Memblaze PBlaze7 7940 Specifications

Memblaze PBlaze7 7940 Gen5 SSD Performance

Testbed

To test the Memblaze PBlaze7 7940 Gen5 SSD, we leveraged the Dell PowerEdge R760 in our test lab. The R760 is a highly versatile 2U rackmount server that supports two 4th generation Intel Xeon processors with configurations that support up to 24 NVMe drives. This general-use server is intended for mixed workloads, databases, and VDI.

Dell PowerEdge R760 Configuration

• Dual Intel Xeon Gold 6430 (32 cores/64 threads, 1.9GHz base)
• 1TB DDR5 RAM
• Ubuntu 22.04

For ultimate flexibility, we also worked with Serial Cables, who supplied us with an 8-bay PCIe Gen5 JBOF for U.2/U.3, M.2, and EDSFF SSD testing. This allows us to test all current and emerging drive types on the same test hardware.

The Memblaze PBlaze7 7940 marks our second Gen5 SSD with this test platform, following our initial go with the KIOXIA CM7 (results of which we will add to the charts below). Our performance comparison also includes the Samsung PM1743 Gen5 SSD. Notably, the 7940 and PM1743 each have a 1 DWPD endurance rating, in contrast to the CM7-V’s 3 DWPD rating.

VDBench Workload Analysis

When benchmarking storage devices, application testing is best, and synthetic testing is second. While not a perfect representation of actual workloads, synthetic tests help baseline storage devices with a repeatability factor that makes it easy to compare apples-to-apples between competing solutions. These workloads offer a range of testing profiles ranging from “four corners” tests and common database transfer size tests to trace captures from different VDI environments.

These tests leverage the common vdBench workload generator, with a scripting engine to automate and capture results over a large compute testing cluster. This allows us to repeat the same workloads across various storage devices, including flash arrays and individual storage devices. Our testing process for these benchmarks fills the entire drive surface with data and then partitions a drive section equal to 25% of the drive capacity to simulate how the drive might respond to application workloads. This differs from full entropy tests, which use 100 percent of the drive and take them into a steady state. As a result, these figures will reflect higher-sustained write speeds.

Profiles:

• 4K Random Read: 100% Read, 128 threads, 0-120% iorate
• 4K Random Write: 100% Write, 128 threads, 0-120% iorate
• 64K Random Read: 100% Read, 128 threads, 0-120% iorate
• 64K Random Write: 100% Write, 128 threads, 0-120% iorate
• 16K Sequential Read: 100% Read, 32 threads, 0-120% iorate
• 16K Sequential Write: 100% Write, 16 threads, 0-120% iorate
• 64K Sequential Read: 100% Read, 32 threads, 0-120% iorate
• 64K Sequential Write: 100% Write, 16 threads, 0-120% iorate
• 4K, 8K, and 16K 70R/30W Random Mix, 64 threads, 0-120% iorate
• Synthetic Database: SQL and Oracle
• VDI Full Clone and Linked Clone Traces

In our first VDBench Workload Analysis, random 4K read, the PBlaze7 7940 recorded a peak of just 1.1 million IOPS (at 183.2 µs), while the KIOXIA CM7 more than doubled this number with 2.7 million IOPS at a latency of 188.6µs.

For 4K random write, the PBlaze7 7940 delivered much better results with 831K IOPS at 611.7µs, while the KIOXIA CM7 peaked at 948K IOPS with a latency of 537.4µs.

Switching over to 64k sequential workloads, the PBlaze7 7940 exhibited impressive performance. In reads, it peaked at 216K IOPS (13.5GB/s) with 295.4µs in latency. This was noticeably better than the KIOXIA drive, which achieved a peak of 11.4GB/s (182K IOPS) with a latency of 349.5µs in read activity.

In sequential writes, the PBlaze7 7940 edged out the KIOXIA drive, peaking at 4.28GB/s (69K IOPS) at a latency of 928.8µs. The CM7 drive peaked at 4.18GB/s write (67K IOPS) at 951.7µs in latency.

Next up are the 16K tests. In sequential reads, the PBlaze7 7940 demonstrated a peak of just 161K IOPS (2.5GB/s) at 191.3µs, while the KIOXIA CM7 achieved an impressive 352K IOPS (5.5GB/s) at 90.7µs.

