Samsung 990 EVO SSD Specifications

Samsung 990 EVO SSD Performance

For this review, we will be looking at the 2 TB model while comparing it to the following SSDs:

• Samsung 990 PRO
• Crucial T500
• WD Blue SN580

For testing, we use two platforms. The consumer test platform supports PCIe Gen4/Gen5 SSDs. It is sometimes leveraged for lighter consumer-based tests such as BlackMagic DiskSpeed Test, CrystalDiskMark, and our main platform, a Dell PowerEdge R760, which overlaps our enterprise tests. Serial Cables supplied an 8-bay PCIe Gen5 JBOF for U.2/U.3, M.2, and E1.S/E3.S drive testing. This allows us to test all current and emerging drive types on the same test hardware.

Dell PowerEdge R760 Configuration

• Dual Intel Xeon Gold 6430 (32 cores/64 threads, 1.9GHz base)

• 1TB DDR5 RAM

• Ubuntu 22.04

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 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, then partitions a drive section equal to one percent 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 to 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, 64 threads, 0-120% iorate
• 64K Sequential Read: 100% Read, 16 threads, 0-120% iorate
• 64K Sequential Write: 100% Write, 8 threads, 0-120% iorate
• VDI Profiles

Starting with 4K random read, the Samsung 990 EVO SSD drive performed poorly, hitting just over 130 thousand IOPS at 980.1ms. The 990 Pro and T500 were far superior drives here, with the former almost reaching a million IOPS.

4K random writes told a similar story, with the new 990 EVO peaking at 34K IOPS at 3,783ms in latency. Again, the best-performing drive here was the less expensive Samsung 990 Pro.

When switching to the sequential read 64K workload, the 990 EVO took a severe hit in latency, starting at the 25K IOPS mark and eventually hitting 1,131ms at 30K IOPS. The test ended at just 3.22GB/s read at a latency of 548.4ms for last place. Again, the older and less expensive 990 Pro was by far the best drive here.

The 990 EVO again showed very poor latency right out of the gate, hitting a max of 4,754ms. It peaked at 2671ms at 373MB/s write (or 5,874 IOPS) for last among the test SSDs. Again, nothing could come close to the Samsung 990 PRO.

Next, we looked at our VDI benchmarks, designed to tax the drives further. These tests include Boot, Initial Login, and Monday Login. Starting with boot, the Samsung 990 EVO latency increased quickly, ending the test around the 33K IOPS mark at 1,050ms.

Next up is VDI Initial Login. Here, the Samsung 990 EVO demonstrated high latency numbers again and, by the end of the test, hit 2,855.1ms and just 10K IOPS.

Finally, we have the VDI Monday Login benchmark, concluding the test at just 11K IOPS with a latency of 1,450ms.

BootBench

BOOT-BENCH-1 is a workload profile adopted by OCP to profile SSDs designed for server boot duty. While this is intuitively a job for enterprise SSDs, client SSDs are often selected for their performance, capacity, and cost combination. The boot drive issue is germane to hyperscalers and server and storage system providers, as they face similar challenges.

This boot workload executes an intense test plan that fills the entire drive with writes before testing a read-heavy workload sequence. For each test, it performs a 32K random read async operation alongside a 15MiB/s synchronous 128k random write and a 5MiB/s synchronous 128k random write/trim background workload. The script starts with the random-read activity at a 4-job level and scales up to 256 jobs at its peak. The final result is the read-operations performed during its peak run.

The OCP goal for this benchmark is a pass/fail at 60K read IOPS. Most drives we test will far exceed the minimum, but the results are instructive regardless.

The Samsung 990 EVO, unsurprisingly, wasn’t able to complete the test and so received a “DNF” on the current boot leaderboard:

Samsung 990 EVO CrystalDiskMark Speed Test

We performed a lighter-weight test on the Samsung 990 EVO to demonstrate another metric of its speeds. CrystalDiskMark’s higher queue depth (compared to BlackMagic) allows us to showcase the best-case scenario for the drive. Any real-world performance differences would likely be negligible.

Here, the drive recorded sequential transfer speeds exceeding 5.09GB/s read, and 3.75GB/s write, landing on the lower end of the consumer-level SSD spectrum.

Samsung 990 EVO BlackMagic DiskSpeed Test

Lastly, we measured performance inside a Windows 11 environment on our consumer test platform via the popular Blackmagic test. The new Samsung 990 EVO hit just 4.26GB/s read and 3.44GB/s write.

Conclusion

Samsung’s 990 EVO emerges as the latest mainstream SSD offering from Samsung, positioned as an upgrade to the 970 EVO Plus. This drive is designed to balance performance with energy efficiency and interface compatibility. With the usual features like TRIM and SMART support, it aims to offer consumers a viable (albeit basic) SSD experience. However, it falls well short of advancing beyond its predecessor’s capabilities. Moreover, the SSD’s Gen5 classification is also somewhat ambiguous, as it operates on a limited 2 lanes of Gen5 and 4 lanes of Gen4, not fully utilizing the expected speed of Gen5 technology.