AORUS RTX 3090 Gaming box dead + other issues!
Instead of continuing to hijack the pinned thread with the unboxing of the rtx 3080, I decided to post a new one along with some pictures of my issue, hoping that others with similar or different problems/issues might chime in and/or act as a warning for future potential buyers.
As mentioned on Reddit in the eGPU sub, this is my recent experience with two brand new eGPUs, Gigabyte Aorus 3090 gaming boxes.
I received the two eGPUs last week and at first apparently all was fine, that was until I tried to play some games. One was paired to my 2020 16" fully specced MBP, the other one to the wife's 2017 15" MBP. In a game, just a rather soft game, like Dota2, the eGPU would work just fine for a few minutes and afterwards it would crash, screen would go blank, and the eGPU fans would ramp all the way to the max. The only way to recover the eGPU, would be to disconnect the thunderbolt cable and wait for it to recover and connect it back. Sometimes self recovery wouldn't even work, and a reset button push would be needed or even a power cable unplug. At first I thought that was just a one time phenomenon or a rare glitch but I soon found out that it would behave the same, every time.
To try and troubleshoot, I tried the second eGPU on the same laptop and the same thing happened. I then tried a different laptop (wife's) but the same thing happened. Then I thought to try different thunderbolt cable and in different TB ports. Same thing. This kept happening till a couple of nights ago, when one of the two eGPUs, after crashing, it completely died, like no power, nothing. I disconnected the thunderbolt cable and then the power cable. Reconnected the power cable but nothing. Tried it a couple of times even with different power cable in different power outlets. Nothing. It would simply refuse to power on. Not even the pump powers on. Hit the reset button. Nothing. Tried to cycle power it, leaving it unpowered for a while or trying to connect a low powered USB device, like an old mouse to see if it still has power in it and drain it, but no power left in it and no powered delivered to the USB ports, not even plugged in. All of this indicated a PSU issue.
At this point I’m completely disappointed!
I am left with a dead egpu and one that works only for a few minutes every time and so in between constant crashes. Mind you that the one working, takes on just find CUDA tasks, even big tranining models that I tried. Anyway, I'm think those are useless and unreliable products. This is my first Gigabyte gpu and the first egpu watercooled but this is beyond unacceptable. I would like to ask for an RMA and return both products.
Be warned for anyone getting this v1 3000 series Aorus gaming boxes!
Also I decided to take apart the dead eGPU, take some pictures and record a video of the actual PSU plug-in issue!
Some things worthy of notice:
- The big connectivity board that's runs almost all the way from the front to the back of the eGPU.
- Weird PSU form factor! Does anyone know the type of PSU used?
- Tiny PSU cooler that's loud under heavy load (on eGPUs that work anyway).
- The GPU is power connected via some sort fo SATA connectors?
- Nice copper heat pipes and cooling solution!
- Good contacts on inductors and capacitors!
- Tiny case fan that may be a noise factor! Wonder what impact that does on thermals if bypassed!?
- Tried a series of power disconnects, from the 24 pin to the 6 pin and dual 8 pin for the GPU. Ran a couple of power on and resets with any of those configs but to no avail.
- Video of the PSU connect/disconnect: https://vimeo.com/503099485 You can see the PSU fan spinning for a fraction of a second and quickly shutting down. Also in the last 2 seconds of the video you can hear the "bzzzt bzzt" sound which sounds as a short occurs. I say this because the PSU was initially connected in a wall plug that was split to the monitor and whenever I heard the PSU go bzzzt bzzt upon connect/disconnect of the cable, I would see the display powering on and off (loosing power). So maybe it does indeed short!
Sorry to hear you are having issues!
I think that's a Flex PSU, same as Razer Core.
Did you not think of returning before opening up?
2017 13" MacBook Pro Touch Bar
GTX1060 + AKiTiO Thunder3 + Win10
GTX1070 + Sonnet Breakaway Box + Win10
GTX1070 + Razer Core V1 + Win10
Vega 56 + Razer Core V1 + macOS + Win10
Vega 56 + Mantiz Venus + macOS + W10
Thanks for pointing out the PSU type. I searched eBay just to see what's what with this PSU and I was amazed most of the PSUs out there of this form factor are around 300w maximum, with some more server oriented, at 800w or so.
As for returning I still intend to do it but even before opening it up, I'm struggling with the vendor, like a lot! I bought the units from Germany and here's the vendors' reply:
Unfortunately we can not provide you with a RMA-label, because it is only valid in Germany. Furthermore we are only selling this product, therefore if you want to make use of your warranty, we kindly suggest to you to contact the manufacturer.
As you can read in our general terms and conditions for buisness clients, we can take back a product back only if it's unopened and was not used. Otherwise a certain deduction of your bill may occur.
The way I read it, the vendor says that they can't replace it under warranty, and to ask Gigabyte for help, that is if I want a replacement. If I want money back, despite the fact that they have a 30day+ money back policy, clearly visible on their website, they claim this only applies to deals with private citizens, not businesses (as it's my case). And even if that would apply to businesses too, that's still a no no as the "no questions asked" is clearly all BS, as they claim they only accept if "it's unopened and was not used." But no one can actually know if the product is good or is happy with it's performance or even if it works if you are not allowed to even open the package.
Recently, I purchased the same eGPU. The main reason was VR gaming, but when not gaming, I use it to mine Ethereum. Overclocked, it starts at around 125 MH/s, but after a few minutes drops to about 112 MH/s, and at the same time the memory temperature reaches 110 degrees Celsius and the power consumption drops below 300 W, which clearly shows that it is thermal throttling. I am thinking about replacing the thermal pads. Did you by any chance measure the thickness of the thermal pads or can approximately tell how thick they are? They can be seen on the second attached picture, but it is hard to say how thick they are. Thank you in advance.
@andreid2017, Sorry to hear of your troubles, especially with the dodgy dealer. I've just bought one of these last week for VR use. So far it has worked perfectly under heavy load (NeosVR can easily max out a i9-10900K and 3090 GPU) so fingers crossed it holds out. I have a Gigabyte 2080 Ti in a Sonnet eGPU enclosure that has been fine and before that I had a second hand Gigabyte GTX 1080 Gaming Box that was fine too.
When you had it apart, did you notice if there are VRAM chips on the back of the PCB as well as the front? Wondering if all the memory chips are cooled by the plates and water block.
I think it's not good design. There's ram on the back of the card. if gigabyte doesn't watercool them. It's not going to help especially in such a small factor.
If that doesn't work for you, you could get a block that is compatible with the Gigabyte Gaming OC since the PCB is the exact same.