Question Power draw from PCIe slot is much higher than it should be ?

[Szyrs]

I have tested this on a number of motherboards and with different power supplies, so I’m confident that it’s an issue with the gpu itself. I have been building computers for well over 10yrs and have never seen this issue before, nor can I find anyone else posting experience of it online. I’m hoping that somebody does have some insight though…

I have a Gigabyte 1070 Mini (stock settings) that is locked to 25-30fps in benchmarks, at around 35°C.

In Gpu-Z it shows TDP around 260%, with around 388W being drawn from the PCIE slot and around 5W being drawn from the 8pin PSU connector. This is then triggering PerfCap to keep the card in idle.

As far as I’m aware, this should be melting the PCIE slot or at least causing motherboard issues. It does not.

GPU bios is factory and I have reflashed it to make sure.

As I’ve said, the symptoms persist in other rigs I have tested. I have also tried upping the power limits in both Afterburner & Aorus Engine, to no effect.

I have seen cards fail unexpectedly and seen cards catch fire but this is new to me…

Anyone have any ideas about what is going on here? PCIE slots should only supply around 75W…

(Edited for spelling)

 

[CountMike]

I have tested this on a number of motherboards and with different power supplies, so I’m confident that it’s an issue with the gpu itself. I have been building computers for well over 10yrs and have never seen this issue before, nor can I find anyone else posting experience of it online. I’m hoping that somebody does have some insight though…

I have a Gigabyte 1070 Mini (stock settings) that is locked to 25-30fps in benchmarks, at around 35°C.

In Gpu-Z it shows TDP around 260%, with around 388W being drawn from the PCIE slot and around 5W being drawn from the 8pin PSU connector. This is then triggering PerfCap to keep the card in idle.

As far as I’m aware, this should be melting the PCIE slot or at least causing motherboard issues. It does not.

GPU bios is factory and I have reflashed it to make sure.

As I’ve said, the symptoms persist in other rigs I have tested. I have also tried upping the power limits in both Afterburner & Aorus Engine, to no effect.

I have seen cards fail unexpectedly and seen cards catch fire but this is new to me…

Anyone have any ideas about what is going on here? PCIE slots should only supply around 75W…

(Edited for spelling)

Some readings must have been misinterpreted and reversed. As you said it's not possible to draw more than 75W from PCIe slot otherwise there would be no need for added power.

 

[Szyrs]

Some readings must have been misinterpreted and reversed. As you said it's not possible to draw more than 75W from PCIe slot otherwise there would be no need for added power.

It reads the same in HWMonitor as in Gpu-Z

F4353479-FE58-45-CA-96-CD-3206-EA09-BBE6 hosted at ImgBB

Image F4353479-FE58-45-CA-96-CD-3206-EA09-BBE6 hosted in ImgBB

ibb.co

D2007-B0-F-C66-F-4-D6-C-A00-F-F666-F99-A7488 hosted at ImgBB

Image D2007-B0-F-C66-F-4-D6-C-A00-F-F666-F99-A7488 hosted in ImgBB

ibb.co

 

[Tanabata]

You have a maximum of 150W if you have a 8-pin pcie connector. 350W+ coming from the pcie connector is literally impossible, and even if it did, it would have literally melted down.
Best guess is a faulty GPU giving false informations.

 

[Szyrs]

You have a maximum of 150W if you have a 8-pin pcie connector. 350W+ coming from the pcie connector is literally impossible, and even if it did, it would have literally melted down.
Best guess is a faulty GPU giving false informations.

Yeah. I mean I am about to throw it in the bin at this point - but I’d expect the card to actuslly fail if the voltage regulator was burnt out, I’d expect to see visible damage if electrical components had failed and I’d expect to see at least black/green/pink screen or artefacts for all of the above. As it is, it will run benchmarks all day ling without even dropping frames. It just won’t ramp up…

 

[CountMike]

It reads the same in HWMonitor as in Gpu-Z

F4353479-FE58-45-CA-96-CD-3206-EA09-BBE6 hosted at ImgBB

Image F4353479-FE58-45-CA-96-CD-3206-EA09-BBE6 hosted in ImgBB

ibb.co

D2007-B0-F-C66-F-4-D6-C-A00-F-F666-F99-A7488 hosted at ImgBB

Image D2007-B0-F-C66-F-4-D6-C-A00-F-F666-F99-A7488 hosted in ImgBB

ibb.co

Which MB is it in ? Is it true it's an old one with PCIe x16 at Ver 1,1 ? That could explain mixup.

 

[Tanabata]

Yeah. I mean I am about to throw it in the bin at this point - but I’d expect the card to actuslly fail if the voltage regulator was burnt out, I’d expect to see visible damage if electrical components had failed and I’d expect to see at least black/green/pink screen or artefacts for all of the above. As it is, it will run benchmarks all day ling without even dropping frames. It just won’t ramp up…

I'm not an expert, so I can't actually say "Your card is fried, mate, take another one". Sometimes, you can have sensors that are faulty and give false informations while the component is still working fine. It happens, it happens frequently, but that doesn't mean your component is fried. I don't think it's still under warrenty, but if you want to keep that card, maybe try a repair in a shop?

 

[Szyrs]

Which MB is it in ? Is it true it's an old one with PCIe x16 at Ver 1,1 ? That could explain mixup.

Today it’s an MSI Z87 but I’ve tested it in several other boards up to an H-270…

 

[Szyrs]

I'm not an expert, so I can't actually say "Your card is fried, mate, take another one". Sometimes, you can have sensors that are faulty and give false informations while the component is still working fine. It happens, it happens frequently, but that doesn't mean your component is fried. I don't think it's still under warrenty, but if you want to keep that card, maybe try a repair in a shop?

A second hand 1070 is currently around $150. I doubt I could have it repaired for less than that. It does seem a shame though…

 

[Szyrs]

Which MB is it in ? Is it true it's an old one with PCIe x16 at Ver 1,1 ? That could explain mixup.

Under load that changes to v2, even if i set the motherboard to gen3. I was not aware that this could change on the fly but everything else remains. Voltages & power draw do not change under load.

 

[DRagor]

I have a Gigabyte 1070 Mini (stock settings) that is locked to 25-30fps in benchmarks, at around 35°C.

There, low temp means the card is only sipping power and not fully powering. I would guess it is unable to draw power from PCIe connector so it get only about 80W max - which would explain why it refuses to boost and gets low scores.

I have also tried upping the power limits in both Afterburner & Aorus Engine, to no effect.

Which would mean the problem with getting more power is hardware issue.

It reads the same in HWMonitor as in Gpu-Z

Programs can only read card's sensors. If those are broken then programs report BS data.