If you’re a fabless company that relies on external chip manufacturing from companies like TSMC, you’re forced to compete for limited wafer starts, particularly on the bleeding edge nodes. The use of a smaller node dramatically alters the fundamental performance capabilities and even the essential viability of your chip. That's why industry giants like Apple, Qualcomm, AMD and Nvidia all spend billions to secure their supply. If they don’t pay up, then there's no (or fewer) chips for you!
According to Hardware Times, Nvidia is set to pay the exorbitant amount of nearly $10 billion to secure its share of TSMC 5nm capacity. Nvidia is geared up to produce its next generation Ada Lovelace GPU on TSMC's N5 node. Of course this next gen GPU family is likely to be referred to as the RTX 40 series. No one outside of the management of both companies knows extactly how the fees break down or the timeframe, but as noted by Hardware Times, Nvidia spent $9 billion in just the third quarter on inventory and prepayment for future products. This would indicate that Nvidia is already paying through the nose well in advance.
Securing 5nm production is critical to the success of the RTX 40 series. A node shrink can deliver a combination of higher clocks and better power efficiency, and this can make the difference between winning or losing a head to head battle. Though oversimplifying things, It may be that Nvidia’s use of Samsung 8nm is a reason that its products cannot clock as high as 7nm products from AMD, despite using more power.
The other advantage the use of a smaller node offers is better yield. A 300mm wafer can fit more chips on it if the chips themselves are smaller. More chips per wafer equals more profit.
How to buy a
Read more on pcgamer.com