PC shuts down when trying to land on a planet

[Morbad]

Memory fast boot, memory context restore, etc are settings that abbreviate memory calibration and training routines, in the interest of faster boot times, at the cost of stability in edge cases. Probably not the issue here, but as a general thing, I disable these features and enable memory clear (if available) to make sure the standard training is done every POST. On AM4 the difference in boot times is usually in the ballpark of five seconds.

ReBAR is usually a performance enhancement. It generally has little to no impact on stability itself, though it can occasionally cause GPU crashes to display as CPU memory hierarchy errors due to the larger addressable space allocated to the GPU.

Anyway, entering a match/loading a map in Fortnite has many similarities to approaching a planet in EDO...much bursty GPU load and potential for high transient current draw. I'd get a new PSU ASAP and retest with the new unit. Dirty power can easily damage components attached to a defective PSU and hammering the system when you already have reason to suspect a PSU problem is risky.

You won't be able to confirm or rule out your current PSU as the cause without testing with another unit (or test equipment that costs way more than a PSU does).

 

[Suphalak]

I just downgraded to Deepcool PK650D. Unfortunately, there isn't any Corsair in stock and the Seasonic PSU costs like 120 USD in my area, but from what I've heard, Deepcool is a very good company, so I think it won't fail me. What a shame they got sanctioned in the US.

And finally, I can land on the planet without any issues now. Thank you for the support, guys, but I'm not out of the water yet. I'll test Fortnite now and see if everything is in the clear or not.

 

[Morbad]

Most brands selling PSU don't make their own units so it can be hard to tell what's quality or not on brand alone. As far as I can tell DeepCool uses at least three OEMs: CWT, Helly, and Seasonic.

The PK650D looks like it uses a cheaper CWT platform, probably the CSB-A. I could only find one review of this specific unit that provided any useful information:

Обзор блока питания DeepCool PK650D: когда «Gold» не по карману

Сравнительно доступное устройство с 80 PLUS Bronze, способное запитать топовую конфигурацию с одной видеокартой. При этом производитель также старался позаботиться и о его надежности.

ru.gecid.com

There are some other reviews by Aris Mpitziopoulos, who knows his stuff, on units using the same platform:

XPG Pylon 450W - A Great Mainstream Power Supply! - Hardware Busters

Hardware Busters - XPG Pylon 450W - A Great Mainstream Power Supply! - PSUs

hwbusters.com

Corsair CX750M (2021) Power Supply Review

The Corsair CX750M offers much for its price.

www.tomshardware.com

Since the Russian review doesn't seem to have tested hold-up time I have to infer that from these reviews. The 470uF listed on the bulk capacitor of the PK650D, which has the same ratio of capacitance to rated PSU output as the 450w XPG unit...which is borderline on hold-up time.

Other than that, and questionable 3.3v transient response, it's probably entirely fine. Nothing exceptional, but I wouldn't be afraid to plug my parts into it.

 

[GJ51]

2 things I never scrimp on

Monitors

and

PSU's

Good quality stuff can last through several build cycles...

 

[Shurimal]

2 things I never scrimp on

Monitors

and

PSU's

Good quality stuff can last through several build cycles...

Add to that the PC case. A quality one will last 10+ years.

I would also add mechanical keyboards, but these have become so cheap you really don't need to spend a lot of money. Eg Keychron has, like, literally dozens of options for 60...100 €.

 

[Morbad]

2 things I never scrimp on

Monitors

and

PSU's

Good quality stuff can last through several build cycles...

A decent PSU is a must because so much relies upon it. There are definitely points of diminishing returns though and though I do have some very high-end units, I'd probably recommend someone obtain two decent units so they have spare they can test and diagnose with than splurge on a top-end one and not have a fallback.

A good display makes a big difference and usually lasts long enough that it won't depreciate rapidly, but at least a bad display isn't (usually) going to destroy other components.

Add to that the PC case. A quality one will last 10+ years.

How long they last has never been my issue with cases. I swap them all the time because I'm constantly trying to find something less annoying to work within and around, that is also compatible with my space limitations.

