Nvidia releases new $599 GeForce RTX 4070 GPU
Friday, April 14th, 2023 | Posted by Jim Thacker
Nvidia has launched the GeForce RTX 4070, the newest card within the GeForce RTX 40 Collection, the agency’s present technology of shopper GPUs primarily based on its Ada Lovelace GPU structure.
The $599 12GB card, already accessible from retailers, is the lowest-cost Ada Lovelace desktop GPU so far – and, as with different latest GeForce RTX GPUs, is being marketed at CG work in addition to gaming.
The most recent GPU to make use of Nvidia’s Ada Lovelace structure
The GeForce RTX 4070 is the newest GPU primarily based on Nvidia’s new Ada Lovelace structure.
In addition to updates to the playing cards’ RT ray tracing cores and Tensor AI inferencing cores, it includes a new Optical Move Accelerator, utilized by DLSS 3.0, Nvidia’s Deep Studying Tremendous Sampling expertise.
Initially a render upscaling system, DLSS 3.0 additionally makes use of optical circulate to generate complete new frames between these rendered historically, boosting in-game body charges by “as much as 4x”.
Though DLSS is primarily a video games expertise, it does have implications for CG artists, since it’s also now being adopted in visualisation software program like D5 Render.
| Nvidia’s GeForce RTX 40 Collection GPUs | |||||
|---|---|---|---|---|---|
| RTX 4090 | RTX 4080 | RTX 4070 Ti | RTX 4070 | ||
| Structure | Ada Lovelace | Ada Lovelace | Ada Lovelace | Ada Lovelace | |
| CUDA cores | 16,384 | 9,728 | 7,680 | 5,888 | |
| Tensor cores* | 512 | 304 | 240 | 184 | |
| RT cores* | 128 | 76 | 60 | 46 | |
| Base clock (GHz) | 2.23 | 2.21 | 2.31 | 1.92 | |
| Increase clock (GHz) | 2.52 | 2.51 | 2.61 | 2.48 | |
| Compute efficiency FP32 (Tflops)* |
82.6 | 48.7 | 40.1 | 29.2 | |
| GPU reminiscence | 24GB GDDR6X |
16GB GDDR6X |
12GB GDDR6X |
12GB GDDR6X |
|
| TDP | 450W | 320W | 285W | 200W | |
| Launch date | 2022 | 2022 | 2023 | 2023 | |
| MSRP at launch | $1,599 | $1,199 | $799 | $599 | |
*Information taken from third-party web sites.
Key specs for DCC work
With a really helpful value of $599, the GeForce RTX 4070 fills out the mid-range of the GeForce RTX 40 Collection, begun with the discharge of the GeForce RTX 4070 Ti earlier this 12 months.
As you’ll count on, its key specs are a step down from the earlier 40 Collection playing cards, with fewer CUDA and RT cores and a correspondingly decrease FP32 compute efficiency.
Nonetheless, its energy draw is decrease – it’s additionally down on the previous-gen GeForce RTX 3070 and 3070 Ti – and just like the 4070 Ti, it has a good 12GB of reminiscence for GPU rendering,
Nvidia’s comparability chart of the relative efficiency of the GeForce RTX 4070 and the previous-generation GeForce RTX 3070 Ti in graphics purposes. Click on the picture to view it at full measurement.
Benchmarks and efficiency in DCC purposes
Though the launch video for the GeForce RTX 4070 focuses on gaming efficiency, Nvidia has additionally launched a couple of benchmark outcomes for CG work.
Rendering pace in Omniverse, Nvidia’s real-time design and collaboration platform, is 2.8x that of the GeForce RTX 3070 Ti, the equivalently priced previous-generation card.
It’s additionally attention-grabbing to see scores for picture technology with Secure Diffusion: the primary time we’ve seen a generative AI software used as an official GPU benchmark – however not, we suspect, the final.
Pricing and launch dates
The GeForce RTX 4070 is accessible now with a MSRP of $599.
Learn extra concerning the new GeForce RTX 4070 on Nvidia’s product web site
Tags: Ada Lovelace, AI, AI artwork, Ampere, arch viz, benchmark, enhance clock, cg, clock pace, comparability desk, compute efficiency, shopper GPU, CUDA core, D5 Render, DCC, Deep Studying Tremendous Sampling, DLSS, DLSS 3.0, FP32, body technology, recreation artwork, recreation improvement, gaming card, gaming GPU, GDDR6X, GeForce RTX, GeForce RTX 40 Collection, GeForce RTX 4070, GeForce RTX 4070 Ti, GeForce RTX 4080, GeForce RTX 4090, generative AI, GPU, GPU computing, GPU rendering, graphics card, graphics reminiscence, machine studying, NVIDIA, Omniverse, optical circulate, Optical Move Accelerator, efficiency, energy consumption, value, processor pace, ray tracing, launch date, render upscaling, RT core, specs, specs, Secure Diffusion, TDP, Tensor core, vfx, visible results, visualization
