Aurora R12 Specs:
- CPU: i7-11700f (liquid cooled)
- GPU: AW RTX 3080
- RAM: HyperX DDR4 3200MHz 2x8GB
- PSU: 1000W
Been working through this issue for probably the better part of 2 months. I've tried a whole long list of things:
- Fresh OS install
- DDU and reinstall from Nvidia, Dell, Device Manager
- msconfig startup only Windows apps
- Uninstall iCue, AWCC
- TDR Delay changes
- Disabling hardware acceleration
- Setting power to high performance
- Disabling Gsync
- Disabling OC
- Trying XMP1, XMP2, and Disabling it
- Disabling/enabling Intel Speedstep, Speedshift
- MSI Afterburner undervolting, unchecking IO boxes
- New PSU (Tried corsair 1000w from Best Buy)
- New Motherboard (Dell RMA)
- New GPU (Dell RMA)
The mobo and GPU swaps were both from dell support. The new GPU was the "last resort" and it actually resolved the issue, but only for 3-4 days. Now I'm unable to run any game on the tower, and cannot complete any UNIGINE benchmarks either. So....what exactly happens?
- Game with high load kicks off
- GPU fans ramp up
- Screen starts flickering
- PC Freezes
- Screen goes black
- PC drops to idle state - all LEDS still on, keyboard light off, display off. Powers off with a simple press of the power button
It can take up to 20 tries to get the PC to reboot...which is why I'm wondering if it might be temp related. ANYWHO....WhoCrashed shows mainly VIDEO_TDR_ERROR and other GPU/Driver related errors. Event Viewer is mostly inconclusive but will sometimes show video issues too.
I spent some time debugging and monitoring with GPU-Z, and I noticed that PerfCap was showing Pwr/Vrel a ton when under load before the crashes.
As far as temps, CPU is chilling at low 50's, GPU temps spike up to 75-80C under max load. I'm curious if the RMA replacement might have bad thermal paste/pads and some components are overheating. I might take it apart to check in the next few days.
To help confirm it might be GPU I tried a friends 3080 and the PC was running much better, but still felt somewhat unstable at least in Fortnite (FPS drops every 2-3 seconds from 100Hz to 30Hz, some lag). With their GPU, GPU-Z was showing a ton of Vop/Vrel issues moreso than Pwr. FH5 ran fine though, no issues.
So I'm just curious what is causing this. I've tried literally every option I've found online and maybe half a dozen not listed above. I don't want to keep having Dell support send parts out but they don't seem to want to take time to debug my system before just shipping parts out. Would love some help!
Audio crackling/high latency with my old 3080 and RMA'd 3080 (since day 1), but not with friends 3080.
Below is a capture from GPU-Z right before the screen flickers and the PC crashes to black screen/idle.
@Gallifrae wrote:
I'm planning on stressing it a good amount when I get it - my last card started acting up as soon as 1 day after I got it.
That is good news.
But I swear ... how does this happen? One hypothetical but possible scenario is ...
Some user Over-Clocks their nice Nvidia card constantly and permanently damages it ... so it's only 95% good now and tends to fail under stress (likely forever ... as I think they crush-them, not take the time to actually re-solder/replace parts and fix them).
Dells swaps it out for a new one.
The old Nvidia card comes back to Dell. The technician drops it into an appropriate system and stress-tests it for a few minutes and it passes. They have a bunch of high-dollar video-cards to test so they pull-it and they deem it "good", and go-on to the next one.
You get it as your replacement . Since you assumed your "refurbished" video-card is 100% working ... and you are still having problems ... you start looking elsewhere in the system for problems.
Someone might even say "your system is blowing video-cards".
It's just crazy. Just a couple of flakey video-cards in the "refurbished pool" could cause all sorts of ramifications across many different machines and customers.
And all the time, this counts as failure trends and the costs of this Release of machine's Extended Warranty Contract skyrockets (due to all the perceaved failures).
I have always wondered how they refurbish those cards. If they don't really push them hard for an hour or so, it will just end up being a merry go around...
@Vanadiel wrote:
I have always wondered how they refurbish those cards.
Same here.
I could make more guesses (based on my experience in similar shops) but what we used to do isn't really relevant to what Dell does now ( in their shops).
@Gallifrae wrote:
1. Still RMA'd the GPU because it still seems somewhat unstable, but new one is in the spare PC case and all is well. No weird errors, DPC latency, or crashes under load.
2. picked up a MSI z590. dropped it in a spare PC case I had (4000d airflow) and did a "case swap" of sorts (parted over CPU, PSU, RAM, drives). System booted up, was able to successfully
1. Good
2. Sounds similar to the MSI PRO Z690-A WIFI-DDR5 in my latest Intel i9-12900K / RTX-3080 custom-build. After 6 months, it's been rock-solid.
The x590 should be a good board also. Certainly does not have a "limited BIOS".
Well, that is what I did with my R10. Pulled the components and put my own system together with Retail components.
Could not be happier, and I have performed a few upgrades since without any issue.
So after doing the case swap and adding a new motherboard my "R12" has been running seamlessly. I would honestly recommend anyone who has an alienware to budget in a new motherboard/case. The mobo and case design are pretty poor for regulating thermals, especially with some of the bigger card pre-builts (3070 and up). Also, you will definitely not miss the Alienware BIOS and having only 3 fan headers.
I've included photos of my old build/new build just for reference on how much extra space and airflow there are in standard cases. My case swap was to a 4000d Airflow with a z590 Tomahawk motherboard. I was able to reuse the Dell PSU with these extension cables. All other components (CPU, GPU, RAM, and even fans if you wanted) required no extra parts and were integrated easily. I'll also eventually replace the noctua with a chromax...but that's for another day.
