Preface
The FusionServer 2488H V5 is Huawei’s latest 2U 4-socket rack server. It offers an ideal choice for compute-intensive scenarios such as virtualization, high-performance computing (HPC), databases, and SAP HANA. One FusionServer 2488H V5 brings about 32% lower OPEX than two traditional 2U 2-socket rack servers in virtualization scenarios. This article interprets the detailed evaluation report on this server on ZOL.COM.CN (authoritative third-party evaluation media).
Huawei’s FusionServer 2488H V5
Due to the rapid development of cloud computing, big data, and Internet of Things (IoT), requirements in the server market continues to grow. In the mission-critical service field, x86 servers are starting to take a place. Thanks to their excellent RAS features, x86 servers continue to grow in industries such as finance and healthcare. The tested 2488H V5 supports multiple RAS features. In addition to common RAS features such as memory mirroring, memory backup, and automatic link recovery, the server supports advanced RAS features based on next-generation Intel® Xeon® Scalable processors, including refined isolation of memory faults and automatic recovery of CPU errors.
To ensure basic performance consistency with other peer products tested, the FusionServer 2488H V5 also uses two Intel® Gold series processors and adopts the same system environment to test software such as Iometer, CPU-z, CineBench 15, and SiSoftware Sandra.
Notably, the onboard iBMC management module of the 2488H V5 supports management features such as IPMI, SOL, KVM Over IP, and virtual media. The iBMC is a remote server management system that monitors and detects server hardware and driver health status. It is involved in many application scenarios during the entire installation and test process.
Appearance Design
The server is silver in exterior with a handle on the chassis cover for securing the chassis cover. The server provides multiple sockets on the front and rear panels to meet USB, management, display, and network requirements. The product appearance design perfectly combines aesthetics, practicality, and safety. Every detail is refined to build a solid and trustworthy impression. Compared with the previous generation, the server pays more attention to customer experience and builds a brand new appearance.
FusionServer 2488H V5 chassis cover
Looking from exterior, the 2488H V5 provides USB ports both on the front and at the rear, and provides a VGA port, iBMC management module, and hot-swappable power supply units at the rear. The front panel of the server adopts the cellular porosity technology, with a porosity rate of up to 75%, improving air intake efficiency, product energy efficiency, and heat dissipation capability. The front panel is configured with hot-swap drive slots, and the cellular panel in the middle can be replaced with drive slots.
FusionServer 2488H V5 front panel cellular porosity technology
On the front panel of the 2488H V5 server, two front USB ports are provided on the left, and the fault diagnosis LED, health status indicator, UID button, and power switch are provided on the right.
FusionServer 2488H V5 rear panel
On the rear of the chassis, there are two USB 3.0 ports, two power modules, one serial port, one VGA port, two GE network ports, one management network port, and PCIe slots. Specifically, with the iBMC, you can use a computer (or server and display) to connect to the 2488H V5 and log in to the corresponding IP address to implement RAID configuration and management for the server.
Interior Structure
Compared with the previous-generation processor, the computing performance of a single processor on the 2488H V5 is up to 65% higher. In terms of the internal structure, the 2488H V5 with a high-density design incorporates two layers of mainboards in the 2U space. The upper mainboard can be completely removed so that you can maintain the hardware below it. Compared with two traditional 2U 2-socket rack servers, one 2488H V5 can drive down the OPEX by 32% (calculated based on three years of equipment room O&M cost and electricity fee).
Top view of the internal design
Details about the internal fan modules of the server
As for heat dissipation, the 2488H V5 adopts a cellular porosity design for its front panel. Fan modules are also critical for heat dissipation. The 2488H V5 is equipped with four hot-swappable fan modules and allows single fan failures.
Memory module and processors (you can lift the blue handle to remove the entire daughter board)
Power supply module
PCIe slots
The server provides 11 PCIe slots, including riser slots, to support 48 DDR4 DIMMs. A maximum of 25 x 2.5-inch hard drives are supported. Leverages patented Dynamic Energy Management Technology (DEMT), and multiple power-saving measures such as component hibernation, proportional-integral-derivative (PID) algorithm based fan speed tuning, and active-standby power supplies, improving overall energy efficiency by up to 16% without compromising workload performance.
