• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

50 series card?

Status
Not open for further replies.
T

triviumnate

Gawd
Joined
Feb 16, 2005
Messages
852
is it safe to buy a 5070TI for example like is their still a risk of it catching fire? what about a new psu that powers it are those safe to?

not in the loop sorry
 
There have been 5080s also having issues. Recently I even saw an AMD card have a melted connector. Link

The real reason is typically user error, a poor quality/under watt psu and/or faulty cable.

My recipie thus far to avoid this fiasco has been.

1) use my brain
2) always use 1200 or higher wattage psu
3) don’t bend the connector
4) make sure it is pushed all the way in (have clicking sound)
5) check cable to see if it is burning up or room temp/luke warm after 20 rounds of steel nomad stress test at max power
6) undervolt the card

Not sure if this will continue to work for me but the first point definitely has helped. 👀
 
Last edited:
GoldenTiger said:
Millions* of cards, actually. And the melted rate looks big because of forum and YouTube sensationalism. Doesn't take much for it to be kicked off then regurgitated constantly by your average redditor.

It is wildly overblown. When Nvidia had already sold hundreds of thousands of 4090 cards, they said manufacturers had received under 50 melted cards, and all were user error. That's such a tiny percentage as to be a rounding error.
Click to expand...
Post I made in another thread.
 
natej315 said:
is it safe to buy a 5070TI for example like is their still a risk of it catching fire? what about a new psu that powers it are those safe to?

not in the loop sorry
Click to expand...

The only cards that I would have a significant worry about burning down would either be a 4090 with power limits raised above 500W, or a 5090 (which should have just fucking had two of them; it's idiotic that it only has one). With a 5070 Ti, you can definitely get it to burn down, but you would probably have to be rough with the connector. Even under load, they tend to use well below 300W, which is half of what the connector is rated for. That, is half of the thing could be completely faulty and it would still function fine.

Basically the only rule of thumb, if you're still worried: be careful with the connector. Make (double, triple, quadruple) sure it's properly seated, and try to follow best practices on how far away from the card you are before there are any bends in cable. I think for anything below a 4090/5090, it's mostly going to be physical stressors causing a lapse in connectivity for a subset of the pins.

Keep in mind the connector pins are only rated for a certain amount of insertions, too. The same is technically true for 8 pin, but 8 pin has such a high margin of safety that it's generally not an issue. With 8 pin you could keep reusing the same connector for decades. With 12VHPWR, try to keep some mental track of it. I'm using a 12VHPWR->2x8Pin adapter that came with a 5070 Ti, with a rather old Seasonic 850W PSU. There have been absolutely no issues. I made sure to get a wide enough case to allow the adapter some distance out from the plug before bending.
 
Technically even the 150w PCIe connectors had burns. Just didn’t make mainstream news. Why the connector burns has a lot to do with conditions of usage, quality of pins, contact and sustained loads. An average user facing an issue is no different than any other electronic item facing an issue.
 
StoleMyOwnCar said:
The only cards that I would have a significant worry about burning down would either be a 4090 with power limits raised above 500W, or a 5090 (which should have just fucking had two of them; it's idiotic that it only has one). With a 5070 Ti, you can definitely get it to burn down, but you would probably have to be rough with the connector. Even under load, they tend to use well below 300W, which is half of what the connector is rated for. That, is half of the thing could be completely faulty and it would still function fine.

Basically the only rule of thumb, if you're still worried: be careful with the connector. Make (double, triple, quadruple) sure it's properly seated, and try to follow best practices on how far away from the card you are before there are any bends in cable. I think for anything below a 4090/5090, it's mostly going to be physical stressors causing a lapse in connectivity for a subset of the pins.

Keep in mind the connector pins are only rated for a certain amount of insertions, too. The same is technically true for 8 pin, but 8 pin has such a high margin of safety that it's generally not an issue. With 8 pin you could keep reusing the same connector for decades. With 12VHPWR, try to keep some mental track of it. I'm using a 12VHPWR->2x8Pin adapter that came with a 5070 Ti, with a rather old Seasonic 850W PSU. There have been absolutely no issues. I made sure to get a wide enough case to allow the adapter some distance out from the plug before bending.
Click to expand...
5090 doesn’t need two, it just needs a power balancing circuit so it’s not pulling most of the current across one pin.

