Bad FPS (<30) on a 3060ti, help me compare this data!

[sne4k0]

I've been having some major issues with my frame rate the last couple of weeks and have been unable to diagnose what the offending component of my hardware is. Let me preface this post by saying I've ruled out any hardware or driver issues.

The first picture (primo.gif) is when the GPU is functioning normally, the second (2.gif) is when I'm getting poor performance. Some interesting points to compare: GPU Clock, GPU Temperature, GPU Chip Power Draw, PWR_SRC Power Draw, PCIe Slot Power, 8-Pin #1 Power, and GPU Voltage. The symptoms arise while gaming, and don't go away until after a lengthy power down. I'm thinking that the PSU or the Motherboard are the culprits here, but I need someone smarter than me to help me figure it out, or if there are other tests I can run / data I can gather.

Thanks to anyone that takes time to look at this data!

-Nick

 

[Frick]

First, welcome. Second, what is your complete system specs? Thirdly, what specific games are having trouble? Fourthly, 150W over the PCIe slot doesn't sound right. Fifthly, have you done anything to your computer lately, like moved it?

 

[sne4k0]

First, welcome. Second, what is your complete system specs? Thirdly, what specific games are having trouble? Fourthly, 150W over the PCIe slot doesn't sound right. Fifthly, have you done anything to your computer lately, like moved it?

It's an ABS Prebuilit - Intel i7 11700f - GeForce RTX 3060 Ti (gigabyte build) - 1TB M.2 NVMe SSD - 650W PSU - Running Windows 11.

Lately I've been noticing the issue while playing Valheim, but the poor FPS carries over to other games that I might play afterwards. The PC has been moved lately, but that's because I checked all the power connections and moved the GPU to the other PCIe slot (on the advice from a redditor). Moving the GPU got me great performance initially. But whilest playing Valheim this morning I was getting 120 fps, I got up to get a bagel and some coffee and came back to 20 fps.

I find it most interesting how high some of the power draws are when the GPU clock is stuck <500 MHz. Someone else I had spoken to mentioned that the GPU isn't boosting when it's stuck in this state.

Edit: You had mentioned games and I only listed one. Lately I've been playing Raft, Don't Starve Together, and DayZ. DayZ and Valheim are going to be the most demanding on my GPU but I feel like they're still kind of low-end graphically.

 

[shovenose]

Yeah, something seems odd. PCI-E slot power should be 75W or less.

 

[sne4k0]

Yeah, something seems odd. PCI-E slot power should be 75W or less.

Right? It's like something is causing a power drain from the cable and the GPU is trying to compensate by pulling more power from the slot.

 

[metprism]

the problem is that your gpu clock speed is a lot lower in the second image compared to the first image which is 1980 mhz versus 435 mhz. I would recommend trying to put a slight over clock on the core clock of the gpu to see if you can get it to raise the gpu clock in games. If that doesn't work I would recommend you update the motherboard bios and put the gpu back into the first pcie slot of the motherboard and try again. Also you might want to go into the nvidia control panel and put the power management to maximum performance to make sure that your power draw goes to the bios limit of the gpu. There is a chance that your power supply isn't working properly as well since you got this system in a pre built and they tend to cheap out on psu's in those computers even if the rating say's 650 watt's that doesn't mean that it can handle that much load if they give you some no name psu with a 650 watt gold rating.

 

[Count Shagula]

Is your power supply enough to actually power the entire machine? Those are some weird numbers though

 

[Mussels]

View attachment 268022View attachment 268023

I've been having some major issues with my frame rate the last couple of weeks and have been unable to diagnose what the offending component of my hardware is. Let me preface this post by saying I've ruled out any hardware or driver issues.

The first picture (primo.gif) is when the GPU is functioning normally, the second (2.gif) is when I'm getting poor performance. Some interesting points to compare: GPU Clock, GPU Temperature, GPU Chip Power Draw, PWR_SRC Power Draw, PCIe Slot Power, 8-Pin #1 Power, and GPU Voltage. The symptoms arise while gaming, and don't go away until after a lengthy power down. I'm thinking that the PSU or the Motherboard are the culprits here, but I need someone smarter than me to help me figure it out, or if there are other tests I can run / data I can gather.

Thanks to anyone that takes time to look at this data!

