Samsung 9100 PRO vs 990 PRO: Is an Upgrade Worth It on PCIe 4.0?

Your laptop has a Gen4 slot, but the SSD upgrade offers nearly 15GB/s on the box. What would a 9100 PRO change over a 990 PRO, and when is the premium worth paying?

Samsung 9100 PRO vs 990 PRO: Is an Upgrade Worth It on PCIe 4.0?
TLDR
  • Keep a healthy 990 PRO if your Gen4 PC has enough storage and mainly runs games and everyday applications. A faster SSD needs a specific waiting time to eliminate before replacing an already fast drive makes sense.
  • A 9100 PRO works on PCIe 4.0 and retains advantages in some reads. Its Gen5 sequential throughput does not translate into twice the speed for every application, file copy, or game.
  • For a new purchase, compare the premium at the same capacity and with suitable cooling. Frequent large transfers on a Gen5 x4 system strengthen the case for the 9100 PRO; running out of space strengthens the case for more capacity.

Replacing a 990 PRO is different from buying your first drive

The 9100 PRO is the faster Samsung SSD, but that does not automatically make replacing a 990 PRO worthwhile. An upgrade also involves moving data and paying for hardware you do not currently need for capacity. On a PCIe 4.0 machine, the new drive also runs through a slower connection than the one used for its headline speed.

A new build presents a different calculation. If you need an SSD anyway, a small premium for the 9100 PRO buys extra performance potential and the option to move it into a Gen5 system later. That is a stronger argument when the next machine is planned, rather than an unspecified future upgrade. A drive bought for a Gen4 laptop still spends its time in that laptop under Gen4 constraints.

Capacity should follow the files you actually keep. Choosing a smaller Gen5 drive while regularly deleting games or moving project folders leaves the immediate problem unsolved. Compare the capacity you need first, then the additional cost of the faster model at that capacity.

The 2TB 9100 PRO is rated at 14,700MB/s

The 14,800MB/s figure belongs to higher-capacity versions. Samsung specifies 14,700MB/s sequential reads for the 2TB 9100 PRO. The table below uses manufacturer maximums at each drive's supported interface, rather than measurements from a shared test system.

Specification990 PRO 2TB9100 PRO 2TB9100 PRO 4TB
InterfacePCIe 4.0 x4PCIe 5.0 x4PCIe 5.0 x4
Sequential read, up to7,450MB/s14,700MB/s14,800MB/s
Sequential write, up to6,900MB/s13,400MB/s13,400MB/s
NAND / dedicated DRAMTLC / 2GBTLC / 2GBTLC / 4GB
Rated writes for warranty1,200TBW1,200TBW2,400TBW

Sources: Samsung's 990 PRO data sheet, revision 2.0, page 3 and the 9100 PRO data sheet, page 3. The 1TB 9100 PRO is rated at 14,700MB/s reads and 13,300MB/s writes. Both families use the M.2 2280 format, which describes their physical format rather than guaranteeing the connection speed of your slot.

The newer 2TB drive has the same write endurance rating as its predecessor. Both carry a limited warranty ending at five years or their rated writes, whichever comes first. Buying Gen5 does not double that allowance. Nor is TBW a prediction that a drive stops working immediately after reaching it.

Samsung also measures the 9100 PRO's advertised writes with Intelligent TurboWrite active. That fast write buffer operates over a limited transfer size. A long continuous write, especially on a populated drive, is a different purchasing question from the best result in a short sequential test.

What remains faster on a Gen4 connection?

Samsung lists backward compatibility with PCIe 4.0 x4 and 3.0 x4. On Gen4, the interface prevents the 9100 PRO from reaching its advertised Gen5 throughput. However, the interface limit does not make the controller and flash inside two different drives identical.

ComputerBase retested its 4TB 9100 PRO at Gen4. Its high-queue sequential throughput fell close to the 990 PRO, but single-request sequential reads remained roughly 50% faster. Random reads showed little difference. In practical reading and copying, changing that same 9100 PRO between Gen4 and Gen5 produced little difference.

Queue depth describes how many requests are outstanding. A test with multiple requests waiting gives the drive a different workload from handling one at a time. This explains why matching the maximum link throughput does not erase every performance difference. It also explains why a stronger result in one test does not predict an equal reduction in Windows boot time.

For an existing 990 PRO owner, the useful question is which of those operations makes you wait. A specific read-heavy workflow gives a better reason to upgrade than a maximum bandwidth number. For a buyer choosing between similarly priced new drives, the remaining Gen4 advantages are a reason to keep the 9100 PRO on the shortlist.

Copying files and loading models give different answers

A transfer depends on both ends. Moving a large folder from a slower external drive cannot make that source deliver data at the destination SSD's maximum write rate. Copying within one SSD makes that drive handle reads and writes together. File size alone does not describe the task well enough.

