Expert’s Rating
Pros
- Fastest benchmarking NVMe SSD we’ve tested.
- DRAM for fast random operations
- Available in up to 4GB capacities, and soon 8TB
Cons
- Expensive overkill for most users
- A hair off the pace in Windows file transfers
Our Verdict
The fastest Consumer NVMe SSD you can buy — but it ain’t cheap.
Price When Reviewed
This value will show the geolocated pricing text for product undefined
Best Pricing Today
Price When Reviewed
1TB: $200 I 2TB: $300 I 4TB: $550
Best Prices Today: Samsung 9100 Pro
$199.99
There was a time when Samsung was the long-reigning, undisputed performance king of the NVMe SSD market. Recently… Not so much.
Well, the fallen are mighty once more thanks to the company’s 9100 Pro — the fastest NVMe SSD I’ve ever tested according to the benchmark software. Welcome back, Samsung!
Read on to learn more, then see our roundup of the best SSDs for comparison.
What are the 9100 Pro’s features?
The Samsung 9100 is a state-of-the-art, PCIe 5.0 x4 (four-lane) NVMe 2.0 SSD with 1GB of DRAM per terabyte of capacity for primary caching duties. The controller is an in-house Samsung development according to the company, and the NAND is 236-layer(3D/stacked) TLC.
The SSD is available in the 2280 (22mm wide, 80mm long) form factor and currently — 1TB through 4TB capacities. An 8TB version will be available the second half of 2025.
Jon L. Jacobi / Foundry
Most users won’t care that the 9100 requires about half the power of the previous top-of-the-line 990 Pro. But it amounts to longer battery life in laptops, and less heat in any device.
The 9100 Pro is warrantied for five years and rated for 600TBW (terabytes that may be written) per terabyte of capacity. The time span is fine, but the TBW rating, while par for the course, could be a bit more generous given the price. Speaking of which…
How much does the Samsung 9100 Pro cost?
As of this writing the 1TB version of the 9100 Pro is $200, the 2TB capacity is $300, and the 4TB model is $550. Add $20 if you want a heatsink.
Before adding the cost of the latter, make sure you actually need one (most users don’t), or that your system doesn’t already have a heatsink, and if not, can accommodate one. Our testbed’s upside-down secondary PCIe 5.0 M.2 slot forbids heatsinks — it’s a thing.
In terms of the overall SSD market, you’re definitely paying more for the premium performance. 4TB PCIe 4.0 host memory buffer (DRAM-less) NVMe SSDs are available for around half what the 9100 costs. But as you’ll see below, the performance is also premium — if you have the software to take advantage.
In the right system with the right software, the 9100 Pro will run rings around the competition.
How fast is the Samsung 9100 Pro?
The cat’s already out of the bag on this one. The 4TB 9100 Pro that Samsung sent me aced all the synthetic benchmarks I threw at it in terms of sequential throughput: CrystalDiskMark 8, AS SSD 2.0, and ATTO Disk all placed it easily in first place — with a bullet. It also easily outstripped its PCIe 4.0 990 Pro predecessor in all ways.
On the other hand, real-world performance, while well above average, was a just a hair off the fastest I’ve seen. That saw the drive sitting in third place among all the drives I’ve tested, behind the Crucial T705 and Corsair MP700 Pro SE — both of which are also PCIe 5.0 DRAM designs.
Samsung played up the 9100 Pro’s random performance in its press materials, mostly in comparison to the 990 Pro, but it fell a bit short of the competition. In fact, it was well down the charts with multi-queue random operations.
However, there is not a lot of software that uses multiple NVMe queues currently, so the single-queue random performance is what you should really look at. The 9100 Pro was quite good at that.
While the 9100 Pro reigned supreme in CrystalDiskMark 8, it lagged a hair behind its rivals in real-world transfers. Including the 48GB reads and writes shown below. That said, the 2-second difference is right around the margin of error for this test.
Though only 6 seconds slower in the 450GB write, that gap cost the 9100 Pro the overall crown. While lagging in random operations, host memory buffer drives tend to excel at this particular test so the 9100 actually fell to 17th best out of the 52 NVMe SSDs I’ve tested.
Of course, we’re only talking about a 16 second/11 percent lag from the number one drive writing 450GB — the 8TB WD SN850X.
About our real-world transfer tests… They don’t always show off the ultimate capabilities of a drive, only what you’ll see when transferring files using Windows Explorer. Again, that’s largely because Windows uses only a single queue during transfers, negating much of NVMe’s advantage and rendering a vendor’s unique controller tricks basically moot.
I’m hoping for better from Microsoft in the future. It would be nice to finally see real-world I/O that matches that of benchmarks.
Should you buy the Samsung 9100 Pro?
If you have the bucks, absolutely. In the right system with the right software, the 9100 Pro will run rings around the competition. On the other hand, most users not loading large language models or such will be just fine with an HMB design for half the price.
How we test
Drive tests currently utilize Windows 11, 64-bit running on an X790 (PCIe 4.0/5.0) motherboard/i5-12400 CPU combo with two Kingston Fury 32GB DDR5 4800MHz modules (64GB of memory total). Both 20Gbps USB and Thunderbolt 4 are integrated to the back panel and Intel CPU/GPU graphics are used. The 48GB transfer tests utilize an ImDisk RAM disk taking up 58GB of the 64GB of total memory. The 450GB file is transferred from a 2TB Samsung 990 Pro which also runs the OS.
Each test is performed on a newly NTFS-formatted and TRIM’d drive so the results are optimal. Note that in normal use, as a drive fills up, performance may decrease due to less NAND for secondary caching, as well as other factors. This can be less of a factor with the current crop of SSDs with far faster late-generation NAND.
Caveat: The performance numbers shown apply only to the drive we were shipped and to the capacity tested. SSD performance can and will vary by capacity due to more or fewer chips to shotgun reads/writes across and the amount of NAND available for secondary caching. Vendors also occasionally swap components. If you ever notice a large discrepancy between the performance you experience and that which we report, by all means, let us know.
