transient power consumption spikes

Should post this in the news section. This is seriously good info.
 
I watched this last night. Super interesting and does bring up some valid concerns. Going forward regarding the rumored power needs of the next gen GPUs especially. I’m sticking to my principles of living in the 250W range of performance for heat and noise alone.
 
No plans to go 4000 series anytime soon. My 2070 Super still does anything I need gaming wise and I still got a 3070Ti now as well. I'll just deal with these spikes. ;)
 
Just need to add a capacitor bank in another case beside the main case
images.jpeg

:LOL:
 
Would a 600w/800w GPU trip a breaker/blow a fuse regularly in the US (120v 15a/20a) if it had a 2.5x spike? Or is this prevented at the PSU level?
 
Would a 600w/800w GPU trip a breaker/blow a fuse regularly in the US (120v 15a/20a) if it had a 2.5x spike? Or is this prevented at the PSU level?
its prevent by the psu. also, you need to get over 1800w on a single circuit before it will trip a standard breaker*.
*in a modern home with standard electrical
 
Last edited:
its prevent by the psu. also, you need to get over 1800w on a single circuit before it will trip a standard breaker.
Depends on the home. Some area's in the USA have lower circuit draws then some places.

To be honest in some ways this will affect quite a few people if this does happen.
 
Depends on the home. Some area's in the USA have lower circuit draws then some places.

To be honest in some ways this will affect quite a few people if this does happen.
true, just generalizing. i added a *edit...
 
I remember some PSUs having bypass capacitors (parallel to load) on the connector side.
Was an ugly workaround then, I remember I was whining about that.
Guess it's time to embrace modding psu cables and adding big ones for spikes and tiny ones for EMI suppression.
 
  • Like
Reactions: Yiffy
like this
true, just generalizing. i added a *edit...
The thing in the video that got me was how SFX power supplies will have a harder time with those spikes.....As I was thinking of going SFF with my next build next year sometime....Well I might change that thought now.

Guess we will see what the next set of video cards brings.
 
Clickbait crap. There's nothing to be concerned about, and if you just buy good PSUs that are compatible with whatever GPU you buy, you're not going to have any problems.

I'm getting really bored of these scumbag patreon beggars.
 
Clickbait crap. There's nothing to be concerned about, and if you just buy good PSUs that are compatible with whatever GPU you buy, you're not going to have any problems.

I'm getting really bored of these scumbag patreon beggars.
post your testing that proves them wrong or gtfo with your superiority complex and trolling/"edgy" posts.
 
Clickbait crap. There's nothing to be concerned about, and if you just buy good PSUs that are compatible with whatever GPU you buy, you're not going to have any problems.

I'm getting really bored of these scumbag patreon beggars.
How do you know what PSUs are compatible with which GPUs? One of the points that was discussed in the video is that some of these video cards are seeing momentary spikes of double their rated power usage and that their rated power usage is what is being advertised in conjunction with the PSU requirements (e.g. 350W card needs 750W PSU), but if the card is drawing 700W, even for a fraction of a fraction of a second, it can cause a shutdown because there either isn't enough power to keep everything else in the system running or it is triggering the PSUs internal protections.
 
post your testing that proves them wrong or gtfo with your superiority complex and trolling/"edgy" posts.


He literally says in the video transient loads are nothing new and PSU\GPU manufacturers have been blaming each other for the issues for awhile. He's more worried about the future of video card power demands and is just trying to shine some light on an issue that's been happening for over a decade.
 
its prevent by the psu. also, you need to get over 1800w on a single circuit before it will trip a standard breaker*.
*in a modern home with standard electrical
Laser printer fuser heater inrush has been around for decades. A fast acting fuse will open in such conditions. Modern circuit breakers used in homes are designed to allow inrush currents as well as induction motor loads which can draw several times the 15 or 20A rating of the CB without tripping. A computer PSU allowing similar conditions as the aforementioned laser printer to be presented to the wall socket should not be a problem with a circuit breaker. I recall some active power correction circuits would produce a very RF rich 'chirp' upon engaging and cause AFCI (arc fault circuit interrupter) featured CBs to trip. That was annoying!