Even high-end cases tend to have the same annoying trends. Crappy, overly restrictive, stamped fan grills are chief among them; just holes in the case, and the optional use of minimally restrictive wire grills, would be better. Dividing cases into a separate lower PSU chambers is also a frustrating case of aesthetics trumping functionality. So much glass that it compromises air flow and doubles the weight of the case is a close runner up. I usually end up taking a saw to cases I intent to keep for any length of time.

 

[Shurimal]

Even high-end cases tend to have the same annoying trends. Crappy, overly restrictive, stamped fan grills are chief among them; just holes in the case, and the optional use of minimally restrictive wire grills, would be better. Dividing cases into a separate lower PSU chambers is also a frustrating case of aesthetics trumping functionality. So much glass that it compromises air flow and doubles the weight of the case is a close runner up.

Luckily the highly technical reviews by the likes of GamersNexus have resulted in clear airflow improvements in recent years. Fractal Torrent and North are vey good at that, as are many others.

I personally don't mind (tinted) glass, makes seeing POST codes easy when something goes wrong As for weight, I tend to add a kg or two of sound dampening to my cases, anyway.

My pet peeve with PC cases is that most come with fans included, fans which are as a rule just e-waste for me because I replace them with Noctua or Noiseblocker. Just give me an option to buy without fans and drop the price a tenner or so.

 

[Morbad]

Luckily the highly technical reviews by the likes of GamersNexus have resulted in clear airflow improvements in recent years. Fractal Torrent and North are vey good at that, as are many others.

I personally don't mind (tinted) glass, makes seeing POST codes easy when something goes wrong As for weight, I tend to add a kg or two of sound dampening to my cases, anyway.

My pet peeve with PC cases is that most come with fans included, fans which are as a rule just e-waste for me because I replace them with Noctua or Noiseblocker. Just give me an option to buy without fans and drop the price a tenner or so.

Fractal has had some of the least restrictive stamped grills for a while, but I still prefer holes; nothing less restrictive than nothing.

I generally find sound damping material to be counter productive, especially when so many cases are so perforated. With the decline in the number of mechanical drives in most systems fans and coil whine are usually the only significant sources of noise left; it's usually better to leverage the extra ventilation with low-speed fans than to use few, higher speed, fans that are loud enough to benefit from more noise blocking material. Of course if one has an air cooled GPU that gets loud, or excessive coil whine, damping can certainly help.

My main issue has been squeezing sufficient radiator area into the smallest case practical. I've got an ASUS AP201 all cut apart and ready to go for my 9700X setup, but I might try to fit it into a Lian-Li DAN A3 mATX. In either scenario, the CPU will be air cooled, but the RTX 4090 will need at least two radiators if I want to keep coolant temps under control with a 650w power limit on the card. I can manage a 360+240 rad setup in the AP201, but only 2*240s inside the A3 before it gets too cramped. I've been using 2*240s for it in my AM4 setup and with acceptably low noise levels the water can get to 50C...still cools the card better than even gargantuan air coolers, but it softens the tubing and is surely going to reduce pump lifespan, so I don't know if I want to repeat that when I move it to the new system.

Either way, I like having everything as exhaust so the mesh side panels essentially turn into giant, low-restriction, filters and nothing has to directly ingest hot air from anything else.

 

[GJ51]

I find building a new case to be exhausting

It's a total mesh...

 

[cmdr bartford]

I have the same issue as OP and for the life of me I cannot figure out what causes it. Weirdly enough this bug isn't present when playing the game on Linux with Proton, so it's probably an AMD graphical bug in Window.

Since I started having this issue (when I bought my RX 7600XT) I've built an entire different computer around this GPU and the problem still persists but only on Windows. I have also tried playing the game on Windows with DXVK (thinking that maybe it was that that made the game run on Linux) to no avail.

 

[Morbad]

I have the same issue as OP and for the life of me I cannot figure out what causes it. Weirdly enough this bug isn't present when playing the game on Linux with Proton, so it's probably an AMD graphical bug in Window.

Since I started having this issue (when I bought my RX 7600XT) I've built an entire different computer around this GPU and the problem still persists but only on Windows. I have also tried playing the game on Windows with DXVK (thinking that maybe it was that that made the game run on Linux) to no avail.

Have you tried manually disabling double precision floats in AppConfig.xml?

Go to the main game directory (the one where the game's exe is) find this file, open it with any plaintext editor, locate the attribute "PermitNativeDoubles="true"" and change the "true" to "false", then save the file.