Adding two just takes up more space without actually solving the root cause.
 
natej315 said:
the connector can't/ shouldn't bend? or the wire? I have a core P6 case
also 5070VS 5070TI?
Click to expand...
my adapter, 2x8 pin to this 12pin or whatever it is from rtx 4070S.
The cables i.e. wires are bent, to bend just the connector you have to be superman.
With everything below rtx 4090/5090 there should be no problem
i have a medium quality 650w power supply and it works normally

One tip, if for any reason you have to remove the gpu from the pc, do not remove the 12pin cable from the gpu, leave it inside, or leave the adapter or cable in the gpu, unplug it from the psu.
Recently i had to remove the gpu and i forgot and i pulled out the 12pin instead of just unplugging the adapter from the psu.
I overclocked the ram and the computer crashed and i had to remove the cmos battery, and i can only get to the battery if i remove the gpu
 

Attachments

  • IMG_20250824_213650992.jpg
    IMG_20250824_213650992.jpg
    157.3 KB · Views: 0
Last edited:
natej315 said:
the connector can't/ shouldn't bend? or the wire? I have a core P6 case
also 5070VS 5070TI?
Click to expand...
I have a 5090 and a 5080 here and no issues. I have also owned a 5070 and a 5070 TI, and between the two I'd go with the TI card. The further down the list you go the less power consumption each card uses, so less worry for a melting issue.
 
For some peace of mind you can upgrade to a PSU with a thermistor on the cable. I use the ASRock PG-1300G. "TempGuard is a feature in ASRock's new power supply units (PSUs) that includes a temperature probe on the 12V-2×6 power cables. This innovation provides "Cable Over-Temperature Protection," automatically shutting down the power supply if the cable overheats, helping to prevent damage to both the PSU and the connected graphics card."
 
WilyKit said:
5090 doesn’t need two, it just needs a power balancing circuit so it’s not pulling most of the current across one pin.

Adding two just takes up more space without actually solving the root cause.
Click to expand...
Having each connector draw 300 watts instead of 600, which is the max rated capacity of the cable, would absolutely make a huge difference. It also implies some PCB level changes would have to be made to accommodate two connectors, but there are already 5090 designs that have 2 connectors out in the wild for LNO2 overclocking.

I don't care what anyone tells me, having a cable actually running at its maximum possible capacity, out in the wild, is an asinine idea. Also irrelevant to the TC.

natej315 said:
the connector can't/ shouldn't bend? or the wire? I have a core P6 case
also 5070VS 5070TI?
Click to expand...
I'm not sure how many companies are including this as part of their documentation with the 8pin -> 12VHPWR adapters, but when the 4090 launched a lot of them included a slip of paper telling you to not have any sharp bends in the adapter's cabling until you were a certain distance away from the connector. You could probably find that if you googled it. I personally decided to follow that precaution. One thing you could for instance do is tie the connector cables, as they're going out, to an AIO unit's tubing, if you happen to have an AIO. This would give the wiring structural stability and possibly minimize stress on the connector, if you do it right.

Though it's not as big of a deal for this card anyway, with its low TDP.
 
StoleMyOwnCar said:
Having each connector draw 300 watts instead of 600, which is the max rated capacity of the cable, would absolutely make a huge difference. It also implies some PCB level changes would have to be made to accommodate two connectors, but there are already 5090 designs that have 2 connectors out in the wild for LNO2 overclocking.

I don't care what anyone tells me, having a cable actually running at its maximum possible capacity, out in the wild, is an asinine idea. Also irrelevant to the TC.


I'm not sure how many companies are including this as part of their documentation with the 8pin -> 12VHPWR adapters, but when the 4090 launched a lot of them included a slip of paper telling you to not have any sharp bends in the adapter's cabling until you were a certain distance away from the connector. You could probably find that if you googled it. I personally decided to follow that precaution. One thing you could for instance do is tie the connector cables, as they're going out, to an AIO unit's tubing, if you happen to have an AIO. This would give the wiring structural stability and possibly minimize stress on the connector, if you do it right.

