The Truth About NVIDIA s RTX 4090 Adapters: Testing, X-Ray, & 12VHPWR Failures
video description
Date: 2022-11-16
Related videos
Comments and reviews: 15
Todd
I am not sure I agree about the evidence on the damaged cables of partial insertion. That damage could have occurred if the cable was removed while still hot. And even if this was the case it might be that the cable is working itself out. Because there is no 90 degree angle in the cable there is always going to be pressure on that cable and this pressure will be to an angle. The narrower case the worse the pressure will be. Because there are fans in the case and fans on the card there will always be vibration. Without a lockdown the vibrations and pressure on the cable might make the cable work itself out... especially under many heat cycles.
Personally with the amount of power going through such a small connector I would have had a thumbscrew on each side locking the connector in place like on older VGA cables. That way you know it is securely in place and will not vibrate loose over time. I think this was a HUGE design flaw that Nvidia is responsible for and I am afraid when you say it was user error Nvidia will use this as an excuse to not honor their warranties on damaged cards.
Personally I think these cards should never have bene sold. We are push WAY to high a power draw on these cards and even CPU where we will be struggling over time with more and more issues like this. I do not find it exemptible for GPUs and CPUs do push into these realms of power draw... mostly because it can give them a few percentages of extra performance. There should be a cap of power draw that these companies should be required to stay under... likely 250w for GPU and 150 for COU. I think that market has gone the wrong way and far too often and the average computer user has no clue there system is throttling all the time and not actually running at the specs they paid for. Do computers warn you if they are thermal throttling? Nope. Most people are not even aware.
reply
I am not sure I agree about the evidence on the damaged cables of partial insertion. That damage could have occurred if the cable was removed while still hot. And even if this was the case it might be that the cable is working itself out. Because there is no 90 degree angle in the cable there is always going to be pressure on that cable and this pressure will be to an angle. The narrower case the worse the pressure will be. Because there are fans in the case and fans on the card there will always be vibration. Without a lockdown the vibrations and pressure on the cable might make the cable work itself out... especially under many heat cycles.
Personally with the amount of power going through such a small connector I would have had a thumbscrew on each side locking the connector in place like on older VGA cables. That way you know it is securely in place and will not vibrate loose over time. I think this was a HUGE design flaw that Nvidia is responsible for and I am afraid when you say it was user error Nvidia will use this as an excuse to not honor their warranties on damaged cards.
Personally I think these cards should never have bene sold. We are push WAY to high a power draw on these cards and even CPU where we will be struggling over time with more and more issues like this. I do not find it exemptible for GPUs and CPUs do push into these realms of power draw... mostly because it can give them a few percentages of extra performance. There should be a cap of power draw that these companies should be required to stay under... likely 250w for GPU and 150 for COU. I think that market has gone the wrong way and far too often and the average computer user has no clue there system is throttling all the time and not actually running at the specs they paid for. Do computers warn you if they are thermal throttling? Nope. Most people are not even aware.
reply
Zenn
There is a much safer high-power 12 volt connector standard in the RV industry that would solve this problem completely. It's used to run the 12 volt lines from generator to applications and from inverter to batteries. It's 40 years old and hasn't produced a failure that I've been able to confirm. In fact, they are often reused because there is so little wear over time. The failure point is actually the durability of the rubber shielding and even that can be extended by rubbing a penetrating oil over the cables during winterization.
So as awesome as this piece is, you were already starting from a false reality. One created through planned obsolescence to ensure cables like these are replaced. Otherwise the companies manufacturing them won't be able to extract profit from the manufacturing and distribution of said product.
What we have here is a new version of the 1000 hour light bulb. And much like consumerist trash bulbs, they exist along side a century old bulb in a fire house that just keeps on keepin on, back in actual reality. The RV industry has a plethora of high output 12 volt solutions to choose from. Also, mind those electromagnetic fields. If you map them out for this specific type of connector you should discover, as I did, another glaring problem with this design.
