Jump to content

850W PSU, RTX3080 and Kernel-Power ID41 critical error


Recommended Posts

About 3 months ago I was lucky to grab a second hand 3080 card for 999€ in my own city. It replaced my old 1080Ti which went to ebay.

 

I was keeping the same 850W (70A) semi-modular PSU that I acquired when my 1080Ti was needing more amps. Most of the reviews said that 850W is more than enough for a 3080. I put two independent PCI-E cables from PSU.

 

In the last three months, I have experimented occasional sudden shut-off of the PC in the middle of a heated battle always in VR (Index), sometimes after 30 minutes, other after 90 minutes. Normally when the battle was intense. I look at the Event Viewer and it was a critical Kernel-Power error with ID41. My son also experimented that in few occasions with his own games.

 

I read about it and I tried with different versions of BIOS and also deactivating the C-states. Initially I thought that C-states was the cause, but it was not that.

 

About three week ago I installed MFS2020 just to take a look of that. But when I wanted to run it in 4K (all maxed out) the PC shut-off with the same power kernel error. But when I put it in 1080p resolution it worked perfectly. This give me some clues about the origin of the problem: The GPU power drawn. I read more in internet and the PSU could be also the reason of the ID41 error.

 

So, I acquired a brand new 1000W (83A) full modular PSU just to see if it will solve the problem. I have just installed today and launch MFS2020 at 4K and voila!!, the PC didn´t shut-off.

 

So, let´s see if in the future I have not more shut-offs. Just leave here a picture of the old and the new PSUs.

 

pic.thumb.jpg.74f255d2ffd8675e935c5dc7ca0ba5fb.jpg

 

Edited by chiliwili69
  • Like 1
  • Thanks 2
  • Upvote 1
Link to post
Share on other sites
9 hours ago, dburne said:

Always better to have more than needed, not so good to be close or marginal. Also quality of the power supply plays a big part.

 

Well 750 would be more than needed.

Every calculator I’ve used puts it around 650, (lower actually) and that’s with a built-in margin. 

 

Edit - just did another calculation and added in an extra/2nd 24” monitor just for grins..612 Watts.

 

750 should be plenty.

 

Edited by Gambit21
Link to post
Share on other sites

I had the same issues, the RTX 3000 series spike pretty hard in demand especially in VR and with apps that suddenly draw everything *cough* IL-2's loading style *cough* IL-2 in VR *cough*. A less quality 1000W PSU was unstable too..

 

 

Now with Seasonic Prime GX 80 PLUS Gold 850W for my RTX 3090. Problem solved. 

 

 

 

Wer billig kauft, kauft doppelt :coffee:

 

 

 

Link to post
Share on other sites

I have similar issues, although not so frequently.

I have long suspected that my 5700XT GPU spikes on the power-draw and occasionally results in immediate system shut downs. The same power kernal event occurs in the windows log. I am running a PSU with enough juice at 650W, but I don't think it is efficient enough to deal with the heavy spike requirements of newer cards, it is quite an old PSU. It was even worse before I took the 5700XT off a daisy-chain power rail...

Link to post
Share on other sites

I think it's not amount of Wattage available but the stability of the voltage needed during severe power load, good quality components built psu with less power is far better than 1kW cheaper build, sometimes even good brands use not so good components as one would assume. 

Edited by 1PL-Husar-1Esk
  • Like 1
  • Upvote 1
Link to post
Share on other sites

I'm running an ASUS RTX 3080 TUF OC in combination with an i7 8700K @ 4.9GHz and the same thing was happening to me until I upgraded to a Seasonic 850W 80+ Gold PSU last week.

 

The strange thing is, I was running a Seasonic 750W 80+ Gold PSU prior to this with the same exact hardware and never encountered the issue until I jumped from 4.8 to 4.9GHz on my processor. I guess the extra 0.7mV I added to my voltage was the tipping point.

 

Considering I have my rig plugged into a UPS and can measure the wattage, at full system load I'm only pulling around 700W. Perhaps in retrospect, my previous PSU's capacitors could no longer handle the load.

Link to post
Share on other sites

I think this may help you or anyone else running a Nvidia RTX 3080 or 3090, even if you have an 850 W PSU. I installed my EVGA 3080 XC3 Ultra on an ASUS B550-F atx mobo which has very strong VRM, a Sunflower Leadtek III 850W PSU (highly recommended as they make some of the best rated power supplies for other big name companies) with an AMD 3600x originally though now I'm running an AMD Ryzen 5800X with a Noctua cooler, 32GB 3200Mhz (2 sticks) and 3 M.2 SSDs, running it to a 1440 /144fps display using stock DOCP and High performance settings in Windows and the BIOS with no fancy overclocking attempts. I noticed right away I was getting really high temps over 80 C at 1905 Mhz average peaks in flight, and power usage over 320W! I am using MSI Afterburner for everything GPU, but decided to check things in EVGA's Precision X app too where I ran some scans which showed that my card could hit up to 2100Mhz at peak, but with those same temps and the possibility of a temp induced crash. It made me nervous. Then I started to research undervolting when I came across mention of it on some PC gaming forums. I am including some reference urls and videos below that really helped me.

