A100 second-hand prices fall 20% to 30% amid Blackwell and Rubin pressure
11/12/2026 — 11/13, 12:16·1 sources (all)·2 reports (all)
Key takeaways
- A Reddit post reports that A100 GPU resale prices have dropped 20% to 30% recently, driven by the arrival of high-efficiency hardware built…
Background & analysis
On November 12, 2026, a post appeared on Reddit's LocalLLaMA community relaying the experiment log of a developer named Kate. According to the writeup, an 8GB CMP 170HX actually ships with roughly 80GB of HBM onboard, but much of it is disabled through a fuse lock; earlier software-only unlocks had only allowed access to 64GB. Kate reportedly bypassed the fuse lock using voltage fault injection, after which she enabled all the working SMs, counting 120 to 126 in total, versus the 108 NVIDIA enables on a standard A100. In her measurements, the modified card reached 85.9 TFLOP/s in FP16, above the 72.5 TFLOP/s of a retail A100, with the full 80GB usable for inference. NVLink and PCIe Gen4 support were said to come back as well. The post stressed that this is not a casual weekend project: it requires reworked board wiring, a custom STM32 glitcher, and glitching sessions that can take anywhere from 1 to 48 hours. So far the claims rest on this single writeup, with no vendor response or independent verification available yet.
AI-generated from 1 reports · updated 1 day ago
Latest turnAccording to a Reddit user report, secondhand A100 GPU prices have fallen 20% to 30% as sellers rush to offload inventory. The drop is driven by the energy efficiency of newer Blackwell and Rubin architectures and limited inference demand for older cards.
- Heat index
- 76
- All sources
- 1
- All reports
- 2
- First seen
- 1 day ago
Reports on this story headlines open the original
According to a Reddit user report, secondhand A100 GPU prices have fallen 20% to 30% as sellers rush to offload inventory. The drop is driven by the energy efficiency of newer Blackwell and Rubin architectures and limited inference demand for older cards.
Reddit · LocalLLaMAAI score 56
A Reddit post relays Kate's writeup describing how she used voltage fault injection to bypass the fuse lock on an 8GB CMP 170HX, exposing all 80GB of HBM and 120-126 SMs. Her numbers reach 85.9 TFLOP/s FP16 versus 72.5 on a retail A100, with NVLink and PCIe Gen4 restored, though the mod needs board wiring, a custom STM32 glitcher and 1 to 48 hours of work.
Reddit · LocalLLaMAAI score 52
Other stories people are talking about
- Heat index 2236Nvidia stock plunges after Jensen Huang resigns, with retail investors buying the dip35 sources
- Heat index 1463RisingGoogle Opens Gemini 4 Hydrogen to All Users as High-Speed Model21 sources
- Heat index 1429SurgeJensen Huang exits as NVIDIA CEO, leadership vacuum lingers16 sources
- Heat index 785SurgeAnthropic hires gas supply lead to power data center expansion after IPO10 sources
- Heat index 673SurgeMicrosoft races to fix global outage hitting Outlook, Teams and Copilot8 sources
How is heat calculated?About the methodHide
Heat counts how many independent sources covered a story in the last 48 hours: one source counts once no matter how many posts it published, decaying with a 24-hour half-life. What ranks first is what many people are talking about.
This page aggregates public feeds. Headlines and summaries are machine-organized and remain the property of the original authors; verify important facts at the source. If you believe a headline or summary infringes your rights, email the contact address in the footer with the page URL and basis of your claim, and we will remove or replace it after review.
- Surge
- Discussion rising fast
- New
- First report within 6 hours
- Rising
- Still gathering discussion