Specifications
This part unveils the secrecy of the powerful strength of the server from the product configuration and specifications perspectives.
| Form Factor | 2U rack server |
| Processors | Intel® Xeon® Gold 6140 CPU 2.30 GHz |
| Number of processors | 2 or 4 |
| L2 cache | 18 x 1024 KB |
| L3 cache | 25344 KB |
| Chipset | Intel® C622 |
| Memory/Max | 64 GB /3 TB |
| Memory | 48 DIMM slots |
| Expansion slots | 11 PCIe 3.0 slots (including half-height and full-height slots) |
| Network adapter | 2 x 10GE +2 x GE SFP+/BaseT |
| Power supply | 1+1 redundancy; Titanium 750, Platinum 495, Platinum 750, etc. |
| RAID | RAID 0, 1, 5, 6, 10, 50, or 60 |
| Management | Built-in iBMC management module, which supports management features such as IPMI, SOL, KVM Over IP, and virtual media |
| Warranty | Three-year onsite warranty |
The 2488H V5 uses the new-generation Intel® Xeon® Scalable Gold 6140 processors with 18 cores and 36 threads per processor.
| Processors | Name | Intel Processor | |
| Specification | Intel® Xeon® Gold 6140 [email protected] GHz | ||
| Instruction set | MMX,SSE,SSE2,SSE3,SSSE3,SSE4.1,SSE4.2,EM64T,VT-x,AES,AVX,AVX2,FMA3,TSX | ||
| Clock | Core speed | 2298.64 MHz | |
| Frequency | x 92.0 | ||
| Bus speed | 24.99 MHz | ||
| Cache | L1 data | 18 x 32 KBytes 8-way | |
| L1 instruction | 18 x 32 KBytes 8-way | ||
| L2 | 18 x 1024 KBytes 16-way | ||
| L3 | 25344 KBytes 11-way | ||
| Cores | 18 | Threads | 36 |
CPU-Z processor configuration test information
| L1 Data Cache (Cores) | Size | 32 KBytes x 18 |
| Description | 8-way set associative, 64-byte line size | |
| L1 Instruction Cache (Cores) | Size | 32 KBytes x 18 |
| Description | 8-way set associative, 64-byte line size | |
| L2 Cache (Cores) | Size | 1024 KBytes x 18 |
| Description | 16-way set associative, 64-byte line size | |
| L3 Cache (Cores) | Size | 25344 KBytes |
| Description | 11-way set associative, 64-byte line size |
CPU-Z cache configuration test information
Intel® Xeon® Gold Processor Performance Test
To better understand the performance of Intel® Xeon® Gold processors on the server, we have tested the processors using the CineBench 15 and SiSoftware Sandra software.
CineBench R15 test result
The most significant change in the new CineBench R15 version is that it supports only 64-bit operating systems. Compared with the R11.5 version that supports a maximum of 16 cores, the R15 version supports a maximum of 256 logical cores. The new version also enhances the test of the colorizer, anti-aliasing, shadow, lighting, and reflection glossiness, delivering more accurate CPU performance test results.
CineBench R15 also optimizes final scores for the new-architecture processor with multiple cores, making test results more accurate. The optimized test tool enables Intel® Xeon® Gold processors to fully release performance and makes the test results more close to those in actual application environments. Compared with the previous-generation processors, the new-generation processors have obvious advantages in performance. Thanks to the new-generation processors, the FusionServer 2488H V5 scores 8152 cb in CPU value and 61.14x higher MP ratio value in the CineBench R15 test.
The 2488H V5 server delivers much higher performance, thanks to the new architecture and technologies of the new-generation processors. The following are the test results of the new-generation Intel® Scalable processors in the SiSoftware Sandra test.
| SiSoftware Sandra CPU tests |
Arithmetic performance test (the higher the better) | Intel® Xeon® Gold 6140 CPU (2.30 GHz, 18 Cores) | |
| Aggregate Arithmetic Performance | 21 GOPS | ||
| Dhrystone iSSE4.2 | 36.51 GIPS | ||
| Whetstone iSSE3 | 12 GFLOPS | ||
| Multimedia performance test (the higher the better) | Aggregate Multi-Media Performance | 43.15 MPixel/s | |
| Multi-Media Int x32 SSE4.1 | 43.78 MPixel/s | ||
| Multi-Media Float x16 SSE2 | 42.54 MPixel/s | ||
| Multi-Media Double x8 SSE2 | 23.66 MPixel/s | ||
| Encryption and decryption performance test | Cryptographic Bandwidth | 1.07 GB/s | |
| AES256-ECB AES Cryptographic Bandwidth | 1.97 MB/s | ||
| SHA256 AVX Hashing Bandwidth | 0.12 MB/s | ||
The latest SiSoftware Sandra Lite 2011 is used to test the arithmetic performance, multimedia performance, and encryption and decryption performance of the 2488H V5. The test result shows that the arithmetic capability and encryption and decryption capability of the 2488H V5 are excellent, and the multimedia performance is especially outstanding.
