5th Generation Xeon Processors are now available at DigitalOcean

DigitalOcean is proud to announce the release of the fifth generation of Intel’s Xeon Scalable Processors (codenamed Emerald Rapids) to the DigitalOcean premium dedicated fleet. As the newest generation of processors in Intel’s product line, they deliver impressive performance gains over prior generations and will result in significant performance and experience improvements for DigitalOcean users and their customers.

The latest generation of Intel Xeon Scalable processors are currently available for Premium Dedicated CPU Optimized and General Purpose Droplets in the SFO3 and AMS3 data center, with more Droplet types and data centers coming soon.

Discover the performance advantage of Intel 5th generation Xeon Processors

The new designs being introduced are dual-socket Intel Xeon Emerald Rapids systems based on a ten-nanometer chip lithography. This CPU generation provides significantly increased performance per watt thanks to the smaller lithography of the chip (as much as a 17% improvement over Sapphire Rapids and as much as 36% improvement over the Ice Lake CPUs*, which DigitalOcean released in 2021). They enable DDR5 and PCIe Gen5 speeds for memory and component interfacing, and include new Intel acceleration engines on die. Intel’s AMX Advanced Matrix Extensions is a great example—this will allow AI/ML inference workloads with up to 20 million parameters to run on our Xeon Emerald Rapids systems.

DigitalOcean has also chosen 80 PLUS Titanium Grade power supplies for all Server SKUs, which are on average 2% more efficient than Platinum*—reducing our power consumption and bringing our platform in compliance with EU/EC directives around power efficiency. This helps to improve our Power Usage Effectiveness (PUE) and decrease the environmental impacts of our operations.

Compared to the Ice Lake product generation, the Emerald Rapids product generation demonstrates a roughly 30% improvement in throughput per system for most benchmarks, with gains as high as 45% on specific I/O tests that benefit from the increases in both memory access speed and PCIe bus speed*.

Core performance shows a marked improvement over Ice Lake, with our Emerald Rapids solutions benchmarking at significantly higher performance levels per core and overall cores in the system. In the charts below the purple bar represents the new Emerald Rapids solution*.

Testing using Geekbench* reflected similar gains in performance over the Ice Lake generation server designs previously present in the DigitalOcean fleet.

Storage performance also shows a marked improvement over Ice Lake, with random reads and writes both reflecting a significant increase in performance. In the charts below the purple bar represents the new Emerald Rapids solution*.

More speed, same price

The pricing for Droplets has not changed with the introduction of 5th Generation Intel Xeon Scalable processors. This allows you to deliver faster customer experiences without increasing your costs.

Unlock the full potential of your business with DigitalOcean, spin up a Droplet today. New to DigitalOcean? Get started on us with $200 in free credits.

Citations -

1. Intel “Emerald Rapids” Xeon Platinum 8592+ Tested, Shows Improvement over Sapphire Rapids
2. AVX-512 Performance Comparison: AMD Genoa vs. Intel Sapphire Rapids & Ice Lake
3. The FSP Hydro Ti Pro 1000W PSU Review: Titanium Shines for FSP’s Flagship Power Supply
4. Our internal benchmarking test suite is designed to produce a set of benchmark data which can be used to compare between hardware configurations to assess relative performance for comparative analysis. We use Coremark for testing list processing, matrix manipulation, state machine, and CRC. We use TPC-C to simulate a transactional database load at scale. We use Geekbench to evaluate crypto, integer, and floating point workloads on CPUs. We use fio for storage testing, to evaluate the performance of both storage media and the storage resiliency technology that DigitalOcean operates.
5. Geek Bench Testing - This is a storage benchmark we run for comparative analysis between hardware configurations (Read iops: host fio, random reads, 4K block size, iodepth 256 / Write iops: host fio, random writes, 4K block size, iodepth 256)