• The successor of the Gracemont processors is Cresmont, but in comparison, the successor of the Golden Cove processors is Raptor Cove.

Conclusion

The Intel Gracement processors, successor to Tremont and predecessor of Cresmont, come with improved design than its predecessor and support quite a few new and old technologies.

The design and architectural aspects help it significantly to offer a much improved performance in smaller devices and PCs.