The new Ampere GPUs from Nvidia are here and double the AI-supported functions for gamers and content creators. Here is the ball.
After weeks of back and forth and reviewing past product launches, Nvidia finally announced the 3000 series GPUs via a live stream on September 1st.
Nvidia announced three new GPUs during its live streaming event, with the flagship $ 699 RTX 3080 leading the way. The cheaper RTX 3070 offers slightly better performance than Nvidia's best GPU right now – the RTX 2080 Ti – for an incredibly low price of $ 499. After all, the RTX 3090 is a giant, which Nvidia calls "BFGPU" (understood?) And costs an incredible 1499 US dollars. The RTX 3090 is capable of playing 8K games at 60 fps, a previously unheard of and utterly ridiculous combination of resolution and frame rate.
The new GPU range from Nvidia uses the second generation RTX Ampere architecture, which promises faster clock rates, better cooling, advanced AI-supported functions and more efficient ray tracing functions. The live streaming event, presented by CEO Jen-Hsun "Jensen" Huang from his kitchen, mainly focused on the gaming aspect of the new GPUs, with Jen-Hsun hinting at the role of gamers in developing GPUs. Jen-Hsun said:
It's the gamers and their insatiable demand that are the driving force behind the GPU.
All three Founders Edition versions of the GPUs use a new two-axis flow-through cooling system designed to keep the GPU cooler while remaining quieter than the previous models. Ampere GPUs use second generation ray tracing cores and third generation AI tensor cores. The GPUs also use GDDR6-VRAM and the new RTX I / O function, with which Nvidia claims to be able to load game assets directly from the memory into the VRAM and reduce the CPU load when accessing the memory up to twenty times. The specs for each of the GPUs vary, but here is the full data sheet.
Why this is great news for creatives
While the Ampere GPUs are certainly ideal for games thanks to ray tracing – ray tracing makes light in a realistic, lifelike way – they are also ideal for creative people who work in the areas of video editing, visual effects, animation, architecture and other fields, computer graphics require. Video editors using the paid version of DaVinci Resolve or Premiere Pro can leverage a powerful GPU to render videos and support other in-software tasks through hardware acceleration.
While Nvidia professional Quadro GPUs are best suited for creative projects, this new line of consumer ampere GPUs offers incredible performance at a competitive price. The Beat recently published an article on building video editing PCs. We recommended Nvidia's current flagship, the RTX 2080 Ti, in our high-end video editing PC. The RTX 2080 Ti is Adobe approved for use in Premiere Pro and After Effects, making it a reliable option for GPU-accelerated rendering. The problem is that it is too expensive and costs over $ 1000. However, the new, cheaper RTX 3070 for $ 499 offers better performance at less than half the price. When the RTX 3070 goes on sale in September it will be possible to build a great video editing PC for a lot less.
The above video was not pre-rendered. It's a demo that uses real-time ray tracing to simulate the lighting. A previous demo showed the same gameplay running on the current hardware, and the differences are appalling. It's only a matter of time before developers use these GPUs to their full potential to create amazing games, animated movies, and more.
According to Nvidia, the RTX 3000 series GPUs can achieve 3D render times that are up to two times faster than current Nvidia GPUs. For artists and animators with visual effects who use 3D rendering software like Blender Cycles and Maxon Cinema 4D, Nvidia promises to cut waiting times. SheenCity Mars and D5 Render – 3D rendering software for architecture – use Nvidia's new DLSS 2.0 functionality. Deep Learning Super Sampling is a deep learning neural network that RTX GPUs, through their tensor cores – AI processors – can use to render sharp images and high frame rates.
Further functions announced
While many of the features announced at the event were aimed at gamers, some features stood out for content creators. Nvidia Broadcast offers streamers and home workers the opportunity to upgrade their cameras and microphones with various streaming and video conferencing software. Omniverse Machinima allows developers to create their own stories using game assets.
Streamers can now get more out of their setup with Nvidia's broadcast software that uses AI to enhance video and audio. The Nvidia Broadcast App is a plugin that integrates with streaming software like OBS, Twitch Studio, and even video conferencing apps like Zoom.
When streaming or video conferencing, a quiet environment is required, but not always possible. Nvidia Broadcast introduces Noise Removal, a feature that AI uses to remove background noise. In the video above, you can hear the blow dryer go quiet instantly when Noise Removal is on. With noise cancellation enabled, babies crying, barking dogs, and other background noises will no longer interrupt stream or video calls.
The virtual background is another useful feature. Zoom uses similar technology to remove and replace a person's backgrounds. However, Nvidia Broadcast's virtual background uses AI to change the background. This feature not only removes or replaces the background, it can also blur the background. This look is only possible with expensive cameras with large aperture lenses. It's almost impossible to tell the difference between using the virtual background and using a green screen to get the same effect.
Finally, Auto Frame tracks a user's head as they move around the screen, creating the illusion that the camera is moving. The software zooms in on the subject and makes it possible to follow head movements from side to side. Users who enable this will stay in the picture even if they reach for a drink or move away from the computer.
Nvidia Broadcast will be available for download later in September for users with Nvidia RTX GPUs.
Omniverse Machinima is a new tool developed by Nvidia that allows users to create videos using assets from video games. Users can import characters, backgrounds and other elements from video games and animate them to create stunning videos. This type of filmmaking is known as machinima. And while this has been around for years, the tools in Omniverse Machinima will make it easier for others to learn these skills.
The software uses ray tracing, physics, and character animation in real time for a similar simulation. Users can use the software's facial animation AI to import a voice track and fully animate a character's mouth. By capturing a user's movement via webcam, the AI can animate a character's movement. Omniverse Machinima integrates with software such as Maya, Unreal Engine, Photoshop and others. The software is not yet available, but potential users can request early access.
Ampere breaks expectations
A third-party GPU maker may have accidentally leaked the specs of Nvidia's new Ampere GPUs prematurely, but the final reveal seems worth it. Both the RTX 3070 and 3080 are affordable compared to last-generation RTX GPUs and significantly more powerful than even the most powerful consumer GPU currently available.
The datasheet for Ampere GPUs is impressive, but the multitude of other features like Nvidia Broadcast and Omniverse Machinima shows that Nvidia puts the developers at the forefront. While gamers are sure to get a lot out of these formidable GPUs, developers will find ways to push the limits of animation, video production, and 3D rendering.
Until reviewers get their hands on the Ampere GPUs, we don't know how good they actually are, but Nvidia seems to be convinced of their products. Jen-Hsun was referring to a future where holodecks and spaceships will become a reality at the speed of light:
In this future we will look back and find that it started here.
Nvidia and third-party manufacturing partners will be releasing the RTX 3080 and 3090 GPUs in the coming weeks. The affordable yet incredibly powerful RTX 3070 will hit shelves in October.
Cover picture via sdecoret.