Last night, NVIDIA launched their new GPU platform they nickname the Ampere. They also tout that this is their greatest generational leap in performance ever. Welcome to the NVIDIA’s second-generation Ray Tracing GPUs.
For now, they have announced the availability and launch of three new GPUs that is meant to blow even the fastest GeForce RTX 20 series GPU, the GeForce RTX 2080Ti away. We are only talking about the GeForce RTX 3070 mind you. There is still the high-end GeForce 3080 and mind-bending GeForce RTX 3090.
Supposedly the NVIDIA GeForce RTX 30 series are supposed to be about twice as good in performance than the NVIDIA GeForce RTX 20 series while offering nearly twice the power efficiency. That puts the NVIDIA GeForce RTX 3070 about on par with the NVIDIA GeForce RTX 2080 that was launched two years ago. Which also means that the GeForce RTX 3080 is miles away from the GeForce RTX 2080Ti.
What’s New with Ampere?
Just saying that it is faster though, may not mean that much to you. In this case, bear with us a little bit as we go a little more technical than usual in identifying what is new with the new GeForce RTX 30 series cards. NVIDIA’s Jensen Huang also did say in the launch that it might be time for you to upgrade your GPUs if you are still on the GTX platform with this second generation Ray Tracing GPU.
New Multiprocessors
If you think of a GPU, it is actually a mini computer on its own that you dedicate on graphics processing. It sort of works on a piggybacking system on top of your main logic board for your PC. By principle, it is just its own processing unit separate to what the CPU and regular RAM does. In layman terms, the GPU is an additional calculator lodged in your PC.
The multiprocessor that is on the GPU is technically its heart, its cores. The new processor on the NVIDIA GeForce RTX 30 Series is technically the same size as the older NVIDIA GeForce RTX 20 Series cores. That is to say that Ampere did not gain in size for more performance over Turing.
They simply unlocked more power from the chip itself with new technologies and architecture. We will talk about that a little bit more later though. For now, you just have to know that it is now twice as powerful as before, it is twice faster in calculations.
New Generation Ray Tracing Cores
NVIDIA says that new generation consumer graphics standards is their new Ray Tracing technology that is introduced two years ago. The technology, technically, is not new. It has been a technology used in plenty of 3D rendering applications and even machine learning applications.
With the first-generation Ray Tracing cards, that is the GeForce RTX 20 Series GPUs, ray tracing is used to make graphics even more stunning and realistic by correctly predicting reflection and shading models based on virtual lighting positions. That means that graphic artists do not need to predict where the light is coming from or how it interacts with an object, it just simply does. That is why you do not get great mirror reflections, glass reflections, or water textures in older video games up until the late 2018s.
The said technology requires a separate core to the regular GPU cores that we are used to on older GeForce GTX cards though.
The second-generation Ray Tracing core is technically not that much different from the first-generation. In terms of size and build at least, it is about the same. They managed to extract even more performance out of it, they claim twice more performance out of it too. That allows the GPU to render reflections and shades correctly even faster.
In the previous generation Ray Tracing core, they could render shades, reflections, and even physics correctly at about 720p at 60fps. The new generation that was just introduced is supposed to be able to render smoothly at 1440p at 60fps. That is a large leap, about twice according to NVIDIA.
Even Newer Tensor Cores
Artificial Intelligence (A.I.) and machine learning has been a part of NVIDIA’s repertoire for a long time now. They started looking into implementing A.I. cores in their GeForce GTX 10 Series GPUs. They were successful and the cores worked to a certain degree in GTX GPUs.
In the new GeForce RTX architecture though, A.I. took a bigger stage and Deep Learning Super Sampling (DLSS) was born. The first generation DLSS allowed calculations and processing to be offloaded from the GPU and pushed to a different server. That allows graphics to look even better than what the GPU is actually capable of physically. Of course, that takes a lot of processing power.
At the same time the DLSS technology is used as a prediction model as well, to anticipate calculations and render them beforehand. In terms of gaming, the cores allow your games to render at 720p on your own GPU but deliver 1080p or 1440p graphics to your eyes via the display. This is all part of the NVIDIA RTX IO experience as well though.
NVIDIA RTX IO
So, you have watched the demo of the upcoming Sony PlayStation 5. You saw the next generation console load open world maps and transition between places in no time, with no load screens. That technology is achievable through the clever use of SSD and processing power.
That is what NVIDIA RTX IO is going to replicate though. It allows plenty of the processing loads on the CPU and on board RAM from modern SSDs to be offloaded to the GPU itself. Since the GPU, in some cases, have spare processing capacity anyway, you get an improved processing performance for your games anyway and even improved data transfer and read speeds.
That means that game designers can try to build games with huge amount of detail in their maps while not thinking too much about load times. Of course, for lesser CPs, load screens are always going to be an issue in games. NVIDIA’s RTX IO is built to solve that.
There is another benefit to this though. Every other PC builder will tell you that if you want a great performing gaming machine, it has to be a combination of great GPU and CPU. The GPU will handle the 3D stuff, while the CPU will handle the frame generation. That is because most of the frame generation is done by the CPU, and therefore your framerate greatly depends on how much your CPU can handle itself. With RTX IO offloading that load to the GPU, you technically can have a slightly cheaper CPU and still achieve high frame rate in games.
