Failing Soldered RAM: A Linux Recovery Boot or a Board Repair?
An Inspiron owner added 8GB in the spare slot, and the laptop kept crashing anyway. Can Linux boot around the faulty soldered chips long enough to save the files?
Decide whether you need the files or the computer
A laptop with failing soldered memory presents two different jobs: recovering its files and making it dependable again. Linux can sometimes help with the first. Excluding known bad addresses stops the operating system from allocating those regions, potentially allowing a machine that crashes during startup to boot. PassMark describes this as a temporary route to backup and further diagnosis, and recommends replacing faulty memory where possible. PassMark's explanation of memory blacklisting
Before buying another memory module, check whether the files can be reached through removable storage. Soldered RAM does not imply a soldered SSD. Dell's Inspiron 5490 service manual, for example, documents a removable SSD and requires removing the base cover and battery before taking out the drive. A compatible connection to a healthy computer can avoid using the faulty laptop's memory during file access. The appropriate removal procedure depends on the exact model. Dell's SSD procedure
Encryption still applies after a drive is moved. A BitLocker volume may require its 48-digit recovery key following a hardware change. Microsoft explains how to locate a matching key in a personal account, a work or school account, or a saved copy. Replacing a motherboard does not supply a missing decryption key. Find a BitLocker recovery key
Why adding good RAM may leave the crashes unchanged
A first-person report on Dell's community forum illustrates the problem. On January 10, 2026, user DevIndia described an Inspiron 5490 that progressed from Chrome tab crashes to boot loops and failures during early Linux startup. The owner reported memory-test errors, 4GB of soldered memory and an empty expansion slot.
Installing an additional 8GB did not stop the crashes, according to the post. The old memory remained available to the operating system. The owner then used Linux's memmap option to exclude regions and reported successful booting and video playback. That option goes to the Linux kernel; the BadRAM command below goes to GRUB and uses different syntax. In this owner's experience, extra capacity alone did not disable the failing onboard memory. The Inspiron owner's report
The recovery route starts before Ubuntu loads
For the GRUB procedure below, the machine must still run a memory tester and reach the normal GRUB menu for an existing Linux installation. Ubuntu itself need not reach a desktop. GRUB loads the operating system, so its command line provides a place to apply an exclusion first. A machine that freezes before GRUB starts cannot receive that command.
| Observed state | Practical next step |
|---|---|
| Repeated faults at consistent addresses; tester and GRUB run | Collect patterns and attempt an exclusion for one boot. |
| Tester freezes or fault locations keep changing | Investigate the hardware; a stable exclusion list may not describe the fault. |
| GRUB rejects badram under lockdown | This boot path does not permit the command. |
| Only the files are needed and storage is removable | Consider accessing the drive through a healthy machine. |
GNU explicitly prohibits badram under GRUB lockdown. Exclusions also need to survive the handoff: the loaded kernel must obtain its memory map from GRUB. Linux is among the supported kernels, but the actual boot must be checked afterward. This procedure is for a GRUB-based PC installation; Memtest86+ explicitly says that Apple Silicon Macs cannot run it natively. GNU's command conditions · Memtest86+ platform support
Collect an address pattern, not an error-bit value
Prepare the diagnostic USB on a working computer, using a drive whose contents can be erased. Memtest86+ provides a Windows USB Installer for a FAT32 drive; on Linux, its ISO is written to the USB device with an image-writing utility. Select that USB in the faulty machine's startup boot menu, using the key specified in the PC manual. This runs the tester without loading the installed operating system. Memtest86+ downloads and USB instructions
Memtest86+ and PassMark's MemTest86 are separate tools. In Memtest86+, press F1 during the startup pause, enter configuration and select BadRAM patterns as the error reporting mode before running the tests. Patterns accumulate only while that mode is active. Switching to it after a long run in another display mode does not reconstruct every earlier error.
