6x GTX 1080Ti AORUS XTREME EDITION - problem with rig!

Mining
1

Hello,
Once again, I would like advice that will help me to successfully launch a 6x 1080Ti AORUS excavator. Thanks for the last tips you sent - I already have a more powerful power adapter installed. The problem now is that not after firing my two new power supplies - digging is impossible … Most often ends with bluescreen immediately after a series of errors “Thread does not exist” (example from the zcash excavator).
I present detailed specifications and connection. Please give me any tips on what is wrong and where to improve it :slight_smile:

  1. Operating System: Windows 10 Pro ver. The latest (installed through the latest updates [10.0?]
  2. Modular power supply EVGA SUPRNOVA 1300w G2
    THERMALTAKE 1475W G2 semi-modular power supply
  3. AsRock H81 PRO motherboard BTC r2.0 with CPU G1840 Celeron and 4GB RAM

Current connection method:
For powering the motherboard (widest pin), a super PSU connection kit is connected which looks exactly like this:
30cm-ATX-Power-Supply-Adaptor-Cable-font-b-Connector-b-font-for-Mining-24Pin-20-4pin.jpg800×800 49.2 KB

Power supply 1300W connected by plug with wide pin and more wires. An additional plug with two wires is connected to a stronger power supply.
From that 1300 I connected the power supply: CPU, SSD, all risers (3 SATA ports in the power supply - two risers on the cable), two 1080ti cards with 8PIN-> 8PIN cable (that is, 4 VGA ports in the power supply / :wink:
With Thermaltake I supply four remaining 1080Ti (first and second as above two 8PIN-8PIN cables, third: two 8PIN → 8PIN + 6 + 2PIN, fourth power is the fifth card 6 + 2PIN. The cables used are modular … the 4pin target is factory-built and left undisturbed, plus 2x PERIF from the power supply supplies the two MOLEX ports on the motherboard. in the same moment.
I have tried different combinations. This is the only one that allows you to start up your hardware and get to the Desktop from a connected monitor to HDM I (eventually I use TeamViewer).

BIOS Motherboard: LATEST | Set Gen1 and PCI EXPRESS as Primary graphic. Disable unnecessary devices like Onboard audio. The latest 1080Ti drivers (from NVIDIA). Those from AORUS have the same effect. GPU Bios - LATEST!
All were factory-built except for one that was uploaded by me.
All updates are pasted, DDU already used. I have already changed the drivers by manually selecting in the signatures for those signed digitally from microsoft - unsuccessfully! By the time the power adapter came in I had 5 cards connected to 1350 + 750w and digging was not tied to any problems. On the Thermaltake power supply there are diodes, one of which is red (at PG Signal). For every term I read where I can and did not find meaningful posts that would be helpful. Please help where is the error / what is missing / where badly connected / what, why and why :slight_smile:
Thank you for help

2

Are you running nvOC? Sounds like it from what you describe. First thing I would look at is how are your risers powered? Are they powered on their own line, or are you putting multiple risers on a singe power cable? That can make a diff. in 1080 ti stability since the pcie risers will pull as high as 75w each. If you can, put them on different power lines.

If you’re running 6x 1080 tis, depending on brand, you could pull over 300w per card when the card is boosting it’s core clock speed. My aorus 1080 ti xtremes pull 300w consistently in windows… but if you’re on linux such as nvOC you can underpower them to say max 250w to ensure you don’t pull more power than you have.

6x 1080 tis require 1800w to run from the PCIe cables + you need overhead of about 10% or so.
6x risers will pull typically 300w, but max of 450w if they hit the peak 75w allowance for a pcie slot.
add power requirements for SSD, Mobo, CPU of about 150w

With that out of the way, you’re going to want to split your 1080’s (and risers possibly) between the PSUs so that your distribution is weighted based on the available power of each unit.
If you put 3x 1080 and 3x risers on one PSU, that draws 900w. Do that for both PSUs, then pick the more powerful one as main power for mobo/cpu/ssd, and slave the secondary using the 2 pin splitter from your ATX cable.

Also, to the extent possible, only use 1x 8pin power port per cable. That will help distribute the power among the available power rails on the PSU, reducing overdraw problems per rail which will destabilize your system.
When that’s not possible, power a single card from the 2x 8pin cable, rather than splitting the power between cards.

My bet is that your risers are underpowered, as that is the most common problem I’ve seen for instability in mining apps, so try splitting your risers evenly to both PSUs first then see how it runs.

Do NOT set gen1 for PCIe devices… they are NOT gen1 they are gen3 - just leave this on auto. That is true despite the fact that you’re using a 1x pcie riser, which is unrelated to that setting.

Also, on ASrock boards if you’re using nvOC, enable the ‘above 4g’ setting in the bios in addition to disabling the audio device - this is not necessary in windows.

(on closer read I see you’re using Windows - not sure how many GPUs windows can support from one brand, it might max at 5. I highly recommend running a linux based o/s for large numbers of GPUs, as they tend to be more flexible and forgiving, as well as more stable in my experience. If you have no experience in linux, that might be daunting though so start in your comfort zone.)

