Is this a hardware or a software issue?

Hardware failures can look like either layer depending on the symptom. The discriminator is the resolution of the symptom to the recommended fix: if applying the steps under "How to fix it" eliminates the symptom, this is a software/configuration class issue regardless of the hardware it surfaced on. Repeat the fix on a fresh node to confirm.

How long does the fix typically take?

Most resolutions from this entry take 10 to 30 minutes of engineer time once the diagnosis is confirmed. Long fixes come from skipping the diagnosis step and chasing apparent symptoms (validation drift, memory growth, slow throughput) instead of the actual root cause listed above. Apply the fix in a small experiment first, then roll it forward.

Does this affect inference or only training?

Primarily this is a hardware failure observed during training, but the same misconfiguration can leak into downstream evaluation and serving. Treat the prevention checklist as a guardrail: any pipeline that touches a trained checkpoint should re-validate the failure mode before promoting the model.

What's the cheapest way to confirm this is the problem?

Cheapest signal: reproduce the symptom on a 1-GPU toy run with the same configuration. If the toy run exhibits the same behavior, the cause is the configuration rather than the cluster. If only the multi-GPU run is affected, the failure is a distributed interaction, so PyTorch is the most likely layer to inspect.

Can the Denpex platform detect this automatically?

Yes. Denpex correlates the symptoms in this entry against your training logs in real time and surfaces the failure taxonomy, root cause, and exact remediation command on the console. Three full diagnoses are included free during signup, and the encyclopedia entry you are reading is generated from the same signal library that powers the platform.

What related failures should I check next?

Closely related failures cluster around #Silent Data Corruption. After applying this fix, validate against the related entries below to confirm no overlapping cause is masking as this one. Engineers routinely discover that the dominant failure mode is actually a second, sibling failure compounded on top.

Will this recur after the fix?

Applied once with the prevention checklist, this failure does not recur under normal operation. The recurring variant most engineers encounter is the re-emergence of the same symptom on a different model or stack after Silent Data Corruption is upgraded. Pin compatible versions and rerun the prevention smoke test on every dependency change.

Is there a fingerprint I can monitor?

Yes. After days or weeks of stable training, the loss curve suddenly spikes or diverges to NaN. is the highest-signal fingerprint. Wire a metric or alert on that signal (or its absence after the fix) so you can detect the recurrence automatically. The Denpex platform encodes this exact fingerprint and surfaces it directly in the dashboard.

What is the official documentation that covers this?

Reference documentation is listed in the References section at the bottom of this entry. The most authoritative URLs are the upstream framework docs and the issue trackers cited; the Denpex entry synthesizes both into the executive, engineer, and on-call perspectives so each audience can act without reading every linked source.

Does this entry include working code?

Yes. The "Code Examples" section contains copy-paste-ready snippets that target the most common stacks (PyTorch, Hugging Face Trainer, DeepSpeed, Megatron-LM, vLLM/Transformers). Each snippet is annotated with why it works so you can adapt the pattern to your own framework without losing fidelity.

HBM Silent Data Corruption leading to Loss Spikes

Q: What does this failure look like in production?

In production this failure typically appears during the first tens-to-hundreds of training steps: loss curves become non-monotonic, validation metrics plateau above the baseline, and downstream checkpoints either diverge or fail to improve. Engineers frequently misread it as data drift when the underlying mechanism is a transient hardware fault occurred in an unprotected part of the gpu memory pipeline (e.g., a multi-bit flip that bypassed ecc, or a fault in the l1 cache/sram). this resulted in an incorrect matrix multiplication output (e.g., a massive floating-point value). because it wasn't caught by hardware parity checks, the bad data silently poisoned the gradients.. Validating against the symptom list in this entry is the fastest way to confirm the diagnosis.

Q: How does this scale from 1 GPU to 1000 GPUs?

At single-GPU scale the symptom usually looks like slow convergence or a single anomalous loss value. At 64+ GPU scale, distributed interactions amplify a transient hardware fault occurred in an unprotected part of the gpu memory pipeline (e.g., a multi-bit flip that bypassed ecc, or a fault in the l1 cache/sram). this resulted in an incorrect matrix multiplication output (e.g., a massive floating-point value). because it wasn't caught by hardware parity checks, the bad data silently poisoned the gradients. through collectives, and the failure can manifest as a full job abort. Always reproduce the fix at the smallest cluster configuration that still exhibits the symptom before promoting to production scale.

A transient hardware fault occurred in an unprotected part of the GPU memory pipeline (e.g., a multi-bit flip that bypassed ECC, or a fault in the L1 cache/SRAM). This resulted in an incorrect matrix multiplication output (e.g., a massive floating-point value). Because it wasn't caught by hardware parity checks, the bad data silently poisoned the gradients.

Quick answer

A transient hardware fault occurred in an unprotected part of the GPU memory pipeline (e.

