Frustrating Crashes

[Bowza]

I would try a single GPU card from a hardware test

[thebs]

I would try a single GPU card from a hardware test

Seconded. Yank one and run with it for a bit. Then replace it with another with that one for a bit on its own too.

If they both pass without issues running on their own, it will confirm it the SLI config. Whether that's because of SLI, drivers related to it, inadequate power, heat reducing power, etc... root cause is going to be more complicated, and require more tests. But at least it will confirm it's when you have your system in a SLI configuration.

If one fails, then you've narrowed it down to that one card.

BTW, also be sure to swap the PCIe power connectors, and run with one set for one session, then another for another. If you have dual or even triple rail +12V, you could be limiting your power by not using the correct combination of connectors so you get all +12V rails. In fact, I've caught more than one (1) SLI configuration where someone was using a splitter/adapter and it was trying to pull way too much from one +12V rail.

Which begs the further point ...

Most GTX 970 cards require two (2) PCIe power connnectors, even though each connector should tap all +12V rails of the PSU (or at least 2). So, for dual-GTX 970, you'll need four (4) PCIe connnectors. Does your PSU have 4? Or are you using splitters? And if so, from the PCIe cables? Or are you using adapters that connect to the 4-pin Molex ones?

Ideally your PSU should come with four (4) PCIe connectors for SLI. But sometimes they don't. The GTX 970 SLI shouldn't be too bad in the case of only two (2), and splitting. But if you're pulling from Molex ... then you're likely very unbalanced unless you ensure you are using the rails correctly. Or you can just get a single rail +12V PSU and not care. That's an advantage there.

[thebs]

So we're interested in the logs just before that.
I.e., you should export the Event log and send it to one of us. Some of us are long-standing MCPs (from the original NT 3.1 days) and know NT internals in and out.

I may take you up on that.

Been a bit, and I'm used to HP or IBM tools. But you can export the Event Logs. Just be sure to use the options for viewing on another computer.

https://technet.microsoft.com/en-us/lib ... 49339.aspx

[Roger Wilco Jr]

The PSU has a single, high-power, 12V rail and 4 PCIe connectors, and I'm using all of them. The 12V is not dropping when the game is running (same for 5V & 3.3V). So I think I'm fine there. The CPU voltage was a jagged line, but I assume that was normal.

I did run some additional temperature tests while I was checking/recording the voltages. Sitting in the hangar, which seems to make the GFX cards work harder, the card temperatures would vary, with sometimes one being hotter and then the other. I don't know if that was because they were working harder, or I was just catching them with their fans ramping up/down at different stages. In the past, whenever I'd alt-tab out of the game, one card was always the hotter one.

I disabled SLI and pulled one of the cards. In a RES, the card temp is about 57-62C on ultra at 2560x1440. That will maintain 60FPS in space, sometimes dropping into the 50s in a station and into the 40s on a planet. So far there have been no more problems in limited testing. I'll run this way another couple of days and then switch cards (and slots, and connectors).

Regarding the logs, there were about 120 listed, with many having recent dates. I didn't know if there was one particular log that was required, or a range of them. At this point, I think I'll just continue with the testing. If I have another failure, then I can go directly to the event viewer and export just the latest log(s).

Anyway, thanks for all the help and suggestions.

[thebs]

The PSU has a single, high-power, 12V rail and 4 PCIe connectors, and I'm using all of them. The 12V is not dropping when the game is running (same for 5V & 3.3V). So I think I'm fine there. The CU voltage was a jagged line, but I assume that was normal.

That all sounds good, especially having four (4) PCIe connectors. As long as the voltage is not dropping on the +12V, you're likely all good.

I only mentioned the +3.3/+5V issue because I saw it once. The voltage dropped when we were watching the +12V, and I found out the guy had a half-dozen NAND SSDs.

I disabled SLI and pulled one of the cards. In a RES, the card temp is about 57-62C on ultra at 2560x1440. That will maintain 60FPS in space, sometimes dropping into the 50s in a station and into the 40s on a planet. So far there have been no more problems in limited testing. I'll run this way another couple of days and then switch cards (and slots, and connectors).

Yeah, let's see if you have any issues with either card. If not, then I'm wondering if you have another SLI connector you can try when you go back to both? Also, are they both the same brand/model?

Regarding the logs, there were about 120 listed, with many having recent dates. I didn't know if there was one particular log that was required, or a range of them. At this point, I think I'll just continue with the testing. If I have another failure, then I can go directly to the event viewer and export just the latest log(s).

If I'm on the system, I just troll through them. But when it comes to logs from another system, as I mentioned, I'm used to HP and IBM tools. I think this MS TechNet article shows how to archive and send them to someone using a file sharing solution, but don't quote me. Just make sure you check the box that is required to view them on another system.

► Show Spoiler

I've seen some Python out there that provides an "sosreport"-like solution for Windows. Been meaning to try it myself. Someone else probably has a PowerShell equivalent. Whenever anyone has a problem with Linux, I have them generate one. Most distros have adopted it (it comes from the Fedora/Red Hat world) because it's so useful. Gives me the ultimate snapshot of their system, including not just logs, but so many things ... via extensible plugins that can be disabled/enabled. I can even send people a complete tuning profile that usually improves performance, without ever touching the system myself.

Anyway, thanks for all the help and suggestions.

I haven't done anything ... at least not yet.

[Loriath]

I wouldn't rule out a software issue. Have you re-installed Direct X and done a verification on the files? There is a known issue that causes a black screen and if you have FPS enabled, you will see the FPS drop to a crawl. If you wait long enough, it tosses out to an error that it can't read a file after multiple tries. Its a DX issue, so it may be related. I am told this bug is fixed in 2.1 but we will wait and see.

[Roger Wilco Jr]

I wouldn't rule out a software issue. Have you re-installed Direct X and done a verification on the files? There is a known issue that causes a black screen and if you have FPS enabled, you will see the FPS drop to a crawl. If you wait long enough, it tosses out to an error that it can't read a file after multiple tries. Its a DX issue, so it may be related. I am told this bug is fixed in 2.1 but we will wait and see.

No, I've already started the hardware test, but I will keep the FPS counter active during the tests. I imagine 2.1 may drop before I get back to a SLI configuration, and if things are good after that then I guess it will have been the bug. I sure hope so!

[Cmdr Chemtrail]

Does the lock up happen on any other games or when using any other programs ?

[Roger Wilco Jr]

Does the lock up happen on any other games or when using any other programs ?

Any ... other ... games ... ? ... I don't understand the question.

No, I only play ED, read email, do online banking, and maybe edit an occasional video on this computer. I did play American Truck Simulator for a few nights, but got tired of having the steering wheel in the way. The problem has only happened in ED, iirc.

[Cmdr Chemtrail]

Any ... other ... games ... ? ... I don't understand the question.

No, I only play ED, read email, do online banking, and maybe edit an occasional video on this computer. I did play American Truck Simulator for a few nights, but got tired of having the steering wheel in the way. The problem has only happened in ED, iirc.

OK,that's that testing idea out of the window then.