The Bottom Line
Pros
- + Up to 2.8 million IOPS
- + Mixed workload champion
- + Efficiency through speed
- + High endurance
Cons
- - None
Should you buy it?
AvoidConsiderShortlistBuyIntroduction and Drive Details
AI is set to change the world as we know it. There is a mad scramble to create the most powerful AI computing solutions, with GPUs leading the way as the main component of this explosive new technology. However, there is another and just as important component that makes up the AI data pipeline - storage. Faster, more efficient storage is essential for the progression of Artificial Intelligence, and that's what Solidigm's newest SSDs are designed to deliver.
Today, Solidigm unveils its first commercially available PCIe Gen5 SSD, the D7-PS1010 & D7-PS1030 line of Enterprise SSD. This new SSD is specifically designed to accelerate AI data pipelines with up to 50% more AI performance than any Solidigm SSD that has come before:
The D7-PS1010 is a mixed and mainstream data specialist designed to accommodate read-intensive 1-drive write per day (DWPD) workloads. The D7-PS1030 is a mixed and write-centric data specialist designed to deliver up to 3 DWPD endurance for 5 years. Both models are available in 15mm U.2 and 7.5mm E3.S form factors.
As most know by now, SK hynix is the parent company of Solidigm, having acquired the entirety of Intel's SSD business and the Dalian NAND flash manufacturing facility in China on January 4, 2022. With the acquisition, SK hynix instantly became a major player in the enterprise storage business. We mention this because until now, all of Solidigm's enterprise SSDs have essentially been rebranded Intel-developed SSDs, all arrayed with Intel-developed flash technology.
Solidigm's newest enterprise offering is different. Not only is it the first PCIe Gen5 SSD in Solidigm's portfolio, but it is also the first of its kind to be arrayed with SK hynix flash. With double the bandwidth of PCIe Gen4, PCIe Gen5 SSDs require faster flash to fully exploit the next-gen interface. We believe that Intel-developed flash is simply not fast enough to deliver PCIe Gen5 bus saturation, but SK hynix 176L TLC behind a 16-channel controller is. That, in our opinion, is why SK hynix flash powers Solidigm's new PCIe Gen5 SSDs.
To date, this will be the fourth PCIe Gen5 Enterprise offering we've tested. All have claimed up to 14,000 MB/s throughput and up to 2.7-2.8 million IOPS at QD256. Two of the three previous have been able to deliver on sequential throughput, but all have fallen well short of delivering the quoted up to 2.7-2.8 million IOPS at QD256, with one unable to get there at all and the other two requiring 16 threads at QD512 to get there while running on our hardware/software configuration.
However, the 6.4 TB D7-PS1030 U.2 SSD we have on the bench today is different than those we've tested previously on the very same hardware/software configuration. Not only is it significantly faster and more powerful than anything we've tested before, but it can also deliver on all its up to factory specs and then some. It does indeed live up to all its claims.
To begin with, we wanted to see how much random read throughput our test subject could deliver in its FOB state at QD512 with 16-CPU threads, and bam, we were rewarded with a mind-bending 3.33 million IOPS:
Now of course FOB is not the same as steady-state, even when it's pure read, but we believe it serves as evidence that our test subject is fully capable of delivering on its quoted up to 2.8 million IOPS at QD512 in a steady-state with our hardware/software configuration. We did this quick experiment because our steady-state testing tops out at QD256 with 8-CPU threads.
Now, let's dive in and see what the current performance leader can do for you by the numbers.
Specs/Comparison Products
Solidigm offers its D7-PS1010 & D7-PS1030 series at capacity points ranging from 1.6TB -15,36TB across two form factors, including 15mm 2.5-inch U.2 and 7.5mm E3.S. The drive we have in hand is 6.4TB U.2, 16-channel controlled and arrayed with SK hynix V7 512Gb TLC flash.
Solidigm D7-PS 1030 6.4TB NVMe PCIe Gen5 x4 U.2 SSD
Enterprise 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 four steps:
- Secure Erase SSD
- Write the entire capacity of SSD a minimum of 2x with 128KB sequential write data, seamlessly transition to the next step
- Precondition SSD at maximum QD measured (QD32 for SATA, QD256 for PCIe) with the test-specific workload for a sufficient amount of time to reach a constant steady-state, seamlessly transition to the next step
- Run test-specific workload for 5 minutes at each measured Queue Depth, and record results
Benchmarks - Random and Sequential
4K Random Write/Read
We precondition the drive for 16,000 seconds, 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. Steady state is achieved at 2,000 seconds of preconditioning. The average steady-state write performance at QD256 is approximately 860K IOPS.
This is by far the best random write performance curve we've attained to date from any flash-based SSD. Additionally, the drive exceeds spec by 70K IOPS at QD256. Impressive.