Symptom: No Xid errors, CUDA errors, or kernel panics are logged in the system.
Root cause: A transient hardware fault occurred in an unprotected part of the GPU memory pipeline (e.g., a multi-bit flip that bypassed ECC, or a fault in the L1 cache/SRAM).
Recommended fix: Roll back to a known-good checkpoint N/A Once model weights are poisoned by an SDC event, they cannot be recovered. The only solution is to load the last good state.
How Denpex helps: Denpex investigates HBM Silent Data Corruption leading to Loss Spikes using the evidence you provide or your connected workload collects. Earlier rank, host or application evidence is needed to distinguish an initiating failure from a downstream report.

Hardware#Silent Data Corruption

What this failure is

HBM Silent Data Corruption leading to Loss Spikes is a Hardware failure seen during ML training runs. A transient hardware fault occurred in an unprotected part of the GPU memory pipeline (e.g., a multi-bit flip that bypassed ECC, or a fault in the L1 cache/SRAM). This resulted in an incorrect matrix multiplication output (e.g., a massive floating-point value). Because it wasn't caught by hardware parity checks, the bad data silently poisoned the gradients. Common tags: Silent Data Corruption.

Live diagnosis, no signup

Is this what broke your run? Paste your log.

You're reading about HBM Silent Data Corruption leading to Loss Spikes. Paste your own traceback and relevant evidence for an investigation of your workload, with a next action or a specific missing fact. A reference entry does not establish your cause. No account or card for the free diagnosis. Review data handling before submitting sensitive logs.

Before uploading, review cloud data handling and local options.

training_logs.txt

No log to hand? Try one:

Upload or drop a plain-text log

Upload your log file. Large files are processed in your browser. Only selected evidence is submitted when you run diagnosis.

3 free diagnoses/day

Why it happens (the mechanism)

ML engineers will typically assume this is an algorithmic instability, bad data batch, or an excessively high learning rate. They will spend weeks tweaking gradient clipping or analyzing the dataset when the root cause was a transient hardware bit-flip.

What you'll observe

Loss sudden spike to NaN
Gradient norm explosion without any hyperparameter changes

Common symptoms and what they mean

Symptom	Why it happens
After days or weeks of stable training, the loss curve suddenly spikes or diverges to NaN.	A transient hardware fault occurred in an unprotected part of the GPU memory pipeline (e.g., a multi-bit flip that bypassed ECC, or a fault in the L1 cache/SRAM). This resulted in an incorrect matrix multiplication output (e.g., a massive floating-point value). Because it wasn't caught by hardware parity checks, the bad data silently poisoned the gradients.
Checkpoints saved after the spike contain poisoned weights.	A transient hardware fault occurred in an unprotected part of the GPU memory pipeline (e.g., a multi-bit flip that bypassed ECC, or a fault in the L1 cache/SRAM). This resulted in an incorrect matrix multiplication output (e.g., a massive floating-point value). Because it wasn't caught by hardware parity checks, the bad data silently poisoned the gradients.
No Xid errors, CUDA errors, or kernel panics are logged in the system.	A transient hardware fault occurred in an unprotected part of the GPU memory pipeline (e.g., a multi-bit flip that bypassed ECC, or a fault in the L1 cache/SRAM). This resulted in an incorrect matrix multiplication output (e.g., a massive floating-point value). Because it wasn't caught by hardware parity checks, the bad data silently poisoned the gradients.

Which systems are affected

PyTorch
Megatron-LM
DeepSpeed

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.

✓Resume training from the last known good checkpoint using the exact same data seed. If it passes the problematic step smoothly, it was likely SDC.
✓Run PyTorch CP-Bench or similar deterministic matrix multiplication verification suites on the cluster to identify 'silent' faulty GPUs.

Root cause

A transient hardware fault occurred in an unprotected part of the GPU memory pipeline (e.g., a multi-bit flip that bypassed ECC, or a fault in the L1 cache/SRAM). This resulted in an incorrect matrix multiplication output (e.g., a massive floating-point value). Because it wasn't caught by hardware parity checks, the bad data silently poisoned the gradients.

The fix and how to prevent it

Searchable error signature

search key

No Xid errors, CUDA errors, or kernel panics are logged in the system.

Use this text as a lookup key in logs and upstream issue trackers. It is not presented as a captured customer log. Confirm the cause from your own preceding events, versions, configuration and the cited references.

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.

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

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

References

↗Silent Data Corruption in ML Training

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.

Diagnose your logs free Compare a solved incident Browse all failures

Related Hardware errors

Xid 149 NETIR_LINK_DOWN on Blackwell: the NVLink fault that Xid 74 knowledge misses

Hardware · critical

NVSwitch SXid errors: what aborts your CUDA jobs and what is only informational

Hardware · critical

GPU Memory Clock Throttle

Hardware · medium

NVLink Error

Hardware · high