Hash rates are dropping after the first few minutes of a rig running

JayDDee

full member

Activity: 1424

Merit: 225

Quote from: rdluffy on June 24, 2020, 03:43:03 PM

I've mining for a couple of years, and with RX 470, 480, 570 and 580 we needed this plug to rig perform 100%

Have you seen these exact symptoms where GPUs with no attached monitor (or plug) start mining at speed
then slow down after a while? Was it solved by adding a plug?

If not it's a different problem.

I like a higher level of confidence before spending money to "try options". More testing (as I previously suggested) is free.

Edit: I found some discussion that the AMD drivers need one active monitor to initialize properly.
If headless, or the monitor is plugged into the IGPU or a Nvidia GPU a plug may ne necessary.

rdluffy

legendary

Activity: 2366

Merit: 1408

Quote from: JayDDee on June 24, 2020, 03:37:15 PM

Quote from: rdluffy on June 24, 2020, 01:57:21 PM

Those threads are old but the cards are the same, and the problem too,

Different problem, fail to boot or fail to start is not the same as reduced hashrate.

If you are correct it means the 480 requires a dummy plug while the 470 does not.
That's too much of a stretch without some kind of proof. Also why is this problem
only happening now? Both cards have been around for a few years.

"Sleep mode" is just speculation and isn't logical. Lack of a connected monitor is not a
valid reason to put the GPU to sleep.

I don't want to argue with you, I just don't want OP lead down the wrong path.

Oh, be cool, we are not arguing, just talking, but OP said he tried to attach a monitor and the card hashes correctly
I've mining for a couple of years, and with RX 470, 480, 570 and 580 we needed this plug to rig perform 100%

You can look for old threads, I'm really shocked that nobody here heard about these plugs, in 2017 this was very commom, I'm not crazy guys, I'm just an old miner who solved hundreds of problems with rigs Grin

I'm not speculating anything, try to explain to my old rigs that these plug are wothless Grin

And when you face a problem, why not just try options?

JayDDee

full member

Activity: 1424

Merit: 225

Quote from: rdluffy on June 24, 2020, 01:57:21 PM

Those threads are old but the cards are the same, and the problem too,

Different problem, fail to boot or fail to start is not the same as reduced hashrate.

If you are correct it means the 480 requires a dummy plug while the 470 does not.
That's too much of a stretch without some kind of proof. Also why is this problem
only happening now? Both cards have been around for a few years.

"Sleep mode" is just speculation and isn't logical. Lack of a connected monitor is not a
valid reason to put the GPU to sleep.

I don't want to argue with you, I just don't want OP lead down the wrong path.

rdluffy

legendary

Activity: 2366

Merit: 1408

Quote from: JayDDee on June 24, 2020, 01:54:08 PM

Those threads are old and deal with a different problem. In my experience connecting
a monitor will reduce the hashrate for the card. I have one systen right now with a RTX2060
that jumps 40% when the monitor shuts off.

If connecting a monitor helps in this case it could be a secondary effect. A dummy plug may provide
a workaround but I don't think it addresees the main problem.

Since "it worked before" the key is to find everything that changed between when it worked properly
and when the hash rate started dropping.

You have a RTX 2060, I'm not saying all rigs and cards should use dummy plugs, this was a problem only in RX cards, and OP said he have RX 480 8 GB
Those threads are old but the cards are the same, and the problem too, Nvidia works totally different

JayDDee

full member

Activity: 1424

Merit: 225

Those threads are old and deal with a different problem. In my experience connecting
a monitor will reduce the hashrate for the card. I have one systen right now with a RTX2060
that jumps 40% when the monitor shuts off.

If connecting a monitor helps in this case it could be a secondary effect. A dummy plug may provide
a workaround but I don't think it addresees the main problem.

