Nvidia mentioned to be testing an RTX 40 sequence GPU that wants as much as 900 watts
Rumor mill: Nvidia could possibly be making ready a monster GPU that can want two PCIe 5.0 energy connectors; probably a Titan RTX successor primarily based on the Ada Lovelace structure. At least the rumors thus far appear to point much more regular specs for the remainder of the RTX 4000 sequence, so there’s nonetheless hope that players will not want a small energy plant to energy Team Green’s next-generation GPUs.
With the RTX 3090 Ti graphics card, Nvidia squeezed the final little bit of efficiency it might from the Ampere structure by elevating the facility restrict to 450 watts. Independent testing has additionally revealed the cardboard is surprisingly environment friendly when the facility restrict is about to a extra sane 300 watts.
The next-generation Ada Lovelace structure remains to be shrouded in thriller, however that hasn’t stopped the rumor mill from turning hints about what to anticipate. According to infamous Twitter leaker @kopite7kimi, Nvidia could possibly be making ready a monster GPU that can dwarf the RTX 3090 Ti in lots of points.
Specifically, the corporate is alleged to be testing a full-fat AD102 GPU with a TGP of at least 900 watts, increased clock speeds, and 48 gigabytes of GDDR6X reminiscence. The check board reportedly comes with two PCIe 5.0 16-pin energy connectors versus the RTX 3090 Ti which comes with a single connector.
This could possibly be the primary Nvidia Titan product in years or just a sign that transient masses and energy spikes on engineering samples are too excessive to deal with with only one 16-pin energy connector. Leaked layouts for reference RTX 4000 sequence boards solely embody considered one of these, and are in any other case similar to Ampere reference designs. The worst-case situation is one the place the ridiculous April Fools idea from a yr in the past will materialize into an precise product.
Kopite7kimi says the remainder of the RTX 4000 sequence line-up can have related TGP scores to their Ampere predecessors. But extra importantly, the RTX 4080 might include 16 gigabytes of GDDR6X reminiscence whereas the RTX 4070 will sport 12 gigabytes of GDDR6. These are solely preliminary specs and thus topic to modifications down the road, however it could be a welcome enchancment over the RTX 3000 sequence.
The GeForce RTX 4000 sequence lineup is anticipated to interrupt cowl this fall and could possibly be the primary Nvidia GPUs which might be designed with the assistance of AI. In the meantime, PSU producers are preparing for this new technology of power-hungry graphics playing cards. The solely mannequin we have seen in inventory thus far is the Gigabyte GP-UD1000GM at Newegg, with a price ticket of $189.99.