ADATA SSD Data Recovery

ADATA SATA SSD recovery (SU800, SU650, SU760) ranges from $200 for a simple data copy to $1,200–$1,500 for NAND swap with microsoldering. ADATA NVMe recovery (XPG SX8200 Pro, GAMMIX S70 Blade, LEGEND series) ranges from $200 to $1,200–$2,500. Free evaluation, firm quote before paid work, and no data means no charge.

ADATA SATA SSD Pricing (SU800, SU650, SU760)

+$100 rush fee to move to the front of the queue. A donor drive is a matching SSD used for its circuit board. Typical donor cost: $40–$100 for common models, $150–$300 for discontinued or rare controllers.

ADATA NVMe SSD Pricing (XPG, GAMMIX, LEGEND)

+$100 rush fee to move to the front of the queue. A donor drive is a matching SSD used for its circuit board. Typical donor cost: $40–$100 for common models, $150–$300 for discontinued or rare controllers.

ADATA sources controllers from Silicon Motion, InnoGrit, & Realtek. This means ADATA SSDs share failure modes with every other brand using the same controller silicon: Crucial, Kingston, WD, Sabrent, Corsair. Our SSD data recovery lab works the controller down from SMI SM2258XT / SM2259XT FTL panics & SM2262EN boot-loader corruption, not from the brand label up.

The recovery advantage: PC-3000 SSD's Silicon Motion utility covers the SM2258XT, SM2259XT, SM2262EN, & SM2263XT. The same loader injection & FTL reconstruction procedures that recover a Crucial MX500 apply to an ADATA SU800. Controller-level expertise transfers across all Silicon Motion brands.

Two failure categories dominate. Firmware failures cause the controller to lock out, report wrong capacity, or drop the drive's identity string. Hardware failures include shorted voltage regulators, dead PMICs, & NAND cell degradation from write exhaustion. Both require lab-level intervention with firmware repair or board-level microsoldering.

Recovery software works on ADATA SSDs with logical failures only: accidental deletion (with TRIM disabled), partition table corruption, or a formatted volume. The drive must be physically healthy, detected in BIOS with its correct model name & capacity, and responding to read commands.

Software can't fix an SU800 showing 0 bytes in BIOS, an XPG drive that dropped off the NVMe bus, or any ADATA SSD with a dead controller. These failures mean the controller isn't serving data to the operating system. No software running on the OS can talk to hardware the OS can't see.

TRIM is the other boundary. On ADATA SSDs with TRIM enabled (the default on Windows 7+ & on macOS for Apple-shipped SSDs), deleted files are unmapped from NAND within seconds to minutes. The controller unmaps those logical addresses and schedules garbage collection. Once garbage collection completes, no software & no lab can recover that data. If your ADATA SSD has failed (not detected, wrong capacity, or sudden death), power it down & send it for evaluation.

Two ADATA SSD slowdowns get mistaken for each other. One is benign & transient. The other means the drive is failing & you should stop using it now. The distinction matters because continued use of a drive in pre-failure makes the recovery harder & sometimes impossible.

The non-destructive triage rule when symptoms in the right column appear: power the drive down. Do not run CHKDSK on a failing SSD. Do not attempt a firmware update on a drive showing instability; a failed firmware flash on an unstable controller can push the recovery from a firmware tier ($600–$900 SATA, $900–$1,200 NVMe) into the board-repair tier ($450–$600 SATA, $600–$900 NVMe), or brick the drive entirely. To verify the drive is detected at all, move it to a USB-to-SATA or USB-to-NVMe enclosure on another machine & check the BIOS device list. That step is safe; running diagnostic software against the failing drive is not.

ADATA sources controllers from five vendors: Silicon Motion (SATA & NVMe), InnoGrit (NVMe Gen4), Realtek (NVMe Gen3), Maxio Technology (SATA & NVMe), & Phison (PCIe 4.0 first-generation only). Unlike Samsung, which designs its own controllers, ADATA customizes the vendor's reference firmware for branding & warranty telemetry. The underlying silicon, FTL architecture, & diagnostic mode entry points match the vendor's reference design.