Memory Bandwidth and Latency Test
The mainboard of the 2488H V5 integrates 48 DIMM slots. The tested server incorporates four memory controllers, which greatly improve the memory bandwidth.
SiSoftware Sandra Lite 2011 provides the memory bandwidth, memory latency, and cache performance tests. We use this tool to evaluate the memory performance of 2488H V5.
| SiSoftware
Sandra Lite 2011 memory test |
Memory bandwidth test (the higher the better) | Aggregate Memory Performance | 7.33 GB/s | |
| Int Buff’d AVX 128 Memory Bandwidth | 7.34 GB/s | |||
| Float Buff’d 128 AVX Memory Bandwidth | 7.33 GB/s | |||
| Memory latency test (random) | Memory(Random Access) Latency (the lower the better) | 76.8 ns | ||
| Speed Factor (the lower the better) | 57.40 | |||
| Internal Data Cache | 3 clocks | |||
| L2 On-board Cache | 11 clocks | |||
| L3 On-board Cache | 62 clocks | |||
| Memory latency test (linear) | Memory(Linear Access) Latency (the lower the better) | 6.2 ns | ||
| Speed Factor (the lower the better) | 4.6 | |||
| Internal Data Cache | 3 clocks | |||
| L2 On-board Cache | 9 clocks | |||
| L3 On-board Cache | 11 clocks | |||
| Cache and memory test | Cache/Memory Bandwidth (the higher the better) | 50.26 GB/s | ||
| Speed Factor (the lower the better) | 22.70 | |||
The 2488H V5 provides 48 DDR4 DIMM slots. Each processor integrates four memory channels. The tested 2488H V5 server is configured with eight DIMMs and demonstrates excellent memory latency, cache capability, and memory bandwidth.
Drive System I/O Performance Test
The 2488H V5 is configured with four Huawei NVMe SSDs. We use Iometer 1.1.0 to test the maximum read/write performance of the drive system.
To test the maximum I/O performance of the server, we use open source tools to test drive performance. We use Iometer to test the I/O transmission rate and average I/O response time. Iometer contains a workload generator (called Worker in Iometer) that provides a series of standard test components to simulate applications, such as WebServer, FileServer, and OLTP servers. (Note: The test uses a single thread to test one worker.)
Max I/O (read): The test executes 100% read operations on 4 KB data blocks with a randomness rate of 0%. That is, the test executes 100% continuous read operations with a queue depth of 256 to evaluate the maximum read capability of the drive subsystem.
Max I/O (write): The test executes 0% read operations on 4 KB data blocks with a randomness rate of 0%. That is, the test executes 100% continuous write operations with a queue depth of 256 to evaluate the maximum write capability of the drive subsystem.
Iometer read performance test result
Iometer write performance test result
According to the test results, the read and write speeds of the drive subsystem are extraordinarily high and stable. The maximum read speed reaches 20509.498 IOPS, and the maximum write speed reaches 124177.27 IOPS. The read and write I/O speeds are relatively balanced.
Summary
Powered on Intel® Xeon® Scalable Gold processors, the 2488H V5 delivers much higher performance. The 4-socket server provides 48 DIMM slots and a maximum of 3 TB memory capacity. What’s more, the tested 2488H V5 adopts a double-mainboard design and achieves breakthroughs in computing and storage performance. According to the test results, the processors used by the 2488H V5 enable much higher performance, compared with the previous-generation processors. With cutting-edge architecture and manufacturing process, the 4-socket rack server 2488H V5 proves powerful memory and drive expansion capabilities and network I/O capabilities (note that 300 GB Hitachi HDDs are used in the test). Though the 2488H V5 tested is not fully configured with DIMMs, it still demonstrates outstanding performance in the test, perfectly proving the excellence of the server itself.
(Authorized by Lu Chang from ZOL.COM.CN)
The post Test Report: Huawei FusionServer 2488H V5 Rack Server Review appeared first on Huawei Enterprise Blog.
Source: Huawei Enterprise Blog
—