The Tool Desk
Outbyte Driver Updater FREEScan for outdated or missing drivers - takes under a minuteDriver Scan →Outbyte PC Repair FREEClear out junk files and repair common Windows errorsFree Scan →iTechGuides is reader-supported. When you buy through links on our site, we may earn an affiliate commission. As an Amazon Associate I earn from qualifying purchases. Learn more
Netflix’s 2021 engineering work approached 400 Gb/s of encrypted video delivery from one server, but the reported result was about 380 Gb/s—not a universal 400-Gb/s appliance specification. Drew Gallatin’s EuroBSDCon presentation measured roughly 190 Gb/s per ConnectX-6 Dx adapter with production firmware and TLS_OPTIMIZE. The result depended on the whole system: memory and NUMA behavior, PCIe settings, NIC firmware, storage and network placement, and how retransmits were handled.
What Netflix meant by “400 Gb/s per server”
The “400 Gb/s” figure was a rounded engineering target in Drew Gallatin’s 2021 EuroBSDCon talk, “Serving Netflix Video at 400Gb/s on FreeBSD.” The presentation described a prior level of about 200 Gb/s per server and investigated how to approach twice that throughput. Its reported result was approximately 190 Gb/s per adapter, or about 380 Gb/s across two adapters—not a demonstrated, general-purpose 400-Gb/s service level for every Netflix server.
The talk was about serving TLS-encrypted video efficiently, not just installing a particular processor. Its central lesson was that the bottleneck moved as the design changed: host-side encryption strained memory bandwidth; NIC-side encryption reduced that demand but exposed constraints in adapter memory, firmware, and PCIe behavior. Gallatin’s talk appears in the FreeBSD conference record, which describes his work optimizing the FreeBSD kernel and network stack for Netflix content delivery.
Quick wins for a faster PC:
Repair Windows errors before they cause bigger problemsFix Now →Scan for outdated or missing drivers - takes under a minuteDriver Scan →The test server and serving path
The test platform in the presentation combined an AMD EPYC 7502P “Rome” processor with substantial memory, storage, and networking hardware. It ran FreeBSD-current and NGINX; the initial serving path used sendfile(2) and software kernel TLS (kTLS).
#1 Best Overall
- For AMD EPYC 9754 128 Core Bergamo 2.25GHz (100-000001234) EPYC 9004 Series Socket SP5 ZEN4 256MB L3 Bulk / Tray Pack (Unlocked) Server Processor
| Component | Presentation configuration |
|---|---|
| CPU | AMD EPYC 7502P, 32 cores at 2.5 GHz |
| Memory | 256 GB DDR4-3200 across eight channels; estimated 150 GB/s of memory bandwidth |
| Networking | Two Mellanox ConnectX-6 Dx adapters, each with two full-speed 100 GbE ports |
| Storage | 18 Western Digital SN720 2 TB NVMe drives |
| Software | FreeBSD-current and NGINX; sendfile(2) in the serving path |
The system also had 128 PCIe Gen4 lanes. The configuration matters: two adapters offered four 100 GbE ports in aggregate, while the CPU, memory, PCIe fabric, storage, and software all had to keep data moving fast enough to use them. The result is evidence about this tested setup, not a CPU-only benchmark or a direct comparison of current server products.
Why software TLS stopped at 240 Gb/s
With software kTLS, the host CPU encrypts TLS records before sending them. Gallatin’s presentation reported 240 Gb/s for the initial software-kTLS setup and identified memory bandwidth as the limit. This is notable because encryption work does not consume only CPU cycles: data must also move through memory as it is read, encrypted, and sent.
Rank #2
- Dual Processor Support: Supports and includes 2 AMD EPYC processors installed for enhanced computing performance
- Processor Configuration: Features 2 installed AMD EPYC processors for powerful server operations
- AMD Processor Technology: Equipped with AMD processor manufacturer components for reliable performance
- EPYC Processor Type: Utilizes AMD EPYC processor type designed for enterprise-level server applications
- 5th Generation Processing: Powered by 5th Gen AMD EPYC 9115 processors running at 2.60 GHz with hexadeca-core architecture
The talk’s simplified data-flow model equated 400 Gb/s of payload throughput to 50 GB/s and estimated that the software-encryption route would require about 200 GB/s of memory bandwidth. The test machine’s estimated memory bandwidth was about 150 GB/s, so the model illustrates why adding network capacity alone would not solve the problem.
Free tools Windows power users keep installed
One-click scans. No signup required.
NUMA locality added another constraint
On a multi-socket or multi-die system, memory and devices are not equally close to every CPU core. Moving bulk data across the system’s NUMA fabric can compete with ordinary memory accesses. Gallatin summarized the effect this way: “Bulk data congests NUMA fabric and leads to CPU stalls when competing with normal memory accesses.” Keeping data near its storage and network path therefore mattered alongside raw memory capacity. The fabric bandwidth and latency figures discussed in the talk apply to that particular machine.
Rank #3
- High Performance Server: Features an AMD EPYC 7313 processor with a speed of 1.44 GHz and 32 GB of DDR4 memory for fast performance.
- Expandable Storage: Includes an P408i-a storage controller and 8 SFF drive bays for flexible storage options.
- Modern Design: Has a sleek, modern style with a black finish and ergonomic keyboard for comfortable use.
- Easy Setup: Comes with an 800W power supply and pre-installed operating system for quick installation.
