philipma1957
Legendary
Offline
Activity: 4116
Merit: 7851
'The right to privacy matters'
|
|
June 08, 2015, 12:45:00 PM |
|
Good idea. Speaking of S5's - do you know if I can run 2 on a single EVGA SuperNOVA 1300G2 or Seasonic 1250? We have an investor willing to fund a pretty significant hardware purchase, so I've been in touch with Bitmain Tech and they have been very helpful with info.
I'm running my S5s, two per power supply, on a combination of EVGA SuperNOVA 1300G2s and Antec HCP-1300 Platinum 1300W power supplies. They work well on those power supplies, are decently efficient and the PS stays cool (including cool wires) even if I overclock the S5s a bit. My S5s are consistently in the 1.3TH/s to 1.35TH/s range with a slight overtune on these PS units and no issues at all. Denis Thank you! Exactly the info I was looking for. Thanks! Would you recommend one of those over the other or have they both performed similarly? Any experience with the Seasonic 1250? I have not used that model, but Seasonic is a top notch company on PSU's so very good. Both he mentioned are good. Honestly it depends on price if you can get the Antec he mentioned close to the EVGA I would go for the Antec ad get the platinum. But decent pricing differen then I would go with EVGA. Got it - thanks. The EVGA one is $169.99 now @newegg after rebate. Seems like a great price - I just wish it had 8 PCI-E connections like the Seasonic. I have used: Seasonic 1200 plat Seasonic 1000 plat Seasonic 750 plat I have used Evga 1300 g2 I have used Evga 1600 p2 And a lot of others. the evga 1300 g2 is good gear. it has 6 pcie cables 4 with 1 plug and 2 with 2 plugs you can do 2 single cables and 1 double cable to an s-5. use a single left a single right and use a double on the left and right. run at stock speed of 1150 then do the second s-5 the same way. Is this what you were suggesting? Does it look like I did it correctly? The double cables are plugged into the back of the units. My hardware error rate is a bit high, so either the PSU is struggling or I didn't setup the cables the right away. that is the correct way to do it. next question is how high is error rate? .5% or less? and a suggestion or two: 1) do not point the s-5 hot air at the evga 1300 g2 2) if you are mining the s-5's at stock 1150gh try setting the freq 1 tick lower and mine around 1100gh to 1125gh. Once you do the small under clock report back on error rate should be better with a slight underclock. 2 s-5's on the evga 1300g2 are close to max so a slight under clock should improve error rate.
|
|
|
|
|
|
|
|
The forum strives to allow free discussion of any ideas. All policies are built around this principle. This doesn't mean you can post garbage, though: posts should actually contain ideas, and these ideas should be argued reasonably.
|
|
|
Advertised sites are not endorsed by the Bitcoin Forum. They may be unsafe, untrustworthy, or illegal in your jurisdiction.
|
amelen (OP)
|
|
June 08, 2015, 01:35:28 PM |
|
Good idea. Speaking of S5's - do you know if I can run 2 on a single EVGA SuperNOVA 1300G2 or Seasonic 1250? We have an investor willing to fund a pretty significant hardware purchase, so I've been in touch with Bitmain Tech and they have been very helpful with info.
I'm running my S5s, two per power supply, on a combination of EVGA SuperNOVA 1300G2s and Antec HCP-1300 Platinum 1300W power supplies. They work well on those power supplies, are decently efficient and the PS stays cool (including cool wires) even if I overclock the S5s a bit. My S5s are consistently in the 1.3TH/s to 1.35TH/s range with a slight overtune on these PS units and no issues at all. Denis Thank you! Exactly the info I was looking for. Thanks! Would you recommend one of those over the other or have they both performed similarly? Any experience with the Seasonic 1250? I have not used that model, but Seasonic is a top notch company on PSU's so very good. Both he mentioned are good. Honestly it depends on price if you can get the Antec he mentioned close to the EVGA I would go for the Antec ad get the platinum. But decent pricing differen then I would go with EVGA. Got it - thanks. The EVGA one is $169.99 now @newegg after rebate. Seems like a great price - I just wish it had 8 PCI-E connections like the Seasonic. I have used: Seasonic 1200 plat Seasonic 1000 plat Seasonic 750 plat I have used Evga 1300 g2 I have used Evga 1600 p2 And a lot of others. the evga 1300 g2 is good gear. it has 6 pcie cables 4 with 1 plug and 2 with 2 plugs you can do 2 single cables and 1 double cable to an s-5. use a single left a single right and use a double on the left and right. run at stock speed of 1150 then do the second s-5 the same way. Is this what you were suggesting? Does it look like I did it correctly? The double cables are plugged into the back of the units. My hardware error rate is a bit high, so either the PSU is struggling or I didn't setup the cables the right away. that is the correct way to do it. next question is how high is error rate? .5% or less? and a suggestion or two: 1) do not point the s-5 hot air at the evga 1300 g2 2) if you are mining the s-5's at stock 1150gh try setting the freq 1 tick lower and mine around 1100gh to 1125gh. Once you do the small under clock report back on error rate should be better with a slight underclock. 2 s-5's on the evga 1300g2 are close to max so a slight under clock should improve error rate. Thank you! Yes, it was only pointed that way for the picture. Error rate is ~0.5%, but I can't tell if it's power related. I guess I'll play around with it - thanks!
|
|
|
|
philipma1957
Legendary
Offline
Activity: 4116
Merit: 7851
'The right to privacy matters'
|
|
June 08, 2015, 01:43:54 PM |
|
Good idea. Speaking of S5's - do you know if I can run 2 on a single EVGA SuperNOVA 1300G2 or Seasonic 1250? We have an investor willing to fund a pretty significant hardware purchase, so I've been in touch with Bitmain Tech and they have been very helpful with info.
I'm running my S5s, two per power supply, on a combination of EVGA SuperNOVA 1300G2s and Antec HCP-1300 Platinum 1300W power supplies. They work well on those power supplies, are decently efficient and the PS stays cool (including cool wires) even if I overclock the S5s a bit. My S5s are consistently in the 1.3TH/s to 1.35TH/s range with a slight overtune on these PS units and no issues at all. Denis Thank you! Exactly the info I was looking for. Thanks! Would you recommend one of those over the other or have they both performed similarly? Any experience with the Seasonic 1250? I have not used that model, but Seasonic is a top notch company on PSU's so very good. Both he mentioned are good. Honestly it depends on price if you can get the Antec he mentioned close to the EVGA I would go for the Antec ad get the platinum. But decent pricing differen then I would go with EVGA. Got it - thanks. The EVGA one is $169.99 now @newegg after rebate. Seems like a great price - I just wish it had 8 PCI-E connections like the Seasonic. I have used: Seasonic 1200 plat Seasonic 1000 plat Seasonic 750 plat I have used Evga 1300 g2 I have used Evga 1600 p2 And a lot of others. the evga 1300 g2 is good gear. it has 6 pcie cables 4 with 1 plug and 2 with 2 plugs you can do 2 single cables and 1 double cable to an s-5. use a single left a single right and use a double on the left and right. run at stock speed of 1150 then do the second s-5 the same way. Is this what you were suggesting? Does it look like I did it correctly? The double cables are plugged into the back of the units. My hardware error rate is a bit high, so either the PSU is struggling or I didn't setup the cables the right away. that is the correct way to do it. next question is how high is error rate? .5% or less? and a suggestion or two: 1) do not point the s-5 hot air at the evga 1300 g2 2) if you are mining the s-5's at stock 1150gh try setting the freq 1 tick lower and mine around 1100gh to 1125gh. Once you do the small under clock report back on error rate should be better with a slight underclock. 2 s-5's on the evga 1300g2 are close to max so a slight under clock should improve error rate. Thank you! Yes, it was only pointed that way for the picture. Error rate is ~0.5%, but I can't tell if it's power related. I guess I'll play around with it - thanks! each s-5 and each evga 1300 g2 vary a little bit. so your evga 1300 g2 matched with your 2 s-5's may work better with a slight under clock lets say 1100gh. also let the s-5's run for 4 hours or so before you decide the error rates are correct. 2200gh is about 159 usd a month 2300gh is about 166 usd a month but if the error rate drops from .5 to .1 that 7 dollar income loss is more like 5 or 6. and the gear does not break as fast. good luck
|
|
|
|
amelen (OP)
|
|
June 08, 2015, 02:03:07 PM |
|
Good idea. Speaking of S5's - do you know if I can run 2 on a single EVGA SuperNOVA 1300G2 or Seasonic 1250? We have an investor willing to fund a pretty significant hardware purchase, so I've been in touch with Bitmain Tech and they have been very helpful with info.
