Nvidia Clears the Air on GeForce RTX 40-series Power Requirements

GeForce RTX 4090 Founders Edition
GeForce RTX 4090 Founders Edition (Image credit: Nvidia)

One of the biggest questions surrounding Nvidia's GeForce RTX 40-series (Ada Lovelace) graphics cards is whether you need to upgrade your power supply. Nvidia has uploaded a new FAQ (opens in new tab) that sheds some light on Ada's power requirements for future adopters.

Consumers won't need to upgrade their power supplies if the capacity meets Nvidia's wattage recommendations. The GeForce RTX 4090 has a 450W TGP, meaning you need at least a 850W unit. The GeForce RTX 4080 12GB (285W) requires a 700W power supply while the GeForce RTX 4080 16GB (320W) will need a 750W unit. 

So there you have it — as long as your power supply has enough juice and comes from a reputable vendor, you don't have to upgrade just yet. If yours don't comply with any requirements, you probably want to pick up one of the best power supplies.

For the GeForce RTX 40-series graphics card, Nvidia opted to use the new PCIe 5.0 power connector (12VHPWR) to reduce cable clutter and improve airflow inside the case. Some may be concerned about the 12VHPWR's fragile life cycle. However, the 30-cycle specification is the same standard for any PCIe or ATX connectors that have been with us for years, and it's unlikely that the typical user will plug and unplug the connector consistently.

The new ATX 3.0 power supplies come with the 12VHPWR connector. However, older units will have to rely on old-fashioned power adapters. According to Nvidia, GeForce RTX 40-series graphics cards will come with necessary power adapters, allowing existing ATX 2.0 power supplies to feed Ada with conventional 8-pin PCIe power connectors. These adapters can arrive with a three or four 8-pin to 16-pin designs.

There was a leaked email from a PCI-SIG member making rounds on the Internet about a potential overcurrent or overpower hazard with ATX 2.0 power supplies using 12VHPWR connectors. Obviously, the report caused a lot of panic among the enthusiast community. 

However, some publications misinterpreted the PCI SIG's message: what PCI-SIG wanted to warn its members was that there was proof that some 12VHPWR connectors suffered from thermal variance under certain scenarios due to the components used in their production. It's likely that the components didn't conform with PCI SIG's requirements. Nvidia has confirmed that it was an isolated problem with a prototype connector from a specific vendor, and the supplier has since solved the issue.

With the 12VHPWR connector, four signal contacts are just sense wires. The signal contacts are not present with the power adapter, which is why the power adapters included with GeForce RTX 40-series graphics cards sport an active circuit that smartly picks up the power capacity based on the number of 8-pin PCIe connectors and report it to the graphics card. It is very useful since the minimum requirement for the GeForce RTX 4090 and RTX 4080 16GB is three 8-pin PCIe power connectors. So, for example, when the circuit detects four 8-pin PCIe power connectors, it permits the GeForce RTX 4090 to raise the power headroom for overclocking.

Whether to upgrade to an ATX 3.0 power supply depends on personal preference. It's not a big issue if you use a suitable adapter, but some may be against the usage of adapters with high amperage applications. However, it's understandable that not all users can or want to change their power supplies for a new one.

Zhiye Liu
RAM Reviewer and News Editor

Zhiye Liu is a Freelance News Writer at Tom’s Hardware US. Although he loves everything that’s hardware, he has a soft spot for CPUs, GPUs, and RAM.

  • itchyeyes
    The power requirements have stayed the same or reduced while delivering substantially more performance. RTX 4090 draws the same power as RTX 3090 Ti, the RTX 4080 16GB draws 30W less than RTX 3080 Ti, and the RTX 4080 12GB draws 65W less power than RTX 3080 12GB.
    This is such an absolute horseshit statement, I can't believe they have the audacity to print it. Litterally comparing every single one of these cards to the next card up the lineup in the 3000 series.
    Reply
  • peachpuff
    "However, the 30-cycle specification is the same standard for any PCIe or ATX connectors that have been with us for years"

