The Revived RTX 3060 Costs More Than a Faster Card

Two months after the RTX 3060 12GB came back, it lists above a faster RTX 5050. NVIDIA’s $279 billion in purchase commitments explains where the memory went.

A relaunch is supposed to put a cheap card back on shelves. The revived GeForce RTX 3060 12GB has instead spent two months climbing past the newer and faster card sitting next to it. NVIDIA's own quarterly numbers explain the mechanism without setting out to.

The revived RTX 3060 12GB is now the wrong buy

MSI's Ventus 2X 12G OC sold for $329.99 on Newegg when RTX 3060 12GB stock returned to the US in early July. It lists at $479.99 now, about 45 percent up, per VideoCardz's tracking. Tom's Hardware's wider sweep puts ASUS's Dual V2 OC at $500 against a $330 launch price and Gigabyte's Windforce OC Rev 2.0 at $460 against $340 in late June, while the cheapest UK listing sits at £417 and the cheapest German one at €386, against roughly €330 a month earlier. Depending on which launch figure serves as the base, the increase runs somewhere between 35 and 60 percent.

The RTX 5050 settles it. Newegg lists a Gigabyte low-profile 5050 at $299.99, and the ComputerBase 1080p figures VideoCardz relays put the 5050 at 69.0 FPS average against 57.2 for the 3060, about 21 percent ahead. A five-year-old Ampere card is now roughly 60 percent more expensive than a newer card that is faster and carries Multi Frame Generation. Its one surviving argument is 12GB of GDDR6 against 8GB, which counts only where a card runs out of memory before it runs out of shader throughput: high texture settings at 1080p, some UE5 titles, small local model work. Anywhere else, a used or renewed card from the last two generations does the VRAM job for less money.

NVIDIA's commitments explain the cause

Long-term purchase commitments jumped from $119 billion in the previous quarter to $279 billion, and NVIDIA says the bulk of that increase is memory procurement, per Tom's Hardware on the results. Revenue for the quarter ending July 26 was $96.221 billion, with $89.023 billion of it data center hardware. Jensen Huang told analysts that customer forecasts point to demand doubling next year while the supply chain supports about 70 percent growth, and management expects supply to remain the constraint at least through FY2028.

A commitment that size is not inventory caution, it is a company reserving memory it expects to be scarce, years out. Consumer graphics buys what is left, which is the same mechanism behind a 2021 card at $500 and behind the state of the DDR5 aisle. Nothing in those numbers suggests entry-level GPU pricing corrects on a schedule NVIDIA controls, or wants to control.

Anyone who needs a card this month is choosing between an 8GB Blackwell part around $300 and a used 12GB or 16GB card from the previous two generations, because the relaunch that was supposed to sit between them is now beaten on both sides.

Sources