I don't have an RDNA3 part on hand to test with, but this was one of the workarounds for buggy terrain generation (which causes crashes when trying to land on planets), back when there were issues with the RDNA2 driver.

 

[cmdr bartford]

Have you tried manually disabling double precision floats in AppConfig.xml?

Go to the main game directory (the one where the game's exe is) find this file, open it with any plaintext editor, locate the attribute "PermitNativeDoubles="true"" and change the "true" to "false", then save the file.

I don't have an RDNA3 part on hand to test with, but this was one of the workarounds for buggy terrain generation (which causes crashes when trying to land on planets), back when there were issues with the RDNA2 driver.

Thanks for the quick reply, I just tried to do as you instructed and was met with the same thing: Computer completely shuts off and I have to start it again. Checking Windows Event Viewer I am met with an 'Error' stating that the last system shutdown was unexpected but there is nothing else.

At this point I'm quite literally being forced out of playing Elite, because on Linux no matter what distribution I'm in my HOTAS (T16000M) drifts both on the throttle and joystick and no calibration software I found so far worked very well (besides from other flight games that simply don't recognize a few axis at random on the same hotas), and on Windows my HOTAS works perfectly fine with no need for deadzones/recalibration but whenever I want to do Odyssey content I am met with a complete system shutdown.

I feel like at this point the only way I'll be able to keep playing this game is if I end up buying an Nvidia GPU.

 

[Morbad]

Thanks for the quick reply, I just tried to do as you instructed and was met with the same thing: Computer completely shuts off and I have to start it again. Checking Windows Event Viewer I am met with an 'Error' stating that the last system shutdown was unexpected but there is nothing else.

At this point I'm quite literally being forced out of playing Elite, because on Linux no matter what distribution I'm in my HOTAS (T16000M) drifts both on the throttle and joystick and no calibration software I found so far worked very well (besides from other flight games that simply don't recognize a few axis at random on the same hotas), and on Windows my HOTAS works perfectly fine with no need for deadzones/recalibration but whenever I want to do Odyssey content I am met with a complete system shutdown.

I feel like at this point the only way I'll be able to keep playing this game is if I end up buying an Nvidia GPU.

There is more going on than just a GPU/driver incompatibility if your system is shutting down, especially since it's happening across multiple APIs with and without FP64. No errors in the event log other than an unexpected shut down is almost certainly a hardware problem, quite possibly related to power.

I have several AMD GPUs (and quite a few NVIDIA ones) which don't have any issues running this game. Likewise, I know of multiple individuals on this forum who are using RDNA3/7000 series AMD GPUs in this game without any problems. An AMD specific issue that only affects you seems far fetched, especially in light of the symptoms observed.

Terrain generation when entering glide is one of the more demanding parts of this game and is precisely where load on the GPU, CPU, memory and I/O (PCI-E) could be highest. Playing the game in Linux may have enough overhead/abstraction to soften these loads somewhat. Swapping GPU might mask the issue, if the GPU has lower peak power demands, but there is something else going on that should be addressed.

I would try to duplicate this shutdown via other means. What happens when you run OCCT's "Power" stress test?

Conversely, capping GPU power in the Catalyst Control Center (or whatever AMD calls it now) by going to the tuning options and dragging the power slider down to 50-70% may also help confirm, or rule out, a power issue.

If it's not power, it could be another configuration or hardware issue somewhere, but check power first. Knowing your full system specs may also help diagnose things.

 

[cmdr bartford]

There is more going on than just a GPU/driver incompatibility if your system is shutting down, especially since it's happening across multiple APIs with and without FP64. No errors in the event log other than an unexpected shut down is almost certainly a hardware problem, quite possibly related to power.

I have several AMD GPUs (and quite a few NVIDIA ones) which don't have any issues running this game. Likewise, I know of multiple individuals on this forum who are using RDNA3/7000 series AMD GPUs in this game without any problems. An AMD specific issue that only affects you seems far fetched, especially in light of the symptoms observed.

Terrain generation when entering glide is one of the more demanding parts of this game and is precisely where load on the GPU, CPU, memory and I/O (PCI-E) could be highest. Playing the game in Linux may have enough overhead/abstraction to soften these loads somewhat. Swapping GPU might mask the issue, if the GPU has lower peak power demands, but there is something else going on that should be addressed.

