can any tell me how many lanes each level of card uses. eg 1050 1070 1080 1080ti. why i ask is i had 6 cards running on a gigabyte ga990 board. if i have worked it out correctly the board has 42 lanes available. When i try and go to six 1080 cards i wont boot at 99% if it does boot it tells me the cards arent working, and the miner cant see the cards either. i believe someone said the 1080’s use 8 lanes each so it would seem logical 5 cards works leaving 2 lanes for the cpu and other stuff. but this only makes sense if the 1050 and 1070 cards use less than 8 lanes which is what allowed me to run 6 cards before i stuck the 1080’s in. thanks
A GTX 1080 doesn’t use any more lanes than a 1060. So if it worked with 6 cards, say 1060s, then it should work with 6 1080s. While 1080s can use up to 16 lanes each, if they are in an x16 slot. If they are in an x8 or x4, they use that many. If you are using x1 risers, then you are only using 1 pci-e lanes for that card/slot. So I would think you are fine in terms of PCI lanes.
What 1080s do use much more of than other cards is power. So I would look at power first. I think you need at least 1500 watts to run 6 1080s. Do you have that? Tell us your full configuration. What processor, power supply(s), risers, etc. Will it boot up with 6 cards and just not mine? Or will it not boot? I
To troubleshoot some, try running 1 card. Will it boot and mine? If so, switch out the cards until you are sure each card works fine. Then try each slot/riser with a card. After that, you are sure each card and slot is fine. Then try running 2 cards, then 3, then 4.
Sounds more like a power issue (lack of ) than a lane issue.
!000 watt Psu running 4 1080’s
gigabyte fx990-fxa board
Phenom 1100t cpu 3.7 ghz
16gb ripjaws ddr 3 ram
and a second 550 watt psu to run the other 2 1080’s
have swapped and changed so many time i am dizzy. all cards and risers are happy to run in any combo but only 5. as soon as i add the 6th card it either wont POST or if it does it takes an hour or more to load windows 10 and then windows tells me the cards arent working.
I just rolled it back to windows 7 it seems much happier but now the ewbh miner wont run with the latest drivers but nicehash miner will run
arghhh crazy machine
The ew miner is complaining about a DLL file
Probably the Visual Studio 2013 runtime dll. Search for the Redistributable installer package.
If you are worried about the PCI lanes, I think the nvidia control panel will show you how many lanes each card is using.
The windows 7 issue does sound like the visual studio runtime dll’s. Google microsoft plus the exact filename it complains about and it should pull up a link to the Microsoft page you can get it from.
Windows 10 should boot fast. Usually when it doesn’t, it is a driver issue. If you switch back to windows 10, I would forcibly remove (revo uninstaller and cc cleaner) all video drivers. Then reinstall them. If that doesn’t work, I would do a clean boot and add stuff back until you identify the culprit.
Seem like you should have enough power, but maybe someone else would disagree. I run 6 1070s on 900 watts.
if that is supposed to be 1000w psu and that alone is running 4 1080 gpu’s, risers, mobo then yes power issue
A 1000 watt psu might not give you 1000w and running it past 900 watts is probably not a good idea. A 1080 can use up to 180w at full load so even running the cpu and fans should not hit 900w with 4 gpus. Have you looked at the mobo bios? Sometimes you can go into it and set higher lanes. Some have trouble when you pass 4 gpus. If you have not taken a look through the bios, then that could help. Some even let you increase the bus speed which can help. There are many YouTube videos on mobo bios adjustments.
using your 180w base,
4 gpu’s = 720w
Mobo, cpu, HD, ram = 75w to 125w
riser = 50w each x four
only run 3 gpu’s on the 1000w PSU, and the other two on your 550w psu, bet it will run fine. Add the 4th gpu on the 1000w psu, problems start.
It’s possible the 1080 could draw 230+w total. That is usually the range of the ti, but you are right citric, if the person is drawing that as a full load then it could exceed 1000w with all componants. I use a kilowatt meter to test. It’s a good investment. They could try lowering the power to say 80% and see if it will run all 4 cards? It still would put them above 900w potentially. But if all 4 ran, then that would also confirm a power problem under full load?
thanks all for the insights, have looked at all the bits suggested, but to no avail tried more power (3 PSU 2 cards each) changed risers, cards, cables, power/sata/usb and no joy. it just wont boot with a sixth card now.
it runs ok with 5, except i now get an error code 6 on one pci-e port on the board, i know its the board port to blame as i can swap a new card/riser/psu onto that port and it always fails with code 6. I wonder if the board itself is slowly dying.
ok so i couldnt locate a work around for the error code 6 on cuda device 3 but it does run ok on that gpu using the nicehash miner. i think the error code is a timeout (based on the Nvidia web site) so maybe the nicehash miner has more tolerance for PCIE wait times
You must read the mother boards manual and in there it will explain the buss configuration with a logic map. Now going to their web site will prove more useful or even someone likes Toms Hardware review has great inside information on these things.
Remember that everything connected to the mother board uses buss space (band width) and while some share lanes some (if turned on) like USB/sata connections reserve those lanes for that item exclusively.
Also note that MB have header connections for these as well so a good rule of thumb is disable everything except 1 sata connection and one usb and use a docking station for usb.
Definately power issue I use 1200 watt continuous server power supplies to run 4 x 1080ti cards and still put another PS on the risers. I see draw range from 200- 290 watts on the 1080ti cards 1160 watts continuous draw at peak. So 1 your under powered. I run currently 380 x 1080ti’s without problems. Using a 12 x gpu board or my new 24 x gpu boards it’s really all about power, quality connection, and bios. Be sure to put a dab of silicone on the riser motherboard connections (just a dab) to make certain they stay in well along with the usb connection on the riser and the power connector on the riser. I see when these connections get warm/hot they tend to work themselves out a bit and you get loss of hash, a crash or a no post or a mystical new problem that you can’t figure out lol gotta love those “grr what the hell is going on problems.” Also be sure your only using 220v connections for your power supplies I know people on here are going to yell at me that 120v power supplies are just as good but well I don’t think anyone on this forum is running a bigger farm than me right now and I can tell you from trial and error that the 120v power supplies run hotter and less continuous output and they detune as they get hot they also tend to die out sooner than the commercial 220v power supplies. When we switched about a year and a half ago to all 220v 1200 and 3600 watt power supplies we have only had 1 power supply go bad. We have cheap 1 dollar current and volt monitors we bought off aliexpress on all our power supplies to monitor real world real time current and voltage output of the 12v feed and you would be surprised nearly all our 120v power supplies detune when they get hot or they get alot of distortion which also was surprising.