Silicon Motion SATA: SM2258 and SM2259XT

The Silicon Motion SM2258 powers the ADATA SU800, ADATA's long-running mainstream SATA SSD. The SM2258 includes a DRAM buffer for FTL caching, which gives it better endurance than DRAM-less alternatives. The SU900 used the same SM2258 paired with 3D MLC NAND, one of the few mainstream consumer drives to ship with MLC instead of TLC; the DRAM cache & higher MLC endurance keep it out of typical write-amplification panics, but the drive remains vulnerable to Keep BSY state from power-loss FTL flush interruption & recovers through the same PC-3000 SSD Silicon Motion path as the SU800.

Early SU650 production ran the SM2258XT, a DRAM-less variant of the same family. When SM2258 or SM2258XT FTL pages degrade, the controller enters ROM mode & the drive reports 0GB, 1GB, or 1023MB capacity. PC-3000 SSD's Silicon Motion utility loads a safe mode firmware module for both controllers, bypassing the corrupted boot sequence & rebuilding the translation table from NAND residuals.

How a Panicked SM2258XT or SM2259XT Is Recovered on PC-3000 SSD

ADATA's consumer SATA lineup is a moving target at the bill-of-materials level. The SP550 ships with the Silicon Motion SM2256K paired with planar TLC NAND & a DDR3 cache. The SU800 ships with the SM2258 (the SM2258G & SM2258H revisions both appear in retail) paired with Micron 3D TLC. Early SU650 production ran the DRAM-less SM2258XT; later production swapped in the SM2259XT or SM2259XT2 with Intel or SpecTek NAND.

When the FTL on a Silicon Motion SATA drive corrupts past LDPC's correction ceiling, the controller aborts its boot sequence to keep garbage collection from chewing through what's left of the user data. The drive drops its consumer model string. Disk Management, Device Manager, or the host BIOS will instead report the raw silicon descriptor: SM2258XT, SM2259XT, SM2258AB, SM2258AA in ROM Mode, or the generic SMI Factory alias. Reported capacity collapses to a Mask ROM placeholder: 1024MB (sometimes rendered as 1023MB or 1GB), 0GB / 0MB, or occasionally 2MB. The SATAFIRM S11 identifier is not an SMI symptom. That string belongs exclusively to the Phison PS3111-S11 bootROM & will never appear on a Silicon Motion drive; if a drive enumerates as SATAFIRM S11, it is a Phison controller, not an SMI one.

The electrical state behind these BIOS strings is what SSD repair literature calls Keep BSY. The drive negotiates the SATA link & completes the initial COMRESET handshake, then stalls inside FTL initialization. The status register holds BSY=1 & DRDY=0 indefinitely. UEFI POST can hang waiting for the drive to report ready.

Windows Disk Management or macOS Disk Utility can freeze on enumeration. The drive may register in Device Manager for a moment then drop, or never register at all. No LBAs are exposed to the host, so EaseUS, R-Studio, Disk Drill, & every other host-side recovery scanner sees either nothing or a 0-byte uninitialized device. There is no logical volume to scan because the controller never finished telling the host that one exists.

The recovery path for SM2256, SM2258, SM2258XT, SM2259XT, & SM2259XT2 on PC-3000 SSD starts at the PCB itself. Entry into Safe Mode (ACELab calls it Technological Mode) is a hardware step: a pair of diagnostic test points on the board (silkscreened ROM on the Silicon Motion reference layout) are shorted with precision tweezers during the power-on sequence. The short keeps the controller out of the corrupted on-NAND firmware & in its read-only Mask ROM.

Once the controller is parked in Safe Mode, the ACELab Silicon Motion utility injects a loader (LDR) into the controller's SRAM. The loader is a small volatile firmware image that must match three things at once: the controller silicon revision (an SM2258XT loader will not run on an SM2259XT), the NAND lithography printed on the flash dies (SanDisk 64-layer BiCS3 behaves differently from Micron B27A), & the original firmware version baked into the drive.

When the right loader is in place, it strips out TRIM, wear-leveling, & garbage collection so no background activity can touch the NAND during imaging. Because the loader lives in volatile SRAM, a wrong loader cannot brick the drive; cycling power returns the controller to its ROM-mode starting position.

Silicon Motion controllers write XOR-scrambled pages to the NAND. The scrambling keeps charge distribution even across cells & is not AES-256 encryption, but the raw NAND bytes are not directly readable as user data. The loader activates the controller's hardware XOR descrambler & reapplies LDPC error correction with the correct geometry so the pages decode back into plaintext sectors before they are imaged.

