NV 12VHPWR issues revisited

Page 31 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
Toggle sidebar Toggle sidebar
I

igor_kavinski

Lifer
Jul 27, 2020
26,976
18,578
146
They had bad scary headlines with 450W. What do they smoke at Nvidia companywide that they thought, "oh, 575W won't be a problem. We overbuilt this thing for exactly 576W! Surely no developer can push our GPU to go beyond that!"?
 
DrMrLordX

DrMrLordX

Lifer
Apr 27, 2000
22,805
12,848
136
Hail The Brain Slug said:
What? This failure had nothing to do with the cable or installation. The card's VRM went up in flames on the opposite side of the card from the connector.
Click to expand...
Wait, it was a VRM failure? Then why bring it up in this thread (yes I know it was igor, not you)? That's almost certainly on Zotac.
 
coercitiv

coercitiv

Diamond Member
Jan 24, 2014
7,281
17,192
136
Hail The Brain Slug said:
The card's VRM went up in flames on the opposite side of the card from the connector.
Click to expand...
It simply could not wait any longer for the connector to catch fire. It was burning with excitement.

igor_kavinski said:
Guilty as charged, sire.
Click to expand...
Your honor @DrMrLordX , I'll be taking over @igor_kavinski 's defense from here, I'm the forum's appointed defense lawyer a.k.a devi'l advocate. The reason this Zotac card likely burned with open flame and smoke has a lot more in common with the OP than one might think: the same root cause is at play, an oversimplification of power delivery on premium cards.

I present to you expert rant evidence #1:

continued with evidence #2 - if you treat your 12HPWR connector internally as three connectors and split power to three groups of VRM stages, then you can mount fuses for each group.

But if you treat the entire connector as one big blob... who needs fuses anymore!? :D
 
  • Haha
  • Love
Reactions: Elfear, lightmanek, Tlh97 and 1 other person
DrMrLordX

DrMrLordX

Lifer
Apr 27, 2000
22,805
12,848
136
That's interesting. But was that Zotac's ultimiate decision, or was that them cribbing from NV's reference design? I can definitely see how the connector design contributed to that decision . . .
 
coercitiv

coercitiv

Diamond Member
Jan 24, 2014
7,281
17,192
136
DrMrLordX said:
But was that Zotac's ultimiate decision, or was that them cribbing from NV's reference design?
Click to expand...
I don't have an official source for this, as in some doc that says in black and white the new connector must be used like this. For example maybe that Zotac AMP Extreme Infinity is different, but that would be more of an exception to a rule we've observed with many other cards.
 
coercitiv

coercitiv

Diamond Member
Jan 24, 2014
7,281
17,192
136
The first RDNA4 card has fallen. Press F:
www.techpowerup.com

Melting 12VHPWR Connector Claims its First AMD RX 9070 XT Victim

It appears as though AMD has also fallen prey to the melting 12VHPWR connector issues that plagued the NVIDIA RTX 4000 and 5000-series cards. Although very few AMD graphics card makers used the 12VHPWR connector for the Radeon RX 7000 series, adoption rates for the new connection standard...
www.techpowerup.com www.techpowerup.com
 
  • Haha
  • Wow
Reactions: Tlh97, Ranulf, Elfear and 2 others
Hail The Brain Slug

Hail The Brain Slug

Diamond Member
Oct 10, 2005
3,830
3,210
146
coercitiv said:
The first RDNA4 card has fallen. Press F:
www.techpowerup.com

Melting 12VHPWR Connector Claims its First AMD RX 9070 XT Victim

It appears as though AMD has also fallen prey to the melting 12VHPWR connector issues that plagued the NVIDIA RTX 4000 and 5000-series cards. Although very few AMD graphics card makers used the 12VHPWR connector for the Radeon RX 7000 series, adoption rates for the new connection standard...
www.techpowerup.com www.techpowerup.com
Click to expand...
I just got the Taichi as well. I'm not using the el-cheapo included adapter, though.

I checked my input voltage reading under max load at max power in HWINFO compared to my 3x8 pin mercury, its only 0.014v lower. (0.115Vdroop on mercury, 0.129Vdroop on Taichi)

It would be interesting to get people to measure their PSU 12v rail and then read the GPU input voltage (mine reads in hwinfo) and calculate vdroop to see if we can see variance or excessive cases, which would seem to imply poor connection and might predict a melting case.
 
  • Like
Reactions: Elfear
DrMrLordX

DrMrLordX

Lifer
Apr 27, 2000
22,805
12,848
136
coercitiv said:
The first RDNA4 card has fallen. Press F:
Click to expand...
Apparently the card was fine, it was just the cable. But, it seems consistent with problems seen on some 4090s (e.g. one pin was hit harder than the others). Has anyone done a breakdown on the power delivery for this model of 9070XT?
 
Hail The Brain Slug

Hail The Brain Slug

Diamond Member
Oct 10, 2005
3,830
3,210
146
DrMrLordX said:
Apparently the card was fine, it was just the cable. But, it seems consistent with problems seen on some 4090s (e.g. one pin was hit harder than the others). Has anyone done a breakdown on the power delivery for this model of 9070XT?
Click to expand...
Same as every other 12v2x6 card, all the pins are bussed into one single 12v rail
 
  • Like
Reactions: coercitiv
Fallen Kell

Fallen Kell

Diamond Member
Oct 9, 1999
6,187
524
126
Hail The Brain Slug said:
Same as every other 12v2x6 card, all the pins are bussed into one single 12v rail
Click to expand...
And here, endeth the lesson....

This cable and connection can not be trusted as safe without proper overcurrent power protection. Ideally, we would see the protection happen at the power supply side, placing fuses or overcurrent protection circuit on each of the outgoing 12V lines to prevent more than 9 amps that the lines are certified to handle. Alternatively, the cards need to keep each of the pins separate and provide a similar protection circuit on each of the incoming lines.
 
  • Like
Reactions: igor_kavinski
D

Dave3000

Golden Member
Jan 10, 2011
1,513
113
106
I was considering getting the Seasonic Vertex 1200w ATX 3.0 power supply but now waiting for Microcenter to receive the ATX 3.1 version of this power supply and get that one instead. ATX 3.1 does not have the 600W 12VHPWR connector and uses a different kind of connector that replaces it. Also I have an RTX 4090 Founder's Edition that I purchased over 2 years ago and still the 600W connector looks fine to me, no burn marks or anything. However, I'm not using the adapter it came with and using the Seasonic 600W 12VHPWR cable that I purchased separately about a month after purchasing the card, and it's much easier to connect to the card than the adapter it came with. I'm using a Seasonic Focus GX 850W power supply. My connector on the card probably didn't burn out because I have not pushed the card hard enough yet? I'm only using a 1080p 60Hz monitor, yes I know why use a 1080p 60Hz monitor with a card as powerful as a 4090, well because I had bad luck with purchasing monitors as I ended up returning each one due to some issue and now I'm back to this monitor. Also in terms of resale value, the resale value will be low due to this issue with the 12VHPWR connector, as my card does not have the revised version of the 12VHPWR connector? So I'm best just keeping the card and transferring it to my secondary system, if the card is still fine by then, the day I buy a new more powerful card to replace it in the future instead of trying to resell it?
 
Last edited:
I

igor_kavinski

Lifer
Jul 27, 2020
26,976
18,578
146
Dave3000 said:
Also I have an RTX 4090 Founder's Edition that I purchased over 2 years ago
Click to expand...
gamerant.com

Nvidia RTX 4090 FE Shipping With Shorter Sense Pins To Prevent Melting Cables: Report

It seems like Nvidia has quietly started shipping its Founders Edition RTX 4090 cards with an updated 12-pin connector design.
gamerant.com gamerant.com

Maybe you have the better version of FE (don't ask me how to identify it. Maybe look at date of manufacture and if it's after August 2023, good chance it's the improved one?).

Dave3000 said:
I was considering getting the Seasonic Vertex 1200w ATX 3.0 power supply but now waiting for Microcenter to receive the ATX 3.1 version of this power supply and get that one instead.
Click to expand...
www.igorslab.de

Should you buy an ATX v3.1 power supply now or is ATX v3.0 enough? Here is the surprising answer! | igor´sLAB

The big question that many people are asking themselves: Is it still worth buying an ATX v3.0 power supply now? Recently, everyone has been asking me why they should still buy an ATX v3.0…
www.igorslab.de www.igorslab.de

Choose a recommended PSU from the hwbusters link given there and you should be fine.
 
Hail The Brain Slug

Hail The Brain Slug

Diamond Member
Oct 10, 2005
3,830
3,210
146
DrMrLordX said:
Damn. This standard is cancer.
Click to expand...
I think while it is one major issue with the standard, the lack of standards outlining the pin and socket design and tolerances is just as big of a problem, if not moreso since even if they did split the connector up it would not prevent issues from the connector, it would just reduce incidence of damage occurring.

Fixing the connector itself is just as important, if not moreso IMO. Its the actual cause of the problem, splitting the power rails is just a safety mechanism and not solving the root cause.

You can see it in action, certain brands of power supply or cable are using poorly designed sockets and routinely lighting cards up (looking at you, corsair, cablemod, and MSI).
 
You must log in or register to reply here.

TRENDING THREADS

COMPANY
RESOURCES
FOLLOW
Top Bottom