The moral of this story is, consumerism is immortal cancer. It's not gone just because we forget what used to exist.
reply
There is a much safer high-power 12 volt connector standard in the RV industry that would solve this problem completely. It's used to run the 12 volt lines from generator to applications and from inverter to batteries. It's 40 years old and hasn't produced a failure that I've been able to confirm. In fact, they are often reused because there is so little wear over time. The failure point is actually the durability of the rubber shielding and even that can be extended by rubbing a penetrating oil over the cables during winterization.
So as awesome as this piece is, you were already starting from a false reality. One created through planned obsolescence to ensure cables like these are replaced. Otherwise the companies manufacturing them won't be able to extract profit from the manufacturing and distribution of said product.
What we have here is a new version of the 1000 hour light bulb. And much like consumerist trash bulbs, they exist along side a century old bulb in a fire house that just keeps on keepin on, back in actual reality. The RV industry has a plethora of high output 12 volt solutions to choose from. Also, mind those electromagnetic fields. If you map them out for this specific type of connector you should discover, as I did, another glaring problem with this design.
The moral of this story is, consumerism is immortal cancer. It's not gone just because we forget what used to exist.
reply
John
My assumption was the connection where wire meets pin was starting to get damaged because of this connector being at the top of the GPU, and these are really big GPUs and a PC side panel is often going to bend the CRAP out of that cable. I'm not convinced that this doesn't happen at all but what this shows is still based on the same thing. You have a big ass GPU and a power connector mounted at the top where a cable has to go straight down into it. A person puts it in their system, puts the side panel back on and the bending starts to happen right away. Eventually this is loosening the connection due to heat. A GPU still tries to draw the amount of power across those pins but now there's a lot of resistance and it heats up and melts.
I theorized that the solder joint or crimp, depending on how it's connected can get damaged because of the torque put on it and over time that torque, added with the heat from normal operation is enough to start damaging the connection point. This provides that high resistance path that was talked about and over time the connection becomes more damaged, creating more resistance, and then melt.
The only was to simulate this is to seat a cable PROPERLY and then put a LOT of torque on that cable to one side, just like a PC case side panel would do with a case that doesn't have a lot of clearance above the GPU.
reply
My assumption was the connection where wire meets pin was starting to get damaged because of this connector being at the top of the GPU, and these are really big GPUs and a PC side panel is often going to bend the CRAP out of that cable. I'm not convinced that this doesn't happen at all but what this shows is still based on the same thing. You have a big ass GPU and a power connector mounted at the top where a cable has to go straight down into it. A person puts it in their system, puts the side panel back on and the bending starts to happen right away. Eventually this is loosening the connection due to heat. A GPU still tries to draw the amount of power across those pins but now there's a lot of resistance and it heats up and melts.
I theorized that the solder joint or crimp, depending on how it's connected can get damaged because of the torque put on it and over time that torque, added with the heat from normal operation is enough to start damaging the connection point. This provides that high resistance path that was talked about and over time the connection becomes more damaged, creating more resistance, and then melt.
The only was to simulate this is to seat a cable PROPERLY and then put a LOT of torque on that cable to one side, just like a PC case side panel would do with a case that doesn't have a lot of clearance above the GPU.
reply
Michaels
I'll start by saying overall great job. The industry will only be better for all of this. Now; might want to not advise people to wriggle and pull a fully set connector to test it's fully seated as that will then stress and break the connector solder joints to the board as you know there will be some very enthusiastic testing of doing that. Anyway, interesting conclusion, people are just that weak at plugging a plug. Doesn't surprise, they've had to key everything for a reason too, people were blowing up AT gear a lot because they failed at colours in kindergarten, now they are showing they also failed at shapes. Maybe the connector face should be a little higher for easier view in the video card, especially from the back of it which would be a typical view for a user that has an installed card, just notch that damn back plate as it's not like it's taking from the aesthetic from having a gonking big wire bundle sticking from it which evga ironically fixed by having it rear mount that we'll never see in retail. Also another interesting thing is connector wear from socket cycles. The connector still needs a revision, having it plated and very susceptible to being worn off is a huge nope. Making me think to not get an atx 3.0 supply now that has an attached prone to wearing out consumable on it.
