Exceptional fake SSD clone of Samsung 990 Pro is almost impossible to spot — near-identical performance blurs the line between real and fake as AI crunch drives knock-off market

Samsung 990 Pro
(Image credit: Samsung)

They say imitation is the sincerest form of flattery, and since the Samsung 990 Pro is one of the best SSDs, it's easy to see why there are so many clones of the PCIe 4.0 drive on the market. Japanese news outlet Akiba PC Hotline! recently examined one of the latest fake Samsung 990 Pro drives, and the quality behind the clone is both impressive and frightening at the same time.

The fake Samsung 990 Pro features a sticker that closely mimics the original. This is where the technical know-how comes into play. The drive may look authentic on the surface, but a closer inspection reveals drastic discrepancies. For starters, the Samsung 990 Pro leverages the brand's proprietary Pascal SSD controller. The fake uses the Maxio MAP1602 SSD controller, which has a different design and is smaller than the Pascal controller. Tellingly, the rear of the counterfeit SSD even openly states the Maxio model number.

Latest Videos From

Another key distinction lies in DRAM and NAND. The Samsung 990 Pro features LPDDR4 memory, a component absent from the fake due to the DRAM-less nature of the Maxio MAP1602 SSD controller. As for the NAND, the original uses Samsung’s 176-Layer V-NAND TLC flash, while the type in the counterfeit remains uncertain. It’s likely QLC, though, since counterfeiters won't spend extra on TLC for a product whose primary purpose is to bamboozle the buyer.

Fake Samsung 990 Pro SSD Benchmarks

Swipe to scroll horizontally
Header Cell - Column 0

Real Samsung 990 Pro 1TB

Fake Samsung 990 Pro 1TB

SEQ1MQ8T1 Read (MB/s)

7453.34

7255.02

SEQ128KQ8T1 Read (MB/s)

7448.13

7190.50

RND4KQ32T16 Read (MB/s)

4915.30

4885.70

RND4KQ1T1 Read (MB/s)

92.61

84.66

SEQ1MQ8T1 Write (MB/s)

6953.81

6090.03

SEQ128KQ8T1 Write (MB/s)

6946.02

6122.06

RND4KQ32T16 Write (MB/s)

6446.80

3537.48

RND4KQ1T1 Write (MB/s)

310.12

307.63

Typical Samsung 990 Pro lookalikes fall flat when you benchmark them. Some fakes are slower than a USB 2.0 pendrive, while others perform similarly to a PCIe 3.0 SSD. However, the one that Akiba PC Hotline! had in its possession exhibited a solid, very convincing performance. The drive delivered sequential read and write speeds of up to 7,255 MB/s and 6,090 MB/s, respectively, just a bit shy of the Samsung 990 Pro 1TB's rated speeds of 7,450 MB/s and 6,900 MB/s.

If you compare the specifications between the Maxio MAP1602 and the Samsung Pascal, it's evident that the former isn't too far behind in sequential performance. That's likely the reason why counterfeiters specifically choose the Maxio MAP1602. They know that the typical user is unlikely to look beyond the sequential performance numbers. However, the random performance is more pronounced if you look closely, specifically in the random 4K test using 32 queues and 16 threads, where the Maxio MAP1602 delivered just over half the performance of the Samsung Pascal.

Akiba PC Hotline! submitted the Samsung 990 Pro 1TB clone to the H2testw utility and confirmed that the reported capacity was indeed 1TB. At least on paper, the drive wasn’t lying about its size. A 976.6GB write test stretched out to around two hours, with the drive only managing a sluggish 132 MB/s. In a real-world scenario, the genuine Samsung 990 Pro completed the transfer of a 397.2GB video file in just 3 minutes and 33 seconds (1,861 MB/s), while the clone crawled along, taking a staggering 25 minutes and 20 seconds (261 MB/s) to complete the same task. The real drive maintained a sustained write speed of around 1,500 MB/; once the SLC cache was full, the fake drive stuck to 100 MB/s.

The best way to weed out clones from the real ones with Samsung SSDs is to use the Samsung Magician software to authenticate the drive. Furthermore, it doesn't allow you to use the software's features, such as the diagnostic scan and performance optimization, on fake drives.

Spotting Samsung clones has become routine for every generation of Samsung SSDs. Previously, we saw many Samsung 980 Evo and Samsung 980 Pro clones. Some entrepreneurs even invented SSDs that even Samsung didn't know were in its lineup, such as the Samsung 1080 Pro. The takeaway from all this is always to purchase your Samsung SSDs from trusted, reputable retailers. Even if your new drive passes the sniff test, run Samsung Magician for that extra peace of mind, as you can see, clones are getting better by the day.

Google Preferred Source

Follow Tom's Hardware on Google News, or add us as a preferred source, to get our latest news, analysis, & reviews in your feeds.

TOPICS
Zhiye Liu
News Editor, RAM Reviewer & SSD Technician

Zhiye Liu is a news editor, memory reviewer, and SSD tester at Tom’s Hardware. Although he loves everything that’s hardware, he has a soft spot for CPUs, GPUs, and RAM.