I'm running my S5s, two per power supply, on a combination of EVGA SuperNOVA 1300G2s and Antec HCP-1300 Platinum 1300W power supplies. They work well on those power supplies, are decently efficient and the PS stays cool (including cool wires) even if I overclock the S5s a bit. My S5s are consistently in the 1.3TH/s to 1.35TH/s range with a slight overtune on these PS units and no issues at all. Denis Thank you! Exactly the info I was looking for. Thanks! Would you recommend one of those over the other or have they both performed similarly? Any experience with the Seasonic 1250? I have not used that model, but Seasonic is a top notch company on PSU's so very good. Both he mentioned are good. Honestly it depends on price if you can get the Antec he mentioned close to the EVGA I would go for the Antec ad get the platinum. But decent pricing differen then I would go with EVGA. Got it - thanks. The EVGA one is $169.99 now @newegg after rebate. Seems like a great price - I just wish it had 8 PCI-E connections like the Seasonic. I have used: Seasonic 1200 plat Seasonic 1000 plat Seasonic 750 plat I have used Evga 1300 g2 I have used Evga 1600 p2 And a lot of others. the evga 1300 g2 is good gear. it has 6 pcie cables 4 with 1 plug and 2 with 2 plugs you can do 2 single cables and 1 double cable to an s-5. use a single left a single right and use a double on the left and right. run at stock speed of 1150 then do the second s-5 the same way. Is this what you were suggesting? Does it look like I did it correctly? The double cables are plugged into the back of the units. My hardware error rate is a bit high, so either the PSU is struggling or I didn't setup the cables the right away. that is the correct way to do it. next question is how high is error rate? .5% or less? and a suggestion or two: 1) do not point the s-5 hot air at the evga 1300 g2 2) if you are mining the s-5's at stock 1150gh try setting the freq 1 tick lower and mine around 1100gh to 1125gh. Once you do the small under clock report back on error rate should be better with a slight underclock. 2 s-5's on the evga 1300g2 are close to max so a slight under clock should improve error rate. Thank you! Yes, it was only pointed that way for the picture. Error rate is ~0.5%, but I can't tell if it's power related. I guess I'll play around with it - thanks! each s-5 and each evga 1300 g2 vary a little bit. so your evga 1300 g2 matched with your 2 s-5's may work better with a slight under clock lets say 1100gh. also let the s-5's run for 4 hours or so before you decide the error rates are correct. 2200gh is about 159 usd a month 2300gh is about 166 usd a month but if the error rate drops from .5 to .1 that 7 dollar income loss is more like 5 or 6. and the gear does not break as fast. good luck Thanks for the info! What's an acceptable error rate?
|
|
|
|
amelen (OP)
|
|
June 08, 2015, 06:33:00 PM |
|
Here are how the error rates look at each frequency:
400 0.060% 375 0.040% 350 0.030%
Temp's seem to be around the same. I guess 375 is the best one to leave it at?
|
|
|
|
chalkboard17
|
|
June 08, 2015, 08:48:39 PM |
|
Here are how the error rates look at each frequency:
400 0.060% 375 0.040% 350 0.030%
Temp's seem to be around the same. I guess 375 is the best one to leave it at?
I have same miner and same PSU. Unfortunately if you use the two cords of the double cable you will get more hardware error. At least that's what happened to me. If you really are looking for extreme efficiency it could be a idea to use 6 plugs (leave the two extras alone) instead of 8. In the end, you will have to do tests yourself as it depends on other things, too. Leave it on for hours and go figuring what is best for you. Change the cables, use other schemes such as 3 normal cables and 1 that is doubled and so on. My HW at 412.5 right now is 0.0045%. 1 miner, 1 psu, 4 normal cables no double on. 1360gh/s. I can get 1380gh/s with 425 and after that it starts to drop. From your HW I am going to guess after 387.5 you should start to get lower hashrate.
|
|
|
|
amelen (OP)
|
|
June 08, 2015, 08:52:48 PM |
|
Here are how the error rates look at each frequency:
400 0.060% 375 0.040% 350 0.030%
Temp's seem to be around the same. I guess 375 is the best one to leave it at?