Record the complete output and the memory configuration used for the test. Output takes the form badram=F1,M1,F2,M2...: each F is a fault address and each M is an address mask. In 32-bit builds, the ordinary error display's Err Bits field describes incorrect data bits. It is not the address mask to paste into the configuration. PassMark offers a different collection route: MemTest86 Pro version 9 or later can generate BadRAM patterns in its HTML report. Memtest86+ operation and reporting · PassMark's report feature
A memory-test failure can involve the CPU, caches or motherboard as well as RAM. If the machine uses overclocked settings, retesting at standard settings helps narrow that diagnosis. Some tests cannot identify an exact faulty address, while bus stress or Rowhammer failures do not necessarily identify permanently defective cells. Memtest86+ therefore excludes those results from its BadRAM patterns. An empty pattern list does not cancel errors reported elsewhere in the test. Memtest86+ troubleshooting guidance
Read the mask before applying it
An address and mask are not a start address followed by a length. The mask selects address bits to compare with the base address. GRUB filters matching memory pages. A compact pattern can consequently cover healthy locations as well as observed faults; the amount excluded is not simply the number of bytes that failed.
PassMark's documentation example uses 0x7ddf0000,0xffffc000. Its displayed memory map contains 7ddf0000-7ddf3fff : RAM buffer. Counting both endpoints, that particular interval spans 0x4000 bytes, or 16KiB. It is an example for understanding the output, not a value for another computer. Other addresses can also match a mask, so one displayed interval does not establish the total memory lost on an arbitrary system.
Keep the order of every address/mask pair from your own test. Omitting a pair can leave its faulty locations available; broadening a mask can remove usable memory. If you have only individual failing addresses, rerun the tester in its pattern mode instead of borrowing a mask from an unrelated machine. GNU's page-matching definition
Apply the pattern for one boot
At the normal GRUB menu, interrupt the countdown with a key and press c to enter its command line. A password-protected menu requires authorization for command access. Where GRUB has a hidden-menu timeout, pressing Esc or F4, or holding Shift during that interval, can reveal the menu. GRUB menu controls · Hidden-menu behavior
Enter badram, a space, and your own comma-separated address/mask pairs. The shape is badram ADDRESS1,MASK1,ADDRESS2,MASK2; the capitalized words are placeholders. Remove the tester's badram= prefix when forming this command. It belongs at the GRUB prompt, not among Linux kernel parameters.
If GRUB accepts it without error, press Esc to return to the menu and select the existing Ubuntu entry. The filter continues into the kernel when it receives GRUB’s memory map. The configuration on disk has not changed, so another restart requires entering the command again or using the persistent setting below. A refused command or another crash stops this route before file copying can begin.
Booting the diagnostic USB and permitting this GRUB command are separate security decisions. Memtest86+ version 9 uses its own signing key, which the firmware must trust to boot it under Secure Boot. That does not establish whether GRUB lockdown permits badram.
Check the memory map, then make the setting persistent
Once Ubuntu starts, run sudo cat /proc/iomem. Compare the address intervals with the patterns from your test and look for the corresponding excluded regions. PassMark's example shows them as RAM buffer. A smaller total-memory figure alone cannot show whether the particular faulty pages were removed.
For subsequent boots on Ubuntu-based systems, edit /etc/default/grub with administrator privileges and set GRUB_BADRAM="ADDRESS1,MASK1,ADDRESS2,MASK2" using the actual pairs. A line beginning with # is a comment. Save the file, run sudo update-grub, reboot, and inspect /proc/iomem again. This checks the generated boot configuration rather than assuming the saved text took effect. Other distributions may use different regeneration commands.
The memory map answers whether addresses were excluded. It does not establish that files affected by earlier crashes are intact. Open recovered files on a healthy machine and compare them with existing backups where available. Preserve the originals while checking the copies.
What to ask a repair shop to quote
For continued use, ask whether the proposed work replaces identified faulty chips or circuitry, or replaces the entire motherboard carrying the soldered memory. Those are different scopes of work. Ask for the post-repair memory-test results and the warranty attached to that work; a bootable desktop alone does not demonstrate that the underlying fault is gone.
If only the data matters, request a separate assessment of storage access and the credentials needed to unlock it. That keeps the file-recovery job from depending unnecessarily on a full board replacement.