Yeah, Igor’s lab has been proven several times to not be the most reliable place or doesn’t do a good job verifying things before acting like it’s confirmed.
2 Likes
OK, I need to clarify, there is a little bit anger here I seem to sense? I have no clue why…
-
@Sargon I never told not to bend, where did you take this from? Actually I bend properly 90° and intentionally beyond the specs but properly.
-
If you own a case with 20cm width, it is definitely to small for a bending in specs as PCSIG requested.
This is what I have, and I bent carefully and properly, though its ridiuclous, because standard cases cannot be used without not being compliant with the conditions.
@jojon no I did not experience this issue, I added an extra SATA controller card, because 8x SATA is way to little amount of HDD Ports for me.
But it seems perhaps a shared PCIEX line issue. Count the PCIEX lines and the periphery components they use or share, if shared, BIOS or system driver switch and potentially causes the least devices to be seen as disabled. Read the Mainboard description about the SATA drives and potentially shared lines with PCIEX slots, I had sucha mainboard. Even worse, if NVME share such lines, you may have a silent fallback from 4 to 2 lines (I had that). Bit this is all guessing, maybe helpful I hope
- @Sargon I doubt you have read the series of articles AND PICTURES of disasembled connectors about it on Igor´s lab. You are a bit picky. Why the hell?
At least I sincerely disagree with that statement
“Anyway, the upshot of all this is that there is not a problem with the 12vhpwr connectors”
It is an issue that may burn down your 2K card easily without any need. such a weak connector with low voltage (12V) and 30-40A power surge. As an electrician I can understand the calculation and the thermal edge cases, but this is still a consumer device, where one should not seek for the edge case as “new normal”.
Again, in industry, no problem, you have fixed conditions (mostly…) but consumer? Rather not. I told already the bending issue, this is quite rubbish, or “design to cost” against customers already paying premium.
1 Like
I’m not angry, and I apologize if my responses appear that way. My concern is that there is a great deal of misinformation about this issue that has been repeated by the media and all over the internet. And my intent is to help squash the continued spread of that misinformation.
There is no such bending specification for customers that I am aware of. As far as I’ve been able to determine, the idea that there is such a specification comes from misunderstanding of an email that was leaked to the public which was meant for the engineering departments within the PCI-SIG consortium.
I know that Cablemod mirrored it and actually put it into their user documentation. There’s a chart they put up that people point to. But I think Cablemod has just flowed along with the public’s concerns and the media reporting on this issue. It’s not actually a specification that comes from PCI-SIG.
I have actually. And I am not impressed with his methods and results. I do have relevant professional hardware engineering experience, and I don’t think what Igor has done represents proper failure analysis.
Gamers Nexus, by contrast, used proper methodology. And so I don’t find it surprising that their conclusion was identical to nVidia’s conclusion.
Actual testing of the 12vhpwr connector… and not just dissecting it, looking at it, and drawing theoretical conclusions… has proven definitively that this new power connector, despite appearances to the contrary, is actually highly robust.
The only good thing about the media hysteria over melted connectors was that it caused a great deal of public torture testing of the 12vhpwr. This testing showed that sharp bends had no effect. Current in excess of double the 600W rating did not cause failure. Even deliberately sabotaging it by cutting half the wires did not cause it to fail. Tests were performed that were expected to make these things burn, and they consistently did not. The only way to reproduce the melting which has been found has been to have the connector only partially seated, and you can burn any power connector way.
As you say, I am quite picky about this kind of thing. I had concerns as well at first. They do appear to be too small and fragile. It’s hard to believe that they can safely handle 600W. But I have researched this matter a great deal and with interest in my own personal safety since I have a 4090 myself. And I have concluded to my satisfaction that these new power connectors are reliable and safe.
Don’t just believe me though. I recommend that you check into other sources besides Igor’s Lab, especially the Gamers Nexus failure analysis.
1 Like
Thanks for the tips – looks like I have some reading up to do… :7
I am confused.
I’m not angry, and I apologize if my responses appear that way.
and below
As you say, I am quite picky about this kind of thing.
so, it´s emotions, rather than facts.
I never told I rely on Igor´s lab solely, neither I told I fully agreed with him. I do not “spread of that misinformation” but have my own opinion, I still stick to it, the connector is rubbish, crap.
Don´t tell me it´s hysteria, it´s actually affecting me and all others who do not own an oversized tower.
It´s a hidden flaw, but a dangerous one. for standard Desktop cases, You simply cannot mount it properly in according with the reqired conditions, so its related to warranty. The conditions are not explained, but Nvidia is fully responsible for it.
the bending standards are published aren´t they?
You told:
I have actually (read the series of articles). And I am not impressed with his methods and results.
But in fact, You did not read the site you disliked!
"There is no such bending specification for customers that I am aware of. "
from that site you did not like at all, I found even pictures and examples like 4 cm bending radius minimum in an article from 14. of November.
You told:
I do have relevant professional hardware engineering experience
I cannot believe. Even an apprentice in Germany learns about bending radius depending on wire gauge, wire type like “stranded copper wire”, or “single solid wire”.
PCI SIG already made a statement to 12VHPWR and started a review, funny though, see PCI SIG review zone
We want to support the poster of the topic, we should not start a fight who is right about this and that.
Would you please hold back and reduce pushing your opinion pretending as facts?
AWG 16
with a single bend, the inner bending radius should not be less than 0.5 times the core diameter
But only after the ugly coat 
It IS media hysteria. Look at Gamer’s Nexus video, where he goes into extreme detail, properly tests everything, and is able to reliably replicate the melting issue and prove why it is a certain problem and is largely user error and not just crazy cable bends. He has the best video on it and even called out, rightfully, places like Igor’s Lab and Jay2Cents for jumping to conclusions and causing undue panic. Igor was wrong. At most, you could argue it was not the best design mixing with user error to cause some issues, but watch the video. The whole “bend” issue that people pointed to as the smoking gun? That is objectively not the issue.