With the advent of LED and to some extent CFL lighting of yesteryear, the heavy draws presented by the laser printer, copier, et al weren't noticed due to their superior regulation and ability to maintain much more stable lumen output over a variety of input voltages. For those using incandescent lighting, however, the flickering was atrocious! Speaking of incandescent lighting the inrush on them is huge as well. A large 1500W parking lot halogen flood has a huge wolfram filament with very low cold resistance and initial inrush current at 120V can be over 100 amperes! The reason why it does NOT trip any breakers is the duration is extremely short. I'd imagine if the outdoor climate was similar to Pluto's, you may need ancillary hardware to deal with this. ;-)
 
How do you know what PSUs are compatible with which GPUs? One of the points that was discussed in the video is that some of these video cards are seeing momentary spikes of double their rated power usage and that their rated power usage is what is being advertised in conjunction with the PSU requirements (e.g. 350W card needs 750W PSU), but if the card is drawing 700W, even for a fraction of a fraction of a second, it can cause a shutdown because there either isn't enough power to keep everything else in the system running or it is triggering the PSUs internal protections.
You let other people be guinea pigs and buy hardware combinations confirmed to be working by other people on the internet. It's not hard. That's some of the best advice anyone would ever get with building pcs on the internet, yet none of these streamer/vod scumbags will ever just say it.
 
Laser printer fuser heater inrush has been around for decades. A fast acting fuse will open in such conditions. Modern circuit breakers used in homes are designed to allow inrush currents as well as induction motor loads which can draw several times the 15 or 20A rating of the CB without tripping. A computer PSU allowing similar conditions as the aforementioned laser printer to be presented to the wall socket should not be a problem with a circuit breaker. I recall some active power correction circuits would produce a very RF rich 'chirp' upon engaging and cause AFCI (arc fault circuit interrupter) featured CBs to trip. That was annoying!

With the advent of LED and to some extent CFL lighting of yesteryear, the heavy draws presented by the laser printer, copier, et al weren't noticed due to their superior regulation and ability to maintain much more stable lumen output over a variety of input voltages. For those using incandescent lighting, however, the flickering was atrocious! Speaking of incandescent lighting the inrush on them is huge as well. A large 1500W parking lot halogen flood has a huge wolfram filament with very low cold resistance and initial inrush current at 120V can be over 100 amperes! The reason why it does NOT trip any breakers is the duration is extremely short. I'd imagine if the outdoor climate was similar to Pluto's, you may need ancillary hardware to deal with this. ;-)
cool story, no idea what youre rambling about or what your point is.
 
He literally says in the video transient loads are nothing new and PSU\GPU manufacturers have been blaming each other for the issues for awhile. He's more worried about the future of video card power demands and is just trying to shine some light on an issue that's been happening for over a decade.
nothing new no, but not to the extreme we are seeing now. the 1080 went 250 to 400, the 3090 is going 250 to 600+...
 
  • Like
Reactions: heslo
like this
cool story, no idea what youre rambling about or what your point.
My point is the electrical system can take much higher current than the rated overcurrent protection without issue.
This method can also be used to find weaknesses far away from the demarc.
 
You let other people be guinea pigs and buy hardware combinations confirmed to be working by other people on the internet. It's not hard. That's some of the best advice anyone would ever get with building pcs on the internet, yet none of these streamer/vod scumbags will ever just say it.
While I can't argue the point of using the Internet to research these things (it's probably a large part of why we're all here at [H] isn't it?) I guess I'm failing to understand your disdain for "...these streamer/vod scumbags..." or why it is relevant to the conversation at hand.

Industry finger-pointing is also nothing new. We've seen it in the past when a hardware manufacturer doesn't abide by a standard and their products fail to perform properly when used with hardware that does conform to the standard (Atheros WiFi chips come to mind). This just appears to be a current example where the standards for protection circuitry are at odds with the way that video cards are ramping up their power consumption.
 
nothing new no, but not to the extreme we are seeing now. the 1080 went 250 to 400, the 3090 is going 250 to 600+...


