A new leak has revealed the memory bus configurations for the upcoming Nvidia GeForce RTX 50 series of graphics cards.
Nvidia has remained enigmatic over the specifics of what we can expect for the upcoming RTX 50 ‘Blackwell’ GPUs. Most of the information we have comes from industry insiders.
This trend continues with the latest from reliable leaker Kopite7Kimi, who has again provided new information on what we can expect from the RTX 50 series.
Kopite7Kimi provided a list of the five processors for the Blackwell RTX 50 series of graphics cards, along with the bit width of the memory bus and the memory type. The list is as follows:
| Processor | Config | Bus Bit width | Memory type |
| GB202 | 12 x 8 | 512-bit | GDDR7 |
| GB203 | 7 x 6 | 256-bit | GDDR7 |
| GB205 | 5 x 5 | 192-bit | GDDR7 |
| GB206 | 3 x 6 | 128-bit | GDDR7 |
| GB207 | 2 x 5 | 128-bit | GDDR6 |
The GB202 is believed to be the processor that will power the flagship RTX 5090. It will feature 12 GPCs and 8 TPCs per GPC.
Subscribe to our newsletter for the latest updates on Esports, Gaming and more.
This is estimated to result in a total SM count of 192. However, it is likely that Nvidia will cut the memory bus into different configurations depending on the SKU, such as the previously suggested 448-bit bus.
The GB203 strongly resembles the current generation AD103, featuring the same core count and memory bus, only with the added advantage of GDDR7 memory support.
Somewhat surprisingly, Nvidia has allegedly chosen to stick with GDDR6 for the GB207. This is expected to be the chip for the RTX 5060, and this configuration would put it on par with the AD106 from the RTX 40 Ada Lovelace series.
The 60-suffix cards are usually popular in the consumer market, but if this leak is accurate, it may be difficult to justify upgrading to an RTX 5060 from an RTX 4060. Remember, while these leaks come from a reliable source, they could be subject to change until Nvidia officially announced the chips themselves.