PCIe +12v Input Voltage Droop- How Much is Too Much for GPU?

[LeSandwich]

Not exactly a 'fault', but he was speaking to the PSUs ability/quality to 'load compensate'.
It's mere physics that Voltage droops more, the more current is being demanded; the PSU is supposed to have "Load Line Calibration" or a feedback-mechanism to proportionally 'up' the voltage it is supplying (in compensation for load).

ah, so I guess my PSU is just an “unlucky” model or batch, or something of the sort, and doesn’t have the best compensation

although it seems to be common right? or at least, not a big deal considering others say they dip into 11.8v pretty often. ig it depends on their own load amounts though, mine droops at 100w

higher the wattage is past 100, the more it droops obviously. at 100-110w I believe the PCIe rail is only about 11.89-11.899v so it’s not a large sag. when the wattage goes to 120-130w, that’s when it droops to 11.856-11.87 and stuff

0.6/0.7v = 12.05v
0.8/0.9v = 11.92-11.99v
1v+/100w+= 11.88-11.90v
110-120w = ~11.86-11.87v
120-130w = 11.85-11.88v

little breakdown of my voltage behaviours. the bottom 2 can be pretty variable, especially the 110-120w range but that’s the gist of it

 

[joemama]

Have you tried torture tests to see how much those voltages drop at maximum power?
11.8V seems fine depending on how many watts you are drawing since voltage drops as power increases, if it is 11.8V at not even 500W then it is probably a shitty PSU.
You can check out other PSU's reviews on TPU, only some of them drop to 11.8V but only at max loads.

 

[LeSandwich]

Have you tried torture tests to see how much those voltages drop at maximum power?
11.8V seems fine depending on how many watts you are drawing since voltage drops as power increases, if it is 11.8V at not even 500W then it is probably a shitty PSU.
You can check out other PSU's reviews on TPU, only some of them drop to 11.8V but only at max loads.

it’s a Phanteks Revolt Pro 1000w which in itself is not shitty; although as mentioned in the other reply, I might have gotten an unlucky model with poorer load compensation. no visible stability or performance issues as of now

I have a 3070 which can’t even draw 500w alone; combined with my 5800x as well, I don’t believe the total PC is drawing 500w total under gaming loads

I’ve not run torture tests or many ‘stress’ tests because I usually don’t bother (typically unrealistic load).

I’ve have run max settings Heaven Benchmark at 60 and 120fps, as well as uncapped, which pushes the card to max 230w; PCIe droops to 11.80-11.85 under those conditions, so not drastically lower than my absolute minimums in-game

 

[Lost_Troll]

ah, so I guess my PSU is just an “unlucky” model or batch, or something of the sort, and doesn’t have the best compensation

although it seems to be common right? or at least, not a big deal considering others say they dip into 11.8v pretty often. ig it depends on their own load amounts though, mine droops at 100w

higher the wattage is past 100, the more it droops obviously. at 100-110w I believe the PCIe rail is only about 11.89-11.899v so it’s not a large sag. when the wattage goes to 120-130w, that’s when it droops to 11.856-11.87 and stuff

0.6/0.7v = 12.05v
0.8/0.9v = 11.92-11.99v
1v+/100w+= 11.88-11.90v
110-120w = ~11.86-11.87v
120-130w = 11.85-11.88v

little breakdown of my voltage behaviours. the bottom 2 can be pretty variable, especially the 110-120w range but that’s the gist of it

The numbers you are running are fine since they do not fall below 11.750 volts (The ATX spec is 11.500V). The one issue comes into play, is the surge you will get when the GPU or CPU dose a burst of work and spikes (Brown Out's) the rail below the threshold that is need for them to operate if the power supply cannot recover fast enough.

If you run Hardware Info 64 and have it log your game sessions, you will see the spikes and how many microseconds it takes the supply to recover.

 

[Steevo]

I thought the same thing since I assembled another 6 pin connector for my old system but I measured it with a good multimeter and it was more stable than the software showed. The actual voltage can drift according to sensors due to ground potential difference. The sensor measures voltage and requires to the use of the same ground and not that the ground is saturated, but increases in amperage and temperature will cause the apparent resistance to drift

 

[LeSandwich]

The numbers you are running are fine since they do not fall below 11.750 volts (The ATX spec is 11.500V). The one issue comes into play, is the surge you will get when the GPU or CPU dose a burst of work and spikes (Brown Out's) the rail below the threshold that is need for them to operate if the power supply cannot recover fast enough.

are you saying this is something I’m at risk of or are you just explaining what exactly happens when there’s not enough sufficient power/recovery?


sorry if it’s obvious, I’m super tired lol. the GPU, from what I see from the wattages and voltages on HWInfo, don’t spike super harshly or anything. usually it creeps up, and if it does spike up, it tends to drop down over 2-3 seconds, if that matters

 

[taka]

Had the same voltage drops on 12v line on a older Corsair PSU with AWG 18 cables. Changed to a newer PSU with AWG 16 cables and no more drops.
There is a combination of factors like: connectors wear, cable lenght and AWG, PSU line corection, old capacitors and so on.

 

[LeSandwich]

Had the same voltage drops on 12v line on a older Corsair PSU with AWG 18 cables. Changed to a newer PSU with AWG 16 cables and no more drops.
There is a combination of factors like: connectors wear, cable lenght and AWG, PSU line corection, old capacitors and so on.

I don’t think it’s worn capacitors; there’s no peculiar noise to say that coming from the PSU for one, but also the PSU is only a few months old

but I guess it would probably be the cables, as you said, or ground drifting like the other guy suggested.

do you know what your lowest voltage was with those old cables? like the lowest you saw while gaming, if you can recall?

 

[Lost_Troll]

are you saying this is something I’m at risk of or are you just explaining what exactly happens when there’s not enough sufficient power/recovery?

Just enplaning what happens if the supply cannot react fast enough to handle the transient when it happens.

 

[taka]

I don’t think it’s worn capacitors; there’s no peculiar noise to say that coming from the PSU for one, but also the PSU is only a few months old

but I guess it would probably be the cables, as you said, or ground drifting like the other guy suggested.

do you know what your lowest voltage was with those old cables? like the lowest you saw while gaming, if you can recall?

reported by GPU-z 11.4-11.5 from PCIE and 11.6-11.7 from 8 pin, confirmed with a multimetter. that made me change the PSU after 7 years.

If you have a modular PSU and a multimeter you can check 12V at the PSU side and on the 8PIN near the GPU/CPU/MB. My worse voltage drop was at the ATX24 pin connector.

 

[LeSandwich]

reported by GPU-z 11.4-11.5 from PCIE and 11.6-11.7 from 8 pin, confirmed with a multimetter. that made me change the PSU after 7 years.

If you have a modular PSU and a multimeter you can check 12V at the PSU side and on the 8PIN near the GPU/CPU/MB. My worse voltage drop was at the ATX24 pin connector.

oh wow, that reports much worse than mine. GPU-Z never reports the PCIe Voltage at 11.8 unless at full 230w load, it always stops at a minimum of 11.9, meanwhile HWInfo does report 11.8v

I did multiple Heaven Benchmarks since my last reply; I did No V-Sync Uncapped FPS w/ Medium Settings, and again at Max Settings, and tested both on Windowed and Full-Screen

this was my absolute lowest, ironically it was on Medium Settings. Max didn't drop below 11.797; with all the numerous testings, my lowest did seem to be 11.789v and if I kept doing it, I would probably get a different minimum eventually (like maybe 11.785) or something but whatever

important note that during the benchmark, it wasn't consistently dipping below 11.8v, it was in a pretty constant 11.81-11.82v range with the occasional little droop to 11.80; usually near the end or near the middle it would drop to the lowest, which would be something in the 11.7s

 

[AsRock]

Multimeter is best but had this issue twice over the years and seems to be around 11.2v, both times recorded though HWInfo64.

For me at least you tell Seasonic that HWinfo64 is saying it's dropping that low they will ask you to send it in for a replacement if under warranty

 

[LeSandwich]

Multimeter is best but had this issue twice over the years and seems to be around 11.2v, both times recorded though HWInfo64.

For me at least you tell Seasonic that HWinfo64 is saying it's dropping that low they will ask you to send it in for a replacement if under warranty

well I've had people tell me that I'm okay; and considering it doesn't tend to drop near 11.75 at maximum 230-250w load, I'm inclined to believe it should be good.


as mentioned above, I get minimums of 11.79v under full load (not gaming load, just benchmark), but the actual consistent voltage during the entire test is 11.81-11.84v, and it may drop once or twice down to 11.79v, or in that one case, 11.789v


based on what the people have said in the thread, I don't really think that warrants contacting Seasonic/Phanteks, at least that's what I'm understanding. although in your case 11.2v would absolutely be cause for concern, surprised the PSU even still worked at 11.2?

 

[AsRock]

well I've had people tell me that I'm okay; and considering it doesn't tend to drop near 11.75 at maximum 230-250w load, I'm inclined to believe it should be good.


as mentioned above, I get minimums of 11.79v under full load (not gaming load, just benchmark), but the actual consistent voltage during the entire test is 11.81-11.84v, and it may drop once or twice down to 11.79v, or in that one case, 11.789v


based on what the people have said in the thread, I don't really think that warrants contacting Seasonic/Phanteks, at least that's what I'm understanding. although in your case 11.2v would absolutely be cause for concern, surprised the PSU even still worked at 11.2?

Not with your numbers. And as some one else said if it was dropping to much you be crashing, hard crashes like hole system reboots or even complete shutdowns.

 

[ir_cow]

11.8 is good all day long fyi

 

[taka]

From https://www.techpowerup.com/articles/160/images/ATX_2_2.pdf
page 13

My old Corsair VS 550 barely keep up with Ryzen 5600x and a RTX3070, i was drawing 420W under full load.
This was why for me at least the voltage drop was so bad. But never had a restart or freeze, only a wild coil whine system wide. I could even hear the whine in the headphones
After some research i just got the cheapest CWT build power supply aka Thermaltake GF1 750W and no more whine or woltage drop.

 

[LeSandwich]

From https://www.techpowerup.com/articles/160/images/ATX_2_2.pdf
page 13

View attachment 311095

My old Corsair VS 550 barely keep up with Ryzen 5600x and a RTX3070, i was drawing 420W under full load.
This was why for me at least the voltage drop was so bad. But never had a restart or freeze, only a wild coil whine system wide. I could even hear the whine in the headphones
After some research i just got the cheapest CWT build power supply aka Thermaltake GF1 750W and no more whine or woltage drop.

yeah I've found my 3070 is usually pushed to 250w if I go uncapped fps hog wild on benchmarking, and from what I can tell, whenever it exceeds 240w, that's when it drops to 11.79 and such. my GPU has a constant sort of buzzing sound, but it seems to be strange coil whine, since the buzzing kind of morphs into the normal high-pitched buzzing once it hits a higher wattage


although the direct PSU voltages on HWInfo could be shaky, +12v rail is either 11.884 or 11.980, my 3.3v rail is either 3.31v and 3.32v, and my 5v rail is a constant static 4.980v

 

[taka]

I have the ASUS TUF Gaming GeForce RTX 3070 OC and yes the SAP II coils are more vocal than other coils. 220-250W without changhing the power limit.