is your CPU on the supported list on ASUS official website?
https://www.asus.com/supportonly/A8N5X/HelpDesk_CPU/

Hi, I don't think it's the PSU itself.

Both CPU and GPU use a fuckton of juice for their day, like 200+ Watts at some points. But that's still only 17A at 12 V, a 38A PSU should be fine.
The motherboard was acutally released after that CPU so it should also take it with ease.

Vague suggestions:

If at hand, why don't you try with another GPU as a test? Something more modest in power consumption?

Double check all 12 V connectors to the mobo, maybe clean them and treat with contact cleaner.

Does it actually "trip" the PSU or is it merely a reset that lets the PSU power off? If so, try bypassing the 5Vstandby ATX connection at the connector to rule out the PSU as the problem. If it simply locks up or reboots then, it's probably not the PSU's fault.

Maybe you're SOL and it's the board. But seems unlikely.

Tetrium wrote on 2022-03-27, 19:02:

is your CPU on the supported list on ASUS official website?
https://www.asus.com/supportonly/A8N5X/HelpDesk_CPU/

Yep.

Cuttoon wrote on 2022-03-27, 19:41:

Hi, I don't think it's the PSU itself. […]
Show full quote

Hi, I don't think it's the PSU itself.

Both CPU and GPU use a fuckton of juice for their day, like 200+ Watts at some points. But that's still only 17A at 12 V, a 38A PSU should be fine.
The motherboard was acutally released after that CPU so it should also take it with ease.

Vague suggestions:

If at hand, why don't you try with another GPU as a test? Something more modest in power consumption?

Double check all 12 V connectors to the mobo, maybe clean them and treat with contact cleaner.

Does it actually "trip" the PSU or is it merely a reset that lets the PSU power off? If so, try bypassing the 5Vstandby ATX connection at the connector to rule out the PSU as the problem. If it simply locks up or reboots then, it's probably not the PSU's fault.

Maybe you're SOL and it's the board. But seems unlikely.

I could try that, but I only have some incredibly low end stuff to try with like an HD 4550. The 1900XT doesn't consume a lot of power anyway, it's rated for a TDP of 100W and it makes sense, it only has a single 6pin connector.

I'll try to clean them, that's not a bad idea.

How would I be able to differentiate between these? Shutdown seems instant. I'm also not sure how I could bypass the 5V standby, could you give me a hand on it?

All my suggestions were not meant as a solution, but merely to zero in on the problem.
So, eliminate the GPU as a power sink to rule it out as the source of the malfunction, etc.

Garrett W wrote on 2022-03-28, 20:02:

I could try that, but I only have some incredibly low end stuff to try with like an HD 4550. The 1900XT doesn't consume a lot of power anyway, it's rated for a TDP of 100W and it makes sense, it only has a single 6pin connector.

As I wrote, for the time, 100 W was quite a bit of juice. Few years earlier, fans on graphics cards were still a novelty.
Radeon 1xxx was the first generation where they bothered to include TDP in this list:
https://en.wikipedia.org/wiki/List_of_AMD_gra … rocessing_units
It's moderate by today's high end standards but it's bound to be a third or half of the system under load.
So, I don't think it is but if you replaced it, that would mean it's not about the PSUs total ability here.
Maybe use a CPU burn-in app to stress the CPU, something like prime95.

Garrett W wrote on 2022-03-28, 20:02:

How would I be able to differentiate between these? Shutdown seems instant. I'm also not sure how I could bypass the 5V standby, could you give me a hand on it?

Not sure how an ATX system in general or that board in particular is supposed to behave when resetting, etc. Should be documented somewhere.

Bypassing the 5V, ok, that was imprecise. I assumed the process was familiar, sorry.

You actually bypass the board, by connecting not the standby, but the "power on" ATX cable to ground.
https://en.wikipedia.org/wiki/ATX#Power_supply
So, the green one. (The purple 5V standby is always on or the board couldn't do anything.)
That's just what the board does to tell the PSU to turn on.
So, that manual maneuver is very common to activate the PSU to dry test it, check the voltages, without having to involve a board. A paper clip would suffice.
It's also what you do with an adapter to use an ATX PSU on an old AT motherboard:
Re: Would ATX PSU fitt in AT tower
- all it needs is a switch on that green wire.

But to test the board: Improvise a random bit of cable onto the green one in the ATX plug, connect that to any molex or other "ground" or simply to the case metal.
That should keep the power on, no matter what the board does.
To the board, it won't make any difference, just that the PSU will keep running and can't be shut off by shutting down Windows, etc. So, basically, attach everything as usual, use the external power switch on the PSU.