The translator rebuild is where the DRAM-less designs earn their reputation for fragility. Because the SM2258XT & SM2259XT keep their FTL only on the NAND itself, a degraded system block destroys the L2P map outright. With the loader running, PC-3000 SSD reconstructs the translator from the NAND rather than from the drive's own broken copy.

The drive can then be imaged sector by sector. Nothing is written back to the failed SSD; the rebuilt translator stays in volatile RAM on the drive itself, & the original NAND is left read-only.

Two shortcuts show up in online guides for these symptoms. The first is the SMI MPTool (Mass Production Tool), a factory production utility rather than a recovery tool. If the data on the drive matters, send it to a lab before running any vendor mass-production or reinitialization utility against it.

The second is consumer recovery software (EaseUS, R-Studio, Disk Drill, PhotoRec) pointed at a drive in BSY or ROM state. These tools sit above the OS storage driver. They cannot see a drive that refuses to expose LBAs to the SATA host. They scan nothing because there is nothing for them to scan.

Background on the controller family lives at our Silicon Motion architecture page. The mechanics of ROM-mode entry across vendors are covered at firmware panics & ROM mode. A cross-vendor view of controller-level recovery sits at our SSD controllers overview. If an ADATA SU800, SU650, SU750, or SP550 has dropped to a raw SMI descriptor or is hanging the host on POST, the SSD data recovery service page covers pricing, turnaround, & the no data, no recovery fee policy.

Silicon Motion NVMe: SM2262EN, SM2267XT, and SM2269XT

The SM2262EN in the XPG SX8200 Pro is a dual-core ARM NVMe Gen3 controller with DRAM, and PC-3000 SSD supports it. The SM2269XT (Gen4) is supported by the same Silicon Motion utility for safe mode entry & FTL reconstruction. The SM2264 is Silicon Motion's highest-end Gen4 controller with dedicated DRAM cache; it currently lacks a PC-3000 SSD utility module, so a drive built on one relies on board-level hardware repair rather than firmware reconstruction.

The SM2267 & SM2267XT sit a tier below the SM2262 in Silicon Motion's Gen3/Gen4 NVMe line, and the XT suffix is the whole story. The plain SM2267 carries an onboard DRAM cache for its Flash Translation Layer. The SM2267XT drops that DRAM chip & leans on Host Memory Buffer (HMB) instead, borrowing a slice of the host's system RAM over PCIe to hold hot FTL entries. Same naming pattern as the SATA SM2258 vs. SM2258XT split already on this page. The DRAM-less SM2267XT pages its persistent FTL between a small internal SRAM & reserved NAND metadata, so an unclean shutdown during a background FTL update can corrupt mapping data written days earlier, the same retroactive corruption mechanism that bites the DRAM-less SATA parts.

The SM2267XT is supported by PC-3000 SSD's Silicon Motion utility. When one of these drops into a BSY state or stops exposing full capacity after a power loss, the recovery path mirrors the SMI SATA workflow: short the technology test points to park the controller in Safe Mode, inject the matching SRAM loader through Vendor Specific Commands, then rebuild the translator from the NAND in the workstation's RAM. The loader is volatile, so a wrong image can't brick the drive; cycling power returns the controller to ROM mode.

Because ADATA frequently swaps controllers across production runs without changing the model name, we identify the actual silicon on the PCB before selecting a module rather than guessing from the model name. An ADATA NVMe drive built on the SM2267XT recovers through this same Safe Mode loader path. NVMe firmware recovery runs $900–$1,200.

InnoGrit IG5236: Board-Level Repair Only

The InnoGrit IG5236 is an NVMe PCIe Gen4 controller with onboard DRAM. Rossmann does not currently offer in-lab recovery for InnoGrit IG5236. The IG5236 is not in PC-3000 SSD's supported list, so there is no firmware-level loader path for it; recovery on a drive built around one means board-level repair to bring the original controller back to a responsive state, after which it serves data over the standard NVMe interface. NVMe board repair runs $600–$900.

Realtek RTS5763DL: Board-Level Repair Only

A drive on this silicon can drop to 0 bytes or stop enumerating in BIOS after an unclean shutdown.

Rossmann does not currently offer in-lab recovery for Realtek RTS5763DL. No Realtek NVMe controller is in PC-3000 SSD's supported list, so there's no firmware-level utility path. The recovery vector is board-level repair to revive the original controller. If a Realtek DRAM-less NVMe controller turns up on an ADATA PCB, it falls outside PC-3000 SSD support & into the board-repair tier at $600–$900.

