I found 146W of "ghost power" on NVIDIA A100 at 0% utilization. Here's my API to detect it
A recent study revealed that NVIDIA A100 GPUs consume significant power even at 0% utilization, a phenomenon termed 'ghost power.' The tests showed that while standard monitoring tools indicate idle status, the GPUs can draw between 67 to 146 watts. This unnoticed power consumption could lead to substantial financial waste across data centers globally.
- ▪The A100 GPU consistently measured a power draw of 67 to 146 watts while reporting 0% utilization.
- ▪The study found that $58.70 per GPU per year is wasted due to this ghost power, amounting to $58 million annually for one million GPUs.
- ▪No ghost power was detected on the H100 GPU across multiple tests.
DEV.to (Top) files mainly under programming. We currently carry 4,877 of its stories.
Opening excerpt (first ~120 words) tap to expand
try { if(localStorage) { let currentUser = localStorage.getItem('current_user'); if (currentUser) { currentUser = JSON.parse(currentUser); if (currentUser.id === 3935300) { document.getElementById('article-show-container').classList.add('current-user-is-article-author'); } } } } catch (e) { console.error(e); } mikebains41-debug Posted on May 16 I found 146W of "ghost power" on NVIDIA A100 at 0% utilization. Here's my API to detect it #api #infrastructure #monitoring #showdev I ran 35 controlled energy tests on NVIDIA A100 and H100 GPUs on RunPod. Standard monitoring tools missed something critical. When nvidia-smi reports 0% GPU utilization, data center operators assume the GPU is idle. Billing stops. Cost tracking stops. The power draw does not stop.
…
Excerpt limited to ~120 words for fair-use compliance. The full article is at DEV.to (Top).
Discussion
0 commentsMore from DEV.to (Top)