-Nick

that second screenshot is GPU mining, or at least looks like it


If the power results are accurate you've got damaged cables, possibly a damaged ground wire on your PCI-E cabling
upload photos of how its connected - please, no blurry photos or insane closeups

 

[BetrayerX]

Hi!

That thing is not even throttling, just flat-lined to 435 MHz. Temps are low and power draw is high. Could guess on a BIOS issue (mining card as Mussels called it) or a faulty PSU.

 

[sne4k0]

that second screenshot is GPU mining, or at least looks like it


If the power results are accurate you've got damaged cables, possibly a damaged ground wire on your PCI-E cabling
upload photos of how its connected - please, no blurry photos or insane closeups

there's definitely no mining going on. I recently wiped the PC so there is nothing potentially running in the background that could be trying to use my gpu secretly. Based on your comment about the cabling, I'm going to try the other pcie cable from my psu and see if that addresses the issues. I'll take some pics and upload them at that time as well!

Thanks for offering to take a look at it!

Is your power supply enough to actually power the entire machine? Those are some weird numbers though

Should be, it's an Asus Tuf 650w. Not an excess of power but everything else I've read leads me to believe that it's adequate. Although if I start replacing components, this would be the first just because it's the cheapest. I'm going to explore one of the pcie cables being faulty tomorrow.

the problem is that your gpu clock speed is a lot lower in the second image compared to the first image which is 1980 mhz versus 435 mhz. I would recommend trying to put a slight over clock on the core clock of the gpu to see if you can get it to raise the gpu clock in games. If that doesn't work I would recommend you update the motherboard bios and put the gpu back into the first pcie slot of the motherboard and try again. Also you might want to go into the nvidia control panel and put the power management to maximum performance to make sure that your power draw goes to the bios limit of the gpu. There is a chance that your power supply isn't working properly as well since you got this system in a pre built and they tend to cheap out on psu's in those computers even if the rating say's 650 watt's that doesn't mean that it can handle that much load if they give you some no name psu with a 650 watt gold rating.

Thanks, I'll start exploring those options. Turns out the PSU is an Asus Tuf 650w, so at least it's not a no name

 

[80251]

You're getting perfcaps (Vrel and Vop) which means your 3060ti is throttling. Are you overclocking it? I don't think you get Vrel or Vop perfcaps unless you're boosting the voltage.

 

[Sithaer]

The PSU should be ok, I'm running my Asus TuF 3060 Ti with an entry level Seasonic 500W Gold PSU with no issues even if I use the stock settings.
My entire system draws ~300W from the wall and that 11700F shouldn't make that much of a difference to cause such issues with a +150W PSU vs mine. 'TuF PSU is not a no name at least'

You could try an undervolt profile in Afterburner, that can help stabilizing the boost clock and lower the power draw a bit.
Not sure if this will help but it worths a try, I had a card in the past that only worked properly with an undervolt.. 'that was an AMD card tho'

I have mine set up this way and its stable tho it depends on your model and silicon so you might have to play around with it.

 

[pavle]

I'd say it's a faulty graphics card, or GPU-Z reports it totally wrongly. You can't have GPU using 19W and total board power is around 200W, that just doesn't make sense (second GPU-Z screenshot).

 

[sne4k0]

You're getting perfcaps (Vrel and Vop) which means your 3060ti is throttling. Are you overclocking it? I don't think you get Vrel or Vop perfcaps unless you're boosting the voltage.

No overclocking. It's also interesting that when the card isn't throttling that the perfcap is power

I'd say it's a faulty graphics card, or GPU-Z reports it totally wrongly. You can't have GPU using 19W and total board power is around 200W, that just doesn't make sense (second GPU-Z screenshot).

right, and i'll watch it perform normally (first pic) and an indeterminate time (15-90 minutes) later i'll notice my frames drop by 75% and then i get the weird power readings (second pic).

 

[Mussels]

The GPU usage is a now-known nvidia driver bug, you need the new hotfix driver to fix that



Same with GPU clocks getting stuck after quitting games and causing issues - you just need to update your drivers

 

[3x0]

Here's the link for the hotfix driver https://nvidia.custhelp.com/app/answers/detail/a_id/5407
Best use DDU in Safe Mode to remove previous driver

 

[Mussels]

Here's the link for the hotfix driver https://nvidia.custhelp.com/app/answers/detail/a_id/5407
Best use DDU in Safe Mode to remove previous driver

DDU is not needed at all, and often breaks more than it fixes

It is as last resort only, not the first thing to try

 

[pavle]

The GPU usage is a now-known nvidia driver bug, you need the new hotfix driver to fix that...

What would you say about such high power consumption of the whole card while the GPU is just sipping power? It's not like the board has a synchro-mesh chicken plucker on it.

Anyhow - try installing the latest driver (526.xx) and choose "Custom install" and under that tick the "Clean Install" box, so driver itself cleans old installation. You can also deselect "GeForce Experience" if you really don't need it.

 

[izy]

Use the second pcie slot and other 8 pin port from the psu to power your card and see if its working normal.
Edit: if you have dual bios you can try second bios aswell

 

[Calenhad]

I would also like to a picture (or four) of how this computer looks internally. That being said, here are some general thoughts.

You should never move a gpu from the primary pcie slot before you know how the secondary pcie slot(s) are wired in the motherboard. If it is not a custom job (like the junk Dell makes) you can find the motherboard manual online. This should show you which slot is the primary. Then make sure your gpu is in the primary slot. For a quick troubleshooting test you can try another slot, but for any other use always stick to the primary slot. If the slot does not work, then either your motherboard or cpu is broken.

Make sure the 8-pin power connector is properly connected. Asus says that psu has 4x pcie 6+2-pin connectors, I would assume these are two physical cables with two connectors each, for a psu of that wattage. So make sure you try the other cable coming from the psu, not just the other connector on the same cable. If they are pigtails, use the end connector. Do NOT mess around with over-/underclocking, flashing bios, and other stupid ideas like that, unless you are certain you know what you are doing and the computer is no longer under warranty. Because, depending on the age of the computer, I would look into the warranty situation here.

I have to say that second screenshot almost looks bugged honestly. 150W through the pcie slot would probably melt some traces on the motherboard if it were correct.

Do you get similar readings with hwinfo?

And finally, the drivers can mess some of this up. Not the 150W from the pcie slot, if that is a correct reading, but the other issues. Follow pavles advice above, to clean install the latest hotfix driver. A clean install of Windows is not out of the question either.

You're getting perfcaps (Vrel and Vop) which means your 3060ti is throttling. Are you overclocking it? I don't think you get Vrel or Vop perfcaps unless you're boosting the voltage.

This just means the gpu is hitting max voltage before the power limit. It is normal behaviour for modern gpus to boost until it hit one or more limits. This does not have to be the power limit.

 

[DeathtoGnomes]

One of the early steps after updating drivers is to swap out a known working PSU, dont be silly and assume that it cant be a problem, it still can bed the issue. There is a reason Nvidia put out a hotfix update, read the patch notes.

 

[sne4k0]

The GPU usage is a now-known nvidia driver bug, you need the new hotfix driver to fix that



Same with GPU clocks getting stuck after quitting games and causing issues - you just need to update your drivers

Well, I switched to the other pcie cord from the psu, put the gpu back into the 1st pci slot on the motherboard, and installed the hotfix driver. Ran a few games, got great fps, gpu was throttling well. I ran a benchmark test in Warhammer: Vermintide and things were going well, averaged 123 fps. I joined a match and watched my gpu chip power draw drop to 20w and the gpu clock fell to 400 MHz (was 1780 MHz prior)

Attached is the picture of the wiring that you requested.

Thanks for your continued help with this issue!

What would you say about such high power consumption of the whole card while the GPU is just sipping power? It's not like the board has a synchro-mesh chicken plucker on it.

Anyhow - try installing the latest driver (526.xx) and choose "Custom install" and under that tick the "Clean Install" box, so driver itself cleans old installation. You can also deselect "GeForce Experience" if you really don't need it.

I followed all those steps and same issue arose. I also went into nvidia settings and made sure the 3060ti was the selected card and that power settings preferred performance.

I would also like to a picture (or four) of how this computer looks internally. That being said, here are some general thoughts.

You should never move a gpu from the primary pcie slot before you know how the secondary pcie slot(s) are wired in the motherboard. If it is not a custom job (like the junk Dell makes) you can find the motherboard manual online. This should show you which slot is the primary. Then make sure your gpu is in the primary slot. For a quick troubleshooting test you can try another slot, but for any other use always stick to the primary slot. If the slot does not work, then either your motherboard or cpu is broken.