Hope that’s helpful, good luck.

3

Hey.
I have 3x GTX 1080Ti AORUS XTREME EDITION with Win 10, so can share a little bit of info I have.

  1. Recommended PSU 600W: AORUS GeForce® GTX 1080 Ti Xtreme Edition 11G Key Features | Graphics Card - GIGABYTE U.S.A.
  2. I have read that maximum power consumption can reach 325W
  3. maximum power consumption I have seen was 301W in OC mode
  4. I use beQuiet 12000 PCU and it seems to be weak for my 3 cards. Randomly one of cards sets itself to Game mode.

To conclude, 1300W is W-A-A-Y too weak for 6 !!! 1080ti’s. U need smt like this: 6 * 325 (each card) + 50 (CPU and mb) + 150 (reserve) = 2 150W

4

I use a 1300G2 + 1000G2 on my 1080ti rig

5

No, Windows can handle more than 5 GPUs - 6 should not be a problem under Windows 10.

I like using a beefy EVGA coupled with a HP server 1,500 PSU - but the HP server PSU requires an X-adapter board, cables, and a step-up transformer (in the U.S. which doesn’t normally have 240v for residences). I also add a cheap eBay aluminum heat sink to the tops of the HP 1,500 units which seems to help cool them (their tiny fans don’t rev up as much) and I can feel the heat on the heat sink fins. Parallelminer has the pictures and information on these server PSUs. But you can buy the HP 1,500w PSUs from eBay as well for about $75. X-adapters $27 or so by themselves. 5,000W step-up transformer - ELC units for $109 on eBay.

And I do use MSI Afterburner to turn down the power draw on the 1080 Ti’s. Sol/W in the range of 3.4 - 3.5.

6

Thank you very much for your detailed description! I’m very surprised to be involved. I give more details and changes that occurred with subsequent changes in power distribution after the excavator. I finally connected everything from scratch and with more focus and having your eyes fixed on your answer. I managed to get a 3x green led on Thermaltake and the system found six cards. Everything would be ok if not for the fact that the original card was connected 4 pcs to a larger power supply and 2 pcs to the smaller. (Which resulted in a uniform division into GPU groups powered by two power supplies). Cards connected to the larger (4) reached the temp 89-90 degrees, kicked quite a lot, they liked to turn off the overheat … while the two connected to the smaller had a tempo of 70 degrees and a similar value of salt to those hot (all together In the range of 730-760 sol). Connected to a single cable 4 risers (from the same cards that power the card from a larger power supply [on ONE CABLE], and the other one also connected two cables (also connected to it).
I made the changes and aligned the energy distribution on the individual elements with respect to the whole …
Finally it looks like (now / before):
Thermaltake 1485W supplies:
-3x GPU [previous: 4]
-3x riser to the above cards (each with a separate cable [formerly: 4 risers + ssd on one cable from one port of the power supply (cable with five sockets)]

  • CPU, SSD and PERIF (MOLEX 1 of 2 available on the motherboard) [previous same, but both molexes on the motherboard were powered by one cable from this power supply]
    EVGA 1300 G2 supplies schematically the other half:
    3x GPU [previous: 2]
    3x riser [previous: 2]
    1x molex [previous: 0]

I tried to connect it in such a way that as little as possible “load the cable” that is, I used the least developed cable and connected to it ONLY ONE component [to the next I took the next]
This is where I thought it should be due to the full equalization of sol / temp / current consumption - not for the first time I was surprised … All cards except the last one reach temperatures of 85-89oC. When firing EXCAVATOR EQUIHASH - it crashes my memory of lack of memory and everything spills. Digging and SOL values ​​are less stable than the previous configuration. As the excavator joins the teamviever: Due to his requirements, the connection ends with the end of the adventure until I can physically restart it. I replaced 4GB of RAM with 8GB of RAM, but the problem is still there. Can the reason for this be:

  1. MOLEX power supply board from two different power supplies (where previously it was just the stronger?)
  2. The fact of using a much larger number of plugged cables that allowed to reduce the load of one cable so that each cable would only power one cable?
  3. Changes to Gpedit and SYSTEM → ADVANCED by adding the LOCK PAGES IN MEMORY option and trying to enter various values ​​into the shared memory?
  4. NVIDIA driver ownership (latest 384.xx ver?)
    Somewhere I read that someone stabilized it by “undervoltage 1mv” but I can not find it in any way [clocking options voltage is always unavailable [locked?] Has anyone any idea how to set this? Similarly, I also heard that after this setting - the card gets RAM from its own resource (which is powerful anyway), yet it takes from my system, which causes that the amount of RAM is not enough for their operation.
    Which app to use? What key issues in the BIOS can have bad values in my case?
7

How does the stepup transformer work? Does it draws more amps from the circuit to make up for the difference in voltage? Just wondering if I could make that work in my house, since all my wiring is 110v, but I don’t want to do something dangerous to the hardware or wiring.