Map Your ADATA Drive: Controller, PC-3000 Status, Recovery Vector

The controller decides the recovery path, not the ADATA label. Three controller families that share NVMe Gen4 packaging recover three different ways. Match your drive to its controller, then read across to the recovery vector before assuming software or chip-off applies.

Rossmann does not currently offer in-lab recovery for InnoGrit IG5236 or Realtek RTS5763DL; both are board-repair-only. Send the drive for a free evaluation & we confirm the controller on the PCB before quoting. No data, no recovery fee.

XPG GAMMIX S50: Phison NVMe PS5016-E16

ADATA skipped Phison's mainstream PCIe Gen3 turnkey controllers (PS5012-E12, PS5013-E13T) entirely & partnered with Phison only on the PS5016-E16 for the original XPG GAMMIX S50 (the non-Lite variant), one of the first PCIe 4.0 consumer SSDs. The SX8200 Pro, GAMMIX S11 Pro, S40G, Falcon, & Swordfish all use Silicon Motion or Realtek silicon. The S50 Lite uses the Silicon Motion SM2267, not Phison. If a customer ships an SX8200 Pro thinking it's a Phison drive based on community guesses, we identify the actual Silicon Motion SM2262EN on the PCB before quoting.

Realtek NVMe Controllers in ADATA Drives

ADATA's budget NVMe drives, including the SE770G, use Realtek NVMe controllers. Realtek is a newer entrant in the SSD controller market, and PC-3000 SSD does not currently support Realtek NVMe controllers. Recovery from Realtek-based ADATA drives focuses on board-level hardware repair to keep the original controller alive & serving data.

ADATA's SSD lineup spans budget SATA, mainstream NVMe, & high-performance Gen4 drives under both the ADATA & XPG brands. Each model uses a different third-party controller with different failure patterns, encryption implementations, & PC-3000 recovery procedures.

The ADATA SU650, SU630, & SU635 are DRAM-less SATA SSDs. The LEGEND 850 is a DRAM-less NVMe SSD. DRAM-less means there is no dedicated DRAM cache chip on the PCB. The absence of that chip changes how the Flash Translation Layer (FTL) is managed, & it changes what happens when the drive loses power mid-write.

How DRAM-less SATA Controllers Handle the FTL

On an SU800 (SM2258 with DRAM), the entire FTL lives in the onboard DRAM cache & the controller performs address lookups at DRAM speed. On an SU650 (SM2258XT, DRAM-less), the FTL lives in a reserved metadata region of the TLC NAND. The controller uses a small internal SRAM to cache the hottest mapping entries. On a cache miss, the controller stalls the pending I/O, reads the mapping page from NAND, loads it into SRAM, & only then completes the operation. This dual-access penalty hurts 4K random I/O & means the controller is constantly paging FTL fragments back & forth between SRAM & NAND.

How HMB Differs on DRAM-less NVMe

NVMe 1.2 introduced Host Memory Buffer (HMB), which lets a DRAM-less NVMe SSD reach across the PCIe bus via DMA & borrow a slice of the host computer's system RAM. DRAM-less ADATA NVMe controllers use HMB to hold active FTL entries in host RAM instead of NAND. Random I/O recovers most of the speed lost to the DRAM-less design. Two caveats matter for recovery. HMB only holds metadata, not user data. HMB is also unavailable over USB bridges, so a DRAM-less ADATA NVMe drive in a USB enclosure reverts to SATA-DRAM-less-class latency. The persistent FTL still lives on the NAND; HMB is a working cache, not a store of truth.

Power Loss & Retroactive FTL Corruption

DRAM-less SATA drives are unusually fragile in the face of unexpected power loss because the controller is continuously updating FTL pages in NAND during normal operation. If a power cut interrupts an in-flight TLC or QLC program operation, neighboring pages in the same physical cell can lose their voltage margins. The next boot triggers uncorrectable LDPC errors, & the firmware enters ROM mode to protect the NAND from further damage. The drive reports 0 GB, 1 GB, 2 MB, or 1023 MB in BIOS & exposes the raw controller string ("SM2258XT", "SM2259XT") instead of the ADATA model name. On the SM2258XT, the drive may instead hang the host bus with an ATA "Keep BSY" loop. At that point, the logical path between the OS & the NAND is severed. Consumer software cannot reach the data; board-level intervention is the only path forward. SSD firmware corruption recovery starts here.