This is why I recommend against using a 10 year old PSU for a new build.

N

Nightwing749

Limp Gawd
Joined
Nov 27, 2007
Messages
202
cliffs:
- OP patiently builds a new computer at non-scalper prices.
- OP scores a msrp RTX 3080 from Best Buy.
- OP still needed a PSU at non-inflated price, but was eager and decided to use 10yr old PSU
- OP builds system and sets powerlimit to 75% on rtx 3080, measuring 230 watts at load.
- OP runs benchmark, hears pop, computer shuts down and smells smoke
- OP fries RTX 3080, one memory module and possibly motherboard
- OP feelsbadman :(

Specs:
H510 elite, R5 3600, Corsair H80i v2, MSI b550 Gaming Plus, 2x8gb G.Skill TridentZ 3600C16,
ASUS ROG Strix GTX 3080, 512gb Samsung 850 PRO SSD, 650w XFX XXX Edition, APC surge protector

Storytime:
Back in March for my birthday, my wife purchased me a new computer case I've been eyeballing for awhile. She gave me the thumbs up for a new build. While patiently waiting for a 3xxx RTX, I started picking up other computer components. I got lucky two weeks ago and scored a RTX 3080 from a Best Buy drop. However, I wasn't aware it had a 750w PSU recommendation until I picked it up. I thought no big deal, until I saw the current inflated prices in PSU market.

After reading a few other threads, I thought I'll be okay with my 650w PSU until I can find a good deal for a new one. (WRONG!!!)

I built the system, crossed my fingers and powered it on. It booted into BIOS with no hiccups, then installed a fresh copy of Windows 10 PRO. Everything looked good so far. After installing various drivers, updates and software. I set the GPU powerlimit to 75% and measured ~230 watts at load with GPU-Z. With my R5 3600 at stock settings and my GPU limited, I thought I was in the clear. So I ran my first benchmark to test stability and then I heard the *POP!* Computer shuts down and I could smell smoke. I immediately unplugged everything and tried booting again but I got nothing. (R.I.P.)

I was stuck at a black screen, a flashing red light on the middle PCI connector of the 3080 and a solid DRAM led light on the mobo.

Troubleshooting:
-DRAM LED error
First, I tried removing the second ram module from slot 4. And...... The computer now boots past DRAM led error (yay!) but now hangs at VGA led error (boo!). I tried interchanging the two modules, but the mobo won't boot past DRAM error in any other configuration with the second module. I tried all 4 RAM slots, each module by itself and resetting CMOS. The computer will only boot pass the error with that one good module.

-VGA LED error & GPU flashing red light
After a quick google search, I learned the flashing red light meant there's a voltage irregularities. I tried different cables, unplugging the computer, draining the electricity and reseating the GPU. Still nothing. I finally gave up and figured it was the PSU. Fast forward one week later (last night). I received my new Seasonic PX-850 Platinum. Plugged in everything and I'm still at square one. The same blinking light of death. Tried all the PCI connections on the PSU, different cables and a single cable per connector on the GPU. The computer won't boot pass the VGA error. However, if I plug in my GTX 1080ti the computer boots into Windows without issue. So now I'm waiting for Best Buy to open up to see if it's even possible to exchange it or go through RMA.

So long story shorty......

Granted I was not using a top of the line or premium PSU, but it was a rebranded Seasonic.
https://www.tomshardware.com/reviews/power-supply-oem-manufacturer,2913-4.html

I do not recommend using a +10 year old PSU on a new build.
 
Last edited:
Also, I can't get into BIOS on the motherboard anymore. I haven't finish troubleshooting that yet. This just happened last night and I'm now at work (with 5 hours of sleep). I tried resetting the CMOS and flashing the BIOS via USB. The computer boots into Windows without issue with the new psu, one stick of ram and my gtx 1080i. However, when I attempt to boot into bios with the delete key, it hangs at the "Boot" led light on the mobo. I think I'll try removing the os hdd later.
 
Last edited:
Hmm, I wouldnt have expected such a bad failure from a decent PSU, concerning.


To make sure the CMOS is being reset, unplug the PC from the mains and remove the CMOS battery.
If there is a CMOS reset jumper/switch, set that on.
Leave it overnight.


ps
Its wise to quickly test if an old PSU can maintain its voltage, using a multimeter.
When put under a medium load, if 12V drops by near 1% or more, thats a warning sign of what will happen under full load.
Check both rails if the PSU is not a single rail.
Note: Full load will be less Watts than when the PSU was new, it will be easier to max an old PSU out.