 

Needless to say there is some trial and error work involved as each rig and card will display slight variances. I did not follow the instructions from the videos after a bit finding that manually adjusting the numbers to "perfect" whole numbers like -200 Mhz on the MSI Afterburner and forcing the same on the curve plot did not lead to stable setups, so I simply used the numbers as they came up on sliders and the curve plots. I ran the test in SP with a full load of aircraft (16 AC), on the Bodenplatte map where there's a bunch of cities and trees to push the GPU hard. Many of my higher profiles were stuttering, big pauses and crashes/restarts normal. I also increased the fan profile to 80% at 80% GPU usage, 60% @ 60%, etc. which really help temps. Some profiles were reasonably stable, but sooner or later they crashed out so I kept working on it. I now have an extremely stable setup with no noticeable stutter, ultra graphics settings at 143fps average (no grass - just a personal thing) maxed out MSAA, etc; GPU averaging 58 - 65 C at even peak GPU usage and averaging 180-240W power usage with those higher numbers appearing when I zoom in on cities and forests! My undervolting settings are at -197 Mhz (from 2100 the top of my curve), .825 mv and 1860 Mhz. It's been awesome. 1860 Mhz sounds slow compared to 2100 Mhz, but it is not that much lower than the 1905 Mhz and runs flawlessly at very low temps and power usage. At this point I probably could use a 650W PSU safely! Check out the guides below - I found the WCCF tech page and video the easiest to follow and use for starters.

 


Undervolting 3080
The best and easiest to follow: https://wccftech.com/undervolting-ampere-geforce-rtx-3080-hidden-efficiency-potential/ and also the accompanying Youtube video


page-9#post-1044788226

 

 


This ones a bit complicated but seems promising: rtx-3000-series-undervolt-discussion.200


This Reddit post has a lot good input:

I don't know if this will help you, but I think it's definitely worth the time and effort.
 

  • Like 1
Link to post
Share on other sites

Just because the label says 850W it doesn't mean the PSU can sustain 850W load. 

In many cases that number is the maximum peak load, and the sustained load is much lower. 

I learned that a long time ago, when I bought a dual GPU GTX590. The 1000W antec PSU I bought for it wasn't able to run the card, but the Corsair AX 860w that i replaced it with was running like a champ.

 

Glad you figured out the problem. 

Thx.

Link to post
Share on other sites

It's not just overall load, but how many amps each line puts out. Different power supplies might have differences here, note the picture posted of the 850 vs the 1000, and note the 850 is putting out 20A on the 3.3V lines while the 1000 puts out 25A on the same lines. You need to check the cards peak power draw for its required power lines and make sure the supply can provide that.

Link to post
Share on other sites
4 hours ago, grcurmudgeon said:

note the picture posted of the 850 vs the 1000, and note the 850 is putting out 20A on the 3.3V lines while the 1000 puts out 25A on the same lines.

 

I believe the PCI-E power supply connector use the 12V output, so here the difference is 70A vs 83A.  The manufacturer doesn´t give a clue about the max Amps peak.

 

Yesterday I had a 2-3 hours session with IL-2 (this is a luxury for me) and so far not shutdowns. Cross fingers.

 

I also tested MF2020 in 4K (all maxed) and GPU load goes directly to 100% and no problems. With my previous 850W PSU, just after half second flying it shut-off.

 

Most likely all is a question of specific PC configuration and quality of PSUs. The 850W PSU was not a top brand PSU when acquired.And  I didn´t want to buy another 850W PSU of a better brand.

 

The 80 Plus clasification (Bronze, Silver, Gold, Platinium, Titanium) is just about efficiency, but not necesarilly about stability at loads. But probably efficiency and stability are correlated in PSUs.

Link to post
Share on other sites
1 hour ago, chiliwili69 said:

The 80 Plus clasification (Bronze, Silver, Gold, Platinium, Titanium) is just about efficiency, but not necesarilly about stability at loads. But probably efficiency and stability are correlated in PSUs.

Just another marketing tool.  

Edited by Jaws2002
Link to post
Share on other sites
46 minutes ago, Jaws2002 said:

Just another marketing tool.  

Say the guy who put "Corsair AX1000W Titanium PSU" in his PC specs :coffee:

  • Haha 1
Link to post
Share on other sites
32 minutes ago, JG7_X-Man said:

Say the guy who put "Corsair AX1000W Titanium PSU" in his PC specs :coffee:

I had some bad experiences with PSUs in the past, so, forgive me for being a bit paranoid :lol:.

But I still think the "new" Titanium" class was nothing but a marketing trick.

 

The new car stereo, with the dial that goes to eleven.

Link to post
Share on other sites
On 4/11/2021 at 3:07 PM, -Six-Motivated said:

I'm running an ASUS RTX 3080 TUF OC in combination with an i7 8700K @ 4.9GHz and the same thing was happening to me until I upgraded to a Seasonic 850W 80+ Gold PSU last week.

 

The strange thing is, I was running a Seasonic 750W 80+ Gold PSU prior to this with the same exact hardware and never encountered the issue until I jumped from 4.8 to 4.9GHz on my processor. I guess the extra 0.7mV I added to my voltage was the tipping point.

 

Considering I have my rig plugged into a UPS and can measure the wattage, at full system load I'm only pulling around 700W. Perhaps in retrospect, my previous PSU's capacitors could no longer handle the load.

 

Problem with the RTX 3000 Series is the Peak Load <1ms where the power draw alone for the 3080 jump to around 500W. These Peaks can lead to shutdowns in very fast reacting protective circuits (OPP, OCP).

 

I always look at the power draw from my system using Digital Energy Cost Measuring Device but before I buy something new I check the Thermal Design Power (TDP). Had even the 1080 Ti with 450W PSU without problems without overclocking.  

Link to post
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
×
×
  • Create New...