Portable SSD & External Drive Inspection China: Real Capacity, Sustained Write Speed & NAND Grade QC
💾 The Spec Sheet Is a Wish; a Write Test Is a Fact A 1TB portable SSD is one of the few products where the cheap lie is invisible to every inspection method that relies on looking.
💾 The Spec Sheet Is a Wish; a Write Test Is a Fact
A 1TB portable SSD is one of the few products where the cheap lie is invisible to every inspection method that relies on looking. The enclosure is right, the label is right, the PCB is populated, the drive mounts and shows 1TB in Windows — and half the address space is fake or the memory inside is recycled wafer from someone else's dead product. You only find out in month three, when the drive starts dropping files. That is why external storage needs a test-first inspection protocol rather than a cosmetic one, and it is the same discipline behind every pre-shipment inspection in China we run: sample the shipping lot, measure, record, decide. With 2,000+ inspections behind us, storage is the category where "looks fine" has the worst track record.
The good news: almost every one of these faults is detectable in hours with the right sequence, and it is fully reproducible on the same units the factory will ship.
🔬 Eight Checks That Separate Real Capacity from Marketing
| Check | How it is measured | What a failure looks like |
|---|---|---|
| 1. Full-surface write verify | Write the whole drive to capacity, read it back, verify checksums | Errors or dropped data at 480GB on a drive sold as 1TB - patched firmware capacity |
| 2. Sustained write speed | Continuous large-file write for 30-60 GB, not a 1GB burst | Drops from 620 MB/s to 88 MB/s once the SLC cache empties - downgraded NAND |
| 3. NAND grade and type | Controller tool or SMART read-out: power-on hours, host writes, bad-block table, PE cycles | Power-on hours in the thousands on a "new" drive - recycled or ex-farm NAND |
| 4. Link rate negotiation | Measured negotiated speed and cable spec in the same test | A 10Gbps drive capped at 480 MB/s by a Gen1 cable or thin internal wiring |
| 5. Thermal behaviour | Surface and internal temperature at 30 minutes of sustained load | 68 C and a hard throttle long before the rated speed claim |
| 6. Firmware and honesty | Firmware version across the lot, secure erase, wear indicators, encryption claims tested | Three firmware versions in one lot, or an encryption claim that does nothing |
| 7. Physical and mechanical | Solder inspection, connector retention after repeated insertion, drop and shell fit | Cold joints at the connector, cracked shell clips, a drive that disconnects when the cable moves |
| 8. Labels, packing and compliance | Capacity marking, compliance marks, anti-static bag, cable rating, printed FNSKU | Decimal-vs-binary capacity games, missing compliance identity, cable never rated for the claim |
Checks 1, 2 and 3 are the ones that settle arguments. Everything else is quality; those three are whether the product is what the listing says.
🧮 A Real Lot: 2,600 Drives, 1TB Claimed
Case from this year. 2,600 portable 1TB SSDs at $57.50 FOB, roughly $149,500, from a Shenzhen factory quoting 15% under the market. Sampled at n=125 (AQL 2.5 Level II, Ac 7): 9 drives physically capacitated at 512GB — the write-verify run failed at around 480GB with read-back errors on a drive labelled 1TB; 14 fell to 88 MB/s sustained after the cache emptied, with a NAND grade inconsistent with the sample; 7 showed power-on hours in the thousands and non-zero host writes in the SMART data on a supposedly new unit; 6 were capped at 5Gbps by the bundled cable; 5 hit 68 C at thirty minutes; and 4 cartons were missing the USB-C to A cable promised on the box. Outcome: 620 units replaced or re-flashed, labels and cable spec corrected, re-inspection at 1.1% majors, two man-days at $169 per man-day with the report in ~24 hours.
Now the other branch. An 8% return rate on 2,600 drives is roughly 208 units — about $11,960 in refunds before removal fees, plus the reviews that say "lost my wedding photos". For a storage product the review damage is permanent in a way a scratched panel is not, because the failure mode is data loss and buyers describe it loudly.
🧪 Why One Pass/Fail Number Is Not Enough Here
Capacity fraud and NAND substitution are module-clustered: they come from one batch of memory or one firmware flash, so they cluster inside specific cartons or serial ranges. A single accept number hides that. We handle it three ways: run the full-surface capacity test on the whole sampled set rather than a subset, segment the sample by carton and serial range so any cluster is visible, and where the value justifies it, do a 100% capacity flash-and-verify at the factory bench. Sampling math still governs the rest — check the sample size for your lot size with the AQL calculator before you write the inspection brief.
Write the protocol into the PO as well: single NAND source declared, firmware version locked for the lot, cable rating stated, and the balance released against the report. Storage buyers who do that get a product; buyers who do not get a warning letter from Amazon and a warehouse full of returns.
📦 What to Put in the Inspection Brief
Five lines do most of the work: declared capacity and NAND type with the source named; sustained write requirement after cache exhaustion; cable and interface spec that must be verified on the unit; thermal limit at thirty minutes of load; and the accessory list that must be counted per carton. Add your own units where the market is cruel — if you sell to videographers, raise the sustained-write bar; if you sell to schools, test the shell. Coverage and pricing for this kind of functional inspection are published on our pricing page and the service scope is on our inspection services page. Running a storage order now? Send us the SKU, lot size and ship date and we will tell you what the sampling plan should look like.
❓ FAQs
How do you detect a fake-capacity SSD?
By writing the drive to full capacity and verifying the read-back with checksums instead of trusting what the operating system reports. Patched controllers can advertise 1TB while the physical memory is far smaller; the failure appears as errors or lost data once you pass the real capacity point.
What sustained write speed should I require?
Set it as a floor after the SLC cache empties, not as a peak figure - for example a minimum of 300 MB/s across a continuous 30-60 GB transfer for a mainstream USB 3.2 Gen2 drive. Peak numbers from a 1GB burst are unrepresentative and are how throttled drives pass casual testing.
How can you tell if the NAND is recycled or downgraded?
Read the controller-level data: power-on hours, total host writes, factory bad-block counts and in some cases the NAND ID reported by the controller tool. A new drive should show near-zero host writes; thousands of hours on a sealed retail unit points to recycled or ex-farm memory.
Should I test every unit or just a sample?
For capacity and firmware integrity, test the whole sampled set and consider 100% capacity verification when the order value is high or the supplier is new. For cosmetic and mechanical items, a standard AQL sample is enough - these faults are spread randomly rather than clustered by carton.
How much does a functional inspection of electronics cost?
Our rate is $169 per man-day inside major manufacturing regions, with reports back in about 24 hours. Drive capacity and speed testing is time-consuming per unit, so plan the sample size against the time budget - one man-day covers a solid functional set for most small-lot orders.
Frequently asked questions
How do you detect a fake-capacity SSD?
By writing the drive to full capacity and verifying the read-back with checksums instead of trusting what the operating system reports. Patched controllers can advertise 1TB while the physical memory is far smaller; the failure appears as errors or lost data once you pass the real capacity point.
What sustained write speed should I require?
Set it as a floor after the SLC cache empties, not as a peak figure - for example a minimum of 300 MB/s across a continuous 30-60 GB transfer for a mainstream USB 3.2 Gen2 drive. Peak numbers from a 1GB burst are unrepresentative and are how throttled drives pass casual testing.
How can you tell if the NAND is recycled or downgraded?
Read the controller-level data: power-on hours, total host writes, factory bad-block counts and in some cases the NAND ID reported by the controller tool. A new drive should show near-zero host writes; thousands of hours on a sealed retail unit points to recycled or ex-farm memory.
Should I test every unit or just a sample?
For capacity and firmware integrity, test the whole sampled set and consider 100% capacity verification when the order value is high or the supplier is new. For cosmetic and mechanical items, a standard AQL sample is enough - these faults are spread randomly rather than clustered by carton.
How much does a functional inspection of electronics cost?
Our rate is $169 per man-day inside major manufacturing regions, with reports back in about 24 hours. Drive capacity and speed testing is time-consuming per unit, so plan the sample size against the time budget - one man-day covers a solid functional set for most small-lot orders.