NVIDIA held its planned next gen GPU reveal event on Tuesday, and there was a lot of new stuff to unpack. Coming into the event, there was a lot of speculation going on about the specs, capabilities, and creds of the new, next generation of NVIDIA GPU’s. Needless to say, we were among the many who speculated on the cards, as is only natural in the run-up to such a hotly anticipated release. Some of the rumors were right, some were sort of right. In the end, there was a lot more to talk about then just GPU’s, and we are going to go through a lot of it right here.
The NVIDIA next gen event was hosted by none other than NVIDIA CEO Jensen Huang. He took us through an brief history of the NVIDIA family, highlighting some of the groundbreaking work that NVIDIA has done in AI and collaborative computing, and highlighting that Huang notes that, “Modern GPU’s (are) the engine of large industries from design, cloud AI, to scientific computing. But it is the gamers and their insatiable demand that is the driving force of the GPU.” Huang talked about how gamers, “Pooling their GPUs to create the largest distributed computer ever, a million gamers united to counter-strike the COVID-19 coronavirus.” Ultimately, the collaboration resulted in 2.8 exaflops, or as Huang points out – “five times the processing power of the worlds most powerful supercomputer,” and ends by thanking the combined force.
There was a lot of technical details on the next gen gear, and a lot of great renderings that the viewers were treated to, but the first “BIG NEWS” out of the NVIDIA GeForce event was that Fortnite is (finally) turning RTX support on, which means that now both Minecraft and Fortnite will support RTX video, and Epic released a quick trailer during the NVIDIA presentation announcing the coming addition. No indication of just how soon Fortnite will support RTX, as all that was given was a “Ray Tracing and DLSS Coming Soon” splash at the end of the trailer. However, it might coincide with the release of the new 3080 card, which will land on store shelves Sept 18th.
NVIDIA also announced “NVIDIA Reflex” that will be coming along soon with the next generation of software. This is a new esports technology that is intended to improve the latency across the system and GPU by up to 50% – this means that in scenarios where split second reaction time is required (Huang sites and demonstrates and example from Valorant), for the typical gamer, with about a 150ms reaction time, may have a narrowed window to respond and react. With system latencies on existing 1050-1070 cards ranging in the 120 ms range at the high end to 30ms at the low end, Reflex will cut these times in half. The most exciting part about this announcement is that it is an upgrade in the software package that will be supported by those current graphics cards already hooked up, so this is an upgrade that all NVIDIA users will be able to take advantage of on their existing setups. Reflex is coming in the September release of the GeForce game driver. Unfortunately, as it is a game driver, it is an update that will come from the game developer side, and not something that will be system-wide. The list of games that will release updates with the Reflex package include: Apex Legends, Call of Duty:Modern Warfare, Destiny Two, Fortnite, and Valorant.
Next up, NVIDIA has designed a new “insanely fast and beautiful display,” as Huang put it. This NVIDIA next gen display is a 360Hz GSync IPS display, which which features a built in latency analyzer. The player connects directly into the display with the mouse, and the latency analyzer works to compensate for the latency. The displays will be available “in the fall” from Acer, Alienware, Asus, and MSI.
NVIDIA also talked about its new Broadcast suite, which is its new streaming software suite which will debut in September, and will feature new tools features including a noise removal tool, as well as background replacement functions that include a green screen effect that can place the streamer, cut out of the background, on the screen while playing the streamed game in the background, and an auto-frame feature that follows the streamer on a zoomed-in feed (without having to physically reposition the camera.)
NVIDIA Has a new software suite that will be arriving soon for Machinima artists called NVIDIA Omniverse Machinima, which will feature a suite of tools called that will allow artists to bring in video elements from games or other sources, animate faces to match language tracks, integrate accurate physics, and allow an artist to record a performers movements on video and then use that to overlay a character to match the movements and gestures of the performer. Essentially, Omniverse Machinima software will enable users to create high-quality CGI video.
As things moved into the tech side of the presentation, here is a summary of some of featured highlights on the new RTX generation graphics cards:
- GeForce RTX3080 Ampere features three types of processors, just as on Turing – Shader codes, RT Cores, and AI processing (Tensor Cores)
- DLSS 2.0 –
- RTX’s Dedicated RT cores reduce shader core workload, with resulting 60% improvement in latency
- Ampere delivers:
- Shader cores: Ampere = 30 TFLOPS vs Turing =11 TFLOPS
- RT Cores: Ampere = 58 RT TFLOPS vs Turing = 34 TFLOPS
- Tensor (AI) Core: Ampere Tensor = 238 TFLOPS vs Turing = 89 TFLOPS
- 2X Ray/Triangle intersections
- Memory: 10 Gb GDDR6x @ 19Gbps / 780 GB/s – G6X “PAM4” Memory – can transmit 2x the data of G6 memory
- 28 Billion Transistors
- Dual Axial Flowthrough Cooling with independent push/pull fans – Dual airflow across the PBC and separate GPU coolers
NVIDIA Had some side-by-sides to show the performance improvement between the current and next gen hardware. Back in May. NVIDIA debuted a video recording entitled “Marbles at Night,” which was produced from a playable physics game built around the NVIDIA RT technology. It was run in May on a single, top of the line Quadro RTX8000 GPU. NVIDIA produced a new version to show off the significant improvements. We took two pull-out images that NVIDIA showed and put them side by side to demonstrate some of the significant graphic enhancements, which include depth of field and true-to-life textures of materials.
So finally getting down to the business at hand – how much is this going to cost you? Surprisingly, both the 3080 card and its little brother, the 3070, will be a bargain, as will the 3080’s muscle-bound bigger brother, the 3090, which we got a look at as well. See the price to performance chart below that NVIDIA displayed to give a representation of the starting prices of each of these next gen GPU’s in comparison to prior gen prices. The 3080 will debut at starting price of $699 and will be available starting on September 18th.
Not only did we get impressive details on the 3080, but details on the 3070 card as well. Due out in October, many a RTX2080ti owner will likely be upset to learn that their massively expensive cards will be outperformed by the GeForce RTX3070 card at 20 Shader/40 RT/163 Tensor TFLOPS vs the 11/34/89 we referenced earlier. It will sport 8 GB of G6 memory, vs the G6x memory on the 3080 card, and a similar (though not replicated) cooling setup as on the 3080. With a starting price of $499, it is significantly more power at big value.
Then there is GeForce RTX3090 card is a monster, and aptly nicknamed the “BFGPU,” brings 24 Gb of G6x memory, 36 Shader/69 RT/285 Tensor TFLOPS, it is physically massive as well, and runs 60FPS in 8K, and the card will price out at the relative bargain of $1499- Impressive considering the RTX Titan, the card it is taking the place of, can run upwards of $2500 today.
Below is a side-by-side of the basic specs, which clearly demonstrated the huge leap in core density that will be present on the next generation of NVIDIA GPU’s.
Overall, there was a lot of great news, and far more than can be easily packaged in this recap. You can watch the full 40-minute video below and let us know what you think in the comments below!cards/30-series/