NVIDIA GeForce RTX 5000 Series “Blackwell” TDPs Leaked, Entire Lineup Unified with 12+4 Pin Power Connector

[dirtyferret]

via hardwarebusters

In an unexpected reveal, power supply manufacturer Seasonic may have just spilled the power requirements of NVIDIA’s upcoming GeForce RTX 50 Series GPUs, codenamed “Blackwell.” The information surfaced via Seasonic’s PSU wattage calculator, a handy tool for building PC configurations, which now includes entries for the RTX 50 family. If accurate, these leaks offer a sneak peek into the power consumption of NVIDIA’s next-gen cards, revealing both incremental improvements and higher TDPs across the board.

NVIDIA is unifying its power delivery across the entire GPU lineup for the Blackwell generation. All cards, from the entry-level RTX 5050 to the top-tier RTX 5090, will use the 12+4 pin 12V-2×6 connector, updated to comply with PCIe 6.0 CEM specifications. This marks a shift from previous architectures where only higher-end models adopted this power standard.


While this move simplifies cable management and ensures compatibility with future PCIe standards, the increased power demands—particularly for the RTX 5090, which could require up to 500W—raise questions about efficiency and thermal design. Enthusiasts will be eager to see how NVIDIA balances these higher power thresholds with performance gains.

GPU Model​	Blackwell TDP​	Equivalent Ada Lovelace TDP​	TDP Increase​
RTX 5050	100W	N/A	N/A
RTX 5060	170W	115W	+55W
RTX 5070	220W	200W	+20W
RTX 5080	350W	320W	+30W
RTX 5090	500W	450W	+50W

 

[freeagent]

I had to double check the Seasonic page to make sure I am covered, looks like it

My GPU cable says 600w

 

[inquisitor1]

tdp is nuts with gpu today.

 

[Vayra86]

Such a lazy release. Nvidia pushing the power button every other gen does not bode well.

 

[inquisitor1]

Such a lazy release. Nvidia pushing the power button every other gen does not bode well.

which should give a huge opportunity to amd to come up from the rear. intel too. amd did it with ryzen because intel sat on its ass not innovating, just bumping up tdp over time to ridiculous levels

 

[Vayra86]

AMD had its bigger opportunity post-Maxwell but they blew it on Raja. Now? Uphill battle.

AMD can't build a 5090.

 

[dirtyferret]

this is what got me
All cards, from the entry-level RTX 5050 to the top-tier RTX 5090, will use the 12+4 pin 12V-2×6 connector

really, the RTX 5050 & RTX 5060 need a 12+4?

 

[Bobaganoosh]

this is what got me
All cards, from the entry-level RTX 5050 to the top-tier RTX 5090, will use the 12+4 pin 12V-2×6 connector

really, the RTX 5050 & RTX 5060 need a 12+4?

Need? No. But it's easier for them to make the industry move over if all their cards use the same connector. Also, it will likely get just a little bit cheaper per connector with more volume of mid-range cards

 

[lilhasselhoffer]

which should give a huge opportunity to amd to come up from the rear. intel too. amd did it with ryzen because intel sat on its ass not innovating, just bumping up tdp over time to ridiculous levels

AMD had its bigger opportunity post-Maxwell but they blew it on Raja. Now? Uphill battle.

AMD can't build a 5090.

@inquisitor1 , I appreciate your optimism. Unfortunately, I think @Vayra86 is right in this particular case. AMD can't really build a GPU to compete with those upper echelons of performance. They haven't really pushed that for years...and it's partially because of bunches of little bits. That said, I hope their middle ground in the 8000 series competes with the 5070 or better. I think Intel still has higher end (middle to lowest high tier) Battlemage to release...and an effective competition at the $400-$500 range would be absolutely splendid.

For my money, I see Nvidia going the way of Intel with their software driving everything. That'll last for another 1-2 generations, where the LLM based AI runs its course and becomes the next coprocessor...at which point Nvidia will be in the same position they were in (and why they are under investigation) post the crypto boom, and they'll come back to gaming like they never treated gamers as second class options. At that point AMD will compete by not requiring a dying star to power the GPU, and we'll suddenly see everyone love AMD again. My word of the day is Evergreen.

-Edit-

Need? No. But it's easier for them to make the industry move over if all their cards use the same connector. Also, it will likely get just a little bit cheaper per connector with more volume of mid-range cards

While I don't doubt you...I was under the impression that Nvidia only released a limited number of founders cards. All others were/are made by board partners, who can design their own stuff. The requirements then would be defined by the subsequent requirements to board partners...who would not all be buying in volumes which might change the world. You may get a discount on your 16 conductor pinout...but the question is how much more might you get by halving it? Volume discounts are great, but rarely amount to decreasing the required materials by 50%... Heck, even with volume discounts you may only be looking at single digits of reduction. Between 50% less material, and less press setups (assuming injection molding), what is actually going to deliver more bang for your buck?


-Edit end-

 

[ir_cow]

I was hoping for a true 600 watt GPU :/

 

[dirtyferret]

Need? No. But it's easier for them to make the industry move over if all their cards use the same connector. Also, it will likely get just a little bit cheaper per connector with more volume of mid-range cards

I see that side of it but there are not a whole lot of 650w units out there with a 12V-2×6 connector and none that I know of that are 550w or lower. You don't need a 750w unit to run a 100w GPU (or even 170-200w). The PSU will need to step up to the plate in their offerings as well. Sure you can use an adapter but why force people to use an adapter when there really is no need to if you just kept the same connector.

I was hoping for a true 600 watt GPU :/

5090 titan? $3k electrical heater for your home

 

[Dristun]

everyone love AMD again

A lot of people here are ready to love AMD again the moment they'll release a card that's faster than nvidia's flagship (or at least an xx80). Mark my words, you'll even start seeing sentences about "RT being irrelevant" from camp Team Green if AMD get their stuff together and leapfrog Huang.

 

[Vayra86]

this is what got me
All cards, from the entry-level RTX 5050 to the top-tier RTX 5090, will use the 12+4 pin 12V-2×6 connector

really, the RTX 5050 & RTX 5060 need a 12+4?

Yes, because Nvidia said so...

 

[bonehead123]

Yes, because Nvidia said so...

And also because that's what Jacket Man uses on his newest/latest truckloads of jackets, to warm them up on cold mornings, hahahaha

 

[Vayra86]

And also because that's what Jacket Man uses on his newest/latest truckloads of jackets, to warm them up on cold mornings, hahahaha

I hope it gets bent

 

[N/A]

relax N2P will fix the power again. We had 10 and 20 series on the same node, then the odd 30 series which was a paper mining card, and now 40 and 50 same node. Of course to get more performance they have to get bigger and hotter

 

[x4it3n]

I had to double check the Seasonic page to make sure I am covered, looks like it

My GPU cable says 600w

Yeah those new 16-pin ports/cables (ATX 3.1/PCIe 5.1) are able to sustain 600W without any issue whereas the previous ones were more rated as 450W even if they could go "up to 600W" depending on quality...
That said, rating a cable at 600W when its max safe power delivery before melting is 630W doesn't seem very safe to me... 450W seems much better! Let's not forget the 8-pin were rated for 150W but could sustain 300W !

 

[Vayra86]

relax N2P will fix the power again for a single generation We had 10 and 20 series on the same node, then the odd 30 series which was a paper mining card, and now 40 and 50 same node. Of course to get more performance they have to get bigger and hotter

FTFY

Let's reiterate the fact that 10 series x80 ran on 180W TDP
Ada's 'super efficient' 4080 wants 320 TDP. On a 'new node'

The only reason Ada looked good is because Ampere was such a shitshow in efficiency and Turing was just Pascal with shitty RT.

 

[x4it3n]

Such a lazy release. Nvidia pushing the power button every other gen does not bode well.

To be fair they pushed on :
- Ampere because they didn't want AMD to win (AMD were using TSMC 7nm node when Nvidia were using Samsung 8nm aka 10nm Enhanced and much worse than TSMC 7nm)
- Lovelace because they were scared of RDNA3 after RDNA2's success and because RDNA3 was supposed to be much better too (architecture performance + extra shader cores removed (8192 to 6144) to stay efficient), whereas the 4090 rarely hits its 450W TDP @ Stock (I can testify)
- Blackwell because the 5nm Enhanced => 4nm node isn't bringing as much as 3nm could have, therefore power consumption has to go up...even if Blackwell is supposed to be more efficient architecture.

 

[londiste]

Such a lazy release. Nvidia pushing the power button every other gen does not bode well.

What makes it lazy exactly? Or what do you think would be the choice here?