PCWorld's own 9100 PRO testing illustrates the gap between benchmark rankings and file-transfer rankings. Its 450GB transfer originates on a 2TB 990 PRO that also hosts the operating system. Its review measures a complete transfer path; it is not simply a test of how quickly the destination can accept unlimited incoming data.

Local AI offers a more concrete waiting time. StorageReview averaged ten model loads per model, comparing a 2TB 990 PRO with a 4TB 9100 PRO. The capacity difference remains part of this comparison.

Model loading990 PRO 2TB9100 PRO 4TBTime saved
DeepSeek R1 7B2.8758s2.6173s0.2585s
DeepSeek R1 32B5.1080s4.3735s0.7345s

The newer drive loads both models faster, but neither result halves the wait. These numbers measure preparing the model, rather than its token generation rate. Frequently switching models makes loading speed more valuable. Keeping one model resident in memory reduces the number of times a faster SSD can help.

The same distinction matters for games. Installation, updates, loading assets, and rendering frames are different jobs. A sequential bandwidth ratio cannot be used as a frame-rate improvement or a promise that every loading screen becomes half as long.

Efficiency is not a laptop temperature measurement

Samsung's claim of up to 49% better power efficiency has a specific unit. Its 2TB comparison footnote gives sequential read efficiency of 1,221MB/s per watt for the 990 PRO and 1,822MB/s per watt for the 9100 PRO. This expresses throughput relative to power, rather than a 49% reduction in watts.

A more efficient device can use more power while doing more work per second. That ratio does not establish a battery-life increase or a temperature reduction. Laptop results also depend on idle behavior, workload, cooling, and the other components using the battery.

Physical fit is more immediately actionable. The heatsink model data sheet lists a maximum thickness of 8.88mm for 1TB through 4TB. The corresponding bare drives are up to 2.38mm thick. Support for M.2 2280 alone does not establish clearance for a tall factory heatsink.

Choose the version that fits the laptop's specified cooling arrangement. On a desktop, decide whether the motherboard cover or the drive's factory heatsink will provide cooling before ordering. Stacking both is not a useful default installation plan.

How much extra is worth paying?

Compare checkout prices for the same capacity, heatsink configuration, and warranty coverage. A launch MSRP compared with a discounted older drive does not describe today's premium. An imported bare drive and a locally supported heatsink model also represent different purchases.

Your situationPurchase judgment
Gen4 PC, sufficient 990 PRO capacity, everyday useKeep the 990 PRO.
New SSD needed, small premium at equal capacityConsider the 9100 PRO if cooling and fit work.
Storage regularly runs outPrioritize enough capacity over the generation label.
Gen5 x4 system with repeated large data transfersThe 9100 PRO has a stronger justification.

Put the premium against a repeated wait. Suppose equal-capacity drives with matching heatsink and warranty coverage differ by $40 (₩54,800), and your own workload saves two seconds per run. These are calculation assumptions, not retail quotes or measured gains for these SSDs. At one run per working day over 250 days, you save 500 seconds, or 8 minutes 20 seconds. At twenty runs per day, you save 10,000 seconds, or 2 hours 46 minutes 40 seconds.

If you value an hour of that recovered waiting time at $20 (₩27,400), the $40 premium needs two hours of savings: 3,600 runs at two seconds each, or 14.4 runs per working day over a year. That favors a repeated, attended workflow over a game loaded once each evening. A background copy that overlaps other work does not automatically recover paid working time. Dollar conversions here use ₩1,370 per dollar.

The model-loading measurements above saved less than one second per load; they do not establish the assumed two-second gain, or a gain on a Gen4 connection. Use a measured difference for the job and interface you intend to use. If you already own the 990 PRO, the cost is also the new drive's price minus any proceeds from selling the old one, plus migration effort. It is not just the premium between two new drives.

Slot configuration can change what you are paying for. MSI's MAG X870 TOMAHAWK WIFI supports up to Gen5 x4 on M2_1 and M2_2 with Ryzen 9000/7000, Gen4 x2 on M2_3, and Gen4 x4 on M2_4. But M2_2 shares bandwidth with the rear 40Gbps USB-C ports and runs at x2 when populated. Selecting x4 in BIOS disables those USB ports. That trade matters if you also use USB4 storage.

With Ryzen 8000, M2_1 and M2_2 support up to Gen4 x4; M2_2 is unavailable with Ryzen 8500/8300. Populating M2_3 also reduces the separate PCI_E3 expansion slot to x2. Selecting x4 for PCI_E3 disables M2_3. Maximum specifications therefore need to be read alongside the CPU and sharing conditions.

Check the slot table for your CPU, then verify the negotiated generation and lane width after installation. If a copy is slow, check its source as well as its destination. If speed drops during a long write, examine the workload, buffer behavior, and temperature over time. Those observations identify what the upgrade is buying. For most Gen4 users already satisfied with a 990 PRO, the next useful purchase is additional space when they need it.