I have same miner and same PSU. Unfortunately if you use the two cords of the double cable you will get more hardware error. At least that's what happened to me. If you really are looking for extreme efficiency it could be a idea to use 6 plugs (leave the two extras alone) instead of 8. In the end, you will have to do tests yourself as it depends on other things, too. Leave it on for hours and go figuring what is best for you. Change the cables, use other schemes such as 3 normal cables and 1 that is doubled and so on. My HW at 412.5 right now is 0.0045%. 1 miner, 1 psu, 4 normal cables no double on. 1360gh/s. I can get 1380gh/s with 425 and after that it starts to drop. From your HW I am going to guess after 387.5 you should start to get lower hashrate. Great ideas, thanks! I'm also trying to figure out some temp issues. One unit is running 60/60 right now, one at 64/51. Same exact fans - very strange. The unit at 64/51 is getting much higher hardware error rates.
|
|
|
|
chalkboard17
|
|
June 08, 2015, 09:05:44 PM |
|
I also have temp diff between boards. You and I aren't the first people I see running into these problems lately. It's not normal, despite what many people will tell you.
|
|
|
|
amelen (OP)
|
|
June 08, 2015, 09:07:07 PM |
|
I also have temp diff between boards. You and I aren't the first people I see running into these problems lately. It's not normal, despite what many people will tell you.
Strangely enough, I only started seeing the temp differences recently after swapping fans - I'll have to put the stock fan back on for further testing. I'm also not sure how accurate the temp readings are anyway since it's only 1 spot on the board. What difference in temp's are you seeing?
|
|
|
|
chalkboard17
|
|
June 08, 2015, 09:08:50 PM |
|
Air conditioner + ceiling fan + night + nearly winter in southern hemisphere: 412.5 @ 1356 @ 55/47
|
|
|
|
amelen (OP)
|
|
June 08, 2015, 09:33:19 PM |
|
Air conditioner + ceiling fan + night + nearly winter in southern hemisphere: 412.5 @ 1356 @ 55/47
Temp's aren't too bad - although the spread is almost as bad as mine. That's with a stock fan?
|
|
|
|
philipma1957
Legendary
Offline
Activity: 4116
Merit: 7851
'The right to privacy matters'
|
|
June 08, 2015, 10:00:32 PM |
|
Here are how the error rates look at each frequency:
400 0.060% 375 0.040% 350 0.030%
Temp's seem to be around the same. I guess 375 is the best one to leave it at?
yeah the 375 looks pretty good for your machine
|
|
|
|
chalkboard17
|
|
June 08, 2015, 10:02:26 PM |
|
Air conditioner + ceiling fan + night + nearly winter in southern hemisphere: 412.5 @ 1356 @ 55/47
Temp's aren't too bad - although the spread is almost as bad as mine. That's with a stock fan? Yes. Arrived like that to me. Bought directly from bitmain.
|
|
|
|
philipma1957
Legendary
Offline
Activity: 4116
Merit: 7851
'The right to privacy matters'
|
|
June 08, 2015, 11:48:20 PM |
|
try setting the units on a thin mat. i had some silicone sheets.
i set the silicone sheet onto a flat floor then the s-5 onto the sheet.
thus all air pushes through the s-5 no air leaks out of the bottom sides.
some of that air goes up to the top giving 1 or deg cooler.
as long as the boards are under 65 it seems to be okay.
i got 58-62 with good errors
|
|
|
|
amelen (OP)
|
|
June 09, 2015, 03:24:07 PM |
|
try setting the units on a thin mat. i had some silicone sheets.
i set the silicone sheet onto a flat floor then the s-5 onto the sheet.
thus all air pushes through the s-5 no air leaks out of the bottom sides.
some of that air goes up to the top giving 1 or deg cooler.
as long as the boards are under 65 it seems to be okay.
i got 58-62 with good errors
What would you define as good errors? I'm getting 0.05% now, is that considered acceptable?
|
|
|
|
philipma1957
Legendary
Offline
Activity: 4116
Merit: 7851
'The right to privacy matters'
|
|
June 10, 2015, 04:03:47 AM |
|
try setting the units on a thin mat. i had some silicone sheets.
i set the silicone sheet onto a flat floor then the s-5 onto the sheet.
thus all air pushes through the s-5 no air leaks out of the bottom sides.
some of that air goes up to the top giving 1 or deg cooler.
as long as the boards are under 65 it seems to be okay.
i got 58-62 with good errors
What would you define as good errors? I'm getting 0.05% now, is that considered acceptable? yes it is good. just watch it to see if it declines . if it drifts to 0.06 or 0.07 % no worries. If it goes to 0.25% try a slight under clock. there is no big worry unless they go to 1 or 2%
|
|
|
|
amelen (OP)
|
|
June 11, 2015, 02:37:37 PM |
|
try setting the units on a thin mat. i had some silicone sheets.