Sure, that is a bit hairy and experimental - there are some legitimate reasons for a board to shut itself down, like overheating of certain parts. Usually there is a heat sensor in the CPU which would trigger that a preset temperature - oftentimes to be defined in the bios setup.
But by your initial text, you can rule that out. So, if it's the board's problem at all, it should lock up or reset, but not be damaged.

First, I'd say, check and clean. If the problem remains,
2. Change out the GPU, that's less intrusive.
3. If that doesn't help, try the PSU trick.

Good luck!

500W PSU is enough for that configuration.

Might be the same issue my AXi1200 has .. bad OCP circuit, the unit is fine till it gets hit with a spike from the GPU kicking in at which point the OCP on the 12v rail trips, talking to Corsair it was a known issue with some units which is why the AXi 1200 was discontinued and replaced.

I doubt its the exact same issue but it does sound like your PSU is having its OCP tripped so you might want to investigate if it has one common 12v rail or has multiple ones, usually cheaper units use multiple rails so you could move the GPU to its own rail and everything else to the other and see if that helps.

usually cheaper units use multiple rails

Multiple 12V rails are not tied to PSU cheapness.

Thanks for the tips everyone. I'll see if I can use a less powerful GPU. The behavior is somewhat erratic in the sense that I've seen the system shut down with Prime95 in less than 10 minutes and then re-ran it for two hours straight. I'll see if I can pin-point the issue.

The Thermaltake TR2-550 was indeed a split rail and they are known to be somewhat poor PSU AFAIR, so that was my first thought. The CX500 that I own is thankfully single rail and certainly a much more capable PSU.

The Serpent Rider wrote on 2022-03-28, 21:54:

500W PSU is enough for that configuration.

What's your prognosis? Do you reckon it's the mainboard?

1) Try to overclock 4200+. If both PSUs will work fine after that, then it's not PSU or VRM related problem.
2) Try to undervolt 4800+. If shutdowns won't appear after that, then it's probably infamous problem with CPU IHS.

Garrett W wrote on 2022-03-29, 18:27:

Thanks for the tips everyone. I'll see if I can use a less powerful GPU. The behavior is somewhat erratic in the sense that I've […]
Show full quote

Thanks for the tips everyone. I'll see if I can use a less powerful GPU. The behavior is somewhat erratic in the sense that I've seen the system shut down with Prime95 in less than 10 minutes and then re-ran it for two hours straight. I'll see if I can pin-point the issue.

The Thermaltake TR2-550 was indeed a split rail and they are known to be somewhat poor PSU AFAIR, so that was my first thought. The CX500 that I own is thankfully single rail and certainly a much more capable PSU.

The Serpent Rider wrote on 2022-03-28, 21:54:

500W PSU is enough for that configuration.

What's your prognosis? Do you reckon it's the mainboard?

Well, errors tend to be erratic in their nature. 😉

I've enjoyed very basic academic training in electrical engineering so I'm very much a "Ohm's law" type of guy.

In that sense, any TDP 200 W device is supposed to work just dandy with any 200 W power source.

Alas, a "modern" switching PSU and high-end semiconductors are as far removed from the eternal wisdom of Mr. Ohm as it gets.
These items just don't behave in any linear way.

And, thermaltake doesn't have the greatest "quality" ring to my ears, as I grew up in a time before catalytic converters.

But in your case: If I'm not missing something fundamental, (by my google search) that PSU should have a single +12V lane of 38 Amps.
The TDP of you gear should hover around 19 Amps (228 Watts at 12 V).
So, 100 % overhead is a rather safe margin, afaik.

Someone please correct me if I'm missing something here. I did not keep written track of the idiosyncrasies of any platform, ever.
Some need a lot of 5V, some a strong 3.3 V lane but IINVMM, that system should be safely in the "gimme 12 V or death" range.

So, I'd say the mobo is pulling your leg or one of the plugs. But then again - a PSU is by far the cheapest to replace, so fingers crossed. 😁

The Serpent Rider wrote on 2022-03-29, 20:00:

1) Try to overclock 4200+. If both PSUs will work fine after that, then it's not PSU or VRM related problem.
2) Try to undervolt 4800+. If shutdowns won't appear after that, then it's probably infamous problem with CPU IHS.

1) This takes for granted that my 4200+ sample can clock higher, plus I think the L2 cache offset complicates things, I'd need to clock my part to 2.6GHz or so to create a facsimile.

2) This is interesting. I conducted a few tests and will elaborate further below, but what is this infamous problem you allude to? I seem to recall something like that but I can't remember any details off the top of my head.

