PC shuts down when trying to land on a planet

[Suphalak]

Title, every time I try to land on the planet, the PC shuts down.

At first, I thought it was a PSU issue—I tried playing much more demanding games like Ghost of Tsushima on ultra settings and Fortnite but everything is stable.

If I stay in orbit or in space, nothing happens, but as soon as I enter glide mode, it shuts down. I'm not playing on Legacy Horizons, by the way.

CPU: Ryzen 5 5600
Motherboard: Gigabyte A320M-S2H
GPU: Radeon RX7600
RAM: 16 GB
Driver version 24.6.1
PSU is 700 watts

 

[Morbad]

Title, every time I try to land on the planet, the PC shuts down.

Any errors in Event Viewer under Custom Views / Administrative Events?

At first, I thought it was a PSU issue—I tried playing much more demanding games like Ghost of Tsushima on ultra settings and Fortnite but everything is stable.

There are very few games more demanding than Elite: Dangerous...even ones that are much newer and look much better will rarely cause the GPU to pull as much current.

An outright shut down, rather than a crash, does suggest some kind of hardware issue.

If I stay in orbit or in space, nothing happens, but as soon as I enter glide mode, it shuts down. I'm not playing on Legacy Horizons, by the way.

This is a particularly stressful transition due to terrain generation.

Motherboard: Gigabyte A320M-S2H

Probably not the issue, but this is a very low tier board with a borderline VRM, even for a relatively modest CPU like the R5 5600.

GPU: Radeon RX7600
Driver version 24.6.1

Exact brand/model of GPU? How exactly do you have it connected to the PSU?

There have been some issues introduced with the 24.6.1 drivers that may or may not be relevant to Elite (I haven't tested them myself, yet). Have you tried an older driver?

PSU is 700 watts

Exact model? It's entirely possible for a PSU with 700w on the label to not be able to deliver half that, or to have such poor transient response that it kicks off during load changes.

 

[Shurimal]

If your GPU has two power connectors, you should use two separate cables from the PSU. Daisy-chaining two connectors of the same cable will end up your PSU shutting down if the GPU tries to draw more than 150 W from that single cable. I suspect this is the most likely issue, even though 7600 shouldn't draw more than 200 W transients, 160 W continuous (of which subtract 75 W delivered via the PCIe slot directly). I have no idea why PSU makers (even Seasonic!) put two 150 W connectors on a single cable that can't supply 300 W, it's such a dumb thing to do. Other than that, 700 W PSU should be more than plenty for R5 5600 (65 W TDP, IIRC) and RX 7600. My 650 W ten years old Seasonic handles R5 3600X (95 W TDP) and RX 7800 XT just fine

Another thing to check is memory settings. XMP profiles on most older AMD mobos can be flaky, best to set the RAM frequency manually to the XMP specs, leave everything else to Auto, reboot into BIOS and double-check if the automagical settings make sense. Then start the tedious process of tightening timings and testing stability. 1usmus RAM calculator for Ryzen is a great app to see what memory settings you can probably get working stably.

Another thing to bear in mind, AGESA versions higher than 1.2.0.3b (or c) and lower than 1.2.0.C on 3x0 and 4x0 chipsets have a bug that causes memory crashes at even the mildest OC above standard JEDEC specs. Your mobo's mfg, price and feature set don't matter—it will crash under some conditions (for me, loading Unreal Engine based games). The solution (one that worked for me, at least) is to disable "Memory Fast Boot" in BIOS. Don't ask how many hours I spent troubleshooting that little issue

 

[Suphalak]

I haven't touched anything, no overclocking, no XMP enabled. But I do suspect the PSU might be the culprit here.

My PSU is Plenty SPEX700, it's locally produced in Thailand, bought it for around 69 USD. It does have an 80 PLUS certification, though, but I think I'm screwed either way. I've had it for almost six years now and it has never failed me... until now.

It's getting late on my end, I'll try disabling Memory fast boot tomorrow, and I'll get in contact with you soon.