i set the silicone sheet onto a flat floor then the s-5 onto the sheet.
thus all air pushes through the s-5 no air leaks out of the bottom sides.
some of that air goes up to the top giving 1 or deg cooler.
as long as the boards are under 65 it seems to be okay.
i got 58-62 with good errors
What would you define as good errors? I'm getting 0.05% now, is that considered acceptable? yes it is good. just watch it to see if it declines . if it drifts to 0.06 or 0.07 % no worries. If it goes to 0.25% try a slight under clock. there is no big worry unless they go to 1 or 2% Thank you! Will do further testing with some new fans today. I've gotten it pretty stable at 0.05-0.06%. One of the tricks, especially with 2 S5's sharing a single PSU (EVGA G2 1300) has been to put the fans on separate connections to the PSU instead of running off the S5's themselves. That way they are not sharing the power over the same cables. This helped errors come down a good amount - especially with the stock fans that are rated at 2amps each.
|
|
|
|
philipma1957
Legendary
Offline
Activity: 4116
Merit: 7851
'The right to privacy matters'
|
|
July 06, 2015, 02:46:59 AM |
|
try setting the units on a thin mat. i had some silicone sheets.
i set the silicone sheet onto a flat floor then the s-5 onto the sheet.
thus all air pushes through the s-5 no air leaks out of the bottom sides.
some of that air goes up to the top giving 1 or deg cooler.
as long as the boards are under 65 it seems to be okay.
i got 58-62 with good errors
What would you define as good errors? I'm getting 0.05% now, is that considered acceptable? yes it is good. just watch it to see if it declines . if it drifts to 0.06 or 0.07 % no worries. If it goes to 0.25% try a slight under clock. there is no big worry unless they go to 1 or 2% Thank you! Will do further testing with some new fans today. I've gotten it pretty stable at 0.05-0.06%. One of the tricks, especially with 2 S5's sharing a single PSU (EVGA G2 1300) has been to put the fans on separate connections to the PSU instead of running off the S5's themselves. That way they are not sharing the power over the same cables. This helped errors come down a good amount - especially with the stock fans that are rated at 2amps each. how has your gear worked for you over the last few weeks I figured I would follow up.
|
|
|
|
MarkAz
|
|
July 06, 2015, 04:20:37 AM |
|
Great ideas, thanks! I'm also trying to figure out some temp issues. One unit is running 60/60 right now, one at 64/51. Same exact fans - very strange. The unit at 64/51 is getting much higher hardware error rates.
I have 24 S5's that I use as my test setup, and on average I see a 1-2c difference between blades, no real rhyme or reason towards which side is hotter, etc. In almost all cases, if I see a big temperature variance, then I remove the heat sink from the high temperature board, re-apply thermal paste, and this brings them right back in line. Since you just bought yours, if you want to do this make sure to run it by Bitmain before doing so, so it doesn't invalidate your warranty.
|
|
|
|
amelen (OP)
|
|
July 06, 2015, 02:53:41 PM |
|
try setting the units on a thin mat. i had some silicone sheets.
i set the silicone sheet onto a flat floor then the s-5 onto the sheet.
thus all air pushes through the s-5 no air leaks out of the bottom sides.
some of that air goes up to the top giving 1 or deg cooler.
as long as the boards are under 65 it seems to be okay.
i got 58-62 with good errors
What would you define as good errors? I'm getting 0.05% now, is that considered acceptable? yes it is good. just watch it to see if it declines . if it drifts to 0.06 or 0.07 % no worries. If it goes to 0.25% try a slight under clock. there is no big worry unless they go to 1 or 2% Thank you! Will do further testing with some new fans today. I've gotten it pretty stable at 0.05-0.06%. One of the tricks, especially with 2 S5's sharing a single PSU (EVGA G2 1300) has been to put the fans on separate connections to the PSU instead of running off the S5's themselves. That way they are not sharing the power over the same cables. This helped errors come down a good amount - especially with the stock fans that are rated at 2amps each. how has your gear worked for you over the last few weeks I figured I would follow up. I wound up going back to the stock fans but under-volting them to reduce power usage and help with the sound. The error rate is still not great, but seems to be PSU related. The EVGA 1300 is capable of running 2 S5's, but it definitely starts to strain even with a slight over clocking. I'm still playing around with some designs but also waiting to see what the specs on the S7 will look like.
|
|
|
|
|