Though it's not as big of a deal for this card anyway, with its low TDP.
Click to expand...
I didn’t say it won’t make a differ difference. I said it won’t solve the problem. You can still have one wire carrying 90% of the load so now you’re stuck with 2x the space taken up. 2x the cables run and still one pin with load. Put a load balancing circuit and problem solved with one connector.

The issue isn’t the connectors power capacity being exceeded. The issue is not utilizing all the wires that are available. So why double the availability if you’re not going to fix the issue that’s causing you to use a single wire to begin with? Makes no sense

That’s like having a 6 lane freeway with everyone using one lane and your solution to the traffic jam is to add 6 more lanes while ignoring the one lane everyone’s using.
 
Last edited:
KickAssCop said:
There have been 5080s also having issues. Recently I even saw an AMD card have a melted connector.

The real reason is typically user error, a poor quality/under watt psu and/or faulty cable.

My recipie thus far to avoid this fiasco has been.

1) use my brain
2) always use 1200 or higher wattage psu
3) don’t bend the connector
4) make sure it is pushed all the way in (have clicking sound)
5) check cable to see if it is burning up or room temp/luke warm after 20 rounds of steel nomad stress test at max power
6) undervolt the card

Not sure if this will continue to work for me but the first point definitely has helped. 👀
Click to expand...
Hmmm,
1000w p/s

No clicky clicky on connector, more like my bones cracking with the force needed

No yellow showing either on the power supply which also uses this horrid connector and 5090. MSI power supply that came with the card.

Do not undervolt either.

Connector only gets hot when the back plate gets hot enough to fry an egg on it. WTF?

So far so good, I keep an eye on the 5090 side cable connector, touch it periodically, which is utterly ridiculous. Going to put some heat sinks and probably a fan on the backplate so it itself does not melt the connector.
 
Last edited:
noko said:
Hmmm,
1000w p/s

No clicky clicky on connector, more like my bones cracking with the force needed

No yellow showing either on the power supply which also uses this horrid connector and 5090. MSI power supply that came with the card.

Do not undervolt either.

Connector only gets hot when the back plate gets hot enough to fry an egg on it. WTF?

So far so good, I keep an eye on the 5090 side cable connector, touch it periodically, which is utterly ridiculous. Going to put some heat sinks and probably a fan on the backplate so it itself does not melt the connector.
Click to expand...
1000W MSI PSU is one of the 4 PSUs that Linus busted in his testing. I still don’t understand why people skimp on PSU. Difference between a 850W to a 1000W is like 30$ and another 30$ to get a quality 1200W. You pay 750$ or higher for a 5070 Ti or higher and can’t drop another 30-60$? Your money your problems.

Good luck with that. I have seen some MSI yellow connectors melt so just be careful. I have a MSI card but I am using the Lian Li 1200 W PSU cable.

It is ridiculous if your cable is very hot to the touch. Also undervolt as it is free performance.
 

Attachments

  • IMG_8779.png
    IMG_8779.png
    1.3 MB · Views: 0
WilyKit said:
I didn’t say it won’t make a differ difference. I said it won’t solve the problem. You can still have one wire carrying 90% of the load so now you’re stuck with 2x the space taken up. 2x the cables run and still one pin with load. Put a load balancing circuit and problem solved with one connector.

The issue isn’t the connectors power capacity being exceeded. The issue is not utilizing all the wires that are available. So why double the availability if you’re not going to fix the issue that’s causing you to use a single wire to begin with? Makes no sense

That’s like having a 6 lane freeway with everyone using one lane and your solution to the traffic jam is to add 6 more lanes while ignoring the one lane everyone’s using.
Click to expand...

"90% of the load on one wire" would be 300W as opposed to 600W. Which, most (most=almost all) 300W and below cards are having absolutely no issues. I literally said some circuit level changes would be necessary. Obviously one of them is splitting its supply so that it gets a maximum of half from one plug and half from another (or up to 300W from one then overflow to the other, whichever), rather than just idiotically melding them all together into one. I don't see why you're assuming I would just choose the most retarded possible configuration of having two connectors. Likewise, just "solving the root problem" like you said wouldn't do much because they're both the root problem. Part of the problem is that the connector is using its max capacity and any small difference in impedance just causes a huge trainwreck, another part is that the card's design to actually receiving its power from the pins is absolutely unintelligent. I'm pretty damn sure we already have plenty of blueprints for simply splitting supply between two plugs. I think trying to add in circuitry to perfectly balance and detect imbalances across 12 pins sounds a hell of a lot more complicated.

