• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

Suggestions for my recent problems with my 5950X

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.

notarat

Member
Joined
Nov 11, 2010
So as you may or may not know, my 5950X system has had issues with random power-offs since it was built.

The symptoms were not tied to a specific use case, like it only does power offs when gaming, or something like that. They were VERY random. I could (and actually did) binge watch 38 straight hours of a series I like with ZERO problems...

After finishing and letting it sit at idle at the Desktop, it powered off. (that kind of maddening behavior...)

It was suggested that it may be due to the video card 8pins all running on the same rail. I had them connected to VGA 1, 3 and 5

Specs of problem system:
Enthoo Pro 2
5950X
Liquid Devil Ultimate 6900XT
32GB GSKill DDR4 4000
Dark Hero X570
(2) 1TB Gigabyte nvme Drives
(2) Crucial 2TB SSDs
18TB WD HC550 spinner
(2) 10GB Aquantia AC107 NIC Cards
Aqua Computer pump/Res (mfgr/model of Rads isn't applicable to this problem)
(10) Noctua fans (120mm & 140mm)
EVGA SuperNova 1000w G3 PSU

The documentation for the video card recommended at least a 950w PSU

Do you think the PSU I have is just being overwhelmed?

The odd part is in my troubleshooting I replaced the CPU with a 5800X and replaced the Video card with an ASRock OC Formula 6900XT and I replaced the RAM with 16GB of Mushkin DDR4 4000 (so basically the same load, power-wise, minus the extra power required for the 5950X over the 5800X) and the system has been absolutely rock-solid. This is with the same PSU that was in the problem build

Since the 5800X build has been so stable, I decided to make it into its own system. I pulled everything out of the Enthoo Pro 2 case and threw it all into a Fractal Torrent and it's working fantastic. (minus the loss of my Front Panel USB 3...ask me about that, lol)

I am assembling the 5950X into the Enthoo Pro 2 again, but with a new motherboard (EVGA Dark X570) since it has an auxiliary PCIe 6-pin power connector, but I am using the same 1000w PSU (I have 2 EVGA SuperNova 1000's) so do you think the PSU is just not up to snuff?
 
Do you think the PSU I have is just being overwhelmed?
Not a chance. I don't imagine that system to pull more than 500-600W at worst. Maybe 700W stress testing and overclocked?

since it has an auxiliary PCIe 6-pin power connector,
That's for powering the PCIe slots and used for SLI. I don't imagine your 10GbE cards need the supplemental power, but it couldn't hurt anything.

so do you think the PSU is just not up to snuff?
Wattage-wise, it's plenty. If that PSU is functioning properly or not is another story.
 
Not a chance. I don't imagine that system to pull more than 500-600W at worst. Maybe 700W stress testing and overclocked?


That's for powering the PCIe slots and used for SLI. I don't imagine your 10GbE cards need the supplemental power, but it couldn't hurt anything.


Wattage-wise, it's plenty. If that PSU is functioning properly or not is another story.

The only "oddity" in my (problematic) build is that I was using Cable Mods PCIe power Cables instead of the OEM ones that came with the PSU. I already removed those from system in case they were the cause) but I'm not done with the rebuild yet. So, I don't know if they were the root cause, a contributing factor to the problem, or had nothing to do with the problem...

The PSU definitely worked properly with the 5800X/ASRock 6900XT so I left it in the Enthoo case when I moved over the guts of it to the Torrent. Considering the ASRock Video card in the 5800X uses as much power as the Liquid Devil 6900XT I wouldn't think it would have problems with the Liquid Devil...but I don't know and won't know until I get the system built this weekend and test it out.

That's why it's been so frustrating troubleshooting...

The reason I got the Dark X570 mobo was the Aux power connection for the PCIe slots, thinking it may/would help in power delivery to the GPU (in case it was just trying to pull too much and tripping protection circuitry in the PSU)

Here's another question...what would be a good use case to test this new build with? As I said before, the problem occurred with simple stuff like sitting idle, browsing the web, using TinkerCAD and stuff like that. I never experienced an issue when streaming video from my XFinity account or listening to my music or watching movies from my NAS.
 
The reason I got the Dark X570 mobo was the Aux power connection for the PCIe slots, thinking it may/would help in power delivery to the GPU (in case it was just trying to pull too much and tripping protection circuitry in the PSU)
I don't imagine this to be the root cause. There are plenty of boards that run 2 GPUs (@ up to 75W for the slot) without supplemental power. That's there on the EVGA for extreme overclocking of GPUs. I can't imagine 2 10GbE NICs are tripping OCP.

EDIT: The 2 10GbE cards look to run at 6W each...