reply
I'll start by saying overall great job. The industry will only be better for all of this. Now; might want to not advise people to wriggle and pull a fully set connector to test it's fully seated as that will then stress and break the connector solder joints to the board as you know there will be some very enthusiastic testing of doing that. Anyway, interesting conclusion, people are just that weak at plugging a plug. Doesn't surprise, they've had to key everything for a reason too, people were blowing up AT gear a lot because they failed at colours in kindergarten, now they are showing they also failed at shapes. Maybe the connector face should be a little higher for easier view in the video card, especially from the back of it which would be a typical view for a user that has an installed card, just notch that damn back plate as it's not like it's taking from the aesthetic from having a gonking big wire bundle sticking from it which evga ironically fixed by having it rear mount that we'll never see in retail. Also another interesting thing is connector wear from socket cycles. The connector still needs a revision, having it plated and very susceptible to being worn off is a huge nope. Making me think to not get an atx 3.0 supply now that has an attached prone to wearing out consumable on it.
reply
Chris
The question I would ask of Nvidia is whether they are happy about a failure rate (thus far) of one in 1000 to 2000 when the consequence of the failure in the context of these devices being hidden inside a case when failure could occur means it is possible that the failure could have extremely dire consequences including loss of life.
The question I would ask more generally is what can be done with these connectors to mitigate this issue more fully. Very high power electrical connections are a mature field at this point,, and there are plenty of examples of very high power connectors that are incredibly safe and user error resistant. A sensor circuit to ensure proper connection would be a band aid in this sense because positive connections should be required in proper designs before significant current can flow.
reply
The question I would ask of Nvidia is whether they are happy about a failure rate (thus far) of one in 1000 to 2000 when the consequence of the failure in the context of these devices being hidden inside a case when failure could occur means it is possible that the failure could have extremely dire consequences including loss of life.
The question I would ask more generally is what can be done with these connectors to mitigate this issue more fully. Very high power electrical connections are a mature field at this point,, and there are plenty of examples of very high power connectors that are incredibly safe and user error resistant. A sensor circuit to ensure proper connection would be a band aid in this sense because positive connections should be required in proper designs before significant current can flow.
reply
Brian_Hub
Them not being completely at fault for this cable. Doesn't back up any assertion that they aren't incompetent or evil Steve. That's baseless speculation. The size contributes for a need to bend. The socket not seating makes that a huge potential failure point. As you could have rammed it into the hilt. Still not connected. Managed cables. And then had a failure you wouldn't expect. And poor manufacturing could mean bits of metal inside the connectors is also a failure point. All that points to greed. And is it not evil to put your profit above everything else? I mean they didn't rig them to explode. But also. A lot of that is on them. And I'd wager people thinking they seated them and then managing cables, and not checking again. Is more common than, DUR is fine half way in.
reply
Them not being completely at fault for this cable. Doesn't back up any assertion that they aren't incompetent or evil Steve. That's baseless speculation. The size contributes for a need to bend. The socket not seating makes that a huge potential failure point. As you could have rammed it into the hilt. Still not connected. Managed cables. And then had a failure you wouldn't expect. And poor manufacturing could mean bits of metal inside the connectors is also a failure point. All that points to greed. And is it not evil to put your profit above everything else? I mean they didn't rig them to explode. But also. A lot of that is on them. And I'd wager people thinking they seated them and then managing cables, and not checking again. Is more common than, DUR is fine half way in.
reply
NullPoEx
so, I noticed you didn't mention vibration, which is surprising when you rocked the cable out of its inserted position at 7:35.
You have a card with oscillating fans on a card that is usually held in place by nothing more than PCIE slot and a few screws at the extreme end of it, and maybe a kickstand to stop drooping on the other far end. This card is VIBRATING CONSTANTLY when those fans are running. The case also likely has fans, causing the case to vibrate as well at a different speed. If the cable isn't LOCKED into place it will vibrate out of the socket all on its own over time, which makes this cable a ticking time bomb unless you check it every now and then and make sure the cable is firmly set and locked in place.
reply
so, I noticed you didn't mention vibration, which is surprising when you rocked the cable out of its inserted position at 7:35.