    Oh really? 🤔🤔🤔
    Reply
  • spongiemaster
    itchyeyes said:
    This is such an absolute horseshit statement, I can't believe they have the audacity to print it. Litterally comparing every single one of these cards to the next card up the lineup in the 3000 series.
    What are you talking about? The model names are completely irrelevant to the point Nvidia was making. Even without exact performance numbers for the 4000 series, it should be pretty safe to assume what they have said is accurate. More performance with less power used.
    Reply
  • Phaaze88
    This card will need this much wattage... that card needs that much wattage... and quality is once again thrown to the wayside.
    Then reviews show us what power transients are like for these - whether they are worse than Ampere or not.
    Reply
  • FransFebri
    spongiemaster said:
    What are you talking about? The model names are completely irrelevant to the point Nvidia was making. Even without exact performance numbers for the 4000 series, it should be pretty safe to assume what they have said is accurate. More performance with less power used.
    No, it is relevant. The question on that FAQ asking about power requirement compared to last gen, and Nvidia says it stayed the same or reduced (1st point) while delivering more performance (2nd point).
    There is no argument about the 2nd point, 1st point is 100% wrong and misleading.
    Basically if you are now using 3090, you probably want to upgrade to 4090 and not to 4080. If you followed Nvidia PSU recommendation, your old PSU will be 750W, 100W less than Nvidia recommendation. Are you telling me that is accurate?

    Let us move to the not so 4080 tier. Compared to 3080, this 4080 tier is a joke because they use smaller chip for the 16GB variant, and an even smaller chip for 12GB variant. In the 3000 series stack, 4080 12GB would be 3060. At the very least they should call it 4070, but no, they just want more money from it thus called it 4080 12GB. Because of this, comparing it fairly will be difficult. If you just drink Nvidia kool-aid, then yes, Nvidia is kinda right. Kinda because they didn't compare 4080 16GB to 3080 but using the Ti variant. But since they have 2 4080, then I guess it is okay by Nvidia's standard. But if we want a real comparison, 4080 16GB is using similar power to 3080 with the same PSU requirement, thus 3080 owner can upgrade to 4080 16GB without changing their PSU. 4080 12GB should be called at least 4070 and I'm being generous here, then it is 4080 12GB 285W vs 3070 220W GPU.
    Chip size wise, the comparison should be 4080 16GB 320W to 3070 220W and 4080 12GB 285W to 3060 170W.

    Personally I don't mind if the new card is using more power since it does pack more transistor per area, but Nvidia is definitely trying to make it look better than it is (from power requirement perspective, not performance).
    Reply
  • boe rhae
    nVidia screwed up 40xx cards by lying-by-omission about their 4070 being a 4080, insulting EVGA until they dipped, and now this mess. At this point they can talk all they want. If I were to buy one of their cards (I'm probably gonna go Intel or AMD this time though even though I've never tried their cards before), I'd get a new PSU as well. Maybe they're telling the truth but my trust is very low in their brand right now.

    Or better said: It's not really distrust, it's trust (12GB).
    Reply
  • spongiemaster
    FransFebri said:
    No, it is relevant. The question on that FAQ asking about power requirement compared to last gen, and Nvidia says it stayed the same or reduced (1st point) while delivering more performance (2nd point).
    There is no argument about the 2nd point, 1st point is 100% wrong and misleading.
    It isn't. The 3000 series had so many variants bunched together that you can pick lower cards and Nvidia is still right. Don't like comparing the 12GB 3080Ti to a 16GB 4080? Fine, the regular 12GB 3080 12GB has a 350W TGP just like the Ti. No change, but bigger performance advantage for 4080. Don't like that one either? Even the original 10GB 3080 has a 320W TGP, which is equal to the 16GB 4080. Nvidia is still correct. Equal power, 60% more VRAM, better performance. Don't like the 12GB 3080 compared to the 12GB 4080? Looks perfectly logical to me. You can go down two tiers to the 8GB 3070Ti with a 290W TGP which is higher than the 285W for the 12GB 4080. Less power, twice as much VRAM, better performance. Dropping all the way down to the 8GB 3070 would be BS as it uses regular GDDR6 VRAM which uses significantly less power than the the GDDR6X in the 4080. 4090 vs 3090Ti? It was pretty widely reported when the 3090Ti was released, that it was basically a test platform for what the 4090 was going to be. Equal power, $400 cheaper, same amount of VRAM, should be much faster.
    Reply