With technologies such as DirectStorage or BaM, where storage is primarily or exclusively GPU-directed, performance at massive queue depths has become relevant and, in fact, supremely important, and this is exactly where our test subject delivers much more than those that have come before it. We are getting 2.73 million IOPS at QD256, which is a massive 435,000 IOPS more at QD256 than anything we've tested before. Impressive. Additionally, we believe if we can get 2.73 million IOPS at QD256 in a steady state, we should have no problem attaining 2.8 million steady-state IOPS at QD512, where our test subject is spec'd to do so.
8K Random Write/Read
We precondition the drive for 16,000 seconds, 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. Steady state is achieved at 2,000 seconds of preconditioning. The average steady-state write performance at QD256 is approximately 468K IOPS.
We expect 8K random to track pretty much the same as 4K random here, just at a lower rate because it's moving twice the amount of data. The overall performance here is exceptional, coming in better than factory specs and better than 50% of its 4K performance. Again, it is vastly superior to anything in its class that we've tested to date.
Solidigm specs its 6.4TB D7-PS1030 as capable of up to 1,400K IOPS here at QD512. We are hitting a far more impressive 1,605K IOPS at QD256.
128K Sequential Write/Read
We precondition the drive for 6,500 seconds, receiving performance data every second. Steady state for this test kicks in at 0 seconds. Average steady-state sequential write performance at QD256 is approximately 11,390 MB/s.
Our test subject delivers its advertised up to 9,300 MB/s and then some. With our hardware/software configuration, we can extract a whopping 11,396 MB/s steady state 128K sequential write throughput. This is the best 128K sequential write throughput we've ever seen from any enterprise SSD, and 2,100 MB/s higher than its factory spec. Outstanding.
Once again, no issues exceeding the given factory spec of up to 14,500 MB/s at QD128. This is the first time so far that our test subject hasn't beaten everything on our chart at every measured queue depth. Our speedy contender does however win by delivering a new lab record of 14,843 MB/s for single drive sequential throughput at QD32 or higher.
Benchmarks - Workloads
4K 70/30
4K 70/30 is a commonly quoted workload performance metric for Enterprise SSDs.
We precondition the drive for 16,000 seconds, 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. Steady state is achieved at 4,000 seconds of preconditioning. The average steady-state performance at QD256 is approximately 1,600K IOPS.
Again, total domination over anything we've tested to date. We are hitting 1.6 million IOPS at QD256, or more than 100K IOPS over factory spec of 1,500K at QD512. This is 561K IOPS, or 53% more than the closest contender shown on our chart. Incredible.
8K 50/50
Our 8K 50/50 workload is a demanding 8K test with a 50 percent R/W distribution. This application gives a good indication of how well a can perform in a write-heavy workload environment.
We precondition the drive for 16,000 seconds, 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. Steady state is achieved at 5,000 seconds of preconditioning. The average steady-state performance at QD256 is approximately 650K IOPS.
Another across-the-board win for our test subject, culminating in a roughly 150K IOPS lead at QD256. Amazing.
OLTP/Database Server
Our On-Line Transaction Processing (OLTP) / Database workload is a demanding 8K test with a 66/33 percent R/W distribution. OLTP is online processing of financial transactions and high-frequency trading.
We precondition the drive for 16,000 seconds, 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. Steady state is achieved at 4,000 seconds of preconditioning. The average steady-state performance at QD256 is approximately 828K IOPS.
Our test subject is delivering 36% more Database performance than any flash-based SSD we've tested to date. This is performance on another level.
Web Server
Our Web Server workload is a pure random read test with a wide range of file sizes, ranging from 512B to 512KB at varying percentage rates per file size.
We precondition the drive for 16,000 seconds, 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. We precondition for this test with an inverted (all-write) workload, so no relevant information can be gleaned from this preconditioning other than verification of steady-state.
Our test subject manages to stay out front of the 3-DWPD Memblaze P7946 in this supremely taxing all random read test, which is somewhat surprising as the Memblaze drive is arrayed with architecturally superior flash.
Final Thoughts
Throughout our testing, Solidigm's 6.4TB D7-PS1030 demonstrated that it is by far both the fastest and most powerful flash-based SSD we've ever encountered, thereby making it inherently the most efficient as well. Solidigm's first PCIe Gen5 SSD not only outperforms its PCIe Gen4 predecessors in 4-corner testing by at least 50% or more, it also lays the smack-down on both of our PCIe Gen5 3-DWPD comparison SSDs, able to deliver up to 53% more IOPS with a 4K 7030 workload at QD256. That is absolute domination.
Prior to the AI revolution, datacenter SSDs normal operating range would typically never exceed QD32. Now with AI data pipeline storage being directed by GPU, high queue depth performance has become paramount. This is exactly what Solidigm's D7-PS1010 & D7-PS1030 are here to do for you - deliver the most performance where it matters the most, powering your AI infrastructure to new heights.
It's by far the most performant flash-based Enterprise SSD we've encountered. It is also the first PCIe Gen5 enterprise SSD we've tested so far that can deliver as advertised on all fronts and then some. Editor's Choice.