My 3090 system is used with a 10 year old Corsair AX750, derived from a Seasonic X-750.
The following is playing Forza Horizon 4:
Idle 12V measures 12.11V.
Playing the game, voltage decreased to 12.09V at full framerate, vsync off, GPU utilisation 97/98%, max power limited.

Thats a drop of 0.16% under a high load.
The PSU is easily able to handle this load.

The following power use * reported by MSI Afterburner:
* GPU around 390W.
* CPU (10700K @ 5.1GHz) around 35% utilised, around 75W.
I'll add 5% to those for motherboard sensor measurement error = 410W and 79W.
Motherboard and drives will be around 60W
Total power draw is around 550W, 73% of the PSUs 750W rating.
 
I use an old 7-8 year old 650-watt NeoECO Antec. Amazingly, it runs fine, even driving a 280mm cooler and three 140mm fans (RGB) and a power-hungry 10850K chip with GTX 1070 video. I agree though that the PSU is something you do NOT want to go cheap cunt on. I remember using a PSU that an OEM builder in some forum highly recommended because it was only $10 but had strong voltage on the 12 volt rails. So I bought it and after exactly 6 months there was a puff of black smoke that came from my Windows box, and Poof! it (only the PSU though and not the motherboard) was dead. Thinking it was a fluke I bought another of the exact same model/brand. 3 months later there was 2 puffs of black smoke and poof again; it's dead, jim. Only this time it left scorch marks on my abit board but by some miracle it still booted. To quote the character played by Sophie Turner from Game of Thrones: "It's true, I'm a slow learner... but at least I learn!" So I then bought my very first decent PSU - an Antec 350-watt and had it for many years until the NeoECO.
 
Was the problem that your PSU was too old, or was the problem that your PSU was inadequate for your new build?

If it was simply an age issue, it would not have conveniently waited until you changed your hardware in order to fail.

The 1000w PSUs in my main and backup computer are ~14 years old now. Both are still going strong.
 
GotNoRice said:
Was the problem that your PSU was too old, or was the problem that your PSU was inadequate for your new build?

If it was simply an age issue, it would not have conveniently waited until you changed your hardware in order to fail.

The 1000w PSUs in my main and backup computer are ~14 years old now. Both are still going strong.
Click to expand...
Good question. I'm no electronic engineer, however, I've read somewhere that PSU's can become less efficient over time. It was manufactured by a Seasonic, however, it was a bronze version. So it's possible I started with less wiggle room for wear and tear. (I.E. countless hours of FFXI and WoW, lol.)
 
Last edited:
RazorWind said:
If you can't get that 3080 repaired or replaced under warranty, I'd be willing to take a look at it.
Click to expand...
Thanks, I'll keep that in mind. I just got off the phone with BestBuy and they said I need to go to the store. I'm just worried they'll only offer a refund. Plan B is RMA.
 
Last edited:
Dan_D said:
In fairness, PSU's don't normally take out hardware when they fail. That said, it's hardly unheard of.
Click to expand...

Agreed, i doubt it was the PSU that caused the issue. I've run old PSUs with new builds all the time, but I used EVGA, or Corsair PSUs.
 
Haven't heard of PSUs taking out components, but i guess it could happen. Would have thought the PC would just shutdown when it's unable to draw enough power to power a card. Are you sure you didn't forget to plug in some of the 6+2 PCIE connectors? I think i've heard of people smoking/damaging GPUs by forgetting to plug those in.
 
dthree said:
Haven't heard of PSUs taking out components, but i guess it could happen. Would have thought the PC would just shutdown when it's unable to draw enough power to power a card. Are you sure you didn't forget to plug in some of the 6+2 PCIE connectors? I think i've heard of people smoking/damaging GPUs by forgetting to plug those in.
Click to expand...
I'm pretty sure. The card requires three 6+2 PCIE rails and emitts a solid red led if a cable is not connected.
 
NIZMOZ said:
Agreed, i doubt it was the PSU that caused the issue. I've run old PSUs with new builds all the time, but I used EVGA, or Corsair PSUs.
Click to expand...
This is a valid point. It is possible that it actually may have been the graphics card that was defective, precipitating the other failures.

