The rumours surrounding Nvidia's next-gun Ampere GPUs have really started to ramp up. First we saw the leaked images showing the alleged RTX 3080 cooler design, followed by multiple details about the RTX 3000 series and the speculation that these GPUs may feature a “traversal coprocessor”. Now, the latest information seemingly details some technical specifications of the RTX Titan, 3090, and 3080, while a 3DMark Time Spy score belonging to one of the Ampere GPUs has also been uncovered.
Starting with the rumoured specifications of the upcoming Nvidia Ampere GPUs, VideoCardz noticed that two different leakers have shared very similar specifications for these upcoming GPUs. Both @KkatCorgi and @kopite7kimi tweeted what they believe to be the technical specifications of the RTX 3080, 3090 and the new Titan graphics cards.
The RTX 3080 is reportedly using the GA102-200-Kx-A1 GPU, with 4352 CUDA cores and 10 GB of VRAM at 19Gbps, while the RTX 3090 is expected use the GA102-300-A1 GPU, packing 5248 CUDA cores and 12GB of GDDR6X RAM at 21Gbps. The upcoming Titan GPU should utilise the GA102-400-A1, with 5376 CUDA cores and a massive 24GB of VRAM at 17Gbps. The type of VRAM (GDDR6/GDDR6X) used for both the RTX 3080 and the upcoming Titan was not shared.
Image Credit: HardwareLeaks
Regarding the leaked 3DMark Time Spy score of an “Unknown Nvidia Ampere GPU”, rogame from HardwareLeaks spotted a “hidden” database entry detailing an Ampere GPU with a 1935MHz boost clock and 6000MHz memory clock. One thing to note is the odd “6000MHz memory clock” listed in the 3DMark Time Spy database. GDDR6 and HBM2 memory clock values are usually multiplied by 8, which will then result in the “real” memory clock. As per rogame, the displayed 6000MHz memory speed might be “the result of an early/internal driver or a new type of memory that is still unknown to 3dmark detection software”.
This “Unknown Nvidia Ampere GPU” recorded a graphics score of 18257, as seen above. This translates to an improvement of over 30% compared to the stock RTX 2080Ti FE, or 22% when compared to the Nvidia Titan RTX. HardwareLeaks also included a score achieved by KINGPIN through overclocking an EVGA RTX 2080Ti XC with LN2, reaching a 2385MHz boost clock and a 17360Hz memory clock, which beats the Ampere GPU's score by about 2%.
Rogame confirmed the authenticity of this entry, claiming that it belongs to an Nvidia employee. The 1935MHz boost clock may not seem like a big improvement over the current generation of GPUs, but this entry is likely to be an engineering sample so clock speeds could be adjusted between now and the official RTX 3000 launch.
Discuss on our Facebook page HERE.
KitGuru says: What do you make of this Time Spy score? If it belonged to the RTX 3090, would you be happy with a 30% performance improvement over the RTX 2080 Ti?