Whatever, done arguing this shit, it's not even the TC's question. He's not hooking up a 5090.
 
StoleMyOwnCar said:
"90% of the load on one wire" would be 300W as opposed to 600W. Which, most (most=almost all) 300W and below cards are having absolutely no issues. I literally said some circuit level changes would be necessary. Obviously one of them is splitting its supply so that it gets a maximum of half from one plug and half from another (or up to 300W from one then overflow to the other, whichever), rather than just idiotically melding them all together into one. I don't see why you're assuming I would just choose the most retarded possible configuration of having two connectors. Likewise, just "solving the root problem" like you said wouldn't do much because they're both the root problem. Part of the problem is that the connector is using its max capacity and any small difference in impedance just causes a huge trainwreck, another part is that the card's design to actually receiving its power from the pins is absolutely unintelligent. I'm pretty damn sure we already have plenty of blueprints for simply splitting supply between two plugs. I think trying to add in circuitry to perfectly balance and detect imbalances across 12 pins sounds a hell of a lot more complicated.

Whatever, done arguing this shit, it's not even the TC's question. He's not hooking up a 5090.
Click to expand...
No it wouldn’t. What makes you think adding another connector means the load will be split equally between the connectors and unequally between the wires on the connectors? If you’re gonna make circuit level changes then there’s no need for the 2nd connector. I welcome AIBs to add that 2nd connector on their highest end models for those folks who are into extreme overclocking, but to have it as a standard is completely unnecessary
 
KickAssCop said:
1000W MSI PSU is one of the 4 PSUs that Linus busted in his testing. I still don’t understand why people skimp on PSU. Difference between a 850W to a 1000W is like 30$ and another 30$ to get a quality 1200W. You pay 750$ or higher for a 5070 Ti or higher and can’t drop another 30-60$? Your money your problems.

Good luck with that. I have seen some MSI yellow connectors melt so just be careful. I have a MSI card but I am using the Lian Li 1200 W PSU cable.

It is ridiculous if your cable is very hot to the touch. Also undervolt as it is free performance.
Click to expand...
I agree, bought a super flower 1200w but it had the old style 12V connector and not enough pcie 8 pin plugs for adapter and motherboard. Hence using the MSI power supply that came with the 5090. Don’t really want to buy a third power supply. Supplying the 1000w MSI power supply 240v which will help it stay cooler and be more efficient. Still at load that 5090 does draw close to 600w, plus at least 200w more for rest of system does not leave much room.

Cable stays cooler, connector cool at load, until the backplate heats up. Took amp readings of cable which is a nice cable and all are close to each other.
 
When I was looking I couldn’t find a reasonably priced super flower, Silverstone and seasonic. I ended up with Lian Li as it was only one available, in budget and with great reviews. Asrock unfortunately wasn’t ready to ship direct to me so Lian Li it was.

Either way, if you think everything is good then most likely you will be fine. The likelihood is still very low for this connector to bust up and from testing it needs to stay at over 150 C sustained for over 2 hours or so to melt. I doubt most situations will yield that unless you are gaming or doing AI workloads with hours on end and not even checking wtf is happening around you.

As for watts, just undervolt and it should at least drop a 100w easy. I dropped over 150 w on average using 975 mV @ 3000 MHz.
 
KickAssCop said:
Either way, if you think everything is good then most likely you will be fine. The likelihood is still very low for this connector to bust up and from testing it needs to stay at over 150 C sustained for over 2 hours or so to melt. I doubt most situations will yield that unless you are gaming or doing AI workloads with hours on end and not even checking wtf is happening around you.
Click to expand...
If it melts it probably means you have a defective card, cable or PSU unless you fucked up plugging it in. So it'll probably be fine. Just knowing these connectors can melt if you don't plug them in right massively reduces your chances of having a problem. If you're the slightest bit concerned about it you'll double check and make sure it's plugged in properly, so it won't melt unless something is defective.
 
Status
Not open for further replies.
Back
Top