Here's another question...what would be a good use case to test this new build with?
Not sure... a fresh OS install is a start though.
 
The odd part is in my troubleshooting I replaced the CPU with a 5800X and replaced the Video card with an ASRock OC Formula 6900XT and I replaced the RAM with 16GB of Mushkin DDR4 4000 (so basically the same load, power-wise, minus the extra power required for the 5950X over the 5800X) and the system has been absolutely rock-solid. This is with the same PSU that was in the problem build

Since the 5800X build has been so stable, I decided to make it into its own system. I pulled everything out of the Enthoo Pro 2 case and threw it all into a Fractal Torrent and it's working fantastic. (minus the loss of my Front Panel USB 3...ask me about that, lol)

I am assembling the 5950X into the Enthoo Pro 2 again, but with a new motherboard (EVGA Dark X570) since it has an auxiliary PCIe 6-pin power connector, but I am using the same 1000w PSU (I have 2 EVGA SuperNova 1000's) so do you think the PSU is just not up to snuff?
You replaced too many parts in your troubleshooting. Replace one at a time, run the system, and see if the faults happen or not.

If you have any issue in the new 5950X build, RMA the CPU.
 
AMD, especially GFX cards I have noticed have issues with low power states. You could try modifying the power plan and see if that helps. For GFX cards I use MSI AB to disable ULPS and the random driver crashes at idle goes away.
 
Late reply, as I also happen to have a 5950x and 6900xt Red Devil Ultimate, same binned hardware as the Liquid Devil. I'm currently running them off a old 850w EVGA G2 psu and have been rock solid stable running on a slightly overclocked/tweaked 5950x @ 100% cpu load, and stock out of the box/driver settings for the Red Devil, all while leaving the PC running 24/7.

I have seen wattage from the outlet hit 1.1kW on my UPS during spikes, but mainly reside around 800 watts with the cpu loaded 100% and playing MSFS 2020 with most settings maxed. Those 1.1kW spikes, with a 80% Gold effeciency rating psu means it's putting out around 880watts on system spikes, and averaging around 640'ish watts when it's pulling 800watts from the outlet, and I've never had the system shut down or crash due to power issue. Without the 5950x running 100% (I was mining a cpu only coin when it was 100%), I see spikes into the high 700, low 800 watt range from the outlet (600 watts from psu), but it mainly resides around 650'ish watts from the outlet during gaming (520ish watts from psu). Still well within the wattage range the PSU is capable of, so no problems there either.

If you haven't figured out the problem yet, might I suggest just the usual driver updates? I'm currently running the 21.10.02 Radeon drivers, as I have had little to no graphics issues (crashes), and it allow me to keep the PC on for weeks at a time without a reboot. I'm also running the 10_3.08.17.735 chipset driver. I've tried the new Radeon drivers but had too many issues with crashing due to graphics drivers, and I haven't tried the newest chipset drivers either (why break whats working).

Btw, I'm actually looking for a 1000w PSU as I just orderd a EK waterblock for my 6900xt as I'm not happy with the temps during gaming, it hits 105c easily on the GPU hot point and throttles cause of that, and even upping the fan profile doesn't help much, other then greatly increasing the noise. So I'm going to add the GPU to my CPU loop, and with the beter cooling, I shouldn't have any issues playing around with overclocking the GPU, other then the PSU not being able to handle BOTH the cpu and gpu overclocked.

(edit) I just looked over your system setup again and we have suprisingly similar setups.
5950x
6900xt Red Devil Ultimate
Gigabyte x570 Aorus Elite WiFi
16GiBx4 sticks (64GiB) of ram (I forget the model but I think it's G.Skill)
2 Nvme drives (1TiB & 500GiB)
2 SATA SSD (500GiB & 250GiB)
2 4TB spinners
1 10Gb SFP+ NIC
Custom water loop with a D5 pump, 6 fans on 360mm rad
Additional fans in case: 2x140mm, 1x120mm, 1x 200mm, and a tiny 40mm blower for the NIC
And the 850w EVGA G2 powering all that. You really shouldn't have any issues with that 1000w PSU. (/edit)
 
Last edited:
I'm assuming you UPS wattage reading has more than just the PC on it (monitors, etc)? No idea how a 5950X and 6900XT could hit 1.1KW at the wall.. wowzas!
 
I'm assuming you UPS wattage reading has more than just the PC on it (monitors, etc)? No idea how a 5950X and 6900XT could hit 1.1KW at the wall.. wowzas!
No, mainly just the PC and some minor peripherals, like a USB hub. The TV is on a seperate surge protector for now since it alone takes over 200W (65" 4K 120Hz IPS running with full brightness for that HDR eye candy).