New Faster, Nay, Fastest Memory Module in the World
As we mentioned earlier, a GPU is a little bit like a mini computer that piggybacks on top of the main board. It is a calculator that hops on another calculator to make that calculator faster and more powerful. In that sense, the GPU also needs its own memory module, its own RAM unit.
In that, they have worked with Micron to build the world’s fastest discrete graphics memory unit, the GDDR6X. You get up to 1TB/s in bandwidth for any graphics applications. This is where that minimised load times come into play too.
Samsung 8nm Transistors and NVIDIA
Every other tech familiar person knows that there are only a few manufacturers that really know what their doing when it comes to semiconductors and chip making. Samsung is one of those manufacturers in the circle of ‘the best of the best’. The new GeForce RTX 30 Series GPU benefits from Samsung’s know how in chip making.
The new GPU multiprocessing chips are built with 8nm standards and technology by Samsung. Smaller transistors also mean that you can fit more in a certain surface area. More transistors mean more power naturally.
This is technically how NVIDIA managed to achieve double the power from the previous generation GeForce RTX 20 GPUs while still retaining the sizes, just about. Less space and size also means that there is less power requirements from the transistors themselves, which leads to a much higher power efficiency. Hence, double the power, double the efficiency.
Better Technology, Better Experience
With more power, comes more enhancements from NVIDIA. They have done audio before already with RTX noise cancelling, which never cease to amaze us. This time though, they are turning their focus toward competitive gaming and other parts of content creation.
NVIDIA Reflex
It is exactly what it sounds like. It measures reflex. It does not measure your reflexes though.
Instead, NVIDIA Reflex measures latency between your input and the display’s output. From that measurement, NVIDIA’s Reflex algorithm will then optimise the game’s latency. In some sense, it is meant to reduce the latency between your input and the PC’s output.
This impacts eSports titles mostly. In video games with built in NVIDIA Low-Latency Mode, the technology reduces latency by about 50 percent. Every millisecond counts in eSports environment. That is also why NVIDIA’s Reflex Analyser also can calculate and tabulate the time it takes for you input to be translated into output. The Reflex Analyser is coming to most of the new 360Hz NVIDIA G-Sync eSports displays from its partners later this year.
NVIDIA Broadcast
Streaming is a huge business. It is gaining in popularity as well. As such, it would not be right for NVIDIA to ignore the market completely. In the sense of the NVIDIA GeForce RTX 30 Series, the extra power can also help with a single PC stream set up with NVIDIA Broadcast. The A.I. noise cancelling is already clever enough. NVIDIA Broadcast though is much cleverer than that with webcam auto framing and even virtual background effects for your stream.
NVIDIA Omniverse Machinima
Forget trying to create new stories with game engines. Forget spending hours on character creations and animations codes. Now, you could do all that for storytelling via NVIDIA Omniverse Machinima. You can use existing models or assets and even create your own via a webcam, amazingly. This is the power of NVIDIA’s new GeForce RTX 30 Series, and NVIDIA’s new storytelling tool.
Well, then again, this is still an early access tool. You might want to check back on this. If not, you can be a part of it by signing up for early access.
The NVIDIA GeForce RTX 3070, GeForce RTX 3080, and GeForce RTX 3090 is Here!
The first GPUs that comes out from the NVIDIA GeForce RTX 30 Series is the NVIDIA GeForce RTX 3070, a technical replacement of the NVIDIA GeForce RTX 2070. Thanks to the technological advancements, the GeForce RTX 3070 is way more powerful than the GPU it replaces. In certain benchmarks it is more powerful than the GeForce RTX 2080. With up to 8GB of GDDR6X RAM, 4K and 1440p resolution should be comfortable enough.
Then there is the NVIDIA GeForce RTX 3080 which is way more powerful than the current generation GeForce RTX 2080Ti. It just blows the older GPU away. All that, while costing less than the GeForce RTX 2080Ti. It has up to 10GB of GDDR6X RAM to deliver 4K resolution at 60fps.
The daddy of the lot is what NVIDIA refers to as the “BFGPU” – Big Ferocious GPU. We would expand it to a different thing, but that is just us. This is the spiritual successor to the NVIDIA GeForce TITAN RTX, technically. This is supposed to be 10x quieter, and up to 30°C cooler. Of course, it is a big one, occupying three PCIe slots. Within the large body is an enormous 24GB of GDDR6X RAM and performs up to 50% faster than the TITAN RTX. It can easily support 8K resolution up to 60 fps in Control.
The NVIDIA GeForce RTX 3070 is priced at US$ 499 (MYR 2,399*) and will be available in October 2020 onward. The GeForce RTX 3080 is priced at US$ 699 (MYR 3,360*) and will be available 17th September 2020 onward. The GeForce RTX 3090 is priced at US$ 1,499 (MYR 7,208*) and is available 24th September 2020 onward. Keep in mind that these are release dates for the Founders Edition cards. Of course, partner manufacturers will have their custom boards and own clocks later on or the same date.
*local price as per NVIDIA Malaysia website