Is it all the fault of the PSU?

User Support Hub Hardware & Peripherals
1

Not necessarily.

To be honest, I have had a hard time with msfs for the last months. I have constant crashes to reboot, usually after about 15 to 45 minutes.I was just happy to finish 30min short trips without an crash. It started some 6 to 8 months ago and become worsier and worsier over the time.

And I tried everything. I tweak and fiddle my settings, throttle my 3090 to 1590Mhz, improved my cooling system, all without luck.
OK, I have read about PSU weakness here but because I am running an 850W PSU from an renowned brand and have max. loads of around 600 Watts I never thought it could be the PSU.

And so I blamed it all on microsobo.

Running out of ideas and wondering that it can’t get any worse I bought a new 1300W premium PSU.

In a word: All problems gone (crossing fingers). Since I have installed the new PSU I have had 4 flights of more than one hour including a 150min low altitude VFR trip from Canberra to Brisbane with a lot of addons, PG and heavy cloudy live weather. So what folks, as hard to believe, an 850W PSU is to weak for msfs on an 3090.

Well, do I have to apologize to microsobo now? Partly!

I think everyone in the picture has their share, Microsobo for their more and more “offensive” code (the system was running fine until last winter), Nvidia for the graphics card’s power hunger and EVGA for their oversensitive power supply, which can not absorb these power peaks despite a theoretical power reserve of 40%.

In any case, I would like to recommend anyone who has such crash to reboot problems to think about a new power supply with at least 60 or 70% headroom. It would have saved me a lot of worries if I had believed it more.

Lucky flying!

P.S. Australias photogrammetry is awesome! Perth, Adelaide, Melbourne, Sydney, Gold Cost are a pleasure to view.

1 Like
2

I believe when buying a 3080ti the store I got it from recommended a minimum 850 watt PSU so yeah imagine the 3090 needs between that and a 1000 watt I went with a 1000w when I got the 3080ti as not much difference in the price at the time.

3

Saw this video last week. There’s definitely the possibility of issues with even “good” power supplies and higher-tier cards.

2 Likes
4

Well you definitely should have taken a more methodical approach.

  1. You say MSFS has been “crashing” - did you go into Event Viewer and make note of the reason for each crash? What were they - exactly.
  2. You say you have a RTX 3090. Which one? They are not homogeneous and some come OC’d and some are reference architecture, meaning there are different power requirements.
  3. As a rule of thumb 850 Watt is okay for a reference spec’d 3090 but doesn’t leave much overhead for overclocking. 1000 WATT is plenty. 1300 WATT is overkill unless you’re planning on doing Cryptomining or other exotic things with your card.
  4. You could simply have a bad powersupply but again a methodical approach of putting a voltmeter to your PSU would tell you a lot.
  5. Since 99% of other MSFS users are running just fine with a 850 - 1000 watt PSU with 3080 or 3090, you’d have to think that the issue is with your system and not Asobo.
  6. Did you OC your CPU? That’s also going to draw more power at load?
  7. What kind of cooling system do you have to dissipate all that heat? What are your temps under load (CPU and GPU)
  8. What kind of power supply do you have? Did you try moving the GPU power cables to another rail (assuming it’s modular?). Did you notice the PSU fan at high speed under load - struggling to keep up?
  9. Do you have your GPU mounted fan up or fan down?

So no, I would NOT suggest that someone who has MSFS crashes to rush out and replace their PSU unless they’ve done some homework first and have VERIFIED that its a PSU issue.

2 Likes
5

I use PC Part Picker to calculate the potential load on my system. Once armed with that information, a good rule of thumb is to run the PSU about 50% with a “do not exceed” of 60% draw at peak.

I’ve read over the years that most PSUs have a “sweet spot” in that 50-60% range. I’ve followed that advice for a long time now, and have never had a problem. I also install a new PSU on a new build. That’s the one piece of kit I never reuse.

6

For efficiency, yes. There is no reason a high-quality unit can’t regularly use more than 50-60% of its wattage from a reliability standpoint.

I also install a new PSU on a new build. That’s the one piece of kit I never reuse.

Good PSU’s have 5-10 year warranties, so there is no reason at all to replace a PSU still under warranty if its specs are still adequate for the new build.

2 Likes
7

I have one of the infamous flaming Gigabyte 850s at home in it’s box in the closet where it’s safe.

1 Like
8

Moved to #self-service:pc-hardware

9

“The system restarted without first shutting down gracefully. This error can occur when the system has stopped responding or has crashed, or the power has been unexpectedly lost.” Mmhh…that does’nt help.

EVGA XC3 Ultra Gaming Hybrid. Not the pick of the litter but was happy to get one last year. And as naive as I am I thought an PSU from the same brand must be an good idea…

I bought the 1300W with an 4090 in mind :slight_smile: .

I am not an electrician. I can connect an plug to an socket but have not the skills to analyze a power supply. And pointless fumbling around with high currents, mmmh…, probably an bad idea, even at low voltages.

I do and tried a lot.

I end up UC/UV with 270-300W power draw from the GPU. Now its fine again with my good old 1800Mhz/900mV OC/UV

360 AIO for the 11900k blowing out on top, 240 AIO for the GPU blowing in on front, 140mm exaust fan rear. CPU is around 60C, GPU between 70C and 80C depending to the room temps. Memory temp junction around 90C, GPU hotspot around 90C.

Old EVGA Supernova 850 G+, New Seasonic Prime PX 1300. They are both single rail. No fan issues noticed.

It’s watercooled.

Yes, but how? As described above I dont want to try the wet finger method :wink:

10

If you suspect power issues check every connection carefully including on your gpu and especially if you have been doing maintenance. When I built my PC I slightly dislodged a (out of sight) modular cable while tightening a cable tie. My problem only came to light in the MSFS menus otherwise the PC ran perfectly.

11

Hmmmm. I’ve noticed this happening to me a bunch recently. But I built my own computer, so it’s on me if it’s not working right. LOL

So far has only happened with MSFS and the Aerosoft CRJ atm, consistently within 15 minutes of takeoff from KDEN to KCYS. (4x in a row, with a few BIOS checks in between–RAM voltages, OC vs no OC, pump and fan control boosts, etc.) Then flew the PMDG 737 on the same route start up to shut down and all was good. Flew the FBW a few days ago on an out and back ~40 minute flight time flights, and all was good. All of this in VR with this setup:

Intel i9-9900k chilling at 5.0
ASUS ROG Maximus XI Formula
ASUS ROG Strix RTX 3090 OC
Alienware 3418DW 3480x1440 120Hz
GSkill TridentZ 3600MHz CL16-16-16-36
3x 1 TB Samsung Evo 860s (Games and stuff)
1 500 GB Samsung Evo 970 m.2 (Win10 Pro v2004)
1 2 TB Samsung Evo 970 m.2 (FS2020)
Seasonic 1kW Platinum PSU
A TON of EKWB water cooling stuff (pumps, res, fittings, blocks)
Dual hardline loops (my first hardlines)
2x HWLabs Nemesis 560GTX rads
13x Noctua 140mm fans
All jammed in to The Tower 900
Peripherals:
Thrustmaster HOTAS Warthog
Thrustmaster Pendular Pedals
Thrustmaster Airbus Stick / Throttles / flaps / spoilers
Thrustmaster Boeing Yoke and Throttles
Honeycomb Alpha and Beta
Oculus Rift S

My boy was VR flying the default F-18 and said it happened to him too. I flew the air trike, default PC-6 the other day with @OrigBullethead in VR again, and had no problems for ~2-2.5 hours of flight time.

Any time a computer has a hard shutdown, one would usually suspect a hardware failure and/or the PSU, but that’s just a “good hunch” and may not always be true.

If I find anything, I’ll post here. There are a bunch of threads ranging from '20 to '21 about the issue, and some folks claim victory with various fixes, so a bit of forum research may provide a few more tricks to try.

EDIT: I just had a successful flight with the CRJ-900 with an official livery from KDCA to KLGA, zero crashes. I’ve heard that some liveries can really hose up the sim, so I’m guessing that may be the case here. I’ll check the CRJ-700 later with an official livery, but I’m guessing that was the source of the problem, since nothing else has changed.

1 Like
12

Plus pumps. - Need I say more?

13

Nah, even with everything (including peripherals) under full load all the calcs I’ve done show high 800’s low 900’s for wattage. And I don’t run everything full bore all the time. PWM is my friend.

14

Just let the sim idle with the Leonardo MD-82 in VR for the past 3 hours, then flew a quick flight from KCYS to KCOS without issue. :man_shrugging:

1 Like
15

Did you get the “Leonardo long login first go round” I’m waiting that out right now.|

Big updates first MD-82 for me takes forever.

16

I did have the long load too. It is the first time I’ve used the 1.b.95 version so that was on me. Is there a newer version? I haven’t looked in a while.

1 Like
17

Seems to be a regular with the MD-82 and new builds, solid since

18

I fear the 1kw is to weak for your battleship.

From my experience I would recommend to double the calculated load for the psu by now.

Have you seen the GN vblog? The spikes are tremendous. I think you should have a look for an 1600W psu.

Next: Avoid classic OC on the gpu, especially vram OC. Try OC/UV. This stabilize your load.

Avoid excessive xmp settings/ram OC.

Lock your fps. I am fine with 27fps.

Try „balanced“ for power settings in the nvidia cp.

All this together saved msfs for me.

1 Like
19

Hmmm. You may be right.

I haven’t seen it, but will go check it out.

In the past, all the OC’d “from factory” GPUs seemed to work fine for me. I wasn’t much for pushing past whatever they did since I didn’t get a slower card to start with.

Other than needing XMP1 on to bump the voltage to 1.35 and ensuring it was running at 3600 (instead of the MB default of 3200) I haven’t touched the RAM. It was supposed to be the jedec timing I listed tested from the factory. Not sure who thorough manufacturers are with the volume of materials they produce though anymore. I stressed tested them a few months after building and they were fine.

I’ve never bothered with forcing frames in flight simulator. Since FS2004 had chronic “low frames,” and many beastly machines weren’t strong enough for it, I’ve stopping looking at frames in the light of “if it is smooth enough, then I don’t care what the frames are.” But. . .I’ll try locking it to see what happens.

Cool. I will do that.

I’m still a fan. :smiley: I get that the sim is complex and throwing in a ton of official and non-official addons into the mix can cause hiccups, but for me, this has been the most usable (and by far my favorite) sim for civilian flying.

20

Lucky me. Reinstalled flightbeam KDEN and the CRJs, and now all is good.

image
image1920Ă—480 58.2 KB

It was only the CRJs that were causing the hard crashes in and out of KDEN. MD, 737-700, FBW were all fine.

When the system hard reboots, see if you can duplicate exactly where it happened and with which aircraft (including liveries!) A clean remove and reinstall might clear up some issues.