Keep in mind, the 1.1kW are just intermitent spikes, and since it was playing MSFS 2020 @ 4k, most settings maxed with the CPU mining in the background, the PC pretty much as a whole was running 100%. At times, I did see around 950W from the wall steady, though it seemed to be mostly in the 800-850W range, with occasional dips in the low 720's.

Either way, consumer UPS are no longer viable for my build with spikes that high as at most they can only handle 900W. Anything more and IF I get a brown out and the PC tries to pull more then the UPS can handle, it'll shut off either way. I'm just lucky this Cyperpower unit doesn't just shut off even if its not running off the battery, like my APC unit does. I've only seen enterprise/commercial UPS units that can supply more then that when running off battery, but the price jump from a 900w consumer to a 1200w enterprise UPS is just crazy. Been scouring eBay for a used enterprise unit, but the prices are still pretty damn high.
 
Just surprised to hear that. At 1.1KW (~990W actual) that psu should have shut down. No idea how you pulled thar wattage with a 250W loaded cpu and 400W GPU... even in spikes.
 
Last edited:
Just surprised to hear that. At 1.1KW (~990W actual) that psu should have shut down. No idea how you pulled thar wattage with a 250W loaded cpu and 400W GPU... even in spikes.
The EVGA 850 G2 has been tested up to 1000W from what I remember of reviews on it. Most good PSUs can do around 120% of thier rating, give or take a few percent.
And an overclocked CPU/GPU can take alot more then what you'rve stated in transient spikes. @ 100% load on my CPU (no gaming so GPU idle on desktop), I was seeing around 300-350W from the wall, with the occasional 400W readings, on what's supposed to be a 105W TDP chip. Throw in factory overclocked GPU that even AMD recommends a 950W PSU (I know, I too have seen all kinds of recommendations, but 950W was an actual recommendation from AMD themself, along with other values [I don't know if even AMD knows whats a good value]), and the sudden spikes could easily go past what the normal average usage rate is.
 
MY 750 G2 tripped around 850W running an overclocked flagship X299 CPU (pushing 300W) and that r9 295 x2 dual GPU (a 500W GPU stock) overclocked... :rofl:

Floored you're seeing that kind of wattage, even in bursts. I see what you're saying, but it still doesn't add up. I'd love to see a Hwinfo shot showing the CPU and GPU watt peaks!! :)
 
MY 750 G2 tripped around 850W running an overclocked flagship X299 CPU (pushing 300W) and that r9 295 x2 dual GPU (a 500W GPU stock) overclocked... :rofl:

Floored you're seeing that kind of wattage, even in bursts. I see what you're saying, but it still doesn't add up. I'd love to see a Hwinfo shot showing the CPU and GPU watt peaks!! :)
Currently idling on the desktop with only a android VM running in the background and of course all the background apps I have. You can see the currently recorded maxes which is during play, mainly Rust and some MSFS 2020 while I was doing some short tests. I'll dig out the miner and try to load this up like it did before. I'll up the refresh rate for the sensors too as it only updates once a second, while the display on my UPS updates at 4 times a second. I can easily see the spikes on the UPS, but not always on my Kill-a-watt which also displays only once a second.
 

Attachments

  • CPU.jpg
    CPU.jpg
    167.3 KB · Views: 7
  • gpu ups.jpg
    gpu ups.jpg
    123.5 KB · Views: 6
And here's after a quick 7zip loaded up the cpu 100% and running MSFS 2020.

I didn't even think it hit 1kW @ the UPS as I was watching it almost the whole time and didn't see higher then 900-something for a cycle.
 

Attachments

  • cpu 1kW.jpg
    cpu 1kW.jpg
    206.6 KB · Views: 11
  • gpu ups 1kW.jpg
    gpu ups 1kW.jpg
    120.6 KB · Views: 11
Interesting. The only thing that shows such a value is the UPS. Otherwise, you're well under 600W total between the cpu and gpu.

EDIT: No idea why your UPS shows such a dramatic difference compared to the primary parts driving the significant part of that W total. Did you up the refresh rate in Hwinfo? Maybe it didn't capture the GPU spike? Maybe the UPS reading is simply wonky (or I don't know what it's reading and other losses are a part of that?)?

