The fanxiang S660 PCIe 4.0 500GB NVMe M.2 SSD for PS5 with Heatsink is a 0.5TB NVMe-PCIe4 PS5-SSD in new condition. It is in stock at $149.99 - $299.98/TB.
This drive is made by Fanxiang. About this brand → — who makes it, what controller and NAND are inside, and what the warranty actually says. Part of the S660 line → — sourced controller, NAND and cache detail for this exact model.
Before you buy: check the bay and the lanes
U.2 and U.3 drives need a backplane wired for NVMe or an adapter carrying PCIe lanes to the bay - a SAS backplane will not do it. Check your chassis supports NVMe in the slot you plan to use before ordering. NVMe drive prices and form factors
The fanxiang S660 PCIe 4.0 500GB NVMe M.2 SSD for PS5 with Heatsink delivers enterprise NVMe performance at $299.98/TB. Suited to read-intensive cloud workloads and mixed enterprise applications. Verify U.2/U.3 backplane compatibility before ordering.
Against every other drive we have an in-stock price for today, this one is cheaper per tb than 2% of them (2 of 109). The cheapest we can currently buy is $102.25/TB; the middle of the market is $198.00/TB.
These are third-party figures about the retail market as a whole, quoted and dated. They are not our own measurements and they do not describe this specific drive.
Yes. At our last check it was $149.99 ($299.98 per terabyte at its 0.5TB capacity), sold new under ASIN B0BNQBD46W. Amazon sets the final price at checkout.
This listing (B0BNQBD46W, ASIN B0BNQBD46W) carries no renewed, recertified, refurbished or pulled claim in its title, and Amazon reports it as new. On this site that is what "new" means: sold as factory-new. The stored warranty on this row is 5 years, as the listing states it.
Use Case Scenarios
🗄️
OLTP Database Primary Storage
PostgreSQL, MySQL or Oracle requiring low latency and high IOPS.
✓ Excellent fit
NVMe delivers sub-200μs latency vs 5ms for HDD — a 25x improvement that directly reduces query response times. At 1 DWPD verify your daily write volume stays within endurance limits.
🤖
AI/ML Training & Inference
GPU server loading training datasets or serving inference at high throughput.
✓ Excellent fit
NVMe at 6,900MB/s eliminates storage bottlenecks when loading datasets to GPU memory. Consider multiple drives for larger datasets.
💾
Bulk Cold Data Archive
Storing large volumes of infrequently accessed data.
✗ Not ideal
At ~$300/TB, NVMe costs 10x more than enterprise HDD for cold data. Unless you need fast retrieval, HDDs or LTO tape are far more cost-effective.
Maintenance & Troubleshooting
Maintenance Checklist
On installCheck SMART baseline on arrival
Run smartctl -a /dev/sdX and record: Reallocated_Sector_Ct, Power_On_Hours, Spin_Retry_Count. Establishes baseline for future comparison.
MonthlyRun SMART short self-test
smartctl -t short /dev/sdX — takes 1-2 minutes, catches most developing issues. Schedule during off-peak.
QuarterlyRun SMART long self-test
smartctl -t long /dev/sdX — full surface scan, takes 0 hours for 0.5TB. Any test failure is grounds for immediate replacement.
QuarterlyCheck RAID array health
Verify no degraded drives. A degraded array with no hot spare is one failure from data loss. Check mdstat or controller UI.
AnnuallyVerify drive firmware version
Check fanxiang support site for firmware updates. Some versions have known bugs affecting integrity.
Every 3-5 yearsPlan proactive replacement
Enterprise HDDs have higher failure rates after year 4-5. At 0.5TB per drive, budget for replacements before failures occur.
Troubleshooting Guide
Cause: PCIe slot not initialized, missing NVMe driver, BIOS not updated, or U.2 cable issue.
Fix: Check BIOS for NVMe in PCIe enumeration. Update server firmware. Reseat U.2 cable. Verify PCIe bifurcation if using adapter. Test in another slot.