Anyone running a 4090 on sub 1KW PSU?

The 13900KS can pull close to 400 and the 4090 with a 600W power can pull over 500? Even if they both pull 75% of their max isn't that cutting it close for a 1000w?
Are you referring to power spikes when you say 600 watts? Because 4090 has a 450 watt TDP
 
Lots of FUD out here with power supply stuff. I have an AX1600i - which yeah, duh - is all good. But more importantly I know precise power usage. Never more than 700W...RTX 4090 STRIX OC and 7950X CPU.


power_7950x_4090.jpg
 
I run my FE with a 3-plug adapter (450W) and my 5900X on an 850W Seasonic. No power limits to either one. CPU is on water. No issues.

Once it all goes under water I'll switch to the OE Seasonic 600W 12VHPWR from BTOS Integration. Uses 2 x 8-pin plugs and is jumpered for 600W.
 
Alright gents and ladies if there are any... Verdict is in...

Received the card today. The thing is gigantic compared to my 3080. Photo attached, that 3080 was pulled out of an Alienware desktop. I actually had to remove one of the two intake fans on the front of my case to get it to fit.

I got too lazy and impatient to swap my EVGA Supernova G3 750 watter with my spare 850 watt Seasonic Focus PSU that's still new in box. So instead I simply grabbed a 3rd PCIe Cable and attached it to the G3. Card came with a 3x8 pin to 12VHPWR adapter

I know this is very preliminary, but did a couple Time Spy runs. One at default settings and another with 70% power target and both completed successfully with no power related issues. The performance difference was under 3%

Did several rounds of COD as well as a 15 minute furmark burn-in at 100% power target. No issues there either. I did hear the fan on the PSU ramp up about 8 minutes into the burn-in

I know running a PSU this close to its capacity is taboo to a lot of folks in here, myself included TBH but I figured if the 750 wasn't gonna pull it off, I would rather not tear into the system for an extra 100 watts.

Once the early adopter tax wears off or I see a good quality ATX 3.0 PSU posted in the "hot deals" section, i'll upgrade. Unless of course, I run into a reason to do it sooner.

2023-02-02 15.19.05.jpg
 
Alright gents and ladies if there are any... Verdict is in...

Received the card today. The thing is gigantic compared to my 3080. Photo attached, that 3080 was pulled out of an Alienware desktop. I actually had to remove one of the two intake fans on the front of my case to get it to fit.

I got too lazy and impatient to swap my EVGA Supernova G3 750 watter with my spare 850 watt Seasonic Focus PSU that's still new in box. So instead I simply grabbed a 3rd PCIe Cable and attached it to the G3. Card came with a 3x8 pin to 12VHPWR adapter

I know this is very preliminary, but did a couple Time Spy runs. One at default settings and another with 70% power target and both completed successfully with no power related issues. The performance difference was under 3%

Did several rounds of COD as well as a 15 minute furmark burn-in at 100% power target. No issues there either. I did hear the fan on the PSU ramp up about 8 minutes into the burn-in

I know running a PSU this close to its capacity is taboo to a lot of folks in here, myself included TBH but I figured if the 750 wasn't gonna pull it off, I would rather not tear into the system for an extra 100 watts.

Once the early adopter tax wears off or I see a good quality ATX 3.0 PSU posted in the "hot deals" section, i'll upgrade. Unless of course, I run into a reason to do it sooner.

View attachment 546241
Nice.

Thanks for the tests. However, if the difference at 70% power is only 3%, then could it be that your RTX 4090 is already power-bottlenecked?

Even if it doesn't crash the system, it could be that a more powerful PSU could give it more juice, allowing you to actually score better with a PSU upgrade. I dunno if it's true though, so if you upgrade your PSU, go ahead and let us know the results.

And don't forget to update your sig here, now that you have an RTX 4090. ;)
 
Nice.

Thanks for the tests. However, if the difference at 70% power is only 3%, then could it be that your RTX 4090 is already power-bottlenecked?

Even if it doesn't crash the system, it could be that a more powerful PSU could give it more juice, allowing you to actually score better with a PSU upgrade. I dunno if it's true though, so if you upgrade your PSU, go ahead and let us know the results.

And don't forget to update your sig here, now that you have an RTX 4090. ;)
It’s pushing right up to its 450 watt tdp in furmark so, it isn’t starved. Stronger psus don’t increase performance. Doesn’t work that way. If you’re over drawing it just shuts off. It’s been very widely reported and tested that you can run as low as 55% power limit and still retain 90% performance.

sig update incoming
 
Transient load spikes are well controlled on a 4090 vs 3000 series, it will run on an 850w quality PSU with highend CPU. Don't try and overclock both.
 
It’s pushing right up to its 450 watt tdp in furmark so, it isn’t starved. Stronger psus don’t increase performance. Doesn’t work that way. If you’re over drawing it just shuts off. It’s been very widely reported and tested that you can run as low as 55% power limit and still retain 90% performance.

sig update incoming
I suppose what you say sounds right, and thank goodness for it too. I'd rather deal with just buying a new PSU rather than wondering if I'm leaving performance on the table. So I guess it's just a matter of time before I buy an RTX 4090 then, considering I actually am rather sure that my current PSU will be able to handle it, if yours can.

And then my sig can get updated too hahah. Damn this RX 590.... It's a good one; the specialest edition one I could find, with all the pluses, Xes and other cool gimmicks to promise me that it's the best binned chip.

But it's still an RX 590. Tsk.
 
Ya I'm with you on that Goat. I know I can run everything with tuning, but I want to be comfy knowing I am not leaving any performance sacrificed or running close the limit if I unleash Its full power that's why I have a preorder or backorder in for the MSI Meg Ai1300 Platinum PS should be coming into stock next week If Newegg estimate is correct.
 
Three hour War Zone BR session while also on video chat with my squad, no problems at all. Even alt+tabbed a couple times which I guess is what can cause those transient spikes.

I was surprised to see around 88% VRAM utilization. I know lots of modern games like to cash as much assets as possible and it doesn’t actually “need” that much vram but was still surprised to see it so high.

GPU temps in the low 60‘s with average GPU utilization in the high 70s, so I’m being CPU limited here.
 
Three hour War Zone BR session while also on video chat with my squad, no problems at all. Even alt+tabbed a couple times which I guess is what can cause those transient spikes.

I was surprised to see around 88% VRAM utilization. I know lots of modern games like to cash as much assets as possible and it doesn’t actually “need” that much vram but was still surprised to see it so high.

GPU temps in the low 60‘s with average GPU utilization in the high 70s, so I’m being CPU limited here.
I would for sure not be bothered by high VRAM usage. That sort of thing means that tons of stuff is cached, tons indeed, and that will promise you a very nice gaming experience. To actually use all that VRAM is certainly a good thing, especially when paired with the raw beast of an nVidia processor.


If you didn't change your settings in War Zone between GPUs, I'd enjoy knowing the before-and-after average frames-per-second, to see just how powerful the 4090 really is. Differences in VRAM usage might also be an interesting factor to consider.
 
Yeah I’m not worried about it. Just didn’t expect to see it. I did notice there was virtually no new textures popping in with the 4090. Distant textures would get more detailed the closer you get vs full on buildings appearing in the distance the closer you get with a 10GB buffer.
 
Per subject line, anyone? Issues? 4090 on its way and currently running a 850watt Seasonic. I do plan on setting power limit to 60-70% since it appears to have a negligible performance impact but quite a large power impact. Based on some reviews it should pull roughly the same power as a 3080Ti which is marginally thirstier than my current 3080. Seems transient spikes are also far better controlled on 40 series cards than 30 series.

At some point I will get around to upgrading PSU but ATX 3.0/PCIe 5 PSU's are still few and far between and the new Seasonic Vertex units that just popped up on Newegg are all back ordered.
I would first ask, what your current power usage is for your GPU.
Use GPU-Z to monitor the power, mainly total Board Power Draw. Also, set the pcie voltage sensors to record the minimum seen voltage. This is to check if you are already overloading the PSU, which can cause the output voltage to drop. Click in the values field to toggle each one.

100% power target, +150 core, +1500 ram:

1677252964593.png


At 100% power target the most I have ever seen from my 4090 was 447W.
At 90% power it pulls 408W
At 80% power it pulls 371W

I haven't tried 70% yet, but the power target is a hard target. Overclocking does not change the power draw.

Once you see what your current card safely pulls, we can make an educated guess to the required power draw to set on the 4090. You might be able to get away with 80%. At 80% power target I lose 3.5% graphics performance in a timespy extreme bench.
 
I would first ask, what your current power usage is for your GPU.
Use GPU-Z to monitor the power, mainly total Board Power Draw. Also, set the pcie voltage sensors to record the minimum seen voltage. This is to check if you are already overloading the PSU, which can cause the output voltage to drop. Click in the values field to toggle each one.

100% power target, +150 core, +1500 ram:

View attachment 551583

At 100% power target the most I have ever seen from my 4090 was 447W.
At 90% power it pulls 408W
At 80% power it pulls 371W

I haven't tried 70% yet, but the power target is a hard target. Overclocking does not change the power draw.

Once you see what your current card safely pulls, we can make an educated guess to the required power draw to set on the 4090. You might be able to get away with 80%. At 80% power target I lose 3.5% graphics performance in a timespy extreme bench.
I’ve done 100% as well as 106% as well as 106% with a voltage bump across varying loads. 750 watt EVGA is holding up just fine across every load I’ve put on it at those settings. I have the MSI Gaming X Trio which does not have the 600 watt bios that the FE and Strix have. I’ve been logging all the results will post them once I’m back at my desk.
 
I'm using a seasonic 850W power supply with my MSI 4090 liquid and 5800x3d build, and minimal other stuff (about 10 USB devices, 1 nvme SSD). No issues so far.
 
I would first ask, what your current power usage is for your GPU.
Use GPU-Z to monitor the power, mainly total Board Power Draw. Also, set the pcie voltage sensors to record the minimum seen voltage. This is to check if you are already overloading the PSU, which can cause the output voltage to drop. Click in the values field to toggle each one.

100% power target, +150 core, +1500 ram:

View attachment 551583

At 100% power target the most I have ever seen from my 4090 was 447W.
At 90% power it pulls 408W
At 80% power it pulls 371W

I haven't tried 70% yet, but the power target is a hard target. Overclocking does not change the power draw.

Once you see what your current card safely pulls, we can make an educated guess to the required power draw to set on the 4090. You might be able to get away with 80%. At 80% power target I lose 3.5% graphics performance in a timespy extreme bench.
My peak power usage was 460 watts running TimeSpy Extreme with 106% power target and topping out the vcore to 1.1v

I also just swapped out my eVGA Supernova G3 750w with a Seasonic Vertex 1200. Wasn't planning on it, but another thread with some performance anomalies as they relate to power got me curious. I'm now using a native 12vhpwr connection instead of a 3x8pin. I have a feeling i'll regret this purchase. Honestly, I don't expec to see any performance or power consumption difference and the quality of the eVGA PSU was far better than the Seasonic. The cables on the Vertex are stiff and not sleeved. The Supernova cables are sleeved and easily malleable.
 

Attachments

  • 2023-02-24 11.43.39.jpg
    2023-02-24 11.43.39.jpg
    693.8 KB · Views: 1
My peak power usage was 460 watts running TimeSpy Extreme with 106% power target and topping out the vcore to 1.1v

I also just swapped out my eVGA Supernova G3 750w with a Seasonic Vertex 1200. Wasn't planning on it, but another thread with some performance anomalies as they relate to power got me curious. I'm now using a native 12vhpwr connection instead of a 3x8pin. I have a feeling i'll regret this purchase. Honestly, I don't expec to see any performance or power consumption difference and the quality of the eVGA PSU was far better than the Seasonic. The cables on the Vertex are stiff and not sleeved. The Supernova cables are sleeved and easily malleable.

You don't like stiff power cables? Hahah that just means they're thicker. And a thick cable can likely carry more current.
I wouldn't see it as a bad thing at all, even if it's tougher to work with. Try using a blow-dryer to lightly warm the cables before you set them, and you might have better luck if they're just being too stiff.

-

Oh, and make sure to run before and after benchmarks with nothing changed other than the PSU.
 
You don't like stiff power cables? Hahah that just means they're thicker. And a thick cable can likely carry more current.
I wouldn't see it as a bad thing at all, even if it's tougher to work with. Try using a blow-dryer to lightly warm the cables before you set them, and you might have better luck if they're just being too stiff.

-

Oh, and make sure to run before and after benchmarks with nothing changed other than the PSU.
The Seasonic is .1mm thicker. I don't know if that's the actual wire guage, the jacketing or a combination of the two.

Benchmarks are virtually identical. Was actually about a dozen points lower in Time Spy but that's well within the margin of error.
 
My peak power usage was 460 watts running TimeSpy Extreme with 106% power target and topping out the vcore to 1.1v

I also just swapped out my eVGA Supernova G3 750w with a Seasonic Vertex 1200. Wasn't planning on it, but another thread with some performance anomalies as they relate to power got me curious. I'm now using a native 12vhpwr connection instead of a 3x8pin. I have a feeling i'll regret this purchase. Honestly, I don't expec to see any performance or power consumption difference and the quality of the eVGA PSU was far better than the Seasonic. The cables on the Vertex are stiff and not sleeved. The Supernova cables are sleeved and easily malleable.

Another quick update... I'm feeling a little better about this purchase now... There is no added performance and the GPU isn't drawing any more or any less power.... Buuuuut.... The atrocious coil whine I was having with this MSI Gaming X Trio is virtually eliminated. Couldn't tell you if it was the PSU whining all along or if the card is happier with the Seasonic, but it looks like I should probably go back and edit my Newegg review of the card.
 
Another quick update... I'm feeling a little better about this purchase now... There is no added performance and the GPU isn't drawing any more or any less power.... Buuuuut.... The atrocious coil whine I was having with this MSI Gaming X Trio is virtually eliminated. Couldn't tell you if it was the PSU whining all along or if the card is happier with the Seasonic, but it looks like I should probably go back and edit my Newegg review of the card.
The coil whine is actually fixed? From changing the PSU? Oh my goodness. I think there was a thread about this where people simply concluded that coil whine couldn't be fixed. But you say it's gone? Interesting.
 
The coil whine is actually fixed? From changing the PSU? Oh my goodness. I think there was a thread about this where people simply concluded that coil whine couldn't be fixed. But you say it's gone? Interesting.
It appears to be resolved for me anyway. I tested varying load conditions including having furmark rendering 600+ fps
 
Oh jeex is coil whine a thing again? I thought we were done with that BS
 
Oh jeex is coil whine a thing again? I thought we were done with that BS
I’ve never had an issue with coil wine with my personal systems in all my years of PC building until I got this 4090. Luckily that only lasted a couple weeks until I just swapped the PSU
 
Nice.

Thanks for the tests. However, if the difference at 70% power is only 3%, then could it be that your RTX 4090 is already power-bottlenecked?

Even if it doesn't crash the system, it could be that a more powerful PSU could give it more juice, allowing you to actually score better with a PSU upgrade. I dunno if it's true though, so if you upgrade your PSU, go ahead and let us know the results.

And don't forget to update your sig here, now that you have an RTX 4090. ;)

It means he has a golden sample card more likely than not. Frame Chasers (whether you like him or not) tested a bunch of 4090s running Furmark. AFAIK one card pulled like 460W, another pulled 430W, but one of them only pulled 320W, all at 100% power target with out of the box settings. 70% power target would be 315W on a 450W BIOS.
 
I have a seasonic focus gold 850 watt + i9 18500k + asus TUF OC 4090. All at stock settings. Been stress testing it with cp2077 at 4k ultra RT (psycho) settings. I haven’t had a single issue with stability. Seems like the max power draw from gpu when playing is about 350 watts or so.
I’m confidant that 850 watts will be plenty for my setup.
 

Attachments

  • B55E8C03-4442-4DEB-B5D7-E46A0F7873C9.jpeg
    B55E8C03-4442-4DEB-B5D7-E46A0F7873C9.jpeg
    1.3 MB · Views: 0
I have a seasonic focus gold 850 watt + i9 18500k + asus TUF OC 4090. All at stock settings. Been stress testing it with cp2077 at 4k ultra RT (psycho) settings. I haven’t had a single issue with stability. Seems like the max power draw from gpu when playing is about 350 watts or so.
I’m confidant that 850 watts will be plenty for my setup.
Nice! Gives us all hope that we don't need to run out and buy 1000w PSUs
 
Nice! Gives us all hope that we don't need to run out and buy 1000w PSUs
Yea. I was afraid at first. To be honest I did some back to back testing with my 3080 and my 3080 was drawing just as much as my 4090 but running much hotter so I’m not sweating the power anymore.
 

Attachments

  • 2217F074-1BFD-446E-BF61-54DF4C6DE414.jpeg
    2217F074-1BFD-446E-BF61-54DF4C6DE414.jpeg
    244.2 KB · Views: 0
I was curious so I fired up Furmark.

Zotac 4090 / 5800x / water loop

600w from the wall. This is not typical or real world. Running games or F@H pulls down 450ish for me. Absolutely no worries on my Seasonic 850 focus platinum.
 
I have a seasonic focus gold 850 watt + i9 18500k + asus TUF OC 4090. All at stock settings. Been stress testing it with cp2077 at 4k ultra RT (psycho) settings. I haven’t had a single issue with stability. Seems like the max power draw from gpu when playing is about 350 watts or so.
I’m confidant that 850 watts will be plenty for my setup.
Whoa. You got an 18th Gen i9!?
 
Mainly it's if you have a heavily overclocked 13900K that's over 300W and a 4090 pulling more than 450 is when you'll be getting close to 1k power if you have a lot of peripherals in your setup. Most basic non overclocked pcs even with the latest hardware can get away with an 850 and comfy with a 1K.
 
I'm running a Seasonic titanium 850W with zero issues so far in games and benchmarks.

Running 4090 at stock, maxes out to right under 450W
5800x3D uses ~90W most game loads, max 105W
(3) NVME drives
360mm AIO w 9 RGB fans
 
No issues running my 4090 on a Corasair HX850i with a 7700X, (2) NVMe's, 1 SSD, EK AIO + 7 total fans. Debating picking up a newer PSU once there are more options out there and giving my HX to my son to replace his ancient LEPA G900-MA 900 W that has been running happily since late 2011.
 
A quick search found this:
https://hwbusters.com/gpu/nvidia-rtx-4090-detailed-power-analysis-ideal-power-supply/
I paste the image here. The max system peak power is 780W, max system average power 602W.
You can find many other similar results online.

So a decent 750W PSU is good enough.
If you undervolt GPU to 350W, even 650W PSU is good enough.

I smell that the PSU sellers are using this opportunity to play psych game, selling fear & uncertainty. Really no need to play along to buy 1KW PSU.

Results.jpg
 
Running a 1300 watts. Think anyone should be fine around the 850+ mark.

750 will run but if it starts drawing 520-550 watts (depending on card etc) then it might be too much.
 
A quick search found this:
https://hwbusters.com/gpu/nvidia-rtx-4090-detailed-power-analysis-ideal-power-supply/
I paste the image here. The max system peak power is 780W, max system average power 602W.
You can find many other similar results online.

So a decent 750W PSU is good enough.
If you undervolt GPU to 350W, even 650W PSU is good enough.

I smell that the PSU sellers are using this opportunity to play psych game, selling fear & uncertainty. Really no need to play along to buy 1KW PSU.

View attachment 559558
This is a nice chart thanks. What cpu was used and at what clock speed?
 
Been running an MSI 4090 Gaming X Trio/7950X with a Corsair HX850W just fine. I play FS2020/Black Desert/GTAV.
Think the most graphically intensive game I tried was Hogwarts Legacy and also had no issues at all.
 
Back
Top