Right nothing new. Increased power requirements = increased transient load. Might as well start freaking out that in the future we might have to have a dedicated outlet\room for PC's\consoles if they can't figure a way to increase performance while decreasing power consumption and heat. The fact that they put a PSU minimum requirement on the GPU box high enough to cover the transient load should let you know they're aware of it.
 
Right nothing new. Increased power requirements = increased transient load. Might as well start freaking out that in the future we might have to have a dedicated outlet\room for PC's\consoles if they can't figure a way to increase performance while decreasing power consumption and heat. The fact that they put a PSU minimum requirement on the GPU box high enough to cover the transient load should let you know they're aware of it.
That's just it: the minimum PSU that they're recommending is not handling the load.

That's fine but if you exceed the transistor's SOA it's gonna go boom! ;-)
No worries, with the crash of mining there were be a flood of 1.5kW+ PSUs on the market. ;-)
Should we really need 1.5kW+ PSUs for gaming desktops?
 
That's just it: the minimum PSU that they're recommending is not handling the load.


Should we really need 1.5kW+ PSUs for gaming desktops?

It's not that they aren't handling the load it's that they're triggering the shutdown protection in SOME psu's.
 
While I can't argue the point of using the Internet to research these things (it's probably a large part of why we're all here at [H] isn't it?) I guess I'm failing to understand your disdain for "...these streamer/vod scumbags..." or why it is relevant to the conversation at hand.

Industry finger-pointing is also nothing new. We've seen it in the past when a hardware manufacturer doesn't abide by a standard and their products fail to perform properly when used with hardware that does conform to the standard (Atheros WiFi chips come to mind). This just appears to be a current example where the standards for protection circuitry are at odds with the way that video cards are ramping up their power consumption.

It's just a clickbait 30 minute video where he even admits it's nothing new and is basically a non story at the end. It's scummy.
 
You don't
It's just a clickbait 30 minute video where he even admits it's nothing new and is basically a non story at the end. It's scummy.
It may not be something new but it does seem to be something that is happening more often and there is a potential it could be a bigger problem in the future. I happen to like when channels like this buy the equipment and do research into these things. I don't have the time or money to do that kind of thing myself. Why does someone putting out a video like this offend you so much?
 
He does make a good point about compact format PSUs in budget form.

However in the video it is shown several times where you can hear the fans slowing before the system shuts off. That doesn't sound like transient response to me!
Typically those are INSTANT and come with no warning like someone turning off the power. Or lightning! 🙃

I remember when I got my first AXi1500 and was playing with OCP trip points the system would shut off instantly any time Prime95 was loaded or Furmark.

Perhaps they can take the Apple approach (in regard to battery age) and just gimp the card if it detects the power supply is not adequate. I'm sure that would raise a flurry of videos like this too! ;-)
 
He does make a good point about compact format PSUs in budget form.

However in the video it is shown several times where you can hear the fans slowing before the system shuts off. That doesn't sound like transient response to me!
Typically those are INSTANT and come with no warning like someone turning off the power. Or lightning! 🙃

I remember when I got my first AXi1500 and was playing with OCP trip points the system would shut off instantly any time Prime95 was loaded or Furmark.

Perhaps they can take the Apple approach (in regard to battery age) and just gimp the card if it detects the power supply is not adequate. I'm sure that would raise a flurry of videos like this too! ;-)

If voltage is dropping mostly on the gpu's line, then that could cause the gpu driver to crash. The sudden drop in power draw could then cause voltage issues elsewhere, taking the rest of the system down. In that case, while the PSU might be the root cause, the shutdown will have been triggered by software or firmware.
 
its prevent by the psu. also, you need to get over 1800w on a single circuit before it will trip a standard breaker*.
*in a modern home with standard electrical
And that >1800watts would need to last waaaayyyyy longer than the transient spikes before tripping a breaker.
 
