Pages:
Author

Topic: Help Needed (0.05 BTC prize) - page 2. (Read 3561 times)

member
Activity: 92
Merit: 10
December 05, 2016, 02:25:44 PM
#33
Are all of your cards connected to the same server PSU or you are using 2 different psus , one for MB and cpu and other one for GPU power ?  When you said Same for the 10 rigs did you mean that you have 10 completely same rigs and you tried adding this card to every single one of them or what ?

I have HP 1500 watt server PSUs.To feed mother board and one card I also have a corsair CX750 M.

I bought 50 cards.I have got 10 pcs of 5 card rigs.

Either feeding from server psu or from corsair 4 cards works perfectly.However the fifth one does not work.I can see it at the configuration with a yellow label warning.



If your problem card connected in PCI-E x16 slot, try to use x16 to x16 riser
I saw this problem on two different MB (Asrock - first x16, Asus - last x16)

edit:
you can check even without a riser - just put a GPU in PCI-E x16 directly
if problem will be solved, so - just use x16 to x16 riser to connect
sr. member
Activity: 378
Merit: 250
sr. member
Activity: 273
Merit: 250
BD People Are Legend
December 05, 2016, 03:11:22 AM
#31
if it is code 43 in properties then it is for sure you need 6x gpu mod - i wrote before, did you not try? (it is a driver problem that it can run only 4 cards) https://lbr.id.lv/6xgpu_mod/6xGPU.html
legendary
Activity: 918
Merit: 1000
December 05, 2016, 12:49:44 AM
#30
My OS is Windows 10.The photo of the error:

https://postimg.org/image/7sd78g4sz/


sr. member
Activity: 546
Merit: 250
It takes a lot to build but not much to lose
December 05, 2016, 12:28:49 AM
#29
Which windows version?
sr. member
Activity: 395
Merit: 250
December 04, 2016, 11:38:52 PM
#28
And my last question....Why is it recommended to wait an hour?What is the logic behind that?
legendary
Activity: 1274
Merit: 1000
December 04, 2016, 11:05:07 PM
#27
Are you using 2 PSUs for 1 Rig? 1 server psu for GPUs and 1 standard PSU for Mobo's?
If yes then there is timing issue.
Do change your risers and use USB Power risers with big capacitor.
For RX 480 use 16.9.2 and for 390 use 15.12.

Thats it you will be fine Smiley

EDIT: There is no issue in your RAM neither in SSD, Windows10 and in Antivirus.
Just do what i said above and keep your rig shutdown atleast for 1hour or so.
Then you are good to go.


That is also a good point.What are the differences between 16.9.2 and 15.12.x ? Do they have a manual?

15.12 has no series 400 support  , other wise not really much different then 16.9.2 that supports all there cards but some older out dated stuff they no longer support and no one really uses for much  .


http://support.amd.com/en-us/download

the latest driver from amd is

Crimson Edition 16.11.5 Hotfix.
i use it with out issue

my suggestion is try after burner only to see whats on and off and turn off sync setting between cards once it loaded then hit reset for each card it may take time but it might give you a better idea of whats going on ..

and don't use extend over clocking limits, if your mixing cards.
sr. member
Activity: 395
Merit: 250
December 04, 2016, 10:54:03 PM
#26
Are you using 2 PSUs for 1 Rig? 1 server psu for GPUs and 1 standard PSU for Mobo's?
If yes then there is timing issue.
Do change your risers and use USB Power risers with big capacitor.
For RX 480 use 16.9.2 and for 390 use 15.12.

Thats it you will be fine Smiley

EDIT: There is no issue in your RAM neither in SSD, Windows10 and in Antivirus.
Just do what i said above and keep your rig shutdown atleast for 1hour or so.
Then you are good to go.


That is also a good point.What are the differences between 16.9.2 and 15.12.x ? Do they have a manual?
sr. member
Activity: 395
Merit: 250
December 04, 2016, 10:49:57 PM
#25
problem may be about ULPS

ULPS is an ultra low power saving feature which turns off all GPU's except the primary one in use. ULPS can cause crashing / stuttering / reduced scaling when enabled. ULPS is always enabled by default. It can't be fully disabled via software such as MSI Afterburner or Sapphire Trixx (as of 11/01/2016) since AMD Crimson (and some previous CCC versions) changed various reg entries which controlled ULPS. Before following the guide, read the Q & A below.

Solution

1) Click Start - Type regedit.exe and hit enter.
2) In reg editor, press F3 and type EnableULPS, hit enter.

3) Keep pressing F3 until it finds a key called EnableUlps in a sub folder called 0000
4) Double click the reg entry and change value from 1 to 0.

5) Keep pressing F3 until it finds a key called EnableUlps in a sub folder called 0001
6) Double click the reg entry and change value from 1 to 0.


You got an interesting point.But he says the rig works for a time and closes.If  ULPS was the reason it should not be closed down but continue to work on 1 card.Right?
sr. member
Activity: 331
Merit: 250
December 04, 2016, 10:34:44 PM
#24
I am also very much interested in your poblem since it is a little bit awkward.To be more open I am not quiete sure either it is a software problem or a hard ware problem.

Let's think the possible problems and eliminate.


Bios:Updated, pci-e version 1; OK

Molex: If one 4 pin molex is connected to motherboard, it is OK. You do not need two molex for powered risers.

PSU:Your server PSU's are ok.Corsair is enough if it is not the best option.

CPU: Oldies and cheap goodies.

Mother Board:Asrock H81 Pro BTC, well known and cheap still a good option.

RAM:8 Gb. You must be rich:)

SSD:120 Gb SSD.Check the virtual memory.16 Gb is needed.


OS:Windows 10.   HuhHuh I am a linux fan.

GPU:480 Asus ROG 8Gb. What is the difference with other 480 cards.I will focus on that.


My suggestion: You have reported that the rigs work for some time, say 30 minutes and after they go.On the other hand one of the GPU is always undefined. My solution is downclock your rig and work it there.The problem is at your PSU.How?

Your GPU's have 8 pin connections, probably you convert 6 pin to 8 pin with a cable converter.If the cable is manufactured in China the quality is a matter.Just cut the cable connecter and see if it is copper or aluminium.Also read the AWG, the thickness, it must be eighteen or smaller.

Do not overclock , do not hack your GPUS.

If my answer is not satisfactory just PM me.If you provide TW I can detect the error at your system.
sr. member
Activity: 331
Merit: 250
December 04, 2016, 10:19:59 PM
#23
Dear friend you can either use Ethos or Pimp.Which are quiete different from each other for functions but very similar coding.Both of them are linux based systems.So, what I advise you is just purchase one of them and boot your rig using the flash dirves.By the way Ubuntu is also a fork actually (Debien).

The trouble shhoting flow chart is prepared by Hasflux.Here is the link:
http://forum.ethereum.org/discussion/7457/anyone-using-asrock-h81-pro-btc/p2


[INITIAL LINUX TROUBLESHOOTING STEPS]
First let's see if the kernel even sees the hardware. Use the following two commands. Note I add comments with # symbol but they are not intended to be typed in the terminal.

1. Ssh into the machine or if your running it non headless open a new terminal.(all of my rigs run headless and have no x windows running but with linux amd cards you do have to run a basic x server if you desire to kill your cards life span by playing around with over clocking and undervolting. Truthfully your better off using a custom bios on your card for this anyway and not run a memory hard x server.)
2. # We need to update the pciids to the latest list so the commands we use will recognize all of the cards properly. Run the following command...

sudo update-pciids

3. # Now let's see what VGA devices the system is detecting as connected without relying on the drivers

lspci | grep VGA

# there you will see a nice list of all VGA devices. If you see all six then that elemenates the motherboard and linux as the problem. If thar command only shows 5 then try disconnecting all cards and move the card one by one through your pci slots to find the offending pci slot. If it detects each slot fine try it with the next card and continue until you find either the bad pci slot or the bad gpu and proceed to do an RMA with your vendor.

[Driver Trouble Shooting Steps]

Assuming everything passed our pre trouble shooting steps we have elemenated most possible hardware problems but one more may show up as we troubleshoot our 6 GPU issue. Now we need to check the drivers.

1. Run the following command.

clinfo

# Does the command show all your devices from the opencl drivers perspective or is the driver failing to see them all. NOTE: if the clinfo hangs for a long time and does not return anything then one of your cards is bad and you will have to try the command one by one disconnecting individual GPU and proceed to do an RMA once you find the offending hardware.

