Nvidia GeForce RTX 4090, 4080 to Ship Oct. 12 Starting at 9

What’s taking place

Nvidia has introduced its flagship, next-generation gaming graphics playing cards, the RTX 4090 and RTX 4080.

Why it issues

Nvidia-based graphics playing cards are nonetheless the most well-liked ones available on the market, and any new applied sciences that velocity them up considerably are notable.


Goodbye, Ampere; good day, Ada Lovelace: Nvidia revealed its new era of GeForce graphics playing cards at its annual Graphics Technology Conference, that includes the corporate’s biannual improve to the GPU’s structure. Nvidia claims the flagship RTX 4090 can carry out as much as 2 instances higher on raster-heavy video games and as much as 4 instances quicker on totally ray-traced video games than the previous flagship RTX 3090 Ti, whereas the RTX 4080 must be as much as 3 instances quicker than the RTX 3080 Ti. The RTX 4090 is slated to ship on Oct. 12 beginning at $1,599, whereas the RTX 4080 is coming in November beginning at $899; the latter will are available in two variations, 16GB and 12GB.

In addition to Nvidia’s Founders Edition playing cards, variations utilizing the GPUs will likely be obtainable from Nvidia’s ordinary companions (besides EVGA), together with Asus, Gigabyte, MSI, PNY and Zotac. The RTX 3080, 3070 and 3060 will stay within the line, and I’d search for huge reductions on them throughout the vacation procuring season. 


RTX 4090 RTX 4080 (16GB) RTX 4080 (12GB)
Memory 24GB GDDR6X/384 bit 16GB GDDR6X/256 bit 12GB GDDR6X/192 bit
CUDA cores 16,384 9,728 7,680
Boost clock (GHz) 2.52 2.51 2.61
PCIe Gen 4 4 4
Card width 3 slots 3 slots Varies
Power/PSU requirement 450w/850w 320w/700w 285w/700w
Availability Oct. 12 November November
Price Starts at $1,599 Starts at $1,199 Starts at $899

When the RTX 4090 ships, there will likely be some free proof-of-pretty-and-performance DLC: Portal with RTX and a brand new Ray Tracing: Overdrive Mode for Cyberpunk 2077. There will even be a home-built real-time-rendering app, Nvidia Racer RTX. 

A chip rendering with annotations highlighting the Ada Lovelace architectureA chip rendering with annotations highlighting the Ada Lovelace architecture


The chips’ Ada Lovelace structure — it is not clear whether or not it will likely be known as “Ada,” which is shorter, or “Lovelace,” consistent with custom — introduces seven {hardware} upgrades, a few of that are derived from its datacenter sibling, Hopper:

  • Built on a 4nm course of. Optimally, smaller processes make it doable to cram extra silicon into the identical area, with (normally) improved energy and efficiency efficiencies.
  • Third-generation ray-tracing cores double the throughput of calculations for the way the rays work together with the polygons defining the picture, which Nvidia says enhance floating level operation velocity by as much as 2.8X.
  • 4th-generation Tensor cores, which drive Nvidia’s DLSS upscaling software program, permits DLSS to leap from producing particular person pixels to extra environment friendly frame-based algorithms for its new DLSS 3. Over 35 video games and apps are signed as much as incorporate DLSS 3 beginning in October. 
  • Updated Streaming Multiprocessors ship greater than 2X the throughput of the Ampere SMs.
  • Shader Execution Reordering optimizes the handoff of ray-traced knowledge to the SMs for rasterizing, which the corporate claims boosts shader efficiency by as much as 2X and body charges by as much as 25%.
  • Optical Flow Accelerator makes use of pixel movement knowledge to DLSS in an effort to generate new frames; Nvidia says this improves efficiency in CPU-intensive video games (advanced simulation video games, like Microsoft Flight Simulator). 
  • Dual AV1 Encoders ought to enhance on-the-fly video processing high quality for streamers. OBS is predicted to launch a brand new model of its streaming software program in October, and Discord is slated to assist it later this 12 months.

To go along with the brand new processing capabilities, Nvidia launched some new software program algorithms to enhance efficiency as properly.

  • Opacity Micro-Maps create and save masks in a world’s objects upfront, so they do not must be recreated.
  • Displaced Micro-Meshes are meshes of smaller triangles that the brand new RT cores can course of while not having to be saved or preprocessed someplace else; they enhance velocity (as much as 10X, per Nvidia) and require as much as 20X much less video reminiscence in sure instances. 

Source link