- Reliable Connectivity: Offers multiple USB and Ethernet ports for seamless connectivity to other devices.
How NIC-based kTLS changed the data path
NIC kTLS moves TLS record encryption from the host CPU to the network adapter as data is transmitted. In the presentation’s model, this halves the memory-bandwidth requirement compared with host-side software encryption. That is a design benefit, not a guarantee that offloading TLS by itself will produce 400 Gb/s: data still has to traverse memory and PCIe, reach the adapter, and be handled correctly by its hardware and firmware.
The first NIC-offload results were held back by adapter memory and firmware behavior. The presentation describes enabling PCIe Relaxed Ordering and later using production firmware as important parts of improving throughput. With production firmware and TLS_OPTIMIZE enabled, it reported roughly 190 Gb/s per ConnectX-6 Dx adapter—about 380 Gb/s combined.
Rank #4
- HPE ProLiant DL145 Gen11 – P87460-005 – SMART CHOICE MODEL – COMPACT EDGE SOLUTION: Preconfigured and factory-tested for fast deployment and cost efficiency. Includes AMD EPYC 8024P (8 cores, 2.40 GHz), 16GB DDR5 ECC SmartMemory, 2 SFF chassis, 480GB SATA 6G Read Intensive SSD, Broadcom 1GbE OCP NIC, and single 700W Platinum PSU—ideal for IoT gateways, retail POS, and light virtualization.
- PERFORMANCE AND MEMORY – EFFICIENT FOR LIGHT WORKLOADS: The AMD EPYC 8024P delivers 8 cores at 2.40 GHz for edge compute tasks. Includes 16GB DDR5 RDIMM ECC (1x16GB) and supports up to 768GB across six DIMM slots—ideal for small-scale virtualization and real-time analytics.
- STORAGE – READY FOR OS AND DATA Includes one HPE 480GB SATA 6G Read Intensive SSD for quick deployment. Supports additional SFF drives for storage flexibility—perfect for edge workloads and local data storage.
- ENTERPRISE DESIGN – POWER AND CONNECTIVITY: Single 700W Platinum hot-plug power supply ensures reliable power delivery. Broadcom BCM5719 OCP NIC offers four 1GbE ports for edge networking and connectivity.
- SECURITY AND MANAGEMENT – BUILT-IN PROTECTION: HPE iLO6 with Intelligent Provisioning, TPM 2.0, Silicon Root of Trust, and secure boot protect against threats. Compatible with HPE OneView and Compute Ops Management for simplified lifecycle management.
Retransmits reduced the stable throughput
Handling retransmitted data was an operational trade-off. In the configuration described in the talk, moving roughly one-third of connections to software TLS for retransmit handling reduced stable throughput from about 380 Gb/s to about 350 Gb/s. This shows why a peak result and a stable result under real delivery conditions are not interchangeable.
Recommended Free Tools
Gallatin also said further quality-of-experience testing was needed before production use. The presentation therefore documents an engineering result and a remaining validation step; it does not establish that the tested configuration was already deployed as a production-wide 400-Gb/s service.
Best Value
- The processor features Socket AM5 socket for installation on the PCB
- EPYC product line processor for better usability and increased efficiency
- Dodeca-core (12 Core) processor core allows multitasking with great reliability and fast processing speed
- 64 MB of L3 cache memory provides excellent hit rate in short access time enabling improved system performance
- Processor with 3.40 GHz clock speed for reliable and fast execution of instructions to ensure maximum convenience and feasibility
What the platform comparisons do—and do not—show
The presentation also explored an Ampere Altra Q80-30 and an Intel Xeon 8352V configuration. Those results are useful as clues about system bottlenecks, but not as a general ranking of CPU vendors: the tested configurations and enabled features differed, and the Intel NIC-offload result was unavailable because its BIOS blocked PCIe Relaxed Ordering.
| Platform and path | Reported result | Important qualification |
|---|---|---|
| Ampere Altra Q80-30, software kTLS | 180 Gb/s | Result from the configuration in Gallatin’s 2021 presentation. |
| Ampere Altra Q80-30, NIC kTLS | 240 Gb/s, improving to 320 Gb/s | The increase followed enabling PCIe extended tags; it is specific to the tested setup. |
| Intel Xeon 8352V, software kTLS | 230 Gb/s | The presentation identified memory bandwidth as the limit. |
| Intel Xeon 8352V, NIC kTLS | Not available | The BIOS blocked PCIe Relaxed Ordering in the configuration discussed. |
A fair comparison would need to account for memory bandwidth under the same workload, PCIe bandwidth and settings, NUMA placement, TLS implementation, NIC firmware, retransmit behavior, and storage/network topology—not just processor brand or core count.
How the 2021 test relates to Netflix’s Open Connect network
Netflix describes Open Connect as a geographically distributed content delivery network that uses appliances and ISP connections to serve video closer to members. Its current Open Connect appliance page lists AMD among processor vendors and gives high-level operational throughput figures of about 200 Gbps for a storage appliance and about 80 Gbps for a global appliance. Those appliance-class figures do not identify an EPYC model and are not the same thing as Gallatin’s 2021 test result.
Netflix’s 2016 account of working with ISPs said Open Connect carried 100% of Netflix video traffic at that time. It also reported historical single-server throughput of 8 Gbps in 2012 and more than 90 Gbps in 2016. These are dated company-published figures, not current measurements; they provide context for the growth in delivery capacity that motivated later engineering work.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