2. Assuming you do have the AMD Proprietary drivers installed and clinfo worked ok the next step is to make sure the devices were initialized properly by the driver. Use the following command.

sudo aticonfig --initial --adapter=all -f
sudo shutdown -r now

Once the system comes back up open a terminal and a final check on the configuration with the following command

sudo aticonfig --list-adapters

All your adapters should be showing up now and that pesky sixth gpu is ready to mine with or its in the mail at this point for an RMA
legendary
Activity: 918
Merit: 1000
December 04, 2016, 09:51:12 PM
#22
Mother Board: H81 BTC pro
Risers:6 powered usb risers .. version 3
PSU : 1200 Watt Platinium Server PSU for 4 gepu cards, 1500 watt for five gpu cards
CPU :Intel Dual Core 3250
RAM :8 GB, 1800 Mhz
HDDs connected to mobo :120 Gb SSD

Wi Fi connection with internet adapter and ethernet Cable

legendary
Activity: 1084
Merit: 1003
≡v≡
December 04, 2016, 07:29:10 PM
#21
2gablay12
just use win8.1 or linux Wink
I see a winner here  Grin
full member
Activity: 175
Merit: 100
December 04, 2016, 07:20:14 PM
#20
2gablay12
just use win8.1 or linux Wink
full member
Activity: 142
Merit: 100
December 04, 2016, 06:59:37 PM
#19
Plug a monitor into the card that is not showing up and reboot!?
sr. member
Activity: 546
Merit: 250
It takes a lot to build but not much to lose
December 04, 2016, 06:41:05 PM
#18

  I have done so many things...

I have uninstalled all the drivers with DDU.


I have changed the pci-e bios settings to gen1.

I have reloaded only the driver for GPU.

I have connected GPUs one by one.

Well, it worked for half an hour.With 1 GPU missing (yellow alert).

I have checked all the molex cables, powered risers...Everything seems well.

This situation is valid for my other 9 rigs.

Well, I will try Ubuntu as software.If any suggestions pls feel free to PM.The prize is still valid.

Best Regards

What yellow sign?
Can you screenshot it and send?
newbie
Activity: 3
Merit: 0
December 04, 2016, 02:55:07 PM
#17
When going over 4 GPU's you must connect a molex power cord to the motherboard according to H81 pro BTC manual. The connector is between the CPU and the first PCIE 1x slot.
legendary
Activity: 918
Merit: 1000
December 04, 2016, 02:43:03 PM
#16

  I have done so many things...

I have uninstalled all the drivers with DDU.


I have changed the pci-e bios settings to gen1.

I have reloaded only the driver for GPU.

I have connected GPUs one by one.

Well, it worked for half an hour.With 1 GPU missing (yellow alert).

I have checked all the molex cables, powered risers...Everything seems well.

This situation is valid for my other 9 rigs.

Well, I will try Ubuntu as software.If any suggestions pls feel free to PM.The prize is still valid.

Best Regards
sr. member
Activity: 546
Merit: 250
It takes a lot to build but not much to lose
December 04, 2016, 09:48:44 AM
#15
Smiley Already done.It was my very first thing to do.

It seems to be reasoned by DDU.

I will share how I have done ASA I find the solution.

The solution is simple. First remove all gfx cards, all non essential HDDs and anything else attached to mothereboard battery to reset mobo, connect monitor to onboard hdmi and proceed

1) Reinstall battery
2) Boot (if some screen pops up about molex connectors ) hit 'n'
3) once you see bios screen , Go to advanced chip settings and make sure PCI-e versions are set to 'gen 1' and the preferred display out is PCI-e and not onboard GPU. Save the settings and shut down.
4) Use one gfx card , connect VGA/hdmi cable to it
5) attach the usb riser to the card and mobo 'on the first pcie x16 slot on H81BTC pro' and gfx card.
6) boot
7) Install gfx card drivers and shut down
Cool now move the usb riser to first X1-pcie slot and boot.. Shutdown once you get the screen
9) Repeat the process on each of the pci-e 4 connectors .. one at a time (This way you know if some connector is bad too)
10) If all the pci-e connectors are able to show display and boot and let the drivers install.
11) Now that we know all pci-e connectors could identify gfx card , connect pci-e risers to other cards and connect all of them to all 6 pci-e slots.
12) Move the hdmi to gfx card connected to first pci-e x16 and boot.
13) all 6 gfx card should be seen now on the device manager.
14) You DO NOT need registry hack for 6 gfx cards in windows 10.

Let me know what happens
legendary
Activity: 918
Merit: 1000
December 04, 2016, 08:07:59 AM
#14
Smiley Already done.It was my very first thing to do.

It seems to be reasoned by DDU.

I will share how I have done ASA I find the solution.
Pages:
Jump to: