Nvidia has announced the RTX 4000 SFF, the latest professional GPU to use its Ada Lovelace architecture.

The $1,250 small form factor 20GB card, intended for use in mini-desktop systems, will be joined by five new Ada Generation cards for laptops: the RTX 2000, RTX 3000, RTX 3500, RTX 4000 and RTX 5000.

The new cards were announced at GTC 2023 this week, along with updates to Omniverse and a partnership with Shutterstock to develop new text-to-3D AI art tools.

The latest professional GPUs based on Nvidia’s Ada Lovelace architecture
All six of the new cards are based on Nvidia’s current Ada Lovelace GPU architecture, intended to provide “revolutionary performance for ray tracing and AI-based neural graphics”.

It features iterative improvements to all of the cards’ key hardware core types: CUDA cores for general GPU computing, Tensor cores for AI operations, and RT cores for hardware-accelerated ray tracing.

According to Nvidia, all three core types see improvements of “up to 2x” in raw performance over its previous Ampere architecture.

The architecture also features a new Optical Flow Accelerator, used by DLSS 3.0, the latest version of Nvidia’s Deep Learning Super Sampling technology.

Initially a render upscaling system, DLSS is used in real-time visualisation tools like D5 Render to improve viewport frame rates by rendering frames at lower resolution, then up-resing them to display resolution.

In version 3.0, it also becomes a frame generation system, with DLSS 3.0 using optical flow to generate entire new frames between those rendered traditionally, boosting frame rates by “up to 4x”.

*Estimated street price

 
Specifications, benchmark performance and expected price
The new RTX 4000 SFF is considerably less powerful – but also considerably more affordable – than the first Ada Lovelace GPU for professional desktop workstations, the top-of-the-range 48GB RTX 6000.

Its core specs are similar to those of its equivalent Ampere card, the RTX A4000, with identical core counts and FP32 compute performance, although RT core performance rises from 37.4 Tflops to 44.3 Tlfops.

However, at 20GB, its graphics memory capacity is 4GB higher than the Ampere card, increasing the size of 3D scenes that can be rendered without going out of core.

It’s also much more energy-efficient than the RTX A4000, with a maximum power consumption of just 70W – and obviously, as a small form factor dual-slot card, its dimensions are quite different.

Unlike with previous workstation cards, Nvidia hasn’t released any benchmark comparisons for DCC applications, although the launch video embedded at the top of the story shows both Enscape and KeyShot.

 
Five new laptop GPUs bring the RTX Ada Generation cards to mobile workstations
In addition, Nvidia has brought the Ada Lovelace architecture to laptops as well as desktop systems, announcing five new RTX Ada Generation GPUs for mobile workstations.

The new cards – which range from the 8GB RTX 2000 to the 16GB RTX 5000 – supersede the current Ampere generation cards at the mid-to-high end of Nvidia’s professional laptop GPUs.

The remaining cards – the entry-level RTX A500 and A1000 – don’t yet have Ada Generation equivalents.

Pricing and release dates
The RTX 4000 SFF will be available via Nvidia distribution partners like Leadtek, PNY and Ryoyo Electro in April 2023, with a MSRP of $1,250.

The new Ada Lovelace RTX laptop GPUs will become available via system builders including BOXX, Dell, HP and Lenovo later this month.

 
Find full specifications for the RTX 4000 SFF on Nvidia’s website

Tags: Ada, Ada generation, Ada Lovelace, AI inferencing, Ampere, benchmark, comparison table, compute performance, CUDA cores, DCC software, dimensions, DisplayPort, DLSS 3.0, FP32, game art, game development, GDDR6, GPU, GPU compute, GPU computing, GPU memory, GPU rendering, graphics card, graphics memory, GTC 2023, laptop GPU, memory, memory bandwidth, NVIDIA, NVLink, power consumption, price, professional GPU, ray tracing, release date, RT core, RTX, RTX 2000, RTX 3000, RTX 3500, RTX 4000, RTX 4000 SFF, RTX 5000, RTX 6000, RTX A4000, simulation, specifications, TDP, Tensor core, vfx, virtual reality, visual effects, workstation GPU