Places like Hardware unboxed even had an extreme bend on their 4090 cable adapters, horizontally at that which was supposedly a “worst case scenario” and yet never suffered melting.
2 Likes
Yep, so true!
And then add restrictions based on the connector (!!!), defined by supplier and PCI SIG to be added.
And every PSU supplier with his own variant of the cable adapters and cables, so need to know the specs, then you may really judge
long long time ago I learned something like this “bending 4 times diameter, but factor depending on the diameter”. But it is a one time of to bend this way, because copper becomes brittle.
so it can be done properly, as I told before, but still out of spec
The problem for solid wire or (cheaper) thick stranded copper wire is the removal and re-bending where you need to maintain your machine, may it be updating or watercooling stuff etc. This stresses both the cable and the connector.
@Omniwhatever There are lots of Videos out now, related to “user errors” ok. I never refused that. Where comes your assumption, I complain about it. I don´t. But this is no advance, because how could it be such a secure connector cannot be mounted properly.
Again: I exactly told, I bent it… properly. I also can confirm it still works without a flaw… yet.
Still, it is just rubbish, it´s undersized and bad design, a result of “design to cost” only.
I have no panic, neiter panic to spread, but you cannot use such adapters within 20cmm width cases within specs, it´s too small, the bending is harsh and touches the case. This is a risk, even if it can be handled.
But he doesn’t have a 20cm case.
To your bend, for rigid wire that is correct. But it’s flexible. It’s important not to bend directly at the plug, but that’s not possible anyway because of the stupid shrink tubing.
In principle, the plug should have been delivered from the factory as 90 degrees. Well, Molex, the stupids connectors in the world xD
2 Likes
Which os why he only said within a 20cm case. So if your not in that type of case you would gloss over the detail as it is within specification. Much like you would gloss over a gpu spec that says minimal 500w psu if you have a higher capacity psu.
Card is here- just installing. The PCIe Cable has 1 plug on the power side - and 2 on the other side. so i only need 2 cables or really 4 going out of my power supply?
1 Like
Not sure can you post pics? Sone manufactures like Gainward? removed the original 12vhp connector to maybe use something like your describing.
Here… Another Problem i see is that on my power supply i only have free space where peripheral/Sata is - not where pcie is listet- is it ok to plug the pcie in there too-- in the picture they kind of look the same
Edit: I found no clear answer. But one guy sayed to not use one cable that goes into 3 but at least 3 dedicated plugs in the power supply…
So i used the 2 remaining that i have … What worries me a bit is that nvidia says at least 3 – do you guys have 4 dedicated cables? Will the performance or anything else suffer?
Thank you
1 Like
On your Psu/picture you posted, there are 4 available 8 pin outputs? Do not use 6-pin SATA outputs, they have different voltage/pinout (!)
1 Like
I plugged them in there on the right under periphals and Sata…they were 8 pin slots
I Booted up fine, and played a game. I keep the case open for now. Untill I get confirmation
1 Like
You need 4 individual power cables coming from the power supply, one going to each of those 8 pin PCIE connectors on the power adapter. What you’ve shown in the photo is wrong and potentially unsafe as you’ve daisy chained a single power cable into two plugs on the adapter. This is running your white power cables at double their rated capacity.
Those cables can typically handle it anyway which is why you’ve been able to boot it up and play games this way. But it has not been cabled safely and properly.
They look the same, but they’re actually keyed so that you can’t plug your power cables into the wrong ports.
1 Like
Yeah, that’s totally ok. I thought you somehow was able to use a 6-pin or something from the PSU. Using 3 individual cables from PSU is recommended, even though it should not affect performance as long as the 12V rail they are connected to are able to supply enough power. I belive you have the FE? My Gainward came with only 3x8pin though. You will need the 4th cable if you plan on overclocking. Running the card stock you only need 3. Worst case with the splitter is you get higher temperatures in the cable/plug if using one that splits into 2x8pins as you are running them above spec. I’ve been running 2x8 pin @540W for the past year on my 3080TI, nothing cached fire, even though it is not ideal…
Edit: Please report back if you experience the dreaded lag/jitter from the 40x0-series…
1 Like
No Problems with jitter so far. Just played Mafia 2 Highest Resolution in VR good 120-140fps so i can use lukes real vr mod… damm it looks so good… Like traveling in time.
It seems that here are to oppionion … one of you says its ok to use the cable with the 2 plugs the other says its not… I can only run 3 cables… So i could do 2 just one pluged in the card and the last cable plug both ports in?
1 Like
Yes i have the FE
1 Like
Using 3 single cables will give you 100% power limit for the card, aka full performance. If you want to maximize overclock, you would like to increase the power limit for the card above 100%. To do this, it will require a 4th cable to be plugged in, without it you will not be able to pull power above the 450W already available through 3 cables (100%). The 8 pin PCIe connectors are rated for 150W each, even on a daisy chained cable (total of 300W max on that cable). So, if you want to overclock, you can still connect all 3 cables and using a daisy for the 4th. But stock, the card will not require more than 3 cables to perform at 100%.
All 4 plugs connected will allow the card to pull up to max 600W (may only happen if you overclock and raise the power limit manually iirc, but even then I have not seen overclocking pull that much power) Although, with a 850W PSU I would not recommend overclocking it at all. Just run the card with 3 connected, you’ll be fine. There is not much performance left on the table by increasing the power limit to 600W anyway as these cards are limited by voltage way before power limit is hit.
Hope that made sense.
2 Likes