Yes, there are spaces between the numbers. In fact, the string of memory clock settings in my original post is copied and pasted from the last time I tried implement individual memory overclock settings.

I keep a running Word document with my login links and custom configs for various miners, along with my notes concerning each configuration. While it may not be the most eloquent solution, it works very well, and it's not a bad idea to keep a backup of settings regardless of whether smOS eventually introduces a way to store the setting in the dashboard.

I am having troubles getting smOS to apply varying overclocking settings to individual GPUs in a rig. The rig in question has 7 cards, and 6 cards can run all day and night with the memory overclocked to 2200, but one card (GPU1) can only handle 2025. I tried using several variations of the following entry for the memory overclocking settings:

2200, 2025, 2200, 2200, 2200, 2200, 2200

but it always ends up applying 2200 to all cards. I have tried the string of entries both with commas  and without commas, and I have also tried entering "2200, 2025, 2200" with the understanding that the last number would be applied to the remaining cards, but nothing works.

I tried searching the thread to see if anyone else has dealt with his problem, but the thread is simply to voluminous at this point. I also know that others have been able to get this functionality to work because others have posted screenshots of their smOS dashboard that shows varying OC settings for individual cards in a rig.

If it makes a difference, I am running the beta kernel and AMD driver. However, I had the same problem before I upgraded the kernel and driver.

I started running 14 of these of two weeks ago, and they have performed flawlessly:

http://www.ebay.com/itm/2016-New-design-stable-version-PCI-express-1x-to-16x-riser-for-ETH-GPU-mining/281383006276?ssPageName=STRK%3AMEBIDX%3AIT&_trksid=p2057872.m2749.l2649


I placed two orders, and I received both orders within 3 days of placing my order. Also, the seller was willing to give a slight discount on my second, larger order.

I like this particular riser because it has a 6-pin PCIe power connector. Using this riser in conjunction with my preferred PSU (Corsair AX1200i), I can power 7 cards without using a single splitter or questionable connector. Each plug has its own dedicated cable, and all connectors are 8-pin or 6-pin PCIe connectors.

Just to emphasize this point, if I were to rank order the difficulty of tasks involved in getting a rig up an running using smOS, I would place getting smOS to work somewhere between installing a stick of memory on the MOBO and plugging a power cable into the PSU that was a tight squeeze between three other plugs.

Yup, I am using your OS to run 7 cards on my MSI z170a Gaming M5. Once I sorted out out the BIOS settings, I haven't had any problems.

I will PM you my email address so you can see the rig on your end.

Same here with my RX 470 cards. That's why I noted in my post about the MSI z170a board that PCIe speed settings should be left alone.

It's been working with 7 cards for the last 14 hours on smOS using the beta kernel and AMD driver.

I just got my 7th GPU up and running on an MSI Z170A Gaming M5. So far so good, and if you have a local MicroCenter (I forget if you are in the US), they tend to have several open box discounts. The salesperson I spoke with said they tend to get a lot of returns for this board because it is an entry level "gaming board," and a lot of 12 year old kids end up buying the wrong processor. I didn't know how to feel about that. Regardless, they are available for $100 here on a fairly regular basis, and $120 all the time.

If you buy this board, this is all you have to do to get it working with 7 cards:

--upgrade to the latest BIOS
--activate 4G decoding in PCIe BIOS options
--select IGE (i.e., integrated graphics) as the preferred video source upon bootup.
----there are a couple posts floating around where people state you have to select a different PCIe speed. DON'T DO THIS. It will cause all sorts of problems and is unnecessary since the release of the latest BIOS update.


That's it, and you can run 7 cards with it.

After I changed  PEG1 - Max Link Speed from [Gen1] back to [Auto], while leaving the two below changes in place, everything is running fine with 5 GPUs:

PEG0 - Max Link Speed: change from [Auto] to [Gen1]
Above 4G Decoding: [Disabled] to [Enabled]

As such, I am currently running two more GPUs with these two changes than I could run at the beginning of the night. I am hopeful I will be able to run cards off of all 7 PCIe slots with these settings once I get the necessary hardware to do so. I will keep you posted.

The main reason I provided all of the information in my previous post was that I hoped it would assist you with debugging the 300/300 problem others are experiencing. Once I have the necessary hardware to try and run 7 GPUs, I will report back.

On the other hand, if changing the one troublesome BIOS setting back to what caused the problems and then making the changes you suggested in your post would help you with identifying a larger problem, please let me know. I will be happy to oblige.


I am running the stock BIOS on my cards, at least for now. I figured it would be best first to get as many cards running as possible in stock form before I go monkeying with the cards' BIOS. That way, it is easier to track down the source of any problems that may crop up. Once I get everything running, I may modifying the memory timing, but that's the extent of my planned modding.

I am now experiencing the 300/300 problem with gpu0, regardless of how I set up my rig or which PCIe slot I use.

Before this afternoon, everything was cruising along fine for the last week with my 3 x MSI rx 470 gaming 4GB rig. Today, I received risers in the mail, and proceeded to make my 3 gpu rig into a 6 gpu rig. I eventually managed to get it to boot up and everything looked like it was going as planned, but gpu0 was stuck at 4.04 mh/s (with the dashboard indicating it was clocked at 300/300), while the rest of the cards performed as expected.

I began working backwards, thinking I had a hardware problem, but after a couple rounds of unplugging what I thought was gpu0, and upon reboot, gpu0 still operating at 4.xx mh/s with the dashboard indicting the card was clocked at 300/300, I started to suspect otherwise. I finally got down to one card, and after several reboots and moving it around to several different PCIe slots, same thing.

I upgraded to the beta 4.11 kernel, and that didn't help.

After all of that, I started reversing the other changes I made to my rig in anticipation of adding additional cards. I purchased an MSI Z170A Gaming M5 mobo because all of the mobos traditionally favored by miners are scarce and, to put it politely, selling at a premium. There are several people who said they were able to get 7 cards running on this board, and they outlined the changes to the BIOS that were necessary. When I undid those changes, my gpu0 went back to working correctly. The changes to the BIOS consisted of the following to the PCI subsystem settings:

PEG0 - Max Link Speed: change from [Auto] to [Gen1]
PEG1 - Max Link Speed: change from [Auto] to [Gen1]
Above 4G Decoding: [Disabled] to [Enabled]

Through process of elimination (i.e., testing my system with only one card installed directly on the mobo), I determined that changing PEG1 - Max Link Speed from [Auto] to [Gen1] is the culprit. As long as I leave that setting on auto, I can change the other two settings as described above and gpu0 functions properly, but as soon as I change the PEG1 setting, gpu0 goes back to 4.xx mh/s and 300/300 on the dashboard.

To be clear, I didn't type all of this up because I expect someone to help me. To the contrary, this is a hobby to me, and I will work through it. However, it seems like the 300/300 problem has reared its ugly head multiple times without a decisive resolution of the underlying cause. I'm just hoping that the info I provided will help tytanick diagnose the problem.

I am new to running this card. Could you point me in the right direction for the mods you have in mind?

As for my current settings, I, too, was surprised they remained stable at that setting. However, I have always proceeded with the assumption that if the card remains stable over 12+ hours and temps remain in check, it is good to go.

I just brought 6 MSI 470 4GB gaming cards online last night, and they are averaging 26mh/s per card. I am running the cores 1100 MHz and memory at 2200 MHz at power level 4. I am using Claymore v8.1, ETH only.