Recommended Posts

This isn't really a question but more of a whoa...

So I purchased a eGPU occulink dock and card for my system to add a 2nd GPU... Already had a 4090 and moved to 5090 so through I'd use the 4090 to mess with LLM's and stuff..

Never expected to see at idle the GPU pulls 95 watts from the outlet... 95! This reading was taken from a kill-a-watt outlet meter the GPU was the only thing on the dedicated power supply.. I knew 4090's used 450ish watts when at full load but never expected 90 at idle sitting there with it's fans off doing nothing... 

the best part if you go into GPU monitor it claims it's only using 15 watts at idle, yet the kill-a-watt shows 90 being drawn from the wall outlet. at 100% it hits almost 450 watts and the kill-a-watt shows around 500

anyone else use eGPU's?

Link to comment
https://www.neowin.net/forum/topic/1463041-egpu-wattage/
Share on other sites

On 24/01/2026 at 11:11, Mindovermaster said:

I believe that the hardware in that eGPU dock uses power, too. So it's not just the GPU, right?

the eGPU dock only pulls 2 watts, there isn't much in it besids an occulink to PCI-X bridge chip and a few caps... if I put a smaller card in like a 4060 the idle wattage is only 11 watts

Link to comment
https://www.neowin.net/forum/topic/1463041-egpu-wattage/#findComment-599038612
Share on other sites

On 24/01/2026 at 12:54, Jester124 said:

I use one with a 9060 XT and nowhere near that much draw at idle. Mine sits at around 15 to 20 watts when not using it.

are you going by what the GPU says it's using or the GPU by itself plugged into it's own PSU isolated?

This is the GPU isolated to wall power and it pulls 90 watts yet the GPU claims it isn't in software that says 15 watts

Link to comment
https://www.neowin.net/forum/topic/1463041-egpu-wattage/#findComment-599038699
Share on other sites

On 24/01/2026 at 18:39, neufuse said:

are you going by what the GPU says it's using or the GPU by itself plugged into it's own PSU isolated?

This is the GPU isolated to wall power and it pulls 90 watts yet the GPU claims it isn't in software that says 15 watts

Mine is measured off the plug. Under full load though, mine does only ~165w from the card and pulls about ~170w from the plug. So far my unit has very little overhead for power consumption that I see. Mine is pretty basic though with the controller board and a 250w mini atx power supply. Nothing to beefy as I didn't need much to power the 9060 XT.

  • Like 1
Link to comment
https://www.neowin.net/forum/topic/1463041-egpu-wattage/#findComment-599038879
Share on other sites

That's really weird, what make of 4090 is it? Is there a BIOS/firmware update available for it?

Seeing as you confirmed the 4060 was only pulling 11 watts at idle, maybe there's a defect in the 4090.

I remember there was an issue with 7900 XTX's that prevented it from going into idle mode, so it was constantly drawing more power than it needed.

Link to comment
https://www.neowin.net/forum/topic/1463041-egpu-wattage/#findComment-599038893
Share on other sites

On 26/01/2026 at 02:58, Steven P. said:

That's really weird, what make of 4090 is it? Is there a BIOS/firmware update available for it?

Seeing as you confirmed the 4060 was only pulling 11 watts at idle, maybe there's a defect in the 4090.

I remember there was an issue with 7900 XTX's that prevented it from going into idle mode, so it was constantly drawing more power than it needed.

MSI Gaming GeForce RTX 4090

The fans spin non-stop, I don't think it's going idle at all i dont see any bios updates, not official ones at least

Link to comment
https://www.neowin.net/forum/topic/1463041-egpu-wattage/#findComment-599038910
Share on other sites

On 26/01/2026 at 07:36, neufuse said:

MSI Gaming GeForce RTX 4090

The fans spin non-stop, I don't think it's going idle at all i dont see any bios updates, not official ones at least

the fans do spin down when it's in the normal motherboard slot, but on the eGPU slot it seems to be running all the time

Edited by neufuse
Link to comment
https://www.neowin.net/forum/topic/1463041-egpu-wattage/#findComment-599038913
Share on other sites

On 26/01/2026 at 06:44, neufuse said:

the fans do spin off when it's in the normal motherboard slot, but on the eGPU slot it seems to be running all the time

Ok, that a big difference as mine are fully off on idle. Power management is working perfectly. If you disconnect the dock and power it on without being connected, do the fans spin up and stay on? Mine will do its initial boot like you would expect and go into power saving mode turning the fans off. May be a good way to test if it is something with the connection or the dock itself.

Link to comment
https://www.neowin.net/forum/topic/1463041-egpu-wattage/#findComment-599038918
Share on other sites

On 26/01/2026 at 08:18, Jester124 said:

Ok, that a big difference as mine are fully off on idle. Power management is working perfectly. If you disconnect the dock and power it on without being connected, do the fans spin up and stay on? Mine will do its initial boot like you would expect and go into power saving mode turning the fans off. May be a good way to test if it is something with the connection or the dock itself.

If I power on the dock with no occulink cable the fans come on and stay on, same as when its connected with the cable to the host system. The GPU is working fine, I can get it do work as a GPU or use CUDA fine with the occulink dock so it's working, just not sure why at idle it's using so much power.

Edit: ok so my 5090 seems to be similar... take it out of my workstation the system idles at 80 watts (Intel 285K)... put the 5090 back in the system idles at 145 watts... 65 watts for a GPU idle is kinda nuts... heck idling at 80 watts with the iGPU is nuts too (this is power at the wall outlet measured) meanwhile my Mac Mini 4 I bought as a spare system idles at 2.5-3 watts 🤣

Edited by neufuse
Link to comment
https://www.neowin.net/forum/topic/1463041-egpu-wattage/#findComment-599038953
Share on other sites

I believe the card is fine, it sounds like the dock is the issue. Could be a limitation of that dock and not an actual "problem". The only NVIDIA card I could try that was not in use was a GTX 1650 and that performed as expected with my dock as well.

Power management fully working, fans spin down when at idle and total draw was ~13w from the plug. Since the card is pulling ~8w idling and the dock empty draws ~10w when powered on, I would say it is the dock in your case.

 

Link to comment
https://www.neowin.net/forum/topic/1463041-egpu-wattage/#findComment-599038958
Share on other sites

On 26/01/2026 at 16:59, neufuse said:

Edit: ok so my 5090 seems to be similar... take it out of my workstation the system idles at 80 watts (Intel 285K)... put the 5090 back in the system idles at 145 watts... 65 watts for a GPU idle is kinda nuts... 

I am testing the GMKtec AD-GP1 which has a AMD Radeon RX 7600M XT and it idles at around 25W in Windows 25H2.

1768577835_windows_idle.webp

And that is quite a bit lower card than a 5090 😛 

Link to comment
https://www.neowin.net/forum/topic/1463041-egpu-wattage/#findComment-599038965
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
  • Recently Browsing   0 members

    • No registered users viewing this page.
  • Posts

    • The quantum search for Time's origin had an equally mind-boggling conclusion by Sayan Sen Image by Steve Johnson via Pexels A theoretical study from researchers at the University of Surrey suggested that the direction of time may not be fundamentally fixed in certain quantum systems. The work, published in Scientific Reports, examined how the “arrow of time” could emerge from microscopic physics and found that time-reversal symmetry can remain intact even in models used to describe processes such as energy loss and thermalisation. The arrow of time refers to the observed one-way direction from past to future in everyday life. In macroscopic processes, this is easy to see. Spilled milk spreads across a table and does not gather back into a glass, and heat flows from hotter objects to colder ones. These processes shape the common sense idea that time moves in a single direction. However, at the level of fundamental physics, many equations do not prefer a direction of time. Time-reversal symmetry means that the same physical laws can describe a system whether time moves forward or backward. This has made it difficult to explain why irreversible behaviour appears in the large-scale world even when the underlying rules do not require it. Dr Andrea Rocco, Associate Professor in Physics and Mathematical Biology at the University of Surrey, described this contrast: "One way to explain this is when you look at a process like spilt milk spreading across a table, it's clear that time is moving forward. But if you were to play that in reverse, like a movie, you'd immediately know something was wrong – it would be hard to believe milk could just gather back into a glass. However, there are processes, such as the motion of a pendulum, that look just as believable in reverse. The puzzle is that, at the most fundamental level, the laws of physics resemble the pendulum; they do not account for irreversible processes. Our findings suggest that while our common experience tells us that time only moves one way, we are just unaware that the opposite direction would have been equally possible." The study focused on open quantum systems, which are quantum systems that interact with a surrounding environment. This environment, often described as a heat bath, can exchange energy and information with the system. The researchers used this framework to study how a direction of time might appear even when the underlying physics does not enforce one. A key part of the analysis involved the Markov approximation. This is a simplification used in many models where the system is assumed not to retain memory of its past states. The idea is that changes depend only on the current state, not on earlier history. This is commonly used when studying thermalisation, which is the process where a system settles into equilibrium with its environment. The study also used concepts such as master equations, including the Lindblad and Pauli equations, which describe how probabilities of different quantum states change over time. Another related model discussed was quantum Brownian motion, which describes the random-like movement of a quantum particle interacting continuously with its environment. In these descriptions, a “memory kernel” can appear, which is a mathematical term that accounts for how past states influence current behaviour. The researchers found that applying the Markov approximation did not break time-reversal symmetry. Even when the system interacted with an effectively infinite heat bath, the resulting equations of motion remained symmetric in time. This meant that the same mathematical description could, in principle, run forward or backward in time without contradiction. The study further showed that standard frameworks used in open quantum systems, including quantum Brownian motion and master equations like the Lindblad and Pauli forms, could be written in a time-symmetric way. These equations are typically used to describe processes that look irreversible, such as dissipation and thermalisation, but the results suggested they can also be interpreted as allowing evolution in both time directions. Thomas Guff, Research Fellow in Quantum Thermodynamics, said: "The surprising part of this project was that even after making the standard simplifying assumption to our equations describing open quantum systems, the equations still behaved the same way whether the system was moving forwards or backwards in time. When we carefully worked through the maths, we found that this behaviour had to be the case because a key part of the equation, the "memory kernel," is symmetrical in time. We also found a small but important detail which is usually overlooked – a time discontinuous factor emerged that kept the time-symmetry property intact. It’s unusual to see such a mathematical mechanism in a physics equation because it's not continuous, and it was very surprising to see it appear so naturally." The researchers also noted that deriving a one-way arrow of time from time-reversal symmetric microscopic dynamics remains an open problem across fields such as thermodynamics, statistical mechanics, particle physics, and cosmology. Their results suggested that some standard descriptions of irreversible behaviour in open quantum systems may be better understood using a time-symmetric formulation of Markovianity. According to the study, processes such as thermalisation, which are usually treated as irreversible, could in theory be described in a way that allows evolution in either time direction under the same rules. This does not imply that time reversal occurs in everyday life, but rather that the underlying equations do not strictly enforce a single direction. Overall, the findings suggested that the perceived direction of time may emerge from how physical systems are modelled and approximated, rather than from a fundamental asymmetry in the laws themselves. The researchers noted that this perspective could have implications for ongoing work in quantum mechanics, thermodynamics, and cosmology on the origin of time’s arrow. Source: University of Surrey, Nature This article was generated with some help from AI and reviewed by an editor. Under Section 107 of the Copyright Act 1976, this material is used for the purpose of news reporting. Fair use is a use permitted by copyright statute that might otherwise be infringing
    • A bit premature... 100% Marketing. Bizarre.
    • A $300 price hike is insane! No one is going to want to pay that much!
    • Since the 1st one flopped, there is really no reason to make another one. It's just losing money left and right.
  • Recent Achievements

    • Reacting Well
      BizSAR earned a badge
      Reacting Well
    • First Post
      AndreaB earned a badge
      First Post
    • Week One Done
      Huge Trailer earned a badge
      Week One Done
    • Week One Done
      Classifyskilleducation earned a badge
      Week One Done
    • One Month Later
      eurospharma62 earned a badge
      One Month Later
  • Popular Contributors

    1. 1
      +primortal
      581
    2. 2
      +Edouard
      182
    3. 3
      PsYcHoKiLLa
      75
    4. 4
      Michael Scrip
      73
    5. 5
      neufuse
      64
  • Tell a friend

    Love Neowin? Tell a friend!