FlumeIO F5900 7.68TB E1.S Enterprise SSD Review - Powerful and Compact

TweakTown's Rating: 98%

The Bottom Line

FlumeIO's F5900 7.68TB E1.S enterprise SSD is compact, powerful, and ideally suited for AI/ML.

Pros

• Up to 2.7 million IOPS
• E1.S form factor
• 8K random reads
• Low power consumption

Cons

• None

Should you buy it?

AvoidConsiderShortlistBuy

Check Amazon for prices

Check Newegg for prices

Introduction and Drive Details

FlumeIO is a US-based storage company you may not have heard of before. The fledgling company is a client/partner with Memblaze, but an entirely separate entity. Memblaze is a name that we are all familiar with in the enterprise storage space. By partnering with Memblaze, FlumeIO has access to some of the best enterprise storage technology on the planet, giving the upstart company a leg up on much of its competition.

The FlumeIO F5900/F5901 series NVMe SSDs leverage the state-of-the-art PCIe 5.0 interface with NVMe 2.0 support, offering up to 2.5 times the performance and up to 1.5 times the power efficiency per IO compared with conventional PCIe 4.0 SSDs. With random performance of up to 2.8 million IOPS and sequential bandwidth of up to 14,000 MB/s, FlumeIO's F590x Series is ideal for high-performance computing environments.

From what we can gather, the F5900 series is equivalent in terms of performance to the powerful Memblaze P7940 Series we reviewed back in August of last year. This makes sense because the Memblaze P7940 Series and the F5900 Series share similar hardware configurations. As we see it, and as we will demonstrate here today, what really stands out as impressive about both PCIe Gen5 SSDs is their unrivaled write performance.

Faster, more efficient, more secure storage is essential for the progression of Artificial Intelligence, and that's exactly what FlumeIO's line of PCIe Gen5 SSDs are designed to deliver. For enhanced security, the FlumeIO F5900 Series SSDs offer TCG Opal 2.0 and AES 256 data encryption. Power failure protection and full data path protection ensure data integrity. These SSDs deliver up to 970MB/s sequential read performance per watt, providing superior energy efficiency. Operating temperature ranges from 0C to 77C, suitable for various environments.

FlumeIO states that its F5900/F5901 series SSDs are ideal for a wide range of applications, including database management, cloud DC computing, high-performance computing, AI/ML, Big Data, ERP systems, high-frequency trading, online payment systems, CDN, and edge computing. We agree with that assessment and will note that the E1.S formfactor is especially well suited for AI/ML applications, so much so that NVIDIA is now recommending the E1.S formfactor over others.

Okay, now that we are familiarized with the FlumeIO F5900, let's see exactly what the 7.68TB E1.S variant can do for you by the numbers.

Specs/Comparison Products

29

FlumeIO F5900 7.68TB NVMe PCIe Gen5 x4 E1.S SSD

29

29

The FlumeIO F5900/F5901 series is available in multiple form factors, including 2.5-inch U.2, E3.S, E1.S, and HHHL AIC, ensuring compatibility with diverse deployment environments. These SSDs are compatible with major operating systems such as RHEL, SLES, CentOS, Ubuntu, Windows Server, and VMware ESXi. The drive we have in hand is 7.68TB E1.S, Marvell Bravera 16-channel controlled and arrayed with Micron B58R TLC flash.

Test System Specs & Enterprise Testing Methodology

Enterprise SSD Test System

Prior to the AI revolution, datacenter SSDs' normal operating range would typically never exceed QD32. With AI data pipeline storage being directed by GPU, high queue depth data performance has become paramount. Queue depths in the thousands are now commonplace, which is why we've changed our test platform, methodology, and operating system and charted queue depth range from QD256 to QD4096 for random data and up to QD1024 for sequential testing.

Testing Methodology

TweakTown strictly adheres to industry-accepted Enterprise Solid State Storage testing procedures. Each test we perform repeats the same sequence of the following steps:

1. Secure Erase SSD
2. Write the entire capacity of SSD 2x with 128KB sequential write data, seamlessly transition to the next step (sequential testing skips step 3)
3. Precondition SSD by filling the drive twice with 4K or 8K random writes
4. Run test-specific workload with a 30-second ramp up for 5 minutes at each measured Queue Depth, and record results

Best Deals: Memblaze Pblaze 3.84TB SSD

	Today	7 days ago	30 days ago
	Loading...	Loading...	Loading...	Buy
* Prices last scanned on 12/6/2024 at 6:06 pm CST - prices may not be accurate, click links above for the latest price. We may earn an affiliate commission from any sales.