EDIT2: Take the UPS out of the loop and just throw the Kill-A-Watt at the end and see if you come close. I know it only polls every second, but... if you don't see anywhere close to it, I'd still think the UPS is wonky. Or... clamping ammeter with higher resolution/polling? I dunno.... lol. :)
 
Last edited:
1000W from the wall is only 880W being produced by the PSU @ 88% effeciency (just re-reviewed the tech-powerup review of this model, nice it's still available after so many years). The CPU shows 214W, MB 234W, GPU is, I don't really know what to add together but just adding the Core, Memory, SoC, and 2nd Memory entries gets you 337W. That's 785W. Add in all the fans, pump, NIC, drives, and USB peripherals plugged in, and that 800W+ from the PSU (1kW from the wall) is very feasible. If you go ONLY by the CPU and one of the GPU values, then sure, it's well under 600W. Guess I'll return the Corsair RMx I just ordered and get a 650W psu instead. :beer:
 
lol, I think 850W is appropriate for the system. But I think that reading from the UPS is still wonky. The motherboard surely isn't 234W... X570 chipset is ~8W (according to Der8aur). The CPU Package includes the SOC. DDR4 RAM is only a few (think 1.2V JEDEC) to several (think 1.35V XMP) Watts. I'm sure the fans/pump/NIC/HDDs etc are a part of it, but yeah. Just not sold. I don't have to be though, lol. Thanks for humoring me. :)
 
All of our testing is done on an EVGA 750, I don't have the power draw data any longer but never had any concerns with my 5950X and an MSI 6900 XT GXT
 
Interesting. The only thing that shows such a value is the UPS. Otherwise, you're well under 600W total between the cpu and gpu.

EDIT: No idea why your UPS shows such a dramatic difference compared to the primary parts driving the significant part of that W total. Did you up the refresh rate in Hwinfo? Maybe it didn't capture the GPU spike? Maybe the UPS reading is simply wonky (or I don't know what it's reading and other losses are a part of that?)?

EDIT2: Take the UPS out of the loop and just throw the Kill-A-Watt at the end and see if you come close. I know it only polls every second, but... if you don't see anywhere close to it, I'd still think the UPS is wonky. Or... clamping ammeter with higher resolution/polling? I dunno.... lol. :)
Yes, I normally have HWInfo updating @ 1 poll a second, but for those screenshots I posted above, I upped it to 4 times a second (250ms update rate) since that's the highest my UPS updates as well.

I will admit, the Cyberpower UPS I have does not display exact Wattage values. It displays a fixed Wattage value according to a range of wattage. For example, the Kill-A-Watt may display 203W, 204W, 206W, second to second, meanwhile the entire time the Cyperpower unit will display 207W. It won't display anything different untill the Wattage goes out of that 207W range. The next higher Wattage it will display is 216W (usually for anything around 211W-220W), then 224W, then 234W, etc, etc. But it does update/poll 4 times a second, so as I mentioned earlier, I'm able to see those transient spikes because it's polling fast enough to catch them. The Kill-A-Watt on the other hand, only updates/polls onces a second, so unless a transient spike just so happens right when it polls, one would probably never see such a value. If/when those spikes happen, even on the Cyperpower unit it's only on screen for a single refresh, so 1/4 of a second.

Btw, the Kill-A-Watt is connected directly to the PC, after the UPS. The 2 displays are next to each other, and are always close in value to each other. Because of the faster polling rates on the Cyberpower (and the fixed Wattage values it displays), they can display different values, but it's normally no more then ±20'ish Watt difference till the Kill-A-Watt updates.

I'm actually going to be getting a RaspberryPi power monitor kit for this exact reason. Came aross the kit some weeks ago and am going to use it monitor the PC and home, which is when I saw what the Cyperpower UPS was showing when I was gaming, and got me to dig out my Kill-A-Watt from the box it was buried in because I didn't believe the Cyperpower UPS. Sadly, the Kill-A-Watt just doesn't update fast enough for my taste, nor can it keep a log or Max value which is what got me looking for some way to monitor the power usage via an Arduino or RaspberryPi. I too am curious about the power usage when I noticed it some weeks ago, as I too assume that the PSU should have shut off if it's really hitting that much Watts, even if it is only for a couple of ms at a time.
 
Yes, I normally have HWInfo updating @ 1 poll a second, but for those screenshots I posted above, I upped it to 4 times a second (250ms update rate) since that's the highest my UPS updates as well.
...and it still didn't capture w/e was causing the 'spike'. It reads the UPS... the UPS 'caught' it, but no devices are out of line power-wise. I wonder if you left Hwinfo running for a lot longer, that it would catch one of those spikes? Like the GPU says 500/600W.... or something? To me, you're missing a couple of hundred W somewhere to make that come true... I don't know. I believe you're seeing 600W or so (500W between CPU and GPU and 100W for the system) actual system use. But even adding inefficiencies, fans/HDDs/pump etc., just doesn't add up (and it could all be in my head) to 1.1KW.

:chair::soda:
 
Back