Prolific leaker kopite7kimi is continuing to update the alleged specs of the upcoming GeForce RTX 40 series cards. At the start of the month, kopite mentioned that the RTX 4080 will be based on the AD103 GPU, pack 9,728 CUDA cores, 16 GB of GDDR6X VRAM operating at 21 Gbps, and a 256-bit wide memory bus. The leaker also reported the TDP of the card to be 420 W. He then lowered the power draw prediction for the RTX 4070 and the RTX 4080 a few days later. Per the updated specs at the time, the RTX 4080 was slated to consume 320 W. The leaker has once again tweaked his RTX 4080 hardware specs.
Kopite now maintains that the RTX 4080 will have 16 GB GDDR6X running at 23 Gbps rather than 21 Gbps which was rumored last time, resulting in a total memory bandwidth of 736 GB/s. Similarly, the power draw prediction has also gone up a bit with the RTX 4080 now reported to have a total card power of 340 W. Aside from memory speed and power draw updates, kopite stands behind the rumored CUDA core count (9,728), bus width (256-bit), and memory size (16 GB GDDR6X).
Although the latest details have the RTX 4080 enjoying faster memory and consuming more power, the performance is unlikely to see a huge increase. Kopite suggested previously that the RTX 4080’s TimeSpy Extreme score is approximately 15,000, potentially edging out the RTX 3090 Ti.
Seeing that we are still some months away from the official launch of the GeForce RTX 40 “Lovelace” cards, take the hardware details with a grain of salt. Nvidia can change the specs of the RTX 40 boards on short notice.
Let's make some updates.
— kopite7kimi (@kopite7kimi) August 23, 2022
RTX 4080
PG136/139-SKU360
AD103-300-A1
9728FP32
16GB 23Gbps GDDR6X
total card power 340W