Benchmarks - Sequential

128K Sequential Write/Read

29

We precondition the drive using 100 percent sequential 128K writes at QD256 using 1-thread for 2-drive fills, receiving performance data every second. We plot this data to observe the test subject's descent into steady-state and to verify steady-state is in effect as we seamlessly transition into testing at queue depth. A steady-state is achieved after 1-drive fill. The average steady-state 128K sequential write performance at QD256 is approximately 10,300 MB/s.

29

29

As expected, our test subject is well ahead of the competition, only trailing the U.2 Memblaze P7946 by a couple hundred MB/s. Since the F5900 and P7946 employ the same hardware, we believe the difference between the two here comes down to firmware, where we believe the F5900's firmware is likely tuned for less thermal impact due to its more confined form factor.

29

29

As we see it, our test subject is delivering the best overall performance at queue depths of up to 8. At queue depths above eight, the F5900 is running in third place overall. It is delivering as advertised and then some. Impressive.

Benchmarks - Random

4K Random Write/Read

29

We precondition the drive using 100 percent random 4K writes at QD256 for 2-drive fills, receiving performance data every second. We plot this data to observe the test subject's descent into steady-state and to verify steady-state is in effect as we seamlessly transition into testing at queue depth. A steady-state is achieved after 1-drive fill. The average steady-state write performance at QD256 is approximately 470K IOPS.

29

29

The F5900 7.68TB is advertised at up to 500K 4K random write IOPS. We got close at 485K. Good enough.

29

29

More of the same here. The F5900 7.68TB is advertised at up to 2.8 million 4K random read IOPS. We got very close at 2.74 million IOPS. Overall, our test subject is delivering the third best performance curve of the bunch, tracking virtually identical to its cousin the Memblaze P7946. This becomes even more impressive when we consider our test subjects more constrained E1.S form factor.

4K 7030

29

29

We find it especially impressive that our 1-DWPD test subject can hang in there with the 3-DWPD P7946 and best the 3-DWPD CM7-V on the top end. Compare this with the H5100 7.68TB 1-DWPD Gen5 SSD to gain further perspective.

4K 5050

29

29

Introducing more writes into the mix causes further separation between the F5900 and the P7964 which is to be expected. However, we are again surprised at how well our 1-DWPD test subject can deliver in a heavy write environment. It again trounces the H5100 and at the top end is somehow able to keep up with the 3-DWPD CM7-V in this write heavy test environment. Impressive.

8K Random Write/Read

29

We precondition the drive using 100 percent random 8K writes at QD256 for 2-drive fills, receiving performance data every second. We plot this data to observe the test subject's descent into steady-state and to verify steady-state is in effect as we seamlessly transition into testing at queue depth. A steady-state is achieved after 1-drive fill. The average steady-state write performance at QD256 is approximately 260K IOPS.

29

29

We expect 8K random to track pretty much the same as 4K random here, just at a lower IOPS rate because it's moving twice the amount of data. The overall performance here is exceptional, coming in better than factory spec and better than 50% of its 4K peak random write performance.

29

29

This is where the F5900/P7940 series have no equal. Both SSDs track identically with one performance curve covering the other, that is why we see no black line for the P7946. Even the mighty D7-PS1030 cannot keep up. Outstanding.

8K 7030

29

29

8K 7030 is representative of a common database workload, and here we find that unlike what we saw at 4K 7030, the 3-DWPD (Drive Write Per Day) P7946 has a much more substantial advantage over our similarly configured 1-DWPD F5900 test subject. However, just as we saw at 4K 7030, the F5900 7.68TB easily outperforms the 3-DWPD CM7-V and is again delivering almost double the performance of the H5100 7.68TB. Impressive.

8K 5050

29

29

How our 1-DWPD test subject can best the 3-DWPD CM7-V at queue depths of 64 or higher here, we don't know. This is yet again another jaw-dropping demonstration of how well this hardware/firmware combination with 1-DWPD overprovisioning can slice right through a write-intensive workload. We've never seen anything like this from any other 1-DWPD SSD.

Final Thoughts

The FlumeIO F5900 clearly demonstrates that E1.S SSDs, despite their more restrictive form factor, can indeed deliver just as well as U.2 and E3 form factors. The F5900 7.68TB really surprised us with its random write capabilities. This is the first time we've experienced a 1-DWPD class SSD able to keep up with similarly configured 3-DWPD SSDs in 4K mixed workload scenarios. Hey, this drive can even beat most of its 3-DWPD direct competitors when digesting mixed workloads. This is unprecedented in our experience.

Additionally, when it comes to a pure 8K random read workload, the F5900 delivers far more than any other flash-based SSD we've encountered.

29

It does some things we didn't think possible from a 1-DWPD SSD, really separating itself from its competition and doing so from a more restrictive form factor. Impressive. Editor's Choice.