• Guest, The rules for the P & N subforum have been updated to prohibit "ad hominem" or personal attacks against other posters. See the full details in the post "Politics and News Rules & Guidelines."

Ethereum GPU mining?

Page 50 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

n0x1ous

Platinum Member
Sep 9, 2010
2,524
181
106
and people should really stop mining until this mess is straightened out, since it will end up being worthless until the fork is done.
__________________
What for? If miners are converting there eth to btc or usd on a regular basis I dont see a reason to stop. If someone is mining to sit on a ton of eth then yeah but not sure thats what most of us are doing.
 

Elixer

Lifer
May 7, 2002
10,376
762
126
Unless I missed the "fix", this 'flaw' is still present, so, there is nothing stopping anyone from continuing doing what the original crook did.
They only have 27 days to fix this, if indeed they do fix this via a fork, then the value will plummet even more than what it already has done.
 

EightySix Four

Diamond Member
Jul 17, 2004
5,115
49
91
Unless I missed the "fix", this 'flaw' is still present, so, there is nothing stopping anyone from continuing doing what the original crook did.
They only have 27 days to fix this, if indeed they do fix this via a fork, then the value will plummet even more than what it already has done.
They moved it to a sub DAO which is currently locked.

I am not sure if you have noticed but the value is back up...

It is a lose lose situation here, the longer this goes on, the faster the crook(s) can try and cash out, yet, people are still mining,
He can't cash out yet... and I'm not sure why people should stop mining and bring Ethereum to a halt when this is just one dAPP running in Ethereum. Some of us have other stuff going on on the chain 👍
 
Last edited:

techie81

Senior member
Feb 11, 2008
342
0
76

Feld

Senior member
Aug 6, 2015
287
95
101
Question for you guys. I have two MSI 390s mining but the top card is getting to 94c and throttling.

This is the card
http://www.newegg.com/Product/Product.aspx?Item=N82E16814127874&cm_re=msi_390-_-14-127-874-_-Product

Since these cards are wide they are really close too each other.

Can someone recommend a way to get this to work? I was thinking about moving the power supply to the cdrom area and moving the second GPU to the very bottom PCIE slot. Thoughts?
Yeah I have two of those same cards myself. They really aren't designed to be in slots that are only 3 slot widths apart, since the cooler is 2.5 slots thick. The top card just can't get enough cool air. Those temps are really bad - you do not want to be mining at that temp, it's not designed to run that hot and the gpu is going to fry. Definitely try to move the other card to a further slot if possible. I briefly tried running my two cards in that same kind of configuration and though I didn't get into the 90s I think that even 80s is too hot for 24/7 operation. I swapped them into my open frame mining rig where they are spaced much farther apart and moved a pair of Gigabyte 290s with thinner coolers into the desktop pc. The top Gigabyte card now stays below 80 with an undervolt and underclock.
 

Newbian

Lifer
Aug 24, 2008
24,691
713
126
Question for you guys. I have two MSI 390s mining but the top card is getting to 94c and throttling.

This is the card
http://www.newegg.com/Product/Product.aspx?Item=N82E16814127874&cm_re=msi_390-_-14-127-874-_-Product

Since these cards are wide they are really close too each other.

Can someone recommend a way to get this to work? I was thinking about moving the power supply to the cdrom area and moving the second GPU to the very bottom PCIE slot. Thoughts?
Have you tried risers?
 

Elixer

Lifer
May 7, 2002
10,376
762
126
Question for you guys. I have two MSI 390s mining but the top card is getting to 94c and throttling.
Don't know if you want to do this, but, you can place additional fans zipped tied on the card to get more direct airflow. Yeah, looks like crap, but a friend who did this, the temps did go down when they did it. They put 2 120mm fans at slight angles.
 

DrMrLordX

Lifer
Apr 27, 2000
18,204
7,098
136
Question for you guys. I have two MSI 390s mining but the top card is getting to 94c and throttling.

This is the card
http://www.newegg.com/Product/Product.aspx?Item=N82E16814127874&cm_re=msi_390-_-14-127-874-_-Product

Since these cards are wide they are really close too each other.

Can someone recommend a way to get this to work? I was thinking about moving the power supply to the cdrom area and moving the second GPU to the very bottom PCIE slot. Thoughts?
If I were you, I'd take some steps to reduce power usage for those cards to see if I could get the temps down. That might also involve reducing clockspeeds. Have you tried Sapphire TRIXX software yet? Assuming you're using Windows?

Also, I have decided I am not going to participate in the nanopool soft fork vote, since I would much rather support the hard fork.
 
Last edited:

techie81

Senior member
Feb 11, 2008
342
0
76
Thank you all for the replies. DrMrLordX, I am very interested in this, in addition to adding high cfm fans. Would this be like undervolting the card? Do you have a link to show me how to do this? Thanks!
 

Madpacket

Platinum Member
Nov 15, 2005
2,068
326
126
Have you figured out some sort of guidelines you would like to share?

My 390x I run the core at about 1125 and my 290 is running at 1175 but the memory on both is 1125 because I am sucker for believing things I read on the internet.
What I do is pretty straightforward at least on Windows. I don't mine ether with Linux so I can't help there.

This is all from memory and I don't have a PC in front of me to validate terminology, please keep that in mind.

1. Open up GPU-Z, click on the monitoring tab and scroll down (if necessary) until you see memory bandwidth (measured by percentage)
2. Start your mining software
3. Open GPU tweaking application of choice (I prefer MSI Afterburner).
4. Set memory to a baseline 1175Mhz
5. Slowly adjust the GPU core speed to preferred speed.
6. Refer back to step 1 - Watch the memory bandwidth percentages. You want to be sitting around 95% utilization. Adjust memory speed until you get to this value.
7. Watch your reported hashrates on your mining software for a few minutes. Make sure the adjustments you make to memory speed do not drop your Mh speeds.

After changing memory speeds it's important to watch your reported hashrates from the console output. In my experience if you push the bandwidth utilization too far (past 95%) you will see an almost immediate drop in hashing rates. At least on Hynix based cards.

Again each card is a little different, the memory speed settings for Hynix vs Elpidia is a little different. Even the same cards can have different BIOS's with different memory timings enabled that can affect speeds.

The key is to watch both bandwidth utilization and reported hash rates.
 

Madpacket

Platinum Member
Nov 15, 2005
2,068
326
126

Madpacket

Platinum Member
Nov 15, 2005
2,068
326
126
Question for you guys. I have two MSI 390s mining but the top card is getting to 94c and throttling.

This is the card
http://www.newegg.com/Product/Product.aspx?Item=N82E16814127874&cm_re=msi_390-_-14-127-874-_-Product

Since these cards are wide they are really close too each other.

Can someone recommend a way to get this to work? I was thinking about moving the power supply to the cdrom area and moving the second GPU to the very bottom PCIE slot. Thoughts?
I have two of the same MSI 390's mining away in a very well ventilated case (five 140MM fans with no air obstruction, Fractal Define S case) but to get them at an acceptable temperature you'll have to severely undervolt and underclock them (-100mv core, 885 - 925 Mhz core speeds, 1025-1050Mhz memory) and set the fan speeds to around 70%. Note this will give you approx 50Mh speeds but without using PCIe risers this is the best you can do with these extra wide cards unless you take measures like Exlir stated about adding direct airflow. With the above settings my top card that takes all the heat stays below 80C, usually around 76-78C max. The other card stay about 8 degrees cooler. This particular setup has been running for over 3 months without issues.

With a pair of Asus 390's in the same case, and the same number of fans I can push them to 58Mh without issue. The Asus cards are only dual slot but have 3 fans so they're much easier to cool. The MSI cards are still the best built 390's IMHO (Sapphire a close second) but they really need breathing room to operate efficiently.

Really with multiple cards in the same case you should really be sticking to dual slot cards with blower fans or water cooling them.

I really recommend using risers if you can, 60Mh+ easy enough this way.

There's many guides on how to use MSI afterburner to overclock and undervolt your cards, Google is your friend here. If you get stuck or can't find what you need let us know and we'll help out.
 
Last edited:

MrTeal

Diamond Member
Dec 7, 2003
3,121
795
136
Hey Madpacket, did you ever get your H97 running and how's that working for you. I just assembled my new system last night, and I can't seem to get it to recognize more than 4 GPUs regardless of what I try in a system with Win7-64. Configurations of slots and cards doesn't make a difference, I just always get an error 43 on the card in the last PCIe slot. I've run the 6xGPU mod, but that doesn't seem to do anything either.

Interestingly, I have 5 working in another system with an H97 and Win7-64. There's only 4 physical cards there though, one is a 7990.
 

poofyhairguy

Lifer
Nov 20, 2005
14,610
316
126
What I do is pretty straightforward at least on Windows. I don't mine ether with Linux so I can't help there.

This is all from memory and I don't have a PC in front of me to validate terminology, please keep that in mind.

1. Open up GPU-Z, click on the monitoring tab and scroll down (if necessary) until you see memory bandwidth (measured by percentage)
2. Start your mining software
3. Open GPU tweaking application of choice (I prefer MSI Afterburner).
4. Set memory to a baseline 1175Mhz
5. Slowly adjust the GPU core speed to preferred speed.
6. Refer back to step 1 - Watch the memory bandwidth percentages. You want to be sitting around 95% utilization. Adjust memory speed until you get to this value.
7. Watch your reported hashrates on your mining software for a few minutes. Make sure the adjustments you make to memory speed do not drop your Mh speeds.

After changing memory speeds it's important to watch your reported hashrates from the console output. In my experience if you push the bandwidth utilization too far (past 95%) you will see an almost immediate drop in hashing rates. At least on Hynix based cards.

Again each card is a little different, the memory speed settings for Hynix vs Elpidia is a little different. Even the same cards can have different BIOS's with different memory timings enabled that can affect speeds.

The key is to watch both bandwidth utilization and reported hash rates.
Thank you!
 

DrMrLordX

Lifer
Apr 27, 2000
18,204
7,098
136
Thank you all for the replies. DrMrLordX, I am very interested in this, in addition to adding high cfm fans. Would this be like undervolting the card? Do you have a link to show me how to do this? Thanks!
Yes, undervolting is the easiest way to go about reducing power usage. Many people use MSI Afterburner, but I have switched to Sapphire TRIXX since it allows one to reduce voltage by as much as -200 mV instead of -100 mV (which is the limit for Afterburner).

The first thing I recommend is getting the software here:

http://www.sapphiretech.com/catapage_tech.asp?cataid=291&lang=eng

Next I recommend using the "custom" fan profile for your 390s (the one by default is more aggressive than what is set in your card's BIOS). After that, I highly recommend you install and use GPU-z:

https://www.techpowerup.com/downloads/SysInfo/GPU-Z/

It will show you power usage as VDDC Power In. On my system, it is always 75W too low. Anyway, try to get your power to 200W or lower (should show ~125W in VDDC Power In). A good starting point is to lower the card's clockspeed to 900 MHz and then lower voltage until it starts to crash the driver (bear in mind that the voltage setting you pick will always correct +7-8 mV thanks to some odd behavior of TRIXX). Monitor power usage. You should be able to get your power usage to around 75-80W according to GPU-z, which translates to ~150W power draw from your PSU.

I realize that moving GPU clocks to 900 MHz is a big reduction in performance (from maybe 30 MH/s to 25), but your GPUs are running far too hot, and you need to get heat down on both of them for them to survive.

Please note there's no proposal for a hard for (yet). I was confused about this as well (see my post below) but YOU NEED TO VOTE FOR A SOFT FORK first to give time for the developers to create the proposal for a hard fork.

https://www.reddit.com/r/ethtrader/comments/4pe64s/why_are_mining_pools_limiting_miners_to_soft/
Now it makes sense. Thanks!

What OS do most people recommend for Ethereum mining? I was planning on Unbuntu...would this work well? Thanks.
If you are mining with an older card (Hawaii or Pitcairn) you need to use a Linux version that still supports the old Catalyst 15.12 drivers. The open-source radeon drivers do not support multiple GPUs to the best of my knowledge. One of the easiest is Ubuntu 14.04 LTS, though for a small footprint I recommend the Lubuntu variant.

You can use Lubuntu 15.10 if you want newer kernel features, though be warned that support for that OS (security updates, maybe not that big of a deal) will end well before it end for 14.04 LTS. LTS = long term support.

If you mine with something newer, like Fiji or the upcoming Polaris cards, then you can go all the way to 16.04 since you will be using the amdgpu driver, which might support all the features that you want. In fact, for Polaris, that will be your only option, since Polaris will be unsupported by the old Catalyst drivers.

I'm not 100% sure that mining with amdgpu will work exactly as intended, but I'm at least somewhat certain that it will. amdgpu doesn't yet support hawaii/pitcairn so that is not an option.
 

Mercennarius

Senior member
Oct 28, 2015
466
84
46
If you are mining with an older card (Hawaii or Pitcairn) you need to use a Linux version that still supports the old Catalyst 15.12 drivers. The open-source radeon drivers do not support multiple GPUs to the best of my knowledge. One of the easiest is Ubuntu 14.04 LTS, though for a small footprint I recommend the Lubuntu variant.

You can use Lubuntu 15.10 if you want newer kernel features, though be warned that support for that OS (security updates, maybe not that big of a deal) will end well before it end for 14.04 LTS. LTS = long term support.

If you mine with something newer, like Fiji or the upcoming Polaris cards, then you can go all the way to 16.04 since you will be using the amdgpu driver, which might support all the features that you want. In fact, for Polaris, that will be your only option, since Polaris will be unsupported by the old Catalyst drivers.

I'm not 100% sure that mining with amdgpu will work exactly as intended, but I'm at least somewhat certain that it will. amdgpu doesn't yet support hawaii/pitcairn so that is not an option.
Thank you for the great response. I plan on using a single RX 480 and probably eventually add a 2nd RX 480.
 

DrMrLordX

Lifer
Apr 27, 2000
18,204
7,098
136
Okay. Then I would recommend 16.04 if you are going to use Ubuntu at all. There ARE other distros, and some of them might make kernel/driver updates easier, but if you are going to use Ubuntu or Lubuntu then 16.04 is going to be your ticket.
 

reb0rn

Member
Dec 31, 2009
176
27
91
Does not really matter much from the viewpoint of a potential hacker. A hacker would typically short-sell the coin before the attack. The variety of potential attacks is also much larger compared to those, where the attacker directly benefits from "stolen" value.

Doing a roll-back (hard fork) will set much more dangerous precedent. It also sets the margin for acceptable errors in the code/contract higher, as error seem to be "easily" corrected by roll-backs.
There is almost no point in explaining, most ppl lured by greed will do anything to keep the price up and hype the shit out of broken concept!

If ethereum fork its the end of their primary model they promised!
Ethereum is a decentralized platform that runs smart contracts: applications that run exactly as programmed without any possibility of downtime, censorship, fraud or third party interference.
If it pass ethereum is dead, and all ppl defending hard fork to keep price for short time will know it too!
 

Elixer

Lifer
May 7, 2002
10,376
762
126
If ethereum fork its the end of their primary model they promised!
Ethereum is a decentralized platform that runs smart contracts: applications that run exactly as programmed without any possibility of downtime, censorship, fraud or third party interference.
If it pass ethereum is dead, and all ppl defending hard fork to keep price for short time will know it too! !
I would call this exploit fraud, so the initial premise is false.
If you allow fraud, then ethereum is worthless anyway.
It is the same as printing money on a home printer, and then the Feds don't do anything about it when that person goes to make purchases. Watch how fast the currency value will drop.
 

Madpacket

Platinum Member
Nov 15, 2005
2,068
326
126
Hey Madpacket, did you ever get your H97 running and how's that working for you. I just assembled my new system last night, and I can't seem to get it to recognize more than 4 GPUs regardless of what I try in a system with Win7-64. Configurations of slots and cards doesn't make a difference, I just always get an error 43 on the card in the last PCIe slot. I've run the 6xGPU mod, but that doesn't seem to do anything either.

Interestingly, I have 5 working in another system with an H97 and Win7-64. There's only 4 physical cards there though, one is a 7990.
Yes my socket 1150 H97 Anniversary board works great after some troubleshooting.

I was having issues with Windows detecting the cards at first but I solved detection this way.

- Stopped using the onboard GPU and Intel graphics.
- Added "one" Radeon GPU using a riser in the x16 slot, this acts a the primary card which I hook up to a monitor for setup.
- Removed older drivers using latest DDU and installed latest AMD beta drivers.
- Disabled crossfire.
- Ensured the female Molex on the board near the x16 slot was being fed power (this shouldn't have any effect if using powered risers but it seemed to help with detection).
- Added one card on a time and moved the PCIe adapters around if a card wasn't being detected.

Eventually all five cards worked. Now I think one my ports is dead, the third 1x PCIe port on my board doesn't respond to any card with a riser. I couldn't get any card to be recognized in that slot so it's possible you have a dead slot as well, but I wasn't even getting an error message.

I also helped a friend put together a similar rig with the same motherboard and CPU and the same refurbished Radeon cards (he's only using four instead of five cards). He's not having any issues with his board or never complained about card detection but I warned him about the issues I ran into. He's also using Windows 10.

Question, what's this 6x GPU mod you speak of? :)
 
Last edited:

ASK THE COMMUNITY