My GPU is a Gigabyte Radeon RX7600.

I used WhoCrashed and it didn't detect anything, there aren't any errors either. It just shuts down entirely.

If I have to replace my PSU, I think I'll go with Gigabyte UD750GM (Almost a full Gigabyte setup), is this PSU alright?

 

[Morbad]

I haven't touched anything, no overclocking, no XMP enabled. But I do suspect the PSU might be the culprit here.

My PSU is Plenty SPEX700, it's locally produced in Thailand, bought it for around 69 USD. It does have an 80 PLUS certification, though, but I think I'm screwed either way. I've had it for almost six years now and it has never failed me... until now.

80PLUS only validates efficiency, not overall quality, and one of the best ways to hit efficiency targets on borderline PSU platforms is to throw out protection features and other stuff that you'd rather have.

I can't find much info on that that exact model, but from what little I do see, it's probably a stinker.

Also, PSUs don't last forever. Good ones should last quite a while, but capacitors in particular tend to degrade with use. Humidity is also bad for PSUs and boards in general, which may be an issue in south Asia.

I used WhoCrashed and it didn't detect anything, there aren't any errors either. It just shuts down entirely.

Almost certainly the PSU.

If I have to replace my PSU, I think I'll go with Gigabyte UD750GM (Almost a full Gigabyte setup), is this PSU alright?

Gigabyte does not have a good track record with PSUs, but this particular model seems to be ok.

Gigabyte UD750GM 750 W Review - No Lightning This Time

The Gigabyte UD750GM uses a better version of the platform used in the P750GM, with vastly improved protection features. Overall performance is also higher, but still needs more work to be competitive. A lower price might help since the Corsair RM750x—the king of this category—only costs $10 more.

www.techpowerup.com

Gigabyte UD750GM Power Supply Review

Decent, efficient, and with a reasonable price tag.

www.tomshardware.com

Still has mediocre transient performance and relatively sloppy load regulation. It's almost certainly much better than what you've got, but I'd want to make sure there were no better options in the price range you're looking at before buying one.

Also note that your system doesn't need anywhere near a 750w unit. If you have the option between a truly high-quality 500-600w model and a mediocre unit rated for much more, go for the former, even if the price is the same.

 

[Brammmers]

Seasonic all the way.

 

[Shurimal]

Seasonic or bust Though there are good PSU-s from Corsair, BeQuiet and others, too. Don't skimp on the PSU, buy a good one (my go-to is Seasonic Focus) at around 750 to 850 W and unless you build something silly with a 300 W TDP CPU and RX 4090 it'll last you a decade and be one less component you don't need to buy when putting together your next PC, and the PC after that

 

[Suphalak]

That's strange...

I opened up Legacy Horizons and tried to land on a planet, and it worked. Everything was smooth sailing, not a single crash or system shutdown.

However, it doesn't seem like Legacy likes AMD that much, I've read somewhere that it has driver issues which makes the game unplayable. I've encountered it before—the load times can get very long and sometimes it gave me an Orange Sidewinder.

I really want to make sure if the issue is on the PSU, I don't want to waste my money or time and ended up having the same problem again. If the issue is really on the PSU, then my system would shut down no matter the version, right?

Just a note: I don't have Odyssey DLC.

 

[the100thmonkey]

If the issue is really on the PSU, then my system would shut down no matter the version, right?

No, because Horizons (Legacy) and Odyssey (current) use different ways to render the game world. They look different because they are.

Odyssey rendering is, compared to legacy, power hungry - there's a lot more happening on your GPU.

 

[Shurimal]

Go to Adrenaline center, reduce GPU voltage to 1050 mV and power limit to -10%. Or cap framerate to 30. See if either of that helps.

EDIT: Another thing that popped into my mind. E: D uses GPU for planet and terrain generation. In addition to the above, pull the terrain generation GPU slider all the way to the left to reduce the workload as much as possible.