Without moving to a newer manufacturing process there is not all that much efficiency to be found. There have been a couple examples otherwise but these are rare exceptions. There seems to be a reason both Nvidia and AMD are planning Broadwell and RDNA4 respectively to be built on N4P and not N3 or N3E.

The other way would be to limit the power to the same place as previous gen and get whatever performance boost they can get from either architecture changes or smaller process change. The last example of that was basically Turing. I doubt Nvidia would want to try that again.

And it is not only Nvidia who faces that same choice.

 

[Vayra86]

To be fair they pushed on :
- Ampere because they didn't want AMD to win (AMD were using TSMC 7nm node when Nvidia were using Samsung 8nm aka 10nm Enhanced and much worse than TSMC 7nm)
- Lovelace because they were scared of RDNA3 after RDNA2's success and because RDNA3 was supposed to be much better too (architecture performance + extra shader cores removed (8192 to 6144) to stay efficient), whereas the 4090 rarely hits its 450W TDP @ Stock (I can testify)
- Blackwell because the 5nm Enhanced => 4nm node isn't bringing as much as 3nm could have, therefore power consumption has to go up...even if Blackwell is supposed to be more efficient architecture.

I think the more pressing reason is more powah = moar clocks = lower shader count required = lower die size = bigger margins. I don't think AMD was ever in their minds while designing and positioning their stacks. AMD was happy designing their stuff for consoles already for quite a while - and Nvidia had SUPER warming up anyway.

What makes it lazy exactly? Or what do you think would be the choice here?

Without moving to a newer manufacturing process there is not all that much efficiency to be found. There have been a couple examples otherwise but these are rare exceptions. There seems to be a reason both Nvidia and AMD are planning Broadwell and RDNA4 respectively to be built on N4P and not N3 or N3E.

The other way would be to limit the power to the same place as previous gen and get whatever performance boost they can get from either architecture changes or smaller process change. The last example of that was basically Turing. I doubt Nvidia would want to try that again.

And it is not only Nvidia who faces that same choice.

Lazy because its just a power bump and there have been few if any architectural improvements coming our way. The Nvidia I knew and happily bought from truly innovated. What we've got now is a clusterfuck of vaseline smeared brute forced lighting that makes zero sense and is just there to fake the idea of progress. In the meantime, games play like absolute shit and Nvidia GPUs get slaughtered by Lumen just the same. But who knows, maybe Nvidia pulls out a rabbit on that front. Oh yeah, AI... of course.

 

[Lew Zealand]

Looking at the TDPs at the bottom end, this is Nvidia tacitly admitting the 4060 Ti was really a 4060, and the 4060 some interim 4060 LE that should sit above a 4050.

4060 Ti: 160W - 5060 170W
4050 (laptop) 55-95W typ. - 5050 100W

Power percentage increases are similar to the higher tier GPUs.

And FWIW, the desktop 4050 will be released in about a year as a 70W slot-powered GPU to replace the 3050 6GB with +40% performance.

 

[londiste]

Lazy because its just a power bump and there have been few if any architectural improvements coming our way.

Lets wait for the Blackwell architecture whitepaper first
Edit: The graphics one, applicable to RTX5000. There is one already for GB200 but that focuses almost exclusively on AI and these datacenter things have been quite different.

 

[TheinsanegamerN]

tdp is nuts with gpu today.

Why is it nuts? What is the arbitrary number GPUs should not go past, and why?

Lazy because its just a power bump and there have been few if any architectural improvements coming our way. The Nvidia I knew and happily bought from truly innovated. What we've got now is a clusterfuck of vaseline smeared brute forced lighting that makes zero sense and is just there to fake the idea of progress. In the meantime, games play like absolute shit and Nvidia GPUs get slaughtered by Lumen just the same. But who knows, maybe Nvidia pulls out a rabbit on that front. Oh yeah, AI... of course.

This ignores reality, ADA showed a significant bump from ampere in IPC, not just clock and core bumps. IIRC its a 16% increase when adjusted for core and clock count. It's ironic, the behavior you describe fits AMD's rDNA3 far better then Ada.

When hardware T&L came out, it crushed the first few gens of compatible cards. Can you imagine if the forums were as cynical as they are today? PC gaming would have been snuffed out because your geforce 256 couldnt play half life at 300 FPS on max settings.

The way games paly has nothing to do with nvidia, and game developer's inability to optimize their games is also not their fault.

 

[Steevo]

Is this the price we pay for the handfull of RT games to make RT more legitimate?