Strange fan behavior RTX 4080 Founders Edition

[SuperConker]

I was using my computer earlier today and it was working fine, then i decided to restart it after a driver Update (intel i225 network drivers)

Upon restarting i accidentally clicked Shutdown instead of Restart (so i had to power the computer back up again, right after it switched off).

Once back on the Desktop i noticed that the fans on my RTX 4080 (Founders Edition) were suddenly not spinning anymore
(they were spinning normally just before the shutdown/reboot).



I tried running the GPU-Z "Render Test" to see if the fans would come on again, but this did not trigger the fans to spin up
(according to GPU-Z the GPU-core was 33 degrees Celsius).

I thought this was odd so i tried to google the issue, but then Google Chrome did not want to load any web pages for some reason.



After a few minutes i decided to shut the computer down again (and this time i also unplugged the power cord).

Then i pressed and held down the power-button for like 5 seconds to discharge the capacitors in the computer
(just an old habit i have whenever i unplug the power-cord).

Normally lights/LED's on the Motherboard and RAM will flash briefly when doing this (but not this time).



Then i reconnected the power-cord and tried starting the computer back up, but nothing happened.

None of the LED's on the Motherboard lit up like they normally would do for some strange reason,
i tried pressing the Power-button on the case several times (but still, nothing happened).

I also tried flicking the switch on the PSU off and on a few times and tried unplugging/plugging back the power cord for the PSU as-well.

The computer simply did not want to come back to life for whatever reason (no signs of life anywhere inside the case).



After leaving the computer completely unplugged for a few minutes it finally decided to came back to life,
(the LED's on the motherboard were also lit this time, even before pressing the power button).

The 4080-fans were once again spinning like normal, and Google Chrome was working fine again
(GPU-temp was a tiny bit lower too, at 31 degrees Celsius this time).



The main components (CPU/Ram/Motherboard) in the computer is nearly 2 years old (PSU is almost 9 years old) and i have never seen this issue before
(i've only had the RTX 4080 for 11 days, and this happened for the first time today).

I was finally able to google the fan-issue and the only similar issue i could find was this reddit post:
https://www.reddit.com/r/buildapc/comments/153jtfy
It's kind of the same issue as what i experienced today (fans not spinning up and weird behavior in apps/games).

Only difference is the part where my computer would not power up for a while afterwards.

Anyone have any idea what the issue could have been? Never had my computer do this before.

 

[jon_joy_1999]

Sounds like your PSU is going bad. I'd swap it for a known good one, especially since this one is 2 years out of warranty.

 

[Vya Domus]

This just sounds like the fan stop feature, at 33C the fans are not going to spin.

 

[gffermari]

It’s more a motherboard or psu issue rather than a gpu or fan one.
I had the same issue a couple of weeks back - it seemed my motherboard suddenly died or its bios stuck with no apparent reason and never recovered.

 

[Mussels]

The fans shouldnt spin under 60c

I've seen people asking about this on facebook, i get the feeling people have been forcing 'high performance mode' in the NVCP causing higher temps and fan speeds at idle, and the new drivers are reverting that setting (or something along those lines)

Then it sounds like by cutting power to the PC, your CMOS battery has gone flat causing the CMOS to reset and have failed booting attempts (common with certain RAM setups to not like the defaults)
Time and date being wrong will break most websites these days, the error usually mentions that.

The GPU fan will spin up under certain types of load and temperatures, but that's set by the VBIOS itself or possibly software by you. Not the sort of info we have access to



Checked your system specs, if this is your ryzen system then absolutely - 64GB of RAM at 3600 is one of those setups that is going to have issues on BIOS defaults. Enabling XMP will raise the SoC voltage (should be 1.09v on that board, ZenTimings can show you) and default is 0.95v - low enough that your high speed RAM is causing too much strain on the CPUs memory controller despite being at a lower speed

 

[SuperConker]

Sounds like your PSU is going bad. I'd swap it for a known good one, especially since this one is 2 years out of warranty.

The PSU still has 1 year and 4 months left on it's 10-Year warranty, so i hope it's not that.

This just sounds like the fan stop feature, at 33C the fans are not going to spin.

I do have the setting "Prefer maximum performance" turned on in the nVidia Control Panel so they are supposed to spin in my case i think
(they normally seem to be running at around a constant 1100 rpm).

Most of the time they do spin at startup when i press the power button,
but every now and then the system starts with the fans turned off
(even before windows loads, which i am not sure is normal).

It’s more a motherboard or psu issue rather than a gpu or fan one.
I had the same issue a couple of weeks back - it seemed my motherboard suddenly died or its bios stuck with no apparent reason and never recovered.

Yeah, it looks like it could be a Motherboard or PSU issue at this point.

The fans shouldnt spin under 60c

I've seen people asking about this on facebook, i get the feeling people have been forcing 'high performance mode' in the NVCP causing higher temps and fan speeds at idle, and the new drivers are reverting that setting (or something along those lines)

Then it sounds like by cutting power to the PC, your CMOS battery has gone flat causing the CMOS to reset and have failed booting attempts (common with certain RAM setups to not like the defaults)
Time and date being wrong will break most websites these days, the error usually mentions that.

The GPU fan will spin up under certain types of load and temperatures, but that's set by the VBIOS itself or possibly software by you. Not the sort of info we have access to



Checked your system specs, if this is your ryzen system then absolutely - 64GB of RAM at 3600 is one of those setups that is going to have issues on BIOS defaults. Enabling XMP will raise the SoC voltage (should be 1.09v on that board, ZenTimings can show you) and default is 0.95v - low enough that your high speed RAM is causing too much strain on the CPUs memory controller despite being at a lower speed

I do have "Prefer maximum performance" turned on in the nVidia Control Panel so that's probably why they are spinning at less than 60c in my case.
It could be that it's reverting this setting on some boots and not others, but that would be a weird driver-bug (i do have the latest 536.99 version).

I don't think the CMOS was getting reset, as all the BIOS-settings (including the time and date) were still correct after starting the PC again.

The Memory kit i have does work fine at default JEDEC speeds (and also at their rated XMP speeds).

The Memory kit that i have is a 2x32 GB (Dual Rank) kit, so it should not be too stressful on the memory controller
(i had a choice between 4x16 GB Single Rank and 2x32 GB Dual Rank, and went with a 2x32 GB Dual Rank kit).

The SoC seem to be running at anywhere from 1.1750v -1.1813v according to ZenTimings (it keeps alternating between those two voltages).

I also checked the BIOS and my motherboard seem to have set the SoC to 1.2000v.

I left all the voltages on AUTO except for the voltage for the Memory (which i manually set to 1.3500v as that's what it's rated for).

So far it would seem like this could have been a PSU-issue that prevented the PC from powering on, until i left it unplugged for a few minutes
(and most of the answers seem to point to the PSU).

And that perhaps the newest nVidia drivers have some kind of bug that sometimes reverts the "Prefer maximum performance" setting on some system boot/reboots
(like 9 out of 10 times the setting sticks like it should, but every so often the fans are turned off upon Windows startup).

 

[Mussels]

I do have the setting "Prefer maximum performance" turned on in the nVidia Control Panel so they are supposed to spin in my case i think

turn that off. you do not want that on, at the desktop

It could be that it's reverting this setting on some boots and not others, but that would be a weird driver-bug (i do have the latest 536.99 version).

BIOS and windows don't affect each other - driver bugs do not have anything to do with the PC when it's off or booting.

I don't think the CMOS was getting reset, as all the BIOS-settings (including the time and date) were still correct after starting the PC again.