Power FET fails on graphics card -> Draws a zillion amps through power supply -> aged FET(s) or diode in power supply fails -> High-ish voltage goes where it doesn't belong.
 
Times change. The "pull" on any given line is going to be more than it used to be. So, old PSU's could be great, even component wise, but they were designed for a different time.
 
cjcox said:
Times change. The "pull" on any given line is going to be more than it used to be. So, old PSU's could be great, even component wise, but they were designed for a different time.
Click to expand...

You really need to elaborate on what you are talking about here, because frankly your comment makes no sense to me.

cjcox said:
The "pull" on any given line is going to be more than it used to be.
Click to expand...

Says who? Although the power consumption of certain components has increased, hardware configurations vary greatly, and there have been lots of power-hungry setups in the past. I used my current 14-year-old 1000W PSUs to power both Triple-SLI and Quad-Crossfire setups in the past - both setups consumed more power on the 12V rails than a 3090 would. When I upgraded from my previous 6-core Haswell-E (i7-5820K) to my current Ryzen 3900X, power consumption from my CPU actually went down.

My point being that it's a capacity issue. Even if your new system does pull more power than your old system did, that does nothing as far as changing the relevant criteria when determining if your existing PSU will be adequate or not. PSUs have listed amp output for each voltage and each rail (if it's a multi-rail PSU). Either it's enough, or it's not. You don't want to push your PSU to the edge of what it can handle, regardless of if it's a new or old PSU. If it's a really old PSU, it never hurts to allow yourself a bit of extra headroom. If the power supply you are trying to use doesn't supply enough power for the job, that's not a hardware-age issue, nor is it because "times change". It's because of builder incompetence.

cjcox said:
So, old PSU's could be great, even component wise, but they were designed for a different time.
Click to expand...
The last major change in PSU tech (in terms of compatibility) was almost 18 years ago. That is when PSUs shifted the bulk of their output from the 5V rail to the 12V rail, when the ATX12V/EPS12V CPU power connector was introduced, and when 6-pin/8-pin PCI-E power cables were introduced. For any PSU made after that time, there is nothing that would prevent it from working in a new system so long as it supplies the amps needed on the relevant voltage rails.

A 650W PSU is a pretty tiny PSU for a high-end system - and that was true even 10 years ago. If the OP had bought an 850w PSU 10 years ago instead of a 650w, would he have had more success moving it over to his current computer? Who knows, but my bet would have been Yes.
 
GotNoRice said:
You really need to elaborate on what you are talking about here, because frankly your comment makes no sense to me.
Click to expand...
Likely. What I said would date back a couple more years, 12-13 or so. But just because a spec exists, doesn't mean what was available was up to spec (it could still be ATX12V 2.2 or earlier for example, even in 2008 even 2009). But again, likely more than 10 years, so you're correct. But incorrect in assuming we had ATX12V 2.3 always, we did not.
 
The problem here was a PSU that couldnt provide enough throughput. Not sure why its being blamed on how old it is.

Im using a power supply from 2009 to power my 7980xe and RTX 3090.
 
German Muscle said:
The problem here was a PSU that couldnt provide enough throughput. Not sure why its being blamed on how old it is.

Im using a power supply from 2009 to power my 7980xe and RTX 3090.
Click to expand...
It could be a combination of cap aging and being a lower end bronze unit. Typically those were only warrantied for 5 years.
 
German Muscle said:
The problem here was a PSU that couldnt provide enough throughput. Not sure why its being blamed on how old it is.

Im using a power supply from 2009 to power my 7980xe and RTX 3090.
Click to expand...
Power Supplies have thermal load protection to prevent any device from pulling too much. When it does it shuts off. It doesn't fry components.
 
I had an old psu that occasionally would require me to unplug it from the wall to get it to power on. I figured I would hold out as long as I could on buying a replacement. The worse that could happen would be that one day it won't power on, right? Wrong! It popped and took a few hard drives with it. Lesson learned.
 
That's a bottom-tier Seasonic that people buy to power their GTX 560 10 years ago, not a RTX 3080 in 2021. You probably would've been okay powering a RTX 3060 Ti but oof.
 
Zeoclang said:
I had an old psu that occasionally would require me to unplug it from the wall to get it to power on. I figured I would hold out as long as I could on buying a replacement. The worse that could happen would be that one day it won't power on, right? Wrong! It popped and took a few hard drives with it. Lesson learned.
Click to expand...

My Dads PSU did something similarly strange last week.
I had an emergency call from him just over a week ago because his PCs main display wouldnt show anything. He badly needed it working.
The second screen showed the mouse pointer ok.
I helped him over the phone test the TV and connections which were ok but coudlnt solve why there was no display after the PC booted, it was too remote to attend quickly.
Then my brother remembered a similar event when he had that PSU, it stopped functioning similarly once for him also and only worked properly again after it had been unplugged for some time.
He got our Dad to try that and it worked!
Odd.

He's getting a new PC soon.
 
I don't think the PSU's age was the problem here. The 3080 has a TDP of 320W. Even if you try turning it down, it's still going to suck a lot of power... nVidia actually lowballs their PSU requirements, and in my experience you want an extra 100W on top of what they're saying, so I read that as 850W PSU for a 3080. You were 100W UNDER the recommendation, which means you were actually 200W below what you needed to run it safely. There is a possibility you would have overloaded any 650W PSU trying to run a 3080 on it. There aren't a lot of graphics cards that suck that much power.

Another problem is that that's a really bad pairing between the GPU and the motherboard. A 320W GPU probably shouldn't be tossed into a B550 motherboard with merely average VRMs. IMO, you need a motherboard with beefy VRMs when you start dealing with 300W components like a 3080. I actually have a motherboard that has an extra PCI-E connector on the board itself to help make a GPU more stable. It's possible that not only your old PSU caused problems, but also your motherboard not being able to handle the 3080. I actually found 10 years ago that no matter how much I upgraded my power supply... the cheap motherboard that came in one of my prebuilt machines simply couldn't handle the power draw of both a Q6600 and a 560 Ti. It had a very weird design that would have let it run two 8800 GT cards in SLI, but not a single 560 Ti.

My personal opinion? I try really hard to stick to GPUs that draw about 170W. I don't want a card that runs much hotter than that in my system, because it will require a beefy system, insane cooling, and probably die sooner anyway. Unfortunately this generation seems to be a bad watt-sucking one, so you would have to go down to the 3060 to get a reasonable TDP. The 2070 and the 1080 were a lot more reasonable as far as that goes.
 
A lot of older PSUs list peak load instead of max load, too. Or maybe it's the other way...anyway, you need to know both, because they're both important when it comes to peaky devices like GPUs.
 
athenian200 said:
Another problem is that that's a really bad pairing between the GPU and the motherboard. A 320W GPU probably shouldn't be tossed into a B550 motherboard with merely average VRMs. IMO, you need a motherboard with beefy VRMs when you start dealing with 300W components like a 3080. I actually have a motherboard that has an extra PCI-E connector on the board itself to help make a GPU more stable. It's possible that not only your old PSU caused problems, but also your motherboard not being able to handle the 3080.
Click to expand...

A PCI-E slot is only rated to provide 75-watts, and any additional power that your GPU needs would come directly from the 6-pin/8-pin PCI-E power cables. A card should not be pulling more than 75-watts from the PCI-E slot, regardless of how much power the card actually uses. The VRMs on the motherboard should not be a factor when the GPU is getting the majority of it's power directly from the PSU.

The motherboards with a PCI-E power connector on them are to help when you are running multiple GPUs. 75-watts is still the maximum consumption for a PCI-E slot, but if you are running 3 videocards for example, then that is potentially 225w (3x75w). For a single videocard, that's not going to come into play. You don't need a board with an auxiliary PCI-E power connector to run a single videocard, not even a 3090, because it's never going to pull more than 75w from a single slot.
 
GotNoRice said:
A PCI-E slot is only rated to provide 75-watts, and any additional power that your GPU needs would come directly from the 6-pin/8-pin PCI-E power cables. A card should not be pulling more than 75-watts from the PCI-E slot, regardless of how much power the card actually uses. The VRMs on the motherboard should not be a factor when the GPU is getting the majority of it's power directly from the PSU.

The motherboards with a PCI-E power connector on them are to help when you are running multiple GPUs. 75-watts is still the maximum consumption for a PCI-E slot, but if you are running 3 videocards for example, then that is potentially 225w (3x75w). For a single videocard, that's not going to come into play. You don't need a board with an auxiliary PCI-E power connector to run a single videocard, not even a 3090, because it's never going to pull more than 75w from a single slot.
Click to expand...
Yeah, I know that's how it's supposed to work, but the point is... I have seen low-end motherboards struggle with high-end graphics cards before, and seen power supplies that can run a card in one motherboard not able to keep it powered on another. Maybe there's some other reason and it's not the power delivery or the VRM, perhaps I was wrong on why it doesn't work. It could be some other aspect of the design, like lane speed or how many layers the PCB has, I really don't know... but I still am convinced it's not a good idea regardless of the fact that it should theoretically work.
 
A lower class PSU will have lower quality caps and other components so they will wear faster than the premium units. Not a good idea to put an old lower tier PSU, that is below watt requirement, with a 370w or so GPU (Strix cards have a much higher power limit than most other cards on default). That 650w PSU 12v rating probably is at or below the amount of 12v you need for your build without powerlimiting anything. The powerlimiter will keep long term draw in check, but unsure if it will keep a cap on very short peaks.
 
About 10 years ago I bought a used case from eBay. To my surprise, there was a 750w PSU installed in it. Don't remember the brand, it's in my daily driver which I'm typing this on. The case and PSU have been through 3 major upgrades over the last 10 years without the slightest problem. Hopefully it won't self destruct now that I've said this. :p
 
I've been running this for 13 years, BFG 800 watt PSU:

https://www.newegg.com/bfg-tech-es-series-es-800-800w-continuous-40-c/p/N82E16817702007

I really think about it dying one of these days, I always have a spare PSU somewhere to get me running again, but surely it can not keep running much longer? Granted I paid quite a premium for it back in the day and it has never let me down. I leave my PC on 24/7 and it has been on an APC unit it's entire life.

If I could find a video card in stock I would probably replace it.
 
Kurik said:
I've been running this for 13 years, BFG 800 watt PSU:

https://www.newegg.com/bfg-tech-es-series-es-800-800w-continuous-40-c/p/N82E16817702007

I really think about it dying one of these days, I always have a spare PSU somewhere to get me running again, but surely it can not keep running much longer? Granted I paid quite a premium for it back in the day and it has never let me down. I leave my PC on 24/7 and it has been on an APC unit it's entire life.

If I could find a video card in stock I would probably replace it.
Click to expand...

Hope you change the batteries in the APC on a regular basis. They have been known to kill PSU's.
 
I've been using the same pc power and cooling PSU for way over a decade... think i first got it for my dual opteron rig with an AGP geforce back in the day
 
Kurik said:
I've been running this for 13 years, BFG 800 watt PSU:

https://www.newegg.com/bfg-tech-es-series-es-800-800w-continuous-40-c/p/N82E16817702007

I really think about it dying one of these days, I always have a spare PSU somewhere to get me running again, but surely it can not keep running much longer? Granted I paid quite a premium for it back in the day and it has never let me down. I leave my PC on 24/7 and it has been on an APC unit it's entire life.

If I could find a video card in stock I would probably replace it.
Click to expand...

It'll need a capacitor kit eventually. Most of the PSUs I have from that vintage start to behave strangely from the caps wearing out.
 
Kurik said:
I've been running this for 13 years, BFG 800 watt PSU:

https://www.newegg.com/bfg-tech-es-series-es-800-800w-continuous-40-c/p/N82E16817702007

I really think about it dying one of these days, I always have a spare PSU somewhere to get me running again, but surely it can not keep running much longer?
Click to expand...

There are countless examples of electronics lasting 20, 30, 40+ years. There is nothing special about a PSU in general that guarantees a shorter lifespan.

PSUs die for specific reasons, not because of age. Things like inadequate (or failed) cooling, bad capacitors, and improper maintenance such as allowing a PSU to get caked with dust and pet hair, etc.

If your PSU had inadequate cooling or bad capacitors it probably would have experienced a failure already. At the very least, by now, there would be a lot of info out there from others who have experienced failures that would be easy enough to find. It's easy to check yourself to make sure the fan is still working properly.

Only you know the maintenance history of your unit. Dust, pet hair, etc clogging up a unit will cause it to run much hotter and can cause premature failure.

So if it was a good unit to begin with, and you know that you have taken good care of it over the years, then take solace in the fact that it has now stood the test of time. If it was a bad design, it probably would have died already. It doesn't seem to me like there should be much to worry about.
 
You must log in or register to reply here.
Back
Top