Nvidia’s Ada Lovelace GPU generation: $1,599 for the RTX 4090, $899 and above 4080

Zoom / Time to get the checkbook out again, GPU lovers. The RTX 4090 is here (and it’s not alone).

nvidia

After weeks of excitement, Nvidia’s latest PC graphics card, the “Ada Lovelace” generation of RTX 4000 GPUs, is here. Nvidia CEO Jensen Huang debuted two new models on Tuesday: RTX 4090which will start at a whopping $1,599, and RTX 4080which will be launched in two configurations.

The higher-priced card, scheduled for release on October 12, occupies the same high-end segment as Nvidia’s 2020 Megaton RTX 3090 (previously designated by the company as a “Titan” product). Increasing the physical size of the 4090 will require three slots on the computer of your choice. The specs refer to the highest GPU: 16384 CUDA cores (up from 3,090 cores and 10,496 CUDA cores) and 2.52GHz of boost clock (up from 1,695GHz in the 3090). Despite the improvements, the card still runs in the same 450W power envelope as the 3090. It will still allocate 24GB of GDDR6X RAM.

This jump in performance is driven in part by Nvidia’s rumored long jump to the TSMC “4N” process, a new generation of 5nm chips that deliver a huge jump in efficiency from the previous generation of the 8nm process.

RTX 4080, coming in two SKUs.
Zoom / RTX 4080, coming in two SKUs.

nvidia

Meanwhile, the RTX 4080 will follow in November in two SKUs: a 12GB GDDR6X model (192-bit bus) starting at $899, as well as a 16GB GDDR6X model (256-bit bus) starting at $1,199. Depending on how different the specs are between these two models, it appears that Nvidia is releasing two completely different chips under the same “4080” tagline; Traditionally, this much differentiation of Nvidia hardware comes with separate model names (such as the last generation 3070 and 3080). We’re waiting for a tougher confirmation from Nvidia on whether or not the 4080 models share a chipset.

The price of these two will include more CUDA cores (9728, up from the RTX 3080’s 8704), boost clock (2.51GHz, up from 1.71 in the 3080’s), and power draw (320W, like the 3080’s but more than less. The 4080’s 285W memory) . At least for this generation, Nvidia offers 12GB of memory as a baseline 4080 — an efficient handling of much criticism by its memory-hungry RTX 3000 generation of GPUs.

Micromap, micromesh

Both new models include iterative updates to elements of Nvidia’s RTX chipset – RT cores and tensioner cores. Besides, Nvidia has announced updated processes for both its real-time ray tracing processing in 3D graphics and the Deep Learning Super-Sampling (DLSS) upgrade system. The first will be augmented on Lovelace GPUs with two new types of hardware units: a “transparency micro-map engine”, intended to double the performance of raw ray tracing, and a “micro-network engine” to increase the amount of “no-storage-cost” engineering coverage on the rendering front.

The latter has now been bumped into a new version: DLSS 3, which appears to be an exclusive feature of the RTX 4000 series. According to Huang, this system promises to “create new tires [of gameplay] Without involving the game, effectively boosting the performance of both CPU and GPU. “Real-time frame by frame graphics reconstruction is designed to solve problems found in image reconstruction techniques that cannot necessarily model motion vectors of in-game elements such as particles. This new method, along with techniques developed by previous DLSS generations, can be used to recreate Building a massive 7/8 scene pixels – thus dropping computational demands on both GPUs and CPUs.If it works as promised, DLSS could offer a huge improvement on the pixel-by-pixel process that’s been a hit across both DLSS and its competitors—AMD’s FSR 2.0 and Intel’s upcoming XeSS.

On top of those proprietary systems, Nvidia’s latest GPUs will apparently rely on a new process Huang calls “shading execution rearrangement”. While this will improve raw rasterization performance, Huang’s brief description of the system hinges largely on computationally expensive ray tracing workloads. The system will run “2-3 times” the ray tracing performance of the company’s previous Ampere generation of GPUs.

As part of today’s announcements, Nvidia has introduced some titles with upcoming Nvidia RTX improvements, arguably the biggest of them all. Microsoft Flight Simulator 2020. When combined with the image reconstruction system in DLSS 3, MSFS It rendered at frame rates that easily hit 100 seconds in some of the game’s busiest scenes (although raw rasterization of the same scenes on the unnamed Lovelace GPU and PC also showed a solid performance, considering how these could be The game is energy hungry and massive cityscapes). The gallery above was captured in 4K, so you may want to click and zoom to see how DLSS 3 handles finer pixel detail compared to the crisp set of raw pixels and standard temporal anti-aliasing (TAA) on older systems.

Huang has also tried to win over the PC mod scene by showing how a new Nvidia-developed toolkit, dubbed RTX Remix, can be applied to a number of classic games whose mod possibilities are wide open. The best results came from the before and after demonstration applied to The Elder Scrolls III: Morrowind, which used RTX GPU tensor cores to programmatically update game textures and material properties. (see above, or Go to the Nvidia website for more before and after comparisons.) Similar results are expected from file Gate 1 A DLC bundle that Nvidia will release to fans to apply to classic PC games in November. We’d be curious to see head-to-head matches for Nvidia’s auto-tweak results compared to the best stuff from over a decade of community development.

Ars Technica may receive compensation for sales from the links in this post through Affiliate Programs.