Skip to main content

Nvidia finally made a tiny RTX 4000 graphics card (but you probably don’t want it)

After months of massive graphics cards like the RTX 4090, Nvidia is finally slimming things down at its GPU Technology Conference (GTC). The RTX 4000 SFF delivers the Ada Lovelace architecture in a tiny package, but you probably won’t find it sitting among the best graphics cards.

Although the RTX 4000 SFF uses the same architecture in gaming GPUs like the RTX 4080, it’s built for a very different purpose. It uses Nvidia enterprise drivers, and it’s made to power computer-aided design (CAD), graphics design, AI applications, and software development, according to Nvidia. The card takes up two slots and includes a low-profile bracket for cases like the Hyte Y40.

NVIDIA RTX 4000 SFF Ada Generation: Next-Level Performance for Compact Workstations

For specs, it includes 20GB of GDDR6 memory on a 160-bit bus, as well as 6,144 CUDA cores, 192 Tensor cores, and 48 ray tracing cores. Those core specs are slightly lower than the RTX 4070 Ti, but you shouldn’t expect the RTX 4000 SFF to be nearly as powerful. It has solid specs, but it’s limited to only 70 watts of power due to its small form factor nature.

Recommended Videos

Although the RTX 4000 SFF fills a niche for gamers, it isn’t a great buy. Nvidia says PNY and Ryoyo Electro will start selling the GPU in April for $1,250, and showcasing it as a GPU built for professional applications.

Please enable Javascript to view this content

Nvidia is also bringing the Ada Lovelace architecture to laptop creators. We’ve had our first taste of RTX 40-series GPUs in machines like the Asus ROG Zephyrus M16, but Nvidia’s new range is built for creators, not gamers.

The flagship RTX 5000 comes with 16GB of GDDR6 memory, along with 9,728 CUDA cores. Down the stack, both the RTX 4000 and RTX 3500 come with 12GB of GDDR6 memory, but the RTX 4000 features 7,424 CUDA cores, while the RTX 3500 is packed with 5,120. Below those are the RTX 3000 and RTX 2000 with 4,608 and 3,072 CUDA cores, respectively. Both come with 8GB of GDDR6 memory.

Nvidia didn’t point to any specific models, but it said buyers can expect to see workstations packing the new Ada GPUs from HP, Lenovo, and others starting this month. Again, these GPUs use Nvidia’s Studio drivers and are built for professional applications, so expect workstations to cost a premium over gaming-focused machines.

Jacob Roach
Former Digital Trends Contributor
Jacob Roach is the lead reporter for PC hardware at Digital Trends. In addition to covering the latest PC components, from…
The RTX 5060 will be Nvidia’s most important GPU, and I’m worried about it
Two graphics cards sitting on top of each other.

Nvidia just finished revealing its range of new RTX 50-series GPUs, the first of which will arrive in just a couple of weeks. They're some of the best graphics cards ever made, according to Nvidia, and for the flagship RTX 5090 that clocks in at $2,000, I believe the company. Lower down the stack, however, I'm concerned.

For the past couple of years, there's been a growing issue surrounding graphics cards with 8GB of VRAM, which is something we've seen on full display with games like Indiana Jones and the Great Circle. Despite backlash in the previous generation concerning releases like the RTX 4060 Ti, I'm worried that Nvidia will repeat the mistakes of the past when the RTX 5060 inevitably rolls around.
It'll be popular

Read more
Nvidia’s DLSS 4 isn’t what you think it is. Let’s debunk the myths
DLSS 4 in Cyberpunk 2077.

Nvidia stole the show at CES 2025 with the announcement of the RTX 5090, and despite plenty of discourse about the card's $2,000 price tag, it ushers in a lot of new technology. Chief among them is DLSS 4, which brings multi-frame generation to Nvidia's GPUs, offering a 4X performance boost in over 75 games right away when Nvidia's new RTX 50-series GPUs hit the streets.

I've seen way too much misunderstanding about how DLSS 4 actually works, though. Between misleading comments from Nvidia's CEO and a radical redesign to how DLSS works, it's no wonder there's been misinformation floating around about the new tech, what it's capable of, and, critically, what limitations it has.

Read more
Here’s how Nvidia’s CEO defends the RTX 5090’s $2,000 price tag
Nvidia's RTX 5090 sitting at CES 2025.

"When someone would like to have the best, they just go for the best," said Nvidia's CEO Jensen Huang in a Q&A session with media at CES 2025. Huang was speaking on the newly-announced RTX 5090, and its new price tag of $2,000, making it the most expensive desktop graphics card Nvidia has ever released.

It's a new high for Nvidia, but also a bold departure from the rest of the range. The next card down in Nvidia's stack, the RTX 5080, comes in at $1,000 -- half the price of the flagship. Huang suggested that customers don't want to deal in micro-segmentation minutia. "$2,000 is not small money, it's fairly high value," Huang said. "But a lot of customers, they just absolutely want the best."

Read more