Latest 3rd generation Xeon Scalable processor called Ice Lake
Table of Contents
New third-generation Xeon Scalable processor called Ice Lake
New third-generation Xeon Scalable processor called Ice Lake
Processor:
The 10 nanometer-based CPU is delivering up to 40 cores per processor.
The processors also add Intel SGX (Software Guard Extensions) for built-in security, as well as Intel Crypto Acceleration. It includes Intel’s DL Boost for AI acceleration.
-
AI acceleration:
Intel says that with hardware and software optimizations, Ice Lake delivers 74 percent faster AI performance compared with the prior generation. Compared to the third-generation AMD Epyc 7763, Intel says the platform delivers up to 1.5 times higher performance across a broad mix of 20 popular AI workloads. Compared to the Nvidia A100 GPU, it says it delivers up to 1.3 times higher performance on a broad mix of 20 popular AI workloads.
-
Built-in security:
Intel SGX, and security has the technology can protect as much as 1 terabyte of code and data in private memory areas called enclaves.
Ice Lake chips also feature cryptographic acceleration that promises to let the chip deliver both security and performance.
Some of the key differences between Intel’s new 3rd gen Xeon CPUs with Intel’s 2nd gen Xeon Scalable processors. The below table highlights the processor architecture, cache latencies, memory controller, memory latency, and capacity.
| Latency | Intel Xeon Platinum 8380 Processor (Ice Lake) | Intel Xeon Platinum 8280 Processor (Cascade Lake) |
| L1 hit cache, cycles | 5 | 4 |
| L2 hit cache, cycles | 14 | 14 |
| L3 hit cache (same socket) | 21.7 | 20.2 |
| L3 hit cache (remote socket) | 118 | 180 |
| Memory Controller | On die – 8 ch | On Die – 6ch |
| Max DIMM Capability | 2 DPC 3200/2933/2666 New: PMem (SKU dependent) runs at memory channel speed | 1 DPC 2933/2 DPC 2666 (SKU dependent) |
| DRAM read latency local socket | 85 | 81 |
| DRAM read latency remote socket | 139 | 138 |
| Max Memory Capacity per Socket | 6TB (DDR+PMem). 4TB (DDR) | 4.5 TB (DDR+PMem). 3TB (DDR) |
RAM:
The platform supports up to six terabytes of system memory per socket, up to eight channels of DDR4-3200 memory per socket, and up to 64 lanes of PCIe Gen4 per socket.
Ice Lake delivers an average 46 percent performance increase on popular data center workloads
Network:
Intel produces network-optimized “N-SKUs” to support diverse network environments. The new processors deliver on average up to 62 percent more performance on a range of broadly-deployed network and 5G workloads over the prior generation, Intel says. More than 15 major telecom equipment manufacturers and communications service providers are readying POCs and networking deployments with Ice Lake.
Workload:
Ice Lake delivers up to 1.56 times more AI inference performance for image classification than the prior generations, Intel says. This makes it suitable for AI, complex image or video analytics, and consolidated workloads at the edge.
Intel Optane SSD Storage:
Intel increased the capacity of the drive which started at 375GB, tops at 1.5TB. Optane SSDs perfect companion for slower media. we primarily designed for endurance and frequently used in a two-tier architecture. Optane SSDs are often set up to absorb all of the writes in a system, protecting larger, slower media behind it. These can be QLC SSDs for instance, that perform well for read-heavy workloads, but don’t have much endurance or write performance.
vSAN, Azure HCI, StorONE, and many others are well adept at making multi-tier work well.
PMem: Intel Optane Persistent Memory
A way to bridge the gap between volatile DRAM and high-performance SSDs. PMem 200 is quoted to deliver 32% more memory bandwidth than Gen1. New modules are optimized for the new 3rd Gen Intel Xeon Scalable Processors, making the combination, with Intel SSDs.
In the memory bandwidth throughput, with PMem 200 picking up 3,200 MT/s support.
Core count starts at 16 cores and goes up to 40. With the first PMem, users could add 3TB of PMem for a total memory per socket of 4.5TB, now the total is 6TB per socket, with 4TB of PMem 200 being added. The maximum thermal design power dropped from 18W to 15W. And the newest persistent memory comes with eADR, extended Asynchronous DRAM Refresh.
The MemVerge Memory Machine management interface can help administrators through a number of use cases:
Crash recovery:
The snapshot GUI is used to quickly bring back the database and/or troubleshoot the cause. The database log and Memory Machine Dashboard data determine the time of the crash, allowing the admin to select and restore a snapshot that is closest to the crash time. Developers can then use that restored instance for debugging.
Accelerating Animation & VFX with Memory DVR:
Animation:
Artists want to explore different options on a base Maya scene. They load the base scene, apply the changes, and save it as a different project. They can save many separate scenes, but to show these options they must be repetitively reloaded, which takes a long time.
With Memory DVR functionality, load a base scene once, take a snapshot as the basic snapshot, then apply changes and take another snapshot. To apply a different effect, simply restore the snapshot, edit, and take another snapshot. The restore speed of in-memory snapshots is a few seconds compared to minutes for reloading scenes from storage.
Accelerating Genomic Analytics with Memory DVR:
Scientists want to experiment with a machine learning algorithm using different parameter settings. They load the data, set the parameter, run the algorithm, and check out the results. if the results are not good, the data is reloaded, a different set of parameters is applied, and the algorithm is run again.
With Memory DVR functionality, load the data once and take a snapshot. From that point on, if the results are not good, restore the base data and another run with new parameters is done in seconds.
Intel Agilex FPGAs that leverage PCIe Gen4.
Intel Agilex FPGA family leverages heterogeneous 3D system-in-package (SiP) technology to integrate Intel’s first FPGA fabric built on 10nm process technology and 2nd Gen Intel Hyperflex FPGA Architecture to deliver up to 40% higher performance1 or up to 40% lower power1 for applications in Data Center, Networking, and Edge compute.
Workloads:
VDI, databases, AI and machine learning, Cloud, big data, HPC and Analytics.
About us
Navigator System offers end-to-end data center maintenance services that allow you to extend the life of your IT assets.
MANAGED IT SERVICES are
- 24×7 Field Services & Technical Support
- Multi-Vendor Maintenance Programs (NetGuard)
- Cloud & Network Migration Services
- Monitoring & Infrastructure Management
- Consulting & Project Management Services
- Network Security
- Network Lifecycle Management
- Asset Recovery Programs
- Spares Management & Repairs for AMC contracts
- Materials Management & Reverse Logistics for AMC contracts
CONTACT:
Call our Sales Team: +91 9986288377
Email: sales@navigatorsystem.com