You have a card with oscillating fans on a card that is usually held in place by nothing more than PCIE slot and a few screws at the extreme end of it, and maybe a kickstand to stop drooping on the other far end. This card is VIBRATING CONSTANTLY when those fans are running. The case also likely has fans, causing the case to vibrate as well at a different speed. If the cable isn't LOCKED into place it will vibrate out of the socket all on its own over time, which makes this cable a ticking time bomb unless you check it every now and then and make sure the cable is firmly set and locked in place.
reply
laserspike
Great video Steve. That makes a lot of sense. Just one quibble: please think again about saying that a parallel high-resistance path could cause overheating, because I don't see how this makes any sense. Example: given a 5 milliohm path and a 20 milliohm path in parallel, most of the current will flow through the 5 milliohm path and 4 times as much heat will be generated in that resistance (P = V 2/R, where the voltage is the same across all pins). So unless the heat conduction away from the 20 milliohm path is somehow absurdly awful (requiring the thermal contact to be disproportionately worse than the electrical contact; unlikely, no?), the high-resistance path ain't gonna be the one to melt.
reply
Great video Steve. That makes a lot of sense. Just one quibble: please think again about saying that a parallel high-resistance path could cause overheating, because I don't see how this makes any sense. Example: given a 5 milliohm path and a 20 milliohm path in parallel, most of the current will flow through the 5 milliohm path and 4 times as much heat will be generated in that resistance (P = V 2/R, where the voltage is the same across all pins). So unless the heat conduction away from the 20 milliohm path is somehow absurdly awful (requiring the thermal contact to be disproportionately worse than the electrical contact; unlikely, no?), the high-resistance path ain't gonna be the one to melt.
reply
Crushonius
i am over this just call it for what it is a S H I T T Y design . END of story
i would not trust that to be safe in the lifecycle of the card
if its happening with brand new cards and adapters , what will happen when you had it for a couple years and
had to take the card out for cleaning or put it in a new case or computer . no way it would be like leaving candles to burn unattended,
most of the time its fine until you return to find your house burnt down or worse you go to sleep leaving your pc
to render something and thats the last thing you do . sorry who in their right mind would be happy with such a potential threat
reply
i am over this just call it for what it is a S H I T T Y design . END of story
i would not trust that to be safe in the lifecycle of the card
if its happening with brand new cards and adapters , what will happen when you had it for a couple years and
had to take the card out for cleaning or put it in a new case or computer . no way it would be like leaving candles to burn unattended,
most of the time its fine until you return to find your house burnt down or worse you go to sleep leaving your pc
to render something and thats the last thing you do . sorry who in their right mind would be happy with such a potential threat
reply
Alex
Truly amazing job being the most responsible reporters of this issue and being the only ones that I know of to actually root-cause the issue. I hope that your peers will take this to heart and be a bit more responsible in their reporting of issues and withhold judgement until they can actually reproduce the issue. As someone who often works in fixing software defects reported by validation teams, the first, most basic step is being able to reproduce the issue... I should hope that much should've been obvious. If you can't reproduce the issue, then you have no means of finding the root cause.
reply
Truly amazing job being the most responsible reporters of this issue and being the only ones that I know of to actually root-cause the issue. I hope that your peers will take this to heart and be a bit more responsible in their reporting of issues and withhold judgement until they can actually reproduce the issue. As someone who often works in fixing software defects reported by validation teams, the first, most basic step is being able to reproduce the issue... I should hope that much should've been obvious. If you can't reproduce the issue, then you have no means of finding the root cause.
reply
DieselDrax
As much as I've really enjoyed running NVidia GPUs for as long as I can remember, next upgrade I'm going to go with an AMD GPU. My most recent upgrade was when I built a new PC and went from a 1080Ti to an RTX 3080 FE, I went with the RTX because Ooo, better ray-tracing performance but in reality RTX is something I really don't use and often times I don't even remember to check if RTX support exists. On the other hand, the 3080 should get me by for a while still, but if it were to croak then I'll replace it with an RX 7900 XTX and then have an all-AMD system again.
reply
As much as I've really enjoyed running NVidia GPUs for as long as I can remember, next upgrade I'm going to go with an AMD GPU. My most recent upgrade was when I built a new PC and went from a 1080Ti to an RTX 3080 FE, I went with the RTX because Ooo, better ray-tracing performance but in reality RTX is something I really don't use and often times I don't even remember to check if RTX support exists. On the other hand, the 3080 should get me by for a while still, but if it were to croak then I'll replace it with an RX 7900 XTX and then have an all-AMD system again.
reply
Christian
I have another mode of failure to report! On my Gigabyte Gaming OC, the retention clip broke off on the 12vhpwr port the first time I plugged in the adapter that shipped with it. From then on, it was impossible to properly insert the cable as it would not fully seat in the port. Amazing how a millimeter of thin plastic can make the difference. I have plenty of video and photos as I was preparing my own report when GN dropped this. Thanks for all the research and work you've done on this issue! This is quite comprehensive and extremely well done.
reply
I have another mode of failure to report! On my Gigabyte Gaming OC, the retention clip broke off on the 12vhpwr port the first time I plugged in the adapter that shipped with it. From then on, it was impossible to properly insert the cable as it would not fully seat in the port. Amazing how a millimeter of thin plastic can make the difference. I have plenty of video and photos as I was preparing my own report when GN dropped this. Thanks for all the research and work you've done on this issue! This is quite comprehensive and extremely well done.
reply
Cole
This is an amazing, thorough and detailed analysis. Thanks, Steve! Kind of amazed we didn't see more similar behavior with the older nvidia 12-pin connector/adapters. I guess because the orientation of the connector on the old 3000 series FE cards and the more streamline connector without the sense pin having more of a buffer space. Interesting. Makes sense why we didn't really see this with the 3090Ti cards, avoided that sense bulge since most people used the nvidia 12-pin and the adapter was just the 12 pin vs 16-pin and 1 to 1 pinning
reply
This is an amazing, thorough and detailed analysis. Thanks, Steve! Kind of amazed we didn't see more similar behavior with the older nvidia 12-pin connector/adapters. I guess because the orientation of the connector on the old 3000 series FE cards and the more streamline connector without the sense pin having more of a buffer space. Interesting. Makes sense why we didn't really see this with the 3090Ti cards, avoided that sense bulge since most people used the nvidia 12-pin and the adapter was just the 12 pin vs 16-pin and 1 to 1 pinning
reply
Jonathan
My takeaway is it's a poorly designed cable. If it's not obvious when it's fully seated and it takes an unreasonable amount of force to seat the connector then it's a poor design.
Things need to be designed in a way that takes account of how people are likely to interact with them. It needs to make an obvious click when it is seated. It should also be more tolerant of movement as keeping a pc looking tidy and routing cables is kind of a big thing with PC building. Nvidia and the cable designers should know these things.
reply
My takeaway is it's a poorly designed cable. If it's not obvious when it's fully seated and it takes an unreasonable amount of force to seat the connector then it's a poor design.
Things need to be designed in a way that takes account of how people are likely to interact with them. It needs to make an obvious click when it is seated. It should also be more tolerant of movement as keeping a pc looking tidy and routing cables is kind of a big thing with PC building. Nvidia and the cable designers should know these things.
reply
viffer5
It's amazing that a multi-billion dollar company (NVIDIA) can't figure this out, but a handful of dedicated, intelligent, and resourceful computer geeks have a far better idea of what's going on after arguably far more extensive investigation. This goes to show how much Nvidia actually cares when they have way more resources to throw at this issue but have done next to nothing about the problem. Amazing work as always!
reply
It's amazing that a multi-billion dollar company (NVIDIA) can't figure this out, but a handful of dedicated, intelligent, and resourceful computer geeks have a far better idea of what's going on after arguably far more extensive investigation. This goes to show how much Nvidia actually cares when they have way more resources to throw at this issue but have done next to nothing about the problem. Amazing work as always!
reply
Add a review, comment
Other channel videos
