It’s been a while since the release of the RTX 40 series. The RTX 40 series was not that much of a hit. It had its downsides and pricing issues. But nonetheless it’s hopeful that Nvidia will redeem itself once again with the release of the RTX 50 series.
The latest RTX 50-series is based on Blackwell GPU architecture. Even though Nvidia didn’t provide any insights or information about their next gen cards, rumors and speculation had been running riff. Here we will get to know about the rumors and leaks about the upcoming 50 series.
Blackwell Architecture
The architecture is to have FP4 and FP6 support. Nvidia is sticking to the 4nm technology from TSMC with large L2 cache as well as OFA for DLSS 3. They might make a big improvement in the TeraFLOPS like we saw in the 4090 compared with the 3090.
The 4nm tech will use two identical chips by a 10TB/s NV-HBI connection for dual purpose solution. Even though this 4nm technology is meant for commercial usage, the consumers will see similar chips as they don’t vary much from the commercial ones like we have observed in the previous generations.
GDDR7
According to the rumors, Nvidia will ship its newer chips in GDDR7 memory format. Previously they used GDDR6X and GDDR6 which had a speed of 17 to 23 Gbps. GDDR7 sets to break the record with a whopping 36 Gbps speed which is almost 50% faster than GDDR6X memory.
It’s also rumored that GDDR7 will be capable of 40 Gbps transfer. But Nvidia tends to set the speed of their GPU lower as they clocked the 24 Gbps chips from Micron to 21 Gbps in RTX 4090 cards. So it can be speculated that they will set the GDDR7 speed to 32 Gbps or 34 Gbps.
It is rumored that they will ship the cards with a 384 bit interface which will provide a whopping 1728 GB/s bandwidth. Which is more than 70% bump from the RTX 4090. Moreover the 128 bit cards will get 576 GB/s of bandwidth.
The GDDR7 memory will also solve the capacity problem as it will allow for more capacity. It will pack 16 Gb memory as well as 24 Gb. A 512 bit interface will enable a 32 GB RTX 4090. Subsequently they will be able to pack 12 GB on a 129 bit interface.
Possible Specs
RTX 5090 will have 512 bit or 384 bit memory with 128 MB L2 cache. A total of 32 GB of VRAM will be present in the clocking at 2500 MHz. In the case of RTX 5080, it will have 256 bit memory with 64 MB L2 cache. Besides that a total of 24 GB VRAM clocking at 2500 MHz. The RTX 5070 will have 192 bit memory with 18 GB of VRAM. It will have a 48 MB L2 cache.
Launch Date
Nvidia released its last gen RTX 40 series in October of 2024 while revealing new architecture in Marh of 2022. The RTX 30 series appeared in September of 2020 also the RTX 20 series was released in September of 2018. So given the time frame of previous releases we can conclude that the new RTX 50 series will be released somewhere from September to November of this year. Later at the beginning of next year they will release the mid range 5070. Soon after that they will release the starter range and the Ti versions by the end of next year.
Price
The price of the new RTX 50 series GPU will be top grade as they will release the expensive RTX 5090 and 5080. Initially the RTX 5090 will cost nearly $1,999. The RTX 5080 will cost $999. Depending on the memory size the price will vary. Later when the RTX 5070 releases it will have a price tag of $599. The starter RTX 5600 will have a price of $449. Lastly, the basic card on the list RTX 5500 will cost $299.
The RTX cards are becoming more potent with ray tracing becoming heavy in the gaming industry. The developers are taking it seriously as we can see it in the latest Alan Wake game. It will continue to become dominant in the upcoming titles. Nvidia wants to keep its GPU lineup intact and keep on dominating the consumer GPU market. Consumers are willing to fork an extra for their favorite gaming rig if they are able to get the performance they hope for.