Skip to content

GPU Uncorrectable ECC Memory Error (Xid 48/94/95)

GPU HBM memory experienced an uncorrectable ECC error (double-bit error) that the hardware cannot fix. This indicates physical HBM memory degradation requiring GPU drain and potential RMA.

Quick answer

GPU HBM memory experienced an uncorrectable ECC error (double-bit error) that the hardware cannot fix.

GPU Memory#gpu#ecc#memory#hbm#xid#hardware

What this failure is

GPU Uncorrectable ECC Memory Error (Xid 48/94/95) is a GPU Memory failure seen during ML training runs. GPU HBM memory experienced an uncorrectable ECC error (double-bit error) that the hardware cannot fix. This indicates physical HBM memory degradation requiring GPU drain and potential RMA. Common tags: Gpu, Ecc, Memory, Hbm.

Live diagnosis, no signup

Is this what broke your run? Paste your log.

You're reading about GPU Uncorrectable ECC Memory Error (Xid 48/94/95). Paste your own crash log or traceback below and get the real root cause for YOUR run, not this generic entry. No account, no card. Logs are masked at ingress and never saved to account history.

training_logs.txt
No log to hand? Try one:

3 free diagnoses/day

Want 14 days on the Scale plan?

Request an evaluation code. A verified workplace organization activates up to 50 diagnoses a day, alerts, history, and follow-up questions. No credit card or automatic subscription.

Evaluate one incident

Why it happens (the mechanism)

HBM3 memory cells wear out under sustained high temperature and voltage stress, transitioning from correctable (single-bit) to uncorrectable (double-bit) ECC errors. Manufacturing defects in HBM stack may cause early-life uncorrectable ECC failures. Temperature-induced voltage/frequency scaling amplifies timing margin violations in HBM I/O. Taken together, these mechanisms explain why the failure is reproducible, why it tends to surface on specific workloads or scales, and why generic mitigation attempts often fall short without addressing the underlying cause.

What you'll observe

  • GPU memory controller detected a double-bit error that ECC cannot correct
  • Xid 48, 94, or 95 logged to dmesg
  • GPU may trigger row remap to map out defective memory cells
  • Persistent uncorrectable ECC errors indicate physical HBM wear-out

Common symptoms and what they mean

SymptomWhy it happens
dmesg shows Xid 48, 94, or 95 with ECC error contextHBM3 memory cells wear out under sustained high temperature and voltage stress, transitioning from correctable (single-bit) to uncorrectable (double-bit) ECC errors
nvidia-smi -q -d ECC shows non-zero Aggregate Uncorrectable errorsManufacturing defects in HBM stack may cause early-life uncorrectable ECC failures
GPU memory bandwidth may degrade as bad rows are remappedTemperature-induced voltage/frequency scaling amplifies timing margin violations in HBM I/O
NCCL timeouts or CUDA errors may follow as downstream effectsHBM3 memory cells wear out under sustained high temperature and voltage stress, transitioning from correctable (single-bit) to uncorrectable (double-bit) ECC errors

Which systems are affected

  • NVIDIA H100/H200/B100/B200 GPUs with HBM3/HBM3e memory
  • GPUs under sustained high thermal load (near 700W power limit)
  • Long-running training jobs (>7 days continuous)
  • Multi-GPU nodes where one GPU failure drains the entire node

How to confirm this is the problem

Use this checklist to test the hypothesis against a small reproduction. No single line proves the root cause, so preserve the preceding events and compare one variable at a time.

  • Reproduce the failure from a clean checkpoint/seed: the symptom must appear without warm-up state from a previous run.
  • Verified signal present: dmesg shows Xid 48, 94, or 95 with ECC error context
  • Verified signal present: nvidia-smi -q -d ECC shows non-zero Aggregate Uncorrectable errors
  • Verified signal present: GPU memory bandwidth may degrade as bad rows are remapped
  • Verified signal present: NCCL timeouts or CUDA errors may follow as downstream effects
  • A targeted fix from the "How to fix it" section eliminates or substantially reduces the symptom within one validation pass.

The fix and the prevention pattern

The root cause is on this page and stays free. A free account adds the exact remediation steps, saved history, and the fix on every entry in the encyclopedia.

Sign up free. Unlock the full analysis

No credit card. Daily allowance follows verified trust tier. Instant access.

Xid 48 in context

Xid 48 is one of a small set of codes the NVIDIA driver uses to report GPU faults, and the number is most of the diagnosis: it tells you whether you are looking at your own code, the driver, or a board that needs replacing.

Compare every Xid code side by side

Diagnose this failure in VS Code

Select the traceback or open the failed terminal, then run Denpex locally to see the initiating rank, collateral failures, exact fix, and verification command without uploading the log.

Install the free VS Code extension

Root cause

  • HBM3 memory cells wear out under sustained high temperature and voltage stress, transitioning from correctable (single-bit) to uncorrectable (double-bit) ECC errors
  • Manufacturing defects in HBM stack may cause early-life uncorrectable ECC failures
  • Temperature-induced voltage/frequency scaling amplifies timing margin violations in HBM I/O

The fix and how to prevent it

Evaluate Denpex on your own logs

Request a Scale evaluation code. A verified workplace organization activates 14 days with up to 50 diagnoses a day. Every account keeps its current diagnosis allowance and gets a verification path. No card or automatic subscription.

We send a single-use code tied to that address. Static provider and TLD rules do not reject valid addresses. Account trust determines the benefit after signup.

Don't just read the fix, diagnose your run

The encyclopedia tells you what went wrong. Denpex tells you what went wrong in YOUR training run. With your logs, your config, and your stack.