In sequential write 16K, the PBlaze7 7940 showed a peak of 193K IOPS (3.01GB/s) at 80.9µs while the KIOXIA SSD reached 242K IOPS (3.79GB/s) at just 63.9µs.

Now for our mixed read/write profiles, starting with 70/30 4K. The PBlaze7 7940’s peak performance here hit 562K IOPS at 110µs. The KIOXIA CM7 drive expectedly fared better at 881K IOPS at 71.6µs.

In 8k 70/30 test results, the PBlaze7 7940 peaked at 443K IOPS at 143.3µs compared to the KIOXIA CM7’s 597K IOPS at 106.1µs.

In the 70/30 16K profile, the PBlaze7 7940 reached 327K IOPS at 195.6µs, while the KIOXIA CM7 completed the test at 350K IOPS at 181.4µs.

The following two tests are 64k random benchmarks. In reads, the PBlaze7 7940 posted a very impressive and steady 188K IOPS at 170.1µs, which was more than double the KIOXIA drive’s 81K IOPS at 219.2µs (which also took a pretty steep spike in performance at the end of the test).

In 64k random writes, the Memblaze and KIOXIA drives were neck and neck, though the 7940 pulled away at the end with 64K IOPS at 245.1µs in latency. In comparison, the KIOXIA CM7 drive peaked at 63K IOPS with a latency of 247.6µs.

Our next set of tests is our SQL workloads: SQL, SQL 90-10, and SQL 80-20. Starting with the standard SQL workload, the PBlaze7 7940 posted a peak of 429K IOPS before taking a spike in performance at the end, while the KIOXIA CM7 peaked at 382K IOPS with a latency of 83.5µs.

In SQL 90-10, both Gen5 drives had nearly identical performance. Here, the PBlaze7 7940 hit 392K IOPS at the end of the test with a latency of 81.2µs, while the KIOXIA peaked at 397K with a latency of 80.2µs.

With SL 80-20, the PBlaze7 peaked at 366K IOPS with a latency of 87µs. In comparison, the KIOXIA CM7 posted a peak performance of 298K IOPS at 73.5µs.

Next are the Oracle workloads (Oracle, Oracle 90-10, and Oracle 80-20. Starting with the general Oracle workload, the PBlaze7 7940 had a peak performance of 361K IOPS (98µs), while the KIOXIA CM7 peaked at 416K IOPS at 85.1µs.

Looking at Oracle 90-10, the PBlaze7 7940 and KIOXIA CM7 had near identical performance again, with the former peaking at 293K IOPS (74.9µs) while the latter posted a peak performance of 298K IOPS (73.5µs).

Next is Oracle 80-20, where the PBlaze7 7940 peaked at 280K IOPS with a latency of 78.2µs. In comparison, the CM7 peaked at 310K IOPS with a steady latency of 70.6µs.

Next, we switched to our VDI clone test, Full and Linked. VDI Full Clone (FC) Boot 7940 peaked at 360K IOPS with a latency of 97.3µs (though it had a minor blip in performance at the end). In comparison, the KIOXIA CM7 peaked at 335K IOPS with a latency of 103µs.

During VDI FC Initial Login, the PBlaze7 7940 achieved 209K IOPS (141.7µs), while the KIOXIA CM7 drive fell back a little, peaking at 168K IOPS with a latency of 73.2µs before it took a significant spike in performance at the very end of the test.

With VDI FC Monday Login, the PBlaze7 7940’s performance hit 130K IOPS at 121.3µs in latency. In comparison, the KIOXIA CM7 posted an impressive 161K IOPS at a latency of 98.1µs.

For the VDI Linked Clone (LC) Boot, the PBlaze7 7940 ended the test at a very steady 166K IOPS at 96µs in latency, while the CM7 KIOXIA drive achieved pretty much identical results with 166K IOPS and 96.3µs.

In VDI LC Initial Login, the PBlaze7 7940’s performance peaked at 75K IOPS at 104.3µs in latency. In comparison, the KIOXIA CM7 had a stable peak of 96K IOPS at 81.1µs.

For VDI LC Monday Login, the PBlaze7 7940’s performance peaked at 108K IOPS at 146.5µs in latency. In comparison, the CM7 peaked at 106K IOPS with a latency of 119µs (though showing a spike in performance at the very end, which took it to 100K IOPS).

Conclusion