The BIOS recovery boots do not reset the time or date, often they dont even erase settings - they just load the safe settings for that one boot expecting you to fix the issue. Usually theres a popup giving you time to hit a button to enter the BIOS, but it often passes by so fast we cant see it on modern hardware

The Memory kit i have does work fine at default JEDEC speeds (and also at their rated XMP speeds).

The Memory kit that i have is a 2x32 GB (Dual Rank) kit, so it should not be too stressful on the memory controller
(i had a choice between 4x16 GB Single Rank and 2x32 GB Dual Rank, and went with a 2x32 GB Dual Rank kit).

The SoC seem to be running at anywhere from 1.1750v -1.1813v according to ZenTimings (it keeps alternating between those two voltages).

... You realise that CPU is only rated to 3200 with two single rank modules right?
2x2 is slightly easier than 4x1, but neither are supported at 3600. You're overclocking - so you need to tweak voltages to get this stable and working every boot, or you need to drop speeds to what works stable with your hardware.
No, there is no magic fix for this, no CPU board or RAM change guaranteed to work because you're above the officially supported specs of AM4 as a whole. You have to test things and find out yourself what works.

If your SoC is set to 1.20v and it's drooping to 1.175v, i'd say theres some issues with voltage drops there and that may be related.
1.20v may even be too high for that boards SoC VRM to handle and less stable than a lower voltage - 1.20v is the max you want on AM4, not the starting point.

If 1.20v is the auto setting, try 1.15v - this is where the auto could be flip-flopping between different numbers on each boot and failing, forcing itself higher every attempt until it works (this is what expoded 3D CPUs on AM5)

The newer 5800x3D has more detailed specs on AMD's website, but even it doesn't support what you're doing

Yes it can be done - but it's not supported, so theres zero guarantees it'll work without tweaking things. working 99% of the time and cold boot issues or crashes is exactly within that realm.

 

[jon_joy_1999]

The PSU still has 1 year and 4 months left on it's 10-Year warranty, so i hope it's not that.

https://www.corsair.com/newsroom/pr...-1500-watt-80-plus-titanium-psu-now-available says it's only 7 years. However https://help.corsair.com/hc/en-us/a...580970#h_fa4f9672-b385-4d1f-81d9-84ab99de9c97 suggests it could be 10 years, if your PSU is considered to be a AXi series.

 

[SuperConker]

turn that off. you do not want that on, at the desktop


BIOS and windows don't affect each other - driver bugs do not have anything to do with the PC when it's off or booting.


The BIOS recovery boots do not reset the time or date, often they dont even erase settings - they just load the safe settings for that one boot expecting you to fix the issue. Usually theres a popup giving you time to hit a button to enter the BIOS, but it often passes by so fast we cant see it on modern hardware


... You realise that CPU is only rated to 3200 with two single rank modules right?
2x2 is slightly easier than 4x1, but neither are supported at 3600. You're overclocking - so you need to tweak voltages to get this stable and working every boot, or you need to drop speeds to what works stable with your hardware.
No, there is no magic fix for this, no CPU board or RAM change guaranteed to work because you're above the officially supported specs of AM4 as a whole. You have to test things and find out yourself what works.

If your SoC is set to 1.20v and it's drooping to 1.175v, i'd say theres some issues with voltage drops there and that may be related.
1.20v may even be too high for that boards SoC VRM to handle and less stable than a lower voltage - 1.20v is the max you want on AM4, not the starting point.

If 1.20v is the auto setting, try 1.15v - this is where the auto could be flip-flopping between different numbers on each boot and failing, forcing itself higher every attempt until it works (this is what expoded 3D CPUs on AM5)

The newer 5800x3D has more detailed specs on AMD's website, but even it doesn't support what you're doing
View attachment 308772

Yes it can be done - but it's not supported, so theres zero guarantees it'll work without tweaking things. working 99% of the time and cold boot issues or crashes is exactly within that realm.

I have it set to "Prefer maximum performance" because i want to get the most performance out of games
(the card doesn't seem to use any more power while idle on the desktop).

It also helps keep the overall temperature inside the case down since the card idles at 33c instead of 55c.


I thought the Ryzen 5000-series officially supported up to 1800 Mhz (an effective rate of 3600 Mhz) on the memory controller?

The specific Memory kit i have is this kit here:
https://www.corsair.com/us/en/p/mem...ram-3600mhz-c18-memory-kit-cmt64gx4m2c3600c18

It's a slower CL18 kit with the full timings being: 18-19-19-39.


I think the AUTO setting for the SoC just sets a voltage of 1.200 and then the board will try and use whatever it needs to be stable
(which in reality turns out to be somewhere between 1.1750v and 1.1813v).

I could probably set the SoC voltage manually to something like 1.850v (instead of AUTO) just to be safe,
and still be below the maximum recommended of 1.200v.

https://www.corsair.com/newsroom/pr...-1500-watt-80-plus-titanium-psu-now-available says it's only 7 years. However https://help.corsair.com/hc/en-us/a...580970#h_fa4f9672-b385-4d1f-81d9-84ab99de9c97 suggests it could be 10 years, if your PSU is considered to be a AXi series.

The PSU i have does carry a 10-Year warranty according to Corsair's own website:

https://www.corsair.com/us/en/p/psu/cp-9020057-na/ax1500i-digital-atx-power-supply-1500-watt-fully-modular-psu-cp-9020057-na

 

[Mussels]

I have it set to "Prefer maximum performance" because i want to get the most performance out of games

that's not what the setting does
All it does is disable the power saving mode at idle.

Ryzen does not support 3600MT/s, that's just what people called the sweet spot for overclocking.
1.20v is high, almost dangerously so. If it's not needed for stability lower it - you'll find anywhere from 1.10v to 1.15v should work.

It's a maximum, not a starting point. Just because your software sensors say 1.20v doesnt mean it's not getting 1.25v in reality, you don't want higher than is actually needed.

 

[Assimilator]

Honestly this sounds like one of those random temporary blips that sometimes happens with PCs - could very well be cosmic rays. My rule is that unless it happens more than once, or further unusual behaviour occurs, just forget about it.

I have the AORUS PRO AX (little brother of your board) and a 5900X (little brother of your CPU), with a 2x 32GB 3600 CL18 kit (Patriot Viper 4 Steel Series with slightly looser timings than your Corsairs), and never had any problems with stability with XMP applied. I see you're on a beta BIOS, that could be a problem.

I haven't checked my board's SOC voltage, will let you know once I'm in front of that PC (probably this evening). EDIT: 1.18v, I'mma turn that down.

It also helps keep the overall temperature inside the case down since the card idles at 33c instead of 55c.

This is pretty much irrelevant.

I thought the Ryzen 5000-series officially supported up to 1800 Mhz (an effective rate of 3600 Mhz) on the memory controller?

No, it's always been DDR-3200 as @Mussels has said.

 

[Mussels]

No, it's always been DDR-3200 as @Mussels has said.

At best, DDR4-3200

The 5800x3D has the most detailed information on the AM4 lineup, most of the other models we got for presentation slides instead of the AMD website.
This is from the 5800x3D, with the same memory controller as the rest of the Zen3 lineup.

Ryzen 5000 supports 3200 with 2xDR
Ryzen 3000, 3200 2xSR
Ryzen 2000 was 2933 2xSR

More ranks or older CPUs and it goes down as low as 1866 - which sounds low but then again intel was just as low back then too. DDR4 changed far more than DDR3 did in it's lifetime

This was from a gigabyte slide on overclocking ryzen 3000, rather than stock speeds. The "OC speed" is what the boards could support while 'safe speed' was the best they could achieve with cherry picked boards, CPU's and RAM without unsafe voltages, dropping PCI-E to 1.1 and so on.
(It was posted as part of one of their overclocking competition things)