Make sure the 8-pin power connector is properly connected. Asus says that psu has 4x pcie 6+2-pin connectors, I would assume these are two physical cables with two connectors each, for a psu of that wattage. So make sure you try the other cable coming from the psu, not just the other connector on the same cable. If they are pigtails, use the end connector. Do NOT mess around with over-/underclocking, flashing bios, and other stupid ideas like that, unless you are certain you know what you are doing and the computer is no longer under warranty. Because, depending on the age of the computer, I would look into the warranty situation here.

I have to say that second screenshot almost looks bugged honestly. 150W through the pcie slot would probably melt some traces on the motherboard if it were correct.

Do you get similar readings with hwinfo?

And finally, the drivers can mess some of this up. Not the 150W from the pcie slot, if that is a correct reading, but the other issues. Follow pavles advice above, to clean install the latest hotfix driver. A clean install of Windows is not out of the question either.


This just means the gpu is hitting max voltage before the power limit. It is normal behaviour for modern gpus to boost until it hit one or more limits. This does not have to be the power limit.

I posted a picture of the inside of the case below. I'm more than happy to take additional pictures if you can't see what you're looking for!

The GPU is back in the primary pcie slot and I recently tried the other pcie cord from the psu. I currently do not have the end connector attached to the GPU but I'll try that now. (edit: tried it, no change)
I haven't messed around with any overclocking, bios flashing or any other foolish ideas (I think)

That second screenshot is what my GPU looks like most of the time, I watched the gpu operating normally and then fall into it's "bugged" state about an hour ago. It usually occurs about 15-20 minutes after a fresh restart. I also confirmed the values in hwinfo.

I've also got the newest hotfix driver installed and I wiped the computer with a fresh install of windows when I first starting noticing the issue.

Thanks for all the notes! I'm going to send the PC back to ABS under warranty if I can't get it figured out by the end of the week. I was kind of hoping to figure out that the power supply was the main culprit, as that would be faster and cheaper to replace than having to pay the shipping for the whole rig.

 

[Mussels]

Vermintide and things were going well, averaged 123 fps. I joined a match and watched my gpu chip power draw drop to 20w and the gpu clock fell to 400 MHz

This says somethings wrong with your setup and networking, or a game bug
What antivirus/firewall are you using? Can you test with windows setting your network to public vs private? (This is what helped me diagnose the link in my sig with a smart TV causing FPS drops)


PC is a little messy, but overall fine

That second screenshot is what my GPU looks like most of the time, I watched the gpu operating normally and then fall into it's "bugged" state about an hour ago. It usually occurs about 15-20 minutes after a fresh restart. I also confirmed the values in hwinfo.

Kay but show us these values, we need to see what they actually are in HWinfo, not just GPUz

Lasting 15 minutes makes it sound like somethings overheating and seeing your good and bugged ones would answer that - but it can also be malware (some virus mining software waits, and only activates when you're online and running fullscreen 3D gaming)

Reported values could definitely be broken unless you're on the new nVidia hotfix driver as well, they really did screwup some releases there
GeForce Hotfix Driver Version 526.61 | NVIDIA (custhelp.com)

 

[sne4k0]

Here is the GPU-Z and HWinfo side by side, while in-game. Let me know if you were wanting to look at a different data point that I didn't post in there.

I currently don't (or didn't) have any antivirus installed as I wiped the hard drive when I first started noticing issues with the GPU performance. I'm only using Window's Firewall. I also tested the performance with public and private network settings. No change. I did install malwarebytes and scanned twice, once while a game was running. No detections.

I am also using the newest Nvidia Hotfix Driver.

Clearly I'm no expert, but I think we've ruled just about everything other than a hardware issue.

What would you say about such high power consumption of the whole card while the GPU is just sipping power? It's not like the board has a synchro-mesh chicken plucker on it.

This was the main question I was hoping we could get answered.

I'm just going to start packing up the rig to send back for warranty.

Thanks for your help everyone!

 

[pavle]

...This was the main question I was hoping we could get answered.

I'm just going to start packing up the rig to send back for warranty.

Thanks for your help everyone!

HWInfo doesn't show the GPU watts, just the whole board, sadly nvidia has had quite many quality issues with the 30x0 cards, they just do not care. Nothing to do but to replace.