Author

Topic: (2) of (6) TI 1070s just slowed dow (Read 185 times)

full member
Activity: 1178
Merit: 131
August 24, 2018, 12:50:12 AM
#9
Remove problem cards and install in a different system and benchmark them there.  If performance is the same its an issue with the card, if they work fine, then its likely a software issue.
If its hardware, try cleaning the cards and using new thermal paste on the heatsink ( don't use cheap stuff here ).

Seems like the best advice.  I'll add though before you do that, swap the ports with those 2 cards to be sure its not something with the motherboard.  Also swap power connectors.
jr. member
Activity: 88
Merit: 7
August 24, 2018, 12:07:22 AM
#8
Remove problem cards and install in a different system and benchmark them there.  If performance is the same its an issue with the card, if they work fine, then its likely a software issue.
If its hardware, try cleaning the cards and using new thermal paste on the heatsink ( don't use cheap stuff here ).
copper member
Activity: 658
Merit: 101
Math doesn't care what you believe.
August 23, 2018, 05:05:13 PM
#7

Ethos.  Multiple power cycles and reboots prior to posting.
Did you tried to increase the cclock and mclock a little bit just try but still monitor the power draw.
It looks like someone taking your 10% hashrate or it might be other reasons why the hashrate drop by 10%.
Like what will happen to rx400 and rx 500 series.

Quote
Etherum mining hashrate of these series cards will drop by 30% in next couple of months due to DAG Epoch. 4GB cards will be greatly affected by Etherum DAG Epoch. For mining few GB of resource DATA is needed which sould be stored in GPU cache. DAG file size has increased overtime thus lowering the mining hasrate. Amount of block mined is called DAG Epoch. Each time we move to next Epoch DAG file become larger.

Fortunately doing zcash, not Etherum (not a big believer in Etherum in part for the reasons you mentioned).

I did try upgrading to EthOS 1.3.2 just in case there were security issues, but the box is firewalled off with no DNAT access to it - so unlikely its been hacked unless EthOS has been.

Hmmm.  Netstat is showing 2 ESTABLISHED connections, one to my pool, the other to 188.165.195.21:3443, which I'm hoping is the Ethos stat collection site, presuming they have one.

Odd that its only slowing down GPU #2 and GPU #4 (all the cards are EVGA Hybrid's - so water cooled - e.g. no fan issues to worry about (verified by the reported temps)

Just verified:  all cores are running at +200, mem at +700, pwr at 130 - last time I changed any of that was back in May per the configuration file datestamp.
legendary
Activity: 2016
Merit: 1030
Privacy is always important
August 23, 2018, 04:50:28 PM
#6
Are you using windows 10 and did you update it lately? this is sometimes happening to me if there is software updating that running on the background without notification it might be the cause of the hashrate drop.

Did you try to reboot the rig? try to reboot it may solve your issue.

Ethos.  Multiple power cycles and reboots prior to posting.
Did you tried to increase the cclock and mclock a little bit just try but still monitor the power draw.
It looks like someone taking your 10% hashrate or it might be other reasons why the hashrate drop by 10%.
Like what will happen to rx400 and rx 500 series.

Quote
Etherum mining hashrate of these series cards will drop by 30% in next couple of months due to DAG Epoch. 4GB cards will be greatly affected by Etherum DAG Epoch. For mining few GB of resource DATA is needed which sould be stored in GPU cache. DAG file size has increased overtime thus lowering the mining hasrate. Amount of block mined is called DAG Epoch. Each time we move to next Epoch DAG file become larger.
copper member
Activity: 658
Merit: 101
Math doesn't care what you believe.
August 23, 2018, 03:00:40 PM
#5
Are you using windows 10 and did you update it lately? this is sometimes happening to me if there is software updating that running on the background without notification it might be the cause of the hashrate drop.

Did you try to reboot the rig? try to reboot it may solve your issue.

Ethos.  Multiple power cycles and reboots prior to posting.
copper member
Activity: 658
Merit: 101
Math doesn't care what you believe.
August 23, 2018, 03:00:06 PM
#4
Was there anything that happened just before? Any software update of any kind?

No... last time I updated Ethos or the mining software was over a month ago.
legendary
Activity: 2016
Merit: 1030
Privacy is always important
August 23, 2018, 02:59:15 PM
#3
Are you using windows 10 and did you update it lately? this is sometimes happening to me if there is software updating that running on the background without notification it might be the cause of the hashrate drop.

Did you try to reboot the rig? try to reboot it may solve your issue.
newbie
Activity: 32
Merit: 0
August 23, 2018, 02:36:33 PM
#2
Was there anything that happened just before? Any software update of any kind?
copper member
Activity: 658
Merit: 101
Math doesn't care what you believe.
August 23, 2018, 01:53:08 PM
#1
I have a GPU rig that has been running for close to a year now.  Today, (2) of the (6) TI 1070s took a major plunge in effectiveness, dropping from ~500 sols/s to ~300 sols/s.  The other GPUs continue as before:

Code:
   GPU0  50C  85% |  504.3 Sol/s   508.8 Avg   272.7 I/s | 3.92 S/W  131 W |  2.56  100  124
   GPU1  48C  85% |  514.9 Sol/s   506.6 Avg   271.9 I/s | 3.91 S/W  128 W |  2.46  100  135
   GPU2  48C  85% |  339.1 Sol/s   339.4 Avg   176.7 I/s | 2.70 S/W  126 W |  2.25  100  126 +
   GPU3  50C  85% |  502.1 Sol/s   498.9 Avg   267.3 I/s | 3.85 S/W  133 W |  2.26  100  125 ++
   GPU4  49C  85% |  321.1 Sol/s   336.0 Avg   176.0 I/s | 2.72 S/W  119 W |  1.77  100  127
   GPU5  53C  85% |  489.9 Sol/s   490.5 Avg   265.6 I/s | 3.78 S/W  132 W |  3.19  100  131 +
   ============== | 2671.3 Sol/s  2680.2 Avg  1430.2 I/s | 3.48 S/W  771 W | 14.50  100  128

Note GPUs 2 and 4 in the above screenshot.  This represents about a 10% loss in mining efficiency for me.

Any suggestions on how to recover?
Jump to: