• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

Hard power off with 3090 under load

Associate
Joined
3 Aug 2006
Posts
610
Location
London, UK
Not sure if this should go here or in Power Supplies

Picked up a Inno3D GeForce RTX 3090 iChill X3 a couple of weeks ago as part of a new build, but whenever I put it under a heavy load - especially in VR - the whole PC powers off suddenly.

I've got an 850W power supply (Bitfenix Whisper M), greater than the recommended 750W so I thought it would be ok, but do I really need a 1000/1200W PSU instead, or is the card drawing more power than it's supposed to? It hasn't been overclocked

I've run GPU-Z and it seems to be drawing way more than the 40W per rail the PSU is rated for, but perhaps I'm not reading it correctly

amYqV5H.png


Am I right in that I just need a better PSU, or is here anything I might be overlooking?
 
Associate
Joined
6 Dec 2013
Posts
1,785
Location
Nottingham
under load crash can be a sign of the psu, did it do it before you upgraded on the old card first? what was in there prior. ive seen a list by evga where people where running the 3090 on the 750w version of your psu/

what else is in the system?

edit: yea 40a 12v rails might be the issue. 480w if you gpu is reaching or going over in a transient spike, that's causing the hard lock.

you could try to split the load over 2 rails so to speak, run 1 cable on 1 rail and 1 on another which should help, since there are 2 40amp gpu rails. :)
 
Last edited:
Associate
OP
Joined
3 Aug 2006
Posts
610
Location
London, UK
Only 40 AMPS per rail. You want 70A rail(s)+ ideally.
That's what I was thinking, but surely a 750W PSU would be even worse?

under load crash can be a sign of the psu, did it do it before you upgraded on the old card first? what was in there prior. ive seen a list by evga where people where running the 3090 on the 750w version of your psu/

what else is in the system?

edit: yea 40a 12v rails might be the issue.
It's all brand new. There's a Ryzen 5800, two NVME Drives and a WiFi card

Your 850 should be fine how old is the power supply?
are you using 2 separate pcie cables or 1 ?

Two separate cables. Does the same if I have them on the same rail, or two separate rails too.


I'm inclined to agree with you all that the PSU is just underpowered, just wanted to double check that kind of power draw was typical for a 3090 before I fork out for another PSU
 
Associate
Joined
24 Sep 2013
Posts
2,483
Location
Exmouth, Devon
That's what I was thinking, but surely a 750W PSU would be even worse?

Plenty of 750 and dare I say it 650W PSU's with stronger 12V rails than yours. It's not just the wattage, Rail amperage is where it's at. Your Bitfenix may say 750W but it's done across 2 rails of 40A with some switching between them. DUal or single rails are best when they are 70A+. Those EVGA 750W's that someone mentions above are 62A RAils so that's why they are running 3090's on 750's. 850W are 70A, 100W is 90A and, 1200W are 100A rails normally.

I remember reading some time ago that you want to be looking for a PSU with a 70A rail+ for GPU's with more than one 8 pin connector as weaker rails cause instability. You are better off with a single 70A rail than multi 40A rails.
 
Associate
OP
Joined
3 Aug 2006
Posts
610
Location
London, UK
It's not about the power rating, you just want to find one that can sustain peak current the 3090 requires :)

Plenty of 750 and dare I say it 650W PSU's with stronger 12V rails than yours. It's not just the wattage, Rail amperage is where it's at. Your Bitfenix may say 750W but it's done across 2 rails of 40A with some switching between them. DUal or single rails are best when they are 70A+. Those EVGA 750W's that someone mentions above are 62A RAils so that's why they are running 3090's on 750's. 850W are 70A, 100W is 90A and, 1200W are 100A rails normally.

I remember reading some time ago that you want to be looking for a PSU with a 70A rail+ for GPU's with more than one 8 pin connector as weaker rails cause instability. You are better off with a single 70A rail than multi 40A rails.

Good to know, thanks guys.

Given that the GPU-Z data shows the card pulling 70A+ of power on each 8 pin connector (currently set up on a rail each), would an 80-100A single rail PSU be sufficient? Or do I need to be looking for something that can supply two 80-100A rails?
 
Associate
Joined
24 Sep 2013
Posts
2,483
Location
Exmouth, Devon
Some PSU's allow you to change them from multi rail to single rail. DOes that Bitfenix allow you to do that. Bitfenix arent known for their PSU's.

Only other things to test are RAM seated correctly and run at stock speeds (2400Mhz) if you have fast RAM, re-seat the gfx card in the mobo as I think some power is drawn from the mobo across the PCI-E. Yeah recheck ALL connections. Run afterburner to log file (so you can look on reboot) for temperatures on the MOBO, GPU or CPU. They 'could' cause a shutdown - before you buy another PSU.

List out your whole build. Ryzen's and their RAM can be finickety things. What RAM & MOBO?
 
Associate
Joined
1 Jun 2019
Posts
419
Many others have said it already, but from personal experience chances are it's the PSU. I've had Corsair 750W PSUs do this as soon as the GPU was loaded, it was replaced and everything worked fine. When/if you do replace the PSU make sure you use new cables, they could be damaged by now.
 
Associate
OP
Joined
3 Aug 2006
Posts
610
Location
London, UK
Some PSU's allow you to change them from multi rail to single rail. DOes that Bitfenix allow you to do that. Bitfenix arent known for their PSU's.

Only other things to test are RAM seated correctly and run at stock speeds (2400Mhz) if you have fast RAM, re-seat the gfx card in the mobo as I think some power is drawn from the mobo across the PCI-E. Yeah recheck ALL connections. Run afterburner to log file (so you can look on reboot) for temperatures on the MOBO, GPU or CPU. They 'could' cause a shutdown - before you buy another PSU.

List out your whole build. Ryzen's and their RAM can be finickety things. What RAM & MOBO?
I've seen a couple of Corsair ones that do that, but I don't think this Bitfenix one does. I'll double check all the connections and everything inside, but for the most part everything works fine, even under a reasonably heavy load. The problems only really happen if I use VR or anything else particularly intensive
CPU is getting up to about 70C at it's highest, but I'll check in afterburner too in case it's happening when I can't see temps


Build is below. VR Headset is Oculus Quest 2

Gigabyte X570 GAMING X (AMD AM4) DDR4 X570 Chipset ATX Motherboard
MB-57X-GI

Bitfenix Whisper M Series 850W 80 Plus Gold Modular Power Supply
CA-22D-BX

Corsair Force MP600 series 1TB NVMe PCIe Gen4 M.2 Solid State Drive (CSSD-F1000GBMP600)
HD-068-CS

TP-Link Archer TX50E Dual-Band Wireless AX3000 (WiFi 6) Bluetooth 5.0 PCI-E Adapter
NW-222-TP

Inno3D GeForce RTX 3090 iChill X3 24GB GDDR6X PCI-Express Graphics Card
GX-07K-IN

Arctic Freezer 34 eSports White CPU Cooler - 120mm
HS-071-AR

AMD Ryzen 7 5800X Eight Core 4.7GHz (Socket AM4) Processor - Retail
CP-3CB-AM

Patriot Viper Steel 32GB (2x16GB) DDR4 PC4-28800C18 3600MHz Dual Channel Kit (PVS432G360C8K) - Currently running at default speeds
MY-107-PA

WD Blue SN550 2TB NVME M.2 2280 PCIe Gen3 Solid State Drive (WDS200T2B0C)
HD-57N-WD

MSI OPTIX MPG341CQR 34" 3440x1440 VA FreeSync 144Hz 1ms Curved LED Backlit Widescreen Gaming Monitor
MO-00Q-MS
 
Associate
Joined
22 May 2015
Posts
1,689
Location
Manchester
Bitfenix arent known for their PSU's.

You're right, they're not. But to muddy the waters even more, they recently started putting their badge on one or two good quality units. So it's a real minefield when it comes to power supplies. I tend to stick to the Seasonic or Corsair ones as most of those tend to be okay, there are one or two turds by both as well though.

OP, I'd agree with everyone else. Dual 40a rails probably equates to around 65a in real money and if the quality of the supply is questionable, then it won't sustain the load of a 3090. Go for one big single rail. That's what I've always done, this thing I have delivers 83a, just in case I install something power hungry. And if I don't, I have a whisper quiet system.

You can't link to competitors here, but you can get a make and model of something you're looking at and run it by people here. Woodsta seems very knowledgeable on this subject and will perhaps also weigh in with an opinion with any suggestions you make.
 
Associate
Joined
3 Feb 2012
Posts
2,206
Location
Bath
Many others have said it already, but from personal experience chances are it's the PSU. I've had Corsair 750W PSUs do this as soon as the GPU was loaded, it was replaced and everything worked fine. When/if you do replace the PSU make sure you use new cables, they could be damaged by now.

ALWAYS use the cables which come with a new modular PSU - the pinouts vary between manufacturers/models, and reusing the old cables can cause no-go or even damage.
 
Associate
Joined
3 Feb 2012
Posts
2,206
Location
Bath
Had it with vga power cables. Old ones were braided and looked nicer than the new ones... Nothing. Like the mains plug wasn't in.
Changed them over (after asking for advice here!) and ran perfectly like nothing ever happened. I was lucky...
 
Top Bottom