All that is to try and reduce your GPU power consumption. My suspicion is that your PSU can't provide the required 150 W over the GPU power cable (seeing that both Gigabyte RX 7600 models have one 8-pin connector), but maybe only 75 W (the older standard). This means that the ~160 W power consumption of RX 7600 is right on the edge of your system's capability to power it (75 W from PSU, another 75 W from the PCIe slot) and the PSU's overcurrent protection triggers when your GPU tries to pull the maximum power it needs. But it works fine in most situations when the framerate is capped at eg 60 FPS in open space, around stations or in planetary rings because GPU utilization is low then.

 

[Morbad]

However, it doesn't seem like Legacy likes AMD that much, I've read somewhere that it has driver issues which makes the game unplayable. I've encountered it before—the load times can get very long and sometimes it gave me an Orange Sidewinder.

I'm not aware of any serious issues with Legacy and AMD drivers. There was an issue with terrain generation crashing with Orange Sidewinder errors in Odyssey, but that was resolved years ago.

If the issue is really on the PSU, then my system would shut down no matter the version, right?

Not necessarily. There are significant differences between the 3.8.x and 4.0.x renderers.

Eliminating driver (I suggest the Adrenalin 24.5.1 drivers, for now) and other configuration issues would be wise.

You can run something like OCCT's power stress test to see if you can induce any issues.

Odyssey rendering is, compared to legacy, power hungry - there's a lot more happening on your GPU.

On most of my cards, I can actually get Legacy to pull slightly more power. Watts per frame is significantly lower, but frame rate is also significantly higher.

However, I have no realistic way of accurately measuring peak transient power draw. Software averages from on-board monitoring ICs don't have particularly fine temporal granularity and I don't have a quality occiliscope.

My suspicion is that your PSU can't provide the required 150 W over the GPU power cable (seeing that both Gigabyte RX 7600 models have one 8-pin connector), but maybe only 75 W (the older standard). This means that the ~160 W power consumption of RX 7600 is right on the edge of your system's capability to power it (75 W from PSU, another 75 W from the PCIe slot) and the PSU's overcurrent protection triggers when your GPU tries to pull the maximum power it needs.

I'm highly doubtful that the issue is at the cables or connectors, unless something isn't seated correctly (something to double check) or is seriously damaged. Even a single 8-pin PCI-E connector with three 18 AWG +12v wires can safely deliver way more than the 150w spec, sustained. There is a reason why 2x8-pin to 12VHPWR adapters/cables that fail usually fail at the 12VHPWR end...2x8-pin connectors rated for a total of 300w can typically carry more power than a 12VHPWR connector rated for 600w, because the former typically has ~150% margins, while the latter has ~20% margins.

The issue is very probably with transient spikes, likely at the PSU, but possibly at the card itself. Could also be software, but with almost any remotely recent driver, none of the 7000 series cards should be having any unusual issues with terrain generation and even if they did, it wouldn't be causing the system to power off.

 

[GJ51]

What temperatures are you seeing when running ED?

You could just be overheating

 

[Shurimal]

I'm highly doubtful that the issue is at the cables or connectors

Most probably not. I maybe didn't phrase it in the most unambiguous way, but I think it's the questionable PSU itself that's not capable of putting 150 W down the PCIe power cable, even if the cable itself could withstand the 12.5 A.

 

[Suphalak]

I tried undervolting my GPU, and it worked. Well, looks like the issue really is at the PSU after all. The temperature is perfectly fine.

I ran an OCCT test and nothing happened, the system is perfectly stable

I'll have to wait until next week because I still can't afford a proper good PSU, I'll check on you once that's done.

 

[Suphalak]

I tried undervolting my GPU, and it worked, I was able to land on the planet without crashing. Well, looks like the issue really is at the PSU after all. The temperature of the GPU is perfectly fine.

I still find it odd, though, that my GPU hits 100% usage in both Fortnite and Ghost of Tsushima and nothing happens.

