Tom's Hardware published a headline saying a voltage glitching hack turned Nvidia's CMP 170HX from an 8GB cryptomining GPU into a full-fat A100 with 80GB of VRAM. The claim is striking because it describes a cheap mining card being pushed into the memory range of a much pricier accelerator, with the numbers alone telling the story: $2,500 versus $20,000.
That is why people are looking at the Nvidia CMP 170HX now. The headline says the hack does not simply add memory; it carefully blows onboard fuses to unlock the entire VRAM and shader modules, which is the kind of change that alters how the card can be read by the system. In plain terms, the comparison is not about cosmetics. It is about a card that was sold with 8GB being described as if it could be opened up to 80GB through a hardware-level trick.
What is missing is the proof inside the body of the report. The supplied text contains only the headline and repeated access prompts, not the step-by-step method, the tools used, or any independent verification of how the voltage glitching worked. That leaves the central claim standing on its own: the headline asserts a dramatic unlock, but the mechanism behind it is not laid out in the material provided.
For now, the most important fact is the gap itself. If the fuse-blowing method really exposes the full memory array and shader modules, then the CMP 170HX is no longer just a mining GPU with a smaller footprint; it becomes a case study in how far hardware limits can be pushed. The next question is not whether the headline got attention. It is whether the claimed 80GB unlock can be demonstrated cleanly enough to hold up beyond a headline.