Right nothing new. Increased power requirements = increased transient load. Might as well start freaking out that in the future we might have to have a dedicated outlet\room for PC's\consoles if they can't figure a way to increase performance while decreasing power consumption and heat. The fact that they put a PSU minimum requirement on the GPU box high enough to cover the transient load should let you know they're aware of it.
thanks, i know, im the one always telling people to get at least what the gpu recommends and then have idiots tell me i dont know what im talking about and the online psu calc says they only need a 450....
 
I'm glad he's taken the time to research this. We've been seeing even the best of the best power supplies, at and above whats recommended have problems with shutting down with what most attributed to ocp tripping caused by spikes.
 
I thought the video was good. There was some good scientific research and explanation behind the topic and it was presented so even less technical people like myself could understand [most of] it. That said, I don't think everyone needs to go crazy about it, just evaluate and really consider if your PSU is currently enough, far beyond the simple math, and even beyond the suggested wattage for the card in use. It does seem like this is a bigger issue for the top-end cards than mid-range ones, but I could see how even those have the potential for undesired occurrences.
I might have missed if this was mentioned (watched the video on my mobile while at work), but if the card isn't running at full load, it seems like the spikes would be far less significant (not just in amount but in percentage as well) and this is a primary concern more so when running the GPU hard. The games I play keep my GPU at 50-60% load maximum, so either I've lucked out or that's why I've never had my PC cut off due to a power spike or GPU crash.

The thing in the video that got me was how SFX power supplies will have a harder time with those spikes.....As I was thinking of going SFF with my next build next year sometime....Well I might change that thought now.

Guess we will see what the next set of video cards brings.
I would not worry too much about SFX vs ATX, you just need to make sure you get a high quality one and with plenty enough overhead. You shouldn't rule out your SFF build solely because of that issue, if you're able to compensate adequately for it. An extra plus is that many SFF cases now do support ATX PSUs, at least to a certain length.

thanks, i know, im the one always telling people to get at least what the gpu recommends and then have idiots tell me i dont know what im talking about and the online psu calc says they only need a 450....
Actually I remember having a related discussion with you in an older topic and I appreciate the advice. When I upgrade, I'll probably go with the recommended PSU wattage listed +20-30% for the extra overhead.

I think that an extra 20-30% above the GPU manufacturer's recommendation should be a safe amount to where a spike should not be severe enough to cause a shutdown without becoming flat out cost-prohibitive.
 
Gonna have to start putting in something like a meanwell supply for the video card and your atx supply for the rest. As graphical power requirements keep increasing there is going to have to be a next level of design or requirements. Gaming pc in about 15-20 years will need a dedicated 240v circuit if something magical doesnt happen before then imo.
 
This is something that's missing from the ATX spec. If there was a standard curve of % overcurrent vs. min/max time to trip, then there would be no need for finger pointing, because either the GPU or PSU would be out of spec if OCP tripped in normal operation.
 
This is something that's missing from the ATX spec. If there was a standard curve of % overcurrent vs. min/max time to trip, then there would be no need for finger pointing, because either the GPU or PSU would be out of spec if OCP tripped in normal operation.
That would be interesting. I suppose the 80+ certification wouldn't evaluate it since it's not based on efficiency, but there should be some similar certification or at least a rating put out either by the manufacturer if not tested by a third party for the frequency and amplitude of these spikes. Anything to isolate either the PSU or the GPU as the cause could further help to diagnose the problem and have it either corrected or reduced, or panned by consumers.
 
Would using a UPS help this at all? I know that getting one made a frankly amazing change to the stability of my system because my house wiring isn't the best.

Or is this purely an issue within the PSU and power draw after the wall.
 
Would using a UPS help this at all? I know that getting one made a frankly amazing change to the stability of my system because my house wiring isn't the best.

Or is this purely an issue within the PSU and power draw after the wall.
in the system
 
Seems like the easy solution would be to write video BIOS to have current draw limits. I thought they already did, but clearly the limits aren't set correctly in all scenarios.
 
Back
Top