I would try to duplicate this shutdown via other means. What happens when you run OCCT's "Power" stress test?

Conversely, capping GPU power in the Catalyst Control Center (or whatever AMD calls it now) by going to the tuning options and dragging the power slider down to 50-70% may also help confirm, or rule out, a power issue.

If it's not power, it could be another configuration or hardware issue somewhere, but check power first. Knowing your full system specs may also help diagnose things.

My current computer configuration is: AMD 7700X, Radeon RX 7600 XT, 32GB DDR5-6000, MoBo B650 gaming x ax v2 with a 80+ Gold 1000W PSU.

When the problems started my computer was an older i5-11400f with 32GB DDR4 and an RX 580, which got upgraded to the 7600 XT. The problems started immediately on Elite and I thought power might've been the issue since I had a 500W PSU at the time, slowly upgraded my PC parts and only kept the GPU, I really have nothing from the older computer other than the GPU.

Never ran OCCT's Power Stress test, and since I just wiped Windows from my NVMe again I'm not sure when I'll be able to test it next, but I've ran Furmark, Superposition benchmark, cyberpunk/minecraft raytracing (with terrible performance) and multiple other stress tests and I have never had any issues with anything other than Elite crashing while gliding to a planetary surface. I've also tried capping power limits before, undervolting the GPU, removing all sorts of "enhancements" and memory/cpu clocks from the BIOS and it keeps on happening.

 

[Morbad]

When the problems started my computer was an older i5-11400f with 32GB DDR4 and an RX 580, which got upgraded to the 7600 XT. The problems started immediately on Elite and I thought power might've been the issue since I had a 500W PSU at the time, slowly upgraded my PC parts and only kept the GPU, I really have nothing from the older computer other than the GPU.

To clarify, were problems present on the RX 580, or did they only begin once you upgraded to the 7600 XT?

If the GPU is the only common denominator between builds, chances are that the GPU itself is defective. Every application stresses things differently and it's not unusual for one app to reveal an underlying problem, even when other demanding apps run without issue.

If there is no common hardware between configurations that were experiencing the same spontaneous shut down issue...that's much more mysterious.

 

[cmdr bartford]

To clarify, were problems present on the RX 580, or did they only begin once you upgraded to the 7600 XT?

If the GPU is the only common denominator between builds, chances are that the GPU itself is defective. Every application stresses things differently and it's not unusual for one app to reveal an underlying problem, even when other demanding apps run without issue.

If there is no common hardware between configurations that were experiencing the same spontaneous shut down issue...that's much more mysterious.

Nope, The RX580 was fine and the shutdowns while gliding started when I got the RX 7600 XT. I'd say it happens 90% of the time I try to land on a planet, atmosphere types don't seem to affect it since if I tried hard enough to land eventually I would be able to, just in between computer reboots.

I had thought of the possibility of my GPU being bad in a way that only Elite has been able to mess with it so far, but I think I would probably find issues in other DX11 games or experience anything odd outside of what happens in Elite and I really got nothing, specially with the game running on Linux even better than it does on Windows without crashing.

 

[Morbad]

I had thought of the possibility of my GPU being bad in a way that only Elite has been able to mess with it so far, but I think I would probably find issues in other DX11 games or experience anything odd outside of what happens in Elite and I really got nothing, specially with the game running on Linux even better than it does on Windows without crashing.

In terms of how it leverages hardware Elite: Dangerous is probably the second most demanding title I've ever played, aside from an older (and now depreciated) version of Path of Exile...and I play many much newer, much better looking, generally much slower running, titles. Even if it wasn't that demanding in an absolute sense, it just needs to use some area your sample is weak in to expose a defect that wasn't caught in QA. Running the game in Linux through a translation layer to a totally different API on different drivers is a major change in how the game uses the hardware.

That there is at least one other person with a similar card that had similar issues doesn't rule out a sample specific problem with your card. If it were every, or most, Navi 33 based cards, there would be many more complaints.

I would exchange the card, if it's still under warranty. Document the exact issue, in video (so they can't just test it in something that doesn't reveal the problem and send it back to you without doing anything), then file an RMA request, explaining the issue in terms that someone with English as a third language could understand.