So here's what I did, first off I installed the 4200+ back in and set a baseline test to run. Prime95 smallest FFTs for maximum stress and I also ran HWMonitor to monitor temps and CPU-Z (forgot to do so with the 4200+, you'll have to take my word that this is indeed a 4200+). In each case, I made sure to let Prime95 run for at least a couple of hours.

In any case:

Temps are pretty good with stock settings, auto Voltage and Cool n' Quiet enabled. I noticed an interesting quirk unfortunately with Cool n' Quiet, I can't seem to be able to change voltage while it is enabled, it will default to stock. I'm fairly certain that either TMPIN0 and TMPIN2 refer to some motherboard sensor or VRM and CPU package temp (or vice versa) accordingly.

Next up, I installed the 4800+ and ran the same test with stock settings and CnQ enabled. Before that ever took place, the system shutdown just as it was booting Windows XP. I went for a leak and I wasn't there when it shut down so I can't say for certain at which point exactly it was, but I left the room after the system had POSTed. This was after replacing the 4200+ with the 4800+ and making sure that the system had cooled off from prior tests. Here are the results:

Now this is a much bleaker picture. Clearly, my cooling solution and case fans are just not up to the task here, granted I can hardly imagine a scenario in which both cores would be going at it for five hours that I left this test running. Needless to say, I was expecting it to shutdown, but it didn't!

I then turned off CnQ and lowered voltage slightly to 1.35v (I believe it defaults to 1.37v). In reality, max voltage seems to be slightly higher. Here are the results:

Temps certainly are more manageable here, but still not entirely happy with the result. What I find perplexing is that HWMonitor reports that at some point core frequency was as low as 1GHz, which indicates that CnQ was engaged, but I'm certain that I had disabled it (and like I stated, voltage cannot be decreased while it is enabled).

Moving on, I dropped voltage all the way down to 1.3V, which finally made a significant difference.

Now that's more like it! These are pretty safe temps and the case was noticeably cooler to the touch as well.

I attempted to go lower, but I got some interesting results. I initially set voltage to 1.275v and the system could not POST afterwards. Frustrated and utterly bored to open up the case, I left until the next morning, where I decided to try once again and see if it would POST. And it did! But surely the system would be unstable and restart immediately, right? Well... no. It ran Prime95 for a few hours even.

I restarted to attempt to lower the voltage to 1.25V. It would not POST again! I left for a couple of hours, tried again, it POSTed! I attempted to set 1.25V and unfortunately never got it to POST again. What gives here exactly? How can a system that's demonstrably stable enough to run Prime95 for hours be unable to POST at random intervals? Interestingly, the 4200+ seemed completely stable at 1.25V and did not suffer from any POST anomalies.

In any case, I felt that I had collected enough data for the time being. Through my tests, I made a couple of conclusions:

a) Stock A64 X2 4800+ temps with my current cooling setup are worrisome.

b) Dropping the CPU to 1.3V seems to be stable (or at least as stable as stock) and decreases temps significantly (almost by 20C in the case of CPU package).

c) Despite very high temps on stock settings, system does not seem to shutdown as a result of this.

Conclusion c) is a hypothesis for the time being, but I intend to see if it holds true by running the 4800+ stock with UT2004 on a large Onslaught map. I already did so at 1.3V and the results are very good, cores are running about 10C cooler. This is with D3D Override disabled (I usually enable it to enforce V-Sync, it's been enabled every time I got a shutdown) and the system did not shutdown after 20mins of playing. I'd upload the photo of that, but forum imposes a 5 attachment-per-post limit, so I'll leave it for a further post after I've also tried with stock settings.

Thanks for coming to my TEDTalk!

You need to scalp your 4800+.

Yeah, I don't think I'm doing that. Not that I don't appreciate the help, but I'd find it a bit more helpful if I didn't get cryptic suggestions. You alluded to some issue, can you point me somewhere where I can get some more info on that? You obviously have the knowledge, why not share it?

I spent this evening playing a couple of big maps on UT2004 with stock settings and the system did not shutdown. I've had it running Crysis CPU benchmarks for about 3 hours now and it hasn't shutdown. The temps are also quite a bit more in check than they were with Prime95. More results to come.

Re: Socket 939 FX-60/Opteron 185 cooler recommendation for 110W TDP

Re: Socket 939 FX-60/Opteron 185 cooler recommendation for 110W TDP

Thanks for the links, seems like it is inevitable I'll have to pry that sucker open. I'll try running the CPU at 1.3V for a while, see if I get the same issue again.