NVIDIA took two years to release its new series of graphic cards since its RTX series from 2018. With the introduction of ampere architecture last year by NVIDIA, gamers have waited for 3000 series for quite along. These graphic cards are built using ampere architecture and A100 supercomputing, AI graphic cards. Three graphics cards in RTX 3000 series are- RTX 3070, RTX 3080, and RTX 3090.
The similarities between these cards are that it is built using Samsung’s 8nm process technology with 28 billion transistors including many cores in them. These cards feature new shader cores, 2nd generation ray-tracing cores, and 3rd generation tensor cores. They use Microsoft’s new Directstorage API along with NVIDIA’s RTX IO which merges multiple storage technologies. Directstorage of Microsoft loads games directly into GPU without involvement of CPU in between reducing load times and improves the latency.
They have 1GB/s memory bandwidth by the use of GDDR6X from Micron. Dual axial fan design is available in them which gives them ability of 2x cooling performance. They all have HDMI 2.1 which allows it to pass 8k 60Hz and 4k 144Hz signals through single HDMI cable. They are the first discrete GPUs to feature AV1 codec with full support.
The first GPU from RTX 3000 series is 3070 which requires the power of 220W TDP. According to Nvidia, it is faster than RTX 2080 Ti at less than half the price. RTX 3070 comes with 8GB DDR6 VRAM and will be available in October for $499 (~NPR 58,415).
RTX 3080 comes with 30 TFLOPS of shader performance instead of 11 TFLOPS of RTX 2080. There are 58 TFLOPS of ray tracing cores and 238 TFLOPS of tensor cores comparing to 34 TFLOPS of ray tracing cores and 89 TFLOPS in RTX 2080. It has support for 10GB of GDRR6X memory which is more than 2GB of RTX 2080 which runs at 19Gbps. It needs total power of 320TDP. It has the ability to run games in 4k 60fps consistently. It will be available from 17th September at $699 (~NPR 81,828).
The next one to join the lineup in the series is RTX 3090 is a flagship chip of NVIDIA. It requires a total of 350TDP power that is obviously able to handle 24GB of GDDR6X memory. It can run the game in 8k 60fps. It is claimed by NVIDIA that this card is 50% faster than Titan RTX. It will be available from 24th September at the price of $1499 (~NPR 175481).
These cards are coming with three main software features which are reflex, broadcast, and Omniverse Machinima.
NVIDIA Reflex low-latency mode reduces latency by 50 percent in many games like Apex legends, Call of duty, Fortnite, Valorant. The reflex latency analyzer measures the time taken to see the changes in the screen when the input is given using hardware like a mouse.
NVIDIA Broadcast is a plugin that improves the microphone and webcam quality using AI effects accelerated by RTX. It is able to remove background without the use of green background and removes audio noises.
Omniverse Machinima lets you create characters, physics, face, and voice of your own using Webcam and AI of RTX if your game supports taking assets.