An Nvidia RTX 4090

(Image credit: Future)

New photos have surfaced showing what could be Nvidia RTX 5000 prototype designs with far larger coolers than we’ve seen from even leading current-gen Lovelace graphics cards. 

As Videocardz reports, there’s fresh evidence revealing that Nvidia is experimenting with triple and quad-slot coolers for its upcoming Blackwell RTX 50 series.

Add seasoning appropriately here, but should this cooler size be used for the next-generation GPUs, that would make them the thickest and largest Founders Edition cards by a considerable margin.

It’s believed that Nvidia is testing hefty coolers that could cater for a total power usage of 600W. However, a graphics card with such a power draw would need a whopping 1000W PSU for adequate overhead with overclocking. 

In fact, Benchlife tells us that Team Green has four plans in place for GeForce Blackwell coolers ranging between 250W and 600W. From this, we can guess that the lowest wattage is likely to be for the RTX 5060 and the highest-end would be for the RTX 5090. For reference, the RTX 4090 has a 450W TDP, so that’s an increase of 33% on the face of it.

However, let’s not jump to any conclusions yet. Firstly, even if the top-end cooler does support 600W, that doesn’t mean Nvidia (and board makers) will have the Blackwell flagship chugging that much wattage.

And secondly, just because Nvidia is testing out extra-large coolers doesn’t necessarily mean the company will implement them in Founders Edition models, or make them available to partners. These might be concepts that end up being abandoned (or perhaps are even already discarded).

Get the hottest deals available in your inbox plus news, reviews, opinion, analysis and more from the TechRadar team.

Factoring in the RTX 4090’s 450W TDP, it’s unlikely there will be a dramatic jump to 600W unless the GB202 (Blackwell) chip rumored to be inside the RTX 5090 is larger than the AD102 (Lovelace) in the 4090. 

With that said, it’s technically a possibility. The current 12VHWPR power connector used by Nvidia can cope with up to 600W.

However, factoring in the reported issues around melting and faulty adapters with the RTX 4090, it doesn’t seem likely that Team Green will turn the dial up to 11 here. What seems much more likely is that Nvidia is giving plenty of headroom for cooling, with a very capable cooling solution dealing with a lesser wattage to ensure there’s no room for error. All of this is just theorizing, of course, but if this happens, these cards will still be hefty pieces of hardware.

Graphics cards could be getting bigger

This isn’t the first time we’ve heard rumors or had confirmation of a quad-slot cooler, or mighty system power requirements. For reference, the new XFX RX 7900 XTX Phoenix Nirvana is a quad-slot AMD GPU requiring a total system power of 850W with its three PCIe power connectors.

It’s been rumored that the cooler for this graphics card variant was supposed to be used for the now-canceled Navi 4C die. Team Red was experimenting with a competitor to the RTX 4090 before production was halted. Nvidia appears to be experimenting in the same way, with the specs and hardware yet to be confirmed. 

You might also like

Aleksha McLoughlin is an experienced hardware writer. She was previously the Hardware Editor for TechRadar Gaming until September 2023. During this time, she looked after buying guides and wrote hardware reviews, news, and features. She has also contributed hardware content to the likes of PC Gamer, Trusted Reviews, Dexerto, Expert Reviews, and Android Central. When she isn’t working, you’ll often find her in mosh pits at metal gigs and festivals or listening to whatever new black and death metal has debuted that week.

[ For more curated Computing news, check out the main news page here]

The post Nvidia could use massive 600W-capable cooler for RTX 5090 – but don’t panic about flagship GPU being a power hog first appeared on

New reasons to get excited everyday.

Get the latest tech news delivered right in your mailbox

You may also like

Notify of
Inline Feedbacks
View all comments

More in computing