Since "it worked before" the key is to find everything that changed between when it worked properly
and when the hash rate started dropping.

rdluffy

legendary

Activity: 2366

Merit: 1408

Hey I'm trying to help you but you are ignoring Cheesy

In the past we all used dummy plugs to keep the card "awake" and hashing 100%, some rigs need this plug, it's a problem that put your secondaries cards to "sleep"

You can read here and make your own plug ou buy that piece I showed

https://rumorscity.com/2013/12/06/how-to-create-dummy-plugs-for-your-graphics-cards/
https://bitcointalksearch.org/topic/do-you-use-hdmi-dummy-plug-for-your-mining-rig-1853279
https://bitcointalksearch.org/topic/dummy-plugs-multiple-gpus-11595
https://bitcointalksearch.org/topic/ethereum-gpu-mining-with-multiple-cards-need-dummy-plug-1957323

3dOOm

copper member

Activity: 157

Merit: 174

Member of Cryptios - https://cryptios.com/

Try running card one on a time direct to the mb and see if the hash rates changes, after that if possible try run both direct on mb. I had a similar problem back in the days, i had a gpu with problems( didn't found the problem) when was connected to the rig, decreased the hash rates tot all gpu's. After that if the hash rates doesn't change, try replace the risers.

adaseb

legendary

Activity: 3808

Merit: 1723

Quote from: velma on June 23, 2020, 04:44:44 PM

Quote from: adaseb on June 23, 2020, 04:30:51 PM

I've had this problem for years. It basically hashed at like 30MH/s about 95% of the time and then for some reason hashed at 18MH/s or so for 5% of the time. This was with RX 470 and been happening probably since 2017 or so.

Tried different bios, OS, voltages, fan speeds, etc and basically finally gave up trying to investigate why it happens. If someone has it figured out let me put an end to the mystery.

Was it Micron memory?

I'm trying lower overclock settings.

I think 80% or so of my cards were Hynix memory. Some were Elpida and some Samsung. However I actually never checked if the memory type made a difference. Maybe it was the memory type, it was buggy somewhat. No clue.

Since ETH is rallying and if it keeps on rallying past $300 I might boot up some of those rigs and mine again, however right now its not worth the hassle setting everything up when you are paying $0.10 kwh power.

velma

jr. member

Activity: 48

Merit: 1

Quote from: TheBottomTurtle on June 23, 2020, 05:03:33 PM

Quote from: velma on June 23, 2020, 04:23:56 PM

The cards work fine when plugged into a monitor.

Meaning the hashrate did not throttle when the card was plugged into a monitor? If so that would be pretty conclusive that rdluffy was on the right track. Using the dummy plugs in your cards that he suggest (or leaving the cards plugged into monitors) would solve the problem but in my opinion it would be better to find the low-power setting that is causing the cards to throttle, and turn it off. If by "The cards work fine when plugged into a monitor." you mean they produce a proper display but the hashrate continues to throttle you can rule out low-power "sleep" modes as your culprit.

When diagnosing problems like this it's a good idea to remember the cardinal rule of computers: Computers do exactly what you tell them to, not what you want them to. Somehow you've told your miner that it should be throttling down the cards. Maybe you did it explicitly through a setting you chose; more likely you did it implicitly through the software you've installed. Try flashing back to stock bios, using Win10, etc...Once you find a configuration that works see if you can reproduce your problem. That will tell you how to fix the problem in your desired configuration.

Okay. I understand what you mean. I need to get a cord to try that.

Lowering the overclock settings is not helping.

TheBottomTurtle

jr. member

Activity: 55

Merit: 7

Quote from: velma on June 23, 2020, 04:23:56 PM

The cards work fine when plugged into a monitor.

Meaning the hashrate did not throttle when the card was plugged into a monitor? If so that would be pretty conclusive that rdluffy was on the right track. Using the dummy plugs in your cards that he suggest (or leaving the cards plugged into monitors) would solve the problem but in my opinion it would be better to find the low-power setting that is causing the cards to throttle, and turn it off. If by "The cards work fine when plugged into a monitor." you mean they produce a proper display but the hashrate continues to throttle you can rule out low-power "sleep" modes as your culprit.

When diagnosing problems like this it's a good idea to remember the cardinal rule of computers: Computers do exactly what you tell them to, not what you want them to. Somehow you've told your miner that it should be throttling down the cards. Maybe you did it explicitly through a setting you chose; more likely you did it implicitly through the software you've installed. Try flashing back to stock bios, using Win10, etc...Once you find a configuration that works see if you can reproduce your problem. That will tell you how to fix the problem in your desired configuration.

velma

jr. member

Activity: 48

Merit: 1

Quote from: adaseb on June 23, 2020, 04:30:51 PM

I've had this problem for years. It basically hashed at like 30MH/s about 95% of the time and then for some reason hashed at 18MH/s or so for 5% of the time. This was with RX 470 and been happening probably since 2017 or so.

Tried different bios, OS, voltages, fan speeds, etc and basically finally gave up trying to investigate why it happens. If someone has it figured out let me put an end to the mystery.

Was it Micron memory?

I'm trying lower overclock settings.

adaseb

legendary

Activity: 3808

Merit: 1723

I've had this problem for years. It basically hashed at like 30MH/s about 95% of the time and then for some reason hashed at 18MH/s or so for 5% of the time. This was with RX 470 and been happening probably since 2017 or so.

Tried different bios, OS, voltages, fan speeds, etc and basically finally gave up trying to investigate why it happens. If someone has it figured out let me put an end to the mystery.

velma

jr. member

Activity: 48

Merit: 1

I'm watching the console and the hash rate will drop from 27 to 15 then back up to 18 then to 15 then up to 23. It just keeps changing. it starts doing this within the first 15 seconds and the console is indicating both cards are about 65°C. It doesn't seem to be a heat issue.

velma

jr. member

Activity: 48

Merit: 1

On SMOS, I set the target temperature to 75°C. For overclocking, I'm trying, 1050 Core, 1950 Memory and 950 Core undervolt.
So I changed Core undervolt from 900 to 950. It still doing the same thing.

The cards work fine when plugged into a monitor.

<Did you try?

If you don't have this plug, you can connect to another monitor, another TV, just to see if works >>

rdluffy

legendary

Activity: 2366

Merit: 1408

Quote from: velma on June 23, 2020, 01:32:26 PM

Thanks for the replies. I switched from Hive OS to SMOS. Hash rates are still fluctuating. The Watts are also changing. I played with the overclocking settings, but it didn't do much good. This is weird.

But what about the plug I suggested?
Did you try?

If you don't have this plug, you can connect to another monitor, another TV, just to see if works

TheBottomTurtle

jr. member

Activity: 55

Merit: 7

Checking for memory errors is always a good idea, and would explain the drop in hashrate, but not the corresponding drop in power consumption.

I'm not familiar with the details of Hive or SMOS, but given their reputation I would have to assume they have low-power mode disabled by default so I doubt the "sleep" mode mentioned above is the issue.

Did you check the bios settings you changed as I suggested earlier? If you set the max temp. to 66 it would do exactly what you are seeing; An easy mistake to make if you thought you were adjusting fan settings. It would also be worth your time to try INCREASING the voltage a bit. Undervolting too much can cause instability. It's possible that your cards can handle 900mv when idle and cold but not once they start to heat up. I find a lot of my polaris cards do best at 940mv.

velma

jr. member

Activity: 48

Merit: 1

Thanks for the replies. I switched from Hive OS to SMOS. Hash rates are still fluctuating. The Watts are also changing. I played with the overclocking settings, but it didn't do much good. This is weird.

rdluffy

legendary

Activity: 2366

Merit: 1408

It's unlikely to be a temperature problem

When I had a rig with RX 480, 570 and 580 I had to use some HDMI adaptors because cards used to go at a "sleep" mode, hashing very low or even 0mhs

This

When you plug this, your card will not go to sleep mode and will work as if it's connected to a monitor 100% of the time

rikuu

jr. member

Activity: 102

Merit: 2

It is also possible that your GPUs can't handle the memory straps and are producing errors. Check dmesg just in case. If you see any errors, reduce the memory speed or try different memory timings

velma

jr. member

Activity: 48

Merit: 1

I have two cards going right now. They started at 28 MH, but have now dropped to 16 MH, and have 76 W and 43 W usage.then when I refresh it, the 43 W usage changed to 79 W.it's still hashing out around 16 for both cards.

Topic: Hash rates are dropping after the first few minutes of a rig running (Read 326 times)