NVIDIA releases 65 teraflops RTX 5000 for workstations

This Tuesday (08), nvidia Finally announced the new generation of RTX 5000, RTX 4500 and RTX 4000 graphics cards for workstations targeting Generative artificial intelligencecontent creation and delivery.

As much as the green team doesn’t help users much with different numbers to refer to products, all GPUs are part of the Ada Lovelace generation, which also includes GeForce RTX 4000 For daily use. The new models include the software NVIDIA Omniverse Enterprise and AI Enterprise 4.0.

& Ampere ; nbsp;

How is the new GeForce RTX?

Starting with advertising highlights, workstations with Already launched RTX 6000 Reach to be the flagship of the new generation. The board uses an AD102 chip with an amazing 18176 CUDA cores, 2.5GHz frequency, 48GB GDDR6 memory with 384-bit interface, 960GB/s bandwidth, and 300W TDP (Total Design Power). The company says some machines can equip up to four of these GPUs, combining 5,828 teraflops and 192GB of video memory.

Of the “real” new releases, the RTX 5000 stands out the most. The model comes with the same chip as its bigger sibling, but only has 12,800 CUDA cores, a 2.5GHz clock, and 32GB of VRAM with a transfer speed of 578GB/s and a max consumption of 250W. The total teraflops per GPU is 65.3 floating points.

“The NVIDIA RTX 5000 Ada GPU demonstrates NVIDIA’s impressive generational performance improvements – it has greatly increased our efficiency in creating stereo panoramas with Enscape.” […] With the performance boost and large frame buffer of RTX 5000 GPUs, our large and complex models look great in VR, giving our customers a more comfortable contextual experience. Design by architecture firm Lake | Flato

The RTX 4000 gets a full size version

On the other hand, the RTX 4500 comes with a dedicated version of the AD104 chip. This graphics chip has 7680 CUDA cores, a frequency of 2.6GHz and 24GB of VRAM. The memory is connected to a 192-bit bus, which results in a bandwidth of 432 GB/s. The consumption is still relatively high, fixed at 210W, while the power is 39.4 Teraflops.

Using the same chip, the RTX 4000 is one of the more interesting SKUs announced this week. Signal Released already in March, over the GTC, as a small form factor (SFF). The new model has a standard size and more powerful specifications. The core number and amount of memories remains unchanged, but the frequencies increase from 1.56GHz to 2.2GHz. Bandwidth has also been increased from 320GB/s to 360GB/s, as well as consumption from 70W to 130W.

The new RTX for workstations has a very similar design and a blower-style fan to cool the system. On the back there are only DisplayPort 1.4a connections, except for the RTX 4000, which has an additional mini-DisplayPort connector.

Price and availability

The RTX 6000 will be part of the new workstations launching in the next few months, with a value of $6,800 USD (33,500 R$ in direct transfer, excluding taxes).

The RTX 5000 is now available through horse and other distribution partners such as Leadtek, PNY, and Ryoyo Electro for $4,000 USD (R$19,700).

The RTX 4500 and RTX 4000 will soon be available through brands like BOXX, Dalehorse and Lenovo Priced at $2,250 (R$11,000) and $1,250 (R$6,150), respectively.

Trending on Canaltech: