• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

Ethereum GPU mining?

Page 85 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
Erenhardt said:
I'm not sure what this table shows:
RlfhiOP.jpg

Page 1-15 in Users manual.

Doesn't it mean x1 slots are independent and olny x16 are shared and reduce speed if both used?

Also, there is a report of running 4GPUs:
http://www.coinminingrigs.com/budget-mining-rig/
Click to expand...

Sorry, I'm not sure what the IRQ assignments mean... do you have spare risers to test?
 
My second card dropped to 7MH and after increasing voltage to 14MH/s. It happens when you turn off the display.
It is MSI card and it will not let me go below 0.95V, while the sapphire is happy to go down to 0.9V.
After setting both voltages to 0.95 yesterday it no longer throttles. Both run 27MHs for 13hours now.

It looks like an error between Wattman -> bios

suklee said:
Sorry, I'm not sure what the IRQ assignments mean... do you have spare risers to test?
Click to expand...
Will have. Should be here today.

Edit: stuck in Post Office. So I guess tomorrow should be here.
 
Last edited:
Raduque said:
One more question, I am gonna be picking up another cheap 280x next month, would it be a good idea to buy a new modern CPU/Motherboard, or would it be alright to use a Gigabyte k8n-pro-sli board?
Click to expand...

How many cards are you running? Just two? Pretty much any old board will do for a setup like that.
 
Yes. I mine with 4 cards on some really old crap (790X and 790FX) so you should be fine as long as you have the PCI-e slots to drive those cards.
 
thilanliyan said:
^I always had that problem when I had a dedicated mining comp...but I only had to remote in rarely, so the 640x480 didn't bother me.
Click to expand...

640x480 wouldn't bother me if Radeon Settings can display consistently, but even with Chrome Remote Desktop I'm sometimes seeing a blank window with no monitor plugged in: https://www.dropbox.com/s/afmz8fcsllylb4w/Screenshot 2016-08-26 18.11.50.png?dl=0

Ordered some of those HDMI headless adapters recommended by n0x1ous, I hadn't seen this model before, only the 4K version.

PS. Still can't figure out how to fix the weird 4Mh/s bug, already tried DDU + reinstall of drivers.
 
Feld said:
I've been using D-Link DSP-W215 smart plugs. As long as the wi-fi is stable where the rig is set up (I keep my modem and router connected to a UPS), these things are nice for quickly checking on the total power consumption coming through the plug and turning the power on and off remotely. It also lets you monitor the total power consumed in your monthly billing cycle and put in your power cost to show the actual cost of the power consumed. There are other features like being able to schedule when it turns on and off, and temperature monitoring so it automatically powers down if the power socket overheats (you can define the power off temp). I've found it to be quite handy with a rig that likes to crash from time to time when I'm not around.
Click to expand...

How's the energy monitoring on the dlink - is it accurate?
 
Hey anyone here still running Hawaii cards in any significant quantity? I'm having weird calculated hashrate problems with nanopool on all my rigs, and I'm curious as to whether anyone else is having the same problem. It seems like I'm getting a lot of rejected shares now.
 
suklee said:
How's the energy monitoring on the dlink - is it accurate?
Click to expand...
I don't have a kill-a-watt to compare to, but it seems to be quite accurate for measuring power draw from the wall. Monthly totals are right in line with where it should be compared to my electric bill.
 
DrMrLordX said:
Hey anyone here still running Hawaii cards in any significant quantity? I'm having weird calculated hashrate problems with nanopool on all my rigs, and I'm curious as to whether anyone else is having the same problem. It seems like I'm getting a lot of rejected shares now.
Click to expand...
I'm running 6 Hawaii cards and they're going just as fast as ever. I switched from nanopool to alpereum some time ago though.
 
Elfear said:
I know you're probably pretty busy Madpacket, but any info on what you and your friend have discovered with the throttling issue would be very useful for the forum. 🙂
Click to expand...

Hey guys. My apologies. Still busy but I'll try to summarize.

The following is for 26.5Mh per card, you can go higher but in my experience the tradeoff during this heat isn't worth it. This will get you close to 160Mh stable per rig pulling 770 - 810W (+- 40W depending on ASIC rating) from the wall.

In Wattman for each card:

GPU, setting #7 change the following:

- set speed to 1000Mhz
- set GPU VDDC to 880mv

Mem:

Set to 2190Mhz @ 900mv

Fan settings:

Manually set max RPM to 3500
Manually set Min to 60 and Max to 80 for target temp.


Provide adequate airflow over crate. Keep highest ASIC rated cards at the ends of the crate for best air flow, lowest in middle.

With these settings you shouldn't throttle using Claymore 5.3 or higher. Tested over five crates.

Let me know if you still do. Good luck!
 
Feld said:
I'm running 6 Hawaii cards and they're going just as fast as ever. I switched from nanopool to alpereum some time ago though.
Click to expand...

Hmm.

I first noticed the problem when nanopool changed all their mining addresses following their support for ETC and SC. I updated addresses and the high number of rejected shares being reported by eth-proxy dropped to near-zero. About four days ago I noticed my calculated hashrate dropping to maybe 1/2 of what was being reported, and so as an experiment I'm taking a few of my cards off eth-proxy and into getwork mode to see if that fixes it. So far it looks like it has fixed the issue, so I'm wondering if nanopool isn't having some kind of problem with eth-proxy users or what.
 
My card had lost 100mhz somewhere - had been stuck at 970mhz. Rebooted the machine, started mining, driver crashed. Had set the clock to almost 1.3ghz, oops! Defaulted the clock, rebooted again, and was doing 17.8Mh/s. Overclocked my 100mhz to 1170 and picked up 1 Mh/s. Doing 18.6 now.
 
Yakk said:
I believe Nanopool was (maybe still is) under DDoS attack.
Click to expand...

It seems like it might be. They're using CloudFlare DDoS protection.

Raduque said:
My card had lost 100mhz somewhere - had been stuck at 970mhz. Rebooted the machine, started mining, driver crashed. Had set the clock to almost 1.3ghz, oops! Defaulted the clock, rebooted again, and was doing 17.8Mh/s. Overclocked my 100mhz to 1170 and picked up 1 Mh/s. Doing 18.6 now.
Click to expand...

That's odd. Driver crashes can cause odd behavior sometimes.
 
Madpacket said:
Hey guys. My apologies. Still busy but I'll try to summarize.

The following is for 26.5Mh per card, you can go higher but in my experience the tradeoff during this heat isn't worth it. This will get you close to 160Mh stable per rig pulling 770 - 810W (+- 40W depending on ASIC rating) from the wall.

In Wattman for each card:

GPU, setting #7 change the following:

- set speed to 1000Mhz
- set GPU VDDC to 880mv

Mem:

Set to 2190Mhz @ 900mv

Fan settings:

Manually set max RPM to 3500
Manually set Min to 60 and Max to 80 for target temp.


Provide adequate airflow over crate. Keep highest ASIC rated cards at the ends of the crate for best air flow, lowest in middle.

With these settings you shouldn't throttle using Claymore 5.3 or higher. Tested over five crates.

Let me know if you still do. Good luck!
Click to expand...

Thanks! After 5min it's working well so far. 5 of my GPUs are mining at 26.3MH/s and one is at 24.7MH/s for some reason. I'll watch the mining rig over a few hours to see if the throttling is gone.
 
I picked up an 8gb Sapphire Nitro 470 and got it up and running at 25.1 MH with clocks of 1150/2100. The ASIC rating is 85.8% which is the highest of any card I've ever owned - are such high ratings common with Polaris cards? I might be able to reduce the core clocks a little but those settings are 90% saturating the memory controller so I figured it's good enough for the time being. Wattman is capped at 2100 for the memory, and the Asus GPU Tweak II tool flat out won't launch with the card connected so I can't go that route to try raising the limit.

This card replaced a Sapphire Tri-X 290x in the rig that was always overheating and has a dying fan that rattles a lot. I was running that 290x underclocked and getting about 27 MH with it. Now for the minor loss of 2 MH the rig is drawing 140 fewer watts from the wall and is a lot less noisy. Seems like a good tradeoff.
 
Elfear said:
Thanks! After 5min it's working well so far. 5 of my GPUs are mining at 26.3MH/s and one is at 24.7MH/s for some reason. I'll watch the mining rig over a few hours to see if the throttling is gone.
Click to expand...

Double check in Wattman if all settings stuck. Also if some of the cards are still throttling check the ASIC ratings. The higher they are, the worse they throttle. I would drop down the MHz 10 at a time until throttling subsides (try core VDDC first, then memory).

Also reboot after Wattman changes.

Sometimes after making the changes I still experienced some throttling unless I reboot.

Wattman is buggy. Good luck.
 
Last edited:
Ni
Feld said:
I picked up an 8gb Sapphire Nitro 470 and got it up and running at 25.1 MH with clocks of 1150/2100. The ASIC rating is 85.8% which is the highest of any card I've ever owned - are such high ratings common with Polaris cards? I might be able to reduce the core clocks a little but those settings are 90% saturating the memory controller so I figured it's good enough for the time being. Wattman is capped at 2100 for the memory, and the Asus GPU Tweak II tool flat out won't launch with the card connected so I can't go that route to try raising the limit.

This card replaced a Sapphire Tri-X 290x in the rig that was always overheating and has a dying fan that rattles a lot. I was running that 290x underclocked and getting about 27 MH with it. Now for the minor loss of 2 MH the rig is drawing 140 fewer watts from the wall and is a lot less noisy. Seems like a good tradeoff.
Click to expand...

Nice upgrade for mining.

Not sure about aftermarket cards but with reference models the higher rated ASIC cards generally perform worse than the lower rated ones.

By default the higher ASIC rated cards tend to have more conservative default VDDC set on the GPU core. Since they "leak" less they run a bit hotter and for some reason consume a tad more energy.

I don't know if they overclock better with less voltage applied (probably because of less leakage?) or really what advantages higher rated ASICS are supposed to have, but for mining you want lower rated cards in my experience.

Hopefully you find a way to get the memory to overclock, that's where you'll get the biggest gains. 25Mh isn't bad though and totally worth the tradeoff over your 290X.
 
Last edited:
Madpacket said:
Ni


Nice upgrade for mining.

Not sure about aftermarket cards but with reference models the higher rated ASIC cards generally perform worse than the lower rated ones.

By default the higher ASIC rated cards tend to have more conservative default VDDC set on the GPU core. Since they "leak" less they run a bit hotter and for some reason consume a tad more energy.

I don't know if they overclock better with less voltage applied (probably because of less leakage?) or really what advantages higher rated ASICS are supposed to have, but for mining you want lower rated cards in my experience.

Hopefully you find a way to get the memory to overclock, that's where you'll get the biggest gains. 25Mh isn't bad though and totally worth the tradeoff over your 290X.
Click to expand...
Thanks! I used your guide from a few pages back to set it up and it worked like a breeze. The only thing I really did different was set the core clock a bit higher since the 470 is a bit cut down from the 480. FYI, the default idle voltage was 0.825 and 1.0875 at load according to GPUZ. All your suggested voltage changes worked fine, even for the core at 1150 MHz. It runs super cool and quiet even under load, with no throttling seen in the first 24 hours.

I'm sure it's just a matter of time before Asus, MSI, or Sapphire update their utilities to allow better overclocking on the memory. I'll be on the lookout for updates from them, but in the meantime I can't complain about what I've got.
 
You must log in or register to reply here.

TRENDING THREADS

Back
Top