  • Gururu
    "The takeaway from all this is always to purchase your Samsung SSDs from trusted, reputable retailers."

    Unfortunately, not even Amazon services/warehouse can be trusted anymore.
    Reply
  • Dementoss
    Samsung will always be popular with the cloners, as they are both popular and expensive, added to which, there are a lot of people out there who will think, "ooh cheap", and not even consider checking to see if the retailer is genuine and trustworthy.
    Reply
  • Notton
    It looks familiar...
    I haven't seen a Lexar NQ790 under the sticker, but it's presumably the same PCB as the NM790.
    https://www.tomshardware.com/reviews/lexar-nm790-ssd-review
    Reply
  • PEnns
    Did the article mention the price of the Samsung-like fake SSD?? Is it being sold for less or the same price??

    If it's "exceptional" but cheaper, some might accept the fakery!
    Reply
  • Stomx
    I will not buy Samsung even from the Samsung itself. In October or last year I decided to try Samsung SSDs. On the latest Linux Mint 22.2 (Zara), latest AMD server Turin processor and motherboard with latest BIOS, both the latest PCIe4 and PCIe5 4TB Pro Samsung SSDs (bought from Amazon and sold by Amazon itself not from some third party, everything was genuine Samsung) behaved totally irrationally.

    The PCIe5 one heavily corrupted the data, ran slow and stopped working almost in 5 minutes while PCIe4 was kind of OK but in a couple weeks disappeared from the list of drives.

    Most probable reason was the incompatible or older firmware, at least this is what internet suggested to try to fix. But the Samsung Magician software is absent in Linux and so you can not update the firmware if you do not have Windows
    Reply
  • abufrejoval
    What frightens me, is that any failing part these days means I'm loosing an entire machine, because spares are too expensive, unavailable, or risky.
    Reply
  • usertests
    abufrejoval said:
    What frightens me, is that any failing part these days means I'm loosing an entire machine, because spares are too expensive, unavailable, or risky.
    You probably won't lose the RAM.

    Unless it's soldered. 💀
    Reply
  • Kindaian
    It will be sold at similar, maybe some usd less than the original, but just a small within the tolerances of a different supply chain so that it doesn't make it suspect. If it was me, i would sell it more expensive than the original, just because it is available. But it isn't me! And i wouldn't ever do that anyway...
    Reply
  • abufrejoval
    usertests said:
    You probably won't lose the RAM.

    Unless it's soldered. 💀
    Actually I've been very lucky, very few failures overall, mostly mainboards actually, and likely because of the frequent rebuilding: everything was still an add-on card and all 8 AT or "ISA" slots usually filled. Earliest fail was a 80386 mainboard that still had the old style DIN keyboard connector: that broke off and I had to do some brutish resoldering because there was no such thing as USB in those days and without a keyboard a computer was useless. And they easily cost $10.000 in those days.

    Decapitated a capacitor decades later, which had odd effects, I've purged from memory. The main challenges were getting a matching part (too small to properly identify the capacity) and soldering on a mostly SMD board... looks terrible but it still works, though.

    Mostly it's been EE-PROMs going bad and mainboards then becoming functionally dead. A fine laptop didn't survive a CMOS re-config and an X99 Xeon board that failed to POST was thankfully easy enough to replace.

    Never had a CPU fail or burn, my first RAM chips were 16kbit DILs, but I only had marginal parts I had to accomodate timing for, not outright failures. Some extremely early HDDs died, a few 2nd or 3rd generation SATA SSDs did as well.

    Some cheap Chinese Ethernet NICs and SATA adapters died or were DoA when I tried using them months or even years after putting them in stock, while I've operated LSI RAID6 controllers for 15 years 24x7 without issue.

    Mostly, I've always assumed that computer don't die, they just get obsolete. They might have actually died, but by that time I had long passed them on, and they might have moved on further after that: so I never got the sad news for lack of a multi-hop back-channel.

    But I own a few Nvidia GPUs with that awful power design and connector as well as an ASUS AM5 mainboard with 96 GB of DDR5 ECC RAM very close to the RTX 4090.

    I never considered PCs a fire hazard in 40 years of building them... until recently.

    P.S. competely forgot about mobile batteries: Yes, far too many phones and laptops got destroyed by those. Two rather new laptops were killed by Windows 10 deciding it was a good time to look for updates, while suspended to disk in the middle of a flights between Frankfurt and Lyon: fortunately the damage didn't exit the hand luggage. Phones and tablets just decided it was just time to get replaced after two to three years.
    Reply
  • Shiznizzle
    abufrejoval said:
    What frightens me, is that any failing part these days means I'm loosing an entire machine, because spares are too expensive, unavailable, or risky.
    Which is why i kept my back up machine and my parts tester machine. 2 Am4 and 1 Am5 i have.

    This will have to do till AI flops on its belly. It will. It is just a matter of time. Investors wont keep investing in a venture that is not only losing money but not making, much, money.

    OpenAI even went as far as forcing ads on the paid for track which is insanely stupid if you ask me. I refuse to pay and watch ads. No way. Not on any service. Not doing it.
    Reply