I ran an OCCT test and nothing happened, the system is perfectly stable

Regardless, I should get myself a new PSU anyway, I don't want my old one to blow up in my face if it starts having issues like this. I'll have to wait until next week because I still can't afford a proper good PSU, I'll check on you once that's done.

(EDit: Oops, I didn't meant to post twice.)

 

[Morbad]

I still find it odd, though, that my GPU hits 100% usage in both Fortnite and Ghost of Tsushima and nothing happens.

Not all '100%' loads are the same. The only thing these utilization figures measure is the number of non-gated clock cycles in a particular category. Imagine a conveyor belt with bins on it...any bin with anything inside counts toward the number of occupied bins, no matter how big or heavy.

If you want a better idea of how demanding something is, power consumption is a better metric, but as I mentioned above even this isn't perfect if you can't poll fast enough...transients lasing only microseconds to milliseconds can cause errors or shutdowns.

I ran an OCCT test and nothing happened, the system is perfectly stable

Regardless, I should get myself a new PSU anyway, I don't want my old one to blow up in my face if it starts having issues like this. I'll have to wait until next week because I still can't afford a proper good PSU, I'll check on you once that's done.

Unfortunately, the free version of OCCT has a countdown that makes it hard to iterate the test fast enough to reliably trigger transient spikes without tuning a custom test.

There is still a chance the GPU itself is the culprit. Undervolting it in the AMD control panel not only reduces load on the PSU, but on the card's own voltage regulators. It also reduces core clocks, if you don't compensate by jacking up the clock slider, which means that the card can sometimes get more stable with an undervolt, even in the absence of power issues.

Still, it is wise to replace the PSU regardless and testing with a new unit will allow you to confirm or falsify the source of these crashes with much greater certainty. Sometimes I forget how frustrating it can be to try to pinpoint issues without parts on hand to swap out. At least you definitely know it's not software causing the problem.

 

[Suphalak]

So, update. I couldn't even get Fortnite to run, the game keeps crashing every time I get into a match. The PC doesn't shut down, though, the error message tells me it's driver timeout.

It was normal a few days ago, but again, I didn't play the game that much. I only played one or two matches then left.

Another thing to bear in mind, AGESA versions higher than 1.2.0.3b (or c) and lower than 1.2.0.C on 3x0 and 4x0 chipsets have a bug that causes memory crashes at even the mildest OC above standard JEDEC specs. Your mobo's mfg, price and feature set don't matter—it will crash under some conditions (for me, loading Unreal Engine based games). The solution (one that worked for me, at least) is to disable "Memory Fast Boot" in BIOS. Don't ask how many hours I spent troubleshooting that little issue

I'm also not sure if it's the motherboard or not, I have already disabled Memory Fast Boot and this keeps happening. Again, I'm not overclocking anything at all.

I'm really at a loss here, both Fortnite and Elite Dangerous Odyssey (I think) use Unreal Engine 5, and both of them failed to work, but both of them have different symptoms of failure. I'm on a tight budget, so I can only replace one component at a time—that means I could either replace the PSU, GPU, or the Motherboard; I can't do all of them at once. What should I do?

 

[yianniv]

The most likely culprit is the PSU... Can't any of the local shops test it and replace it on-their-premises with another one that you could buy if it solves the problem?

 

[Shurimal]

both Fortnite and Elite Dangerous Odyssey (I think) use Unreal Engine 5

Odyssey uses Fdev's Cobra engine.

RX 7600 crashing with Fortnite seems to be not uncommon. Two probable fixes could be loading default BIOS settings, then enabling resizable BAR; and loading optimized settings in Fortnite. Instead of trying to write out all the things people have tried, I'll just leave Brave search results here.

 

[Suphalak]

Unfortunately, my A320 motherboard can't use resizable BAR, but I'm trying other methods too, such as lowering the clock speed. I'm still in the process of fixing this, though. I'm running a bunch of benchmark programs to make sure my GPU is okay.