Claymore's Dual Ethereum AMD+NVIDIA GPU Miner v15.0 (Windows/Linux)

[citronick]

So my 1300 watt evga 6 pin won't supply enough power to my 480's? I could understand if your using a not powerful enough power supply or not a single rail because that card should be pulling as much as it can straight from the 6 pin correct?

Just seems weird as I would feel like it would keep trying to pull from the 6pin and that would burn out that connection or kick off the power supply..

I just don't understand why it's not burning up the pci slot as it's only rated at like 75 watts or something correct or depending on the board?

Are the powered risers connected straight to the power supply?

6 pin pci-e is rated for 75w, mboard pci-e slot is rated for 75w while the rx 480 is rated for 150w it seems to be drawing slightly more than this @ stock settings. If only Amd had used an 8pin pci-e (rated for 150w) connector it wouldnt be an issue.
Yes it can depend on the board some manufactures design with higher tolerances than others, it may also depend on cost cheaper models for example might has less tolerance to help reduce cost.
Powered risers can be connected via molex directly into the riser board or often are used with sata to molex tails that commonly ship with most usb risers.

I also have a rig with 6 x Nanos OC to 1040/300 with Heliox ROM Mod hashing 175MHs, with same H81 mobo setup. Both are molex and risers powered from PSU. Nanos have same 6-pin connector like the 480s. This Nano rig is also functioning well and hashing for more than a month.

[osnwt]

The 390s draw much much more power than the 480s.

Not that much, actually.

If you looked at the youtube video by TecLab, they measured every supply source for 480 card (PCI-e 12V and 5V, extra 12V) and found that with some GPU tests (and mining is like hardest of them) the 480 card draws around 126W from 12V PCI-e, 92W from extra 12V + some from 5V PCI-e. In total, they had up to 255W in tests instead of promised 150 (75+75), including 65% more from PCI-e 12V than standard allows. No wonder why they burn motherboards.

In that case I see absolutely no sense replacing 390 by 480.

See this with subtitles:
https://www.youtube.com/watch?v=nKcHR1qW3w4

[dumpida]

Were you using powered USB risers? Was any of the card connected directly to the motherboard?

I believe he said all gpus were connected directly to the motherboard. In this case with the rx 480 I doubt powered riser would help much anyway because the card would draw too much power thru the riser and end up melting the molex connected to the riser (which is admittedly preferable to melting the atx connection on your motherboard). Still it seems clear the 6 pin pci-e connector on reference rx 480 is insufficient, I will be waiting for aftermarket cards before venturing into mining with polaris.

You need the Asrock H81 pro btc boards, they use two extra molex connectors on the mother board, and do not use the USB riser which does not connect to the PCIE slot +12V.

[citronick]

The 390s draw much much more power than the 480s.

Not that much, actually.

If you looked at the youtube video by TecLab, they measured every supply source for 480 card (PCI-e 12V and 5V, extra 12V) and found that with some GPU tests (and mining is like hardest of them) the 480 card draws around 126W from 12V PCI-e, 92W from extra 12V + some from 5V PCI-e. In total, they had up to 255W in tests instead of promised 150 (75+75), including 65% more from PCI-e 12V than standard allows. No wonder why they burn motherboards.

In that case I see absolutely no sense replacing 390 by 480.

See this with subtitles:
https://www.youtube.com/watch?v=nKcHR1qW3w4

Looks like 480s for mining will need to run only on stock clocks and should not be overclocked because of this power overdraw risk. Stock settings only gives 24.5MHs for 480, while 390 can deliver 26-27MHs, without OC, and the 390 and Nanos are very overclockable.

My 480 on PCI slot is running stable mining now on Day 2 without doing any OC & undervolting.

[osnwt]

Hi, is there a way to monitor rejected status , I got a lot yesterday. It just need simple stop and start claymore to recover.

Current CDM version is useless for rejects monitoring, it restarts if you have more than 50% of rejects, and if you run fine for 10 days you can easily have 5 days of rejects in a raw. Claymore agreed that it should use another method and I hope he will implement it.

As for the rejects, go away from DwarfPool, it went nuts last weeks with rejects. There are reports that any miners may suddenly start getting rejects with no reason. DwarfPool advocates usually say it gives max profit of all other pools. I switched to ethermine.org and worked for 2 days there. And I definitely may tell you that profit is EXACTLY the same. For my 272MHs rigs I had this on DwarfPool per day: 2.03, 2.05, 2.01, 1.85, 1.87 (WTF?), 2.01.... Plus few hours of rejects with zero income. On Ethermine I have 2.05 per day with no glitches, stats recalculations, rejects etc. And even with 1% fee (Dwarf had 0% until July to compensate some glitches). Ethermine may estimate lesser profit in Ether than DwarfPool shows per day, but in fact I have exactly the same. Keep mining @ Ethermine right now.

[merc84]

... that means powered risers only will not be enough if you OC? Likely need molex powered too, in case of H81 mobo.

I am guessing probably that's why my rig has no issues because the risers are powered via SATA from PSU. And the ASrock H81 Pro BTC board has also molex powered drawing from PSU. With both of them powered, it will give enough juice to 480 when OC.

Meanwhile, with same mobo setup, I have 5 x 390s OC to 1050Mhz/1125MHz giving me 30.5MHs per card, and no issues with that too. Both risers and molex are powered. The 390s draw much much more power than the 480s.

The reason the 390s are fine is because they have sufficient power from the pci-e connectors on the card.

You need the Asrock H81 pro btc boards, they use two extra molex connectors on the mother board, and do not use the USB riser which does not connect to the PCIE slot +12V.

How to tell if the riser u buy connects to the pcie slot +12V? I believe these btc boards were designed when unpowered ribbon risers were the norm, these additional molex connections should be redundant when using powered risers.

I also have a rig with 6 x Nanos OC to 1040/300 with Heliox ROM Mod hashing 175MHs, with same H81 mobo setup. Both are molex and risers powered from PSU. Nanos have same 6-pin connector like the 480s. This Nano rig is also functioning well and hashing for more than a month.

The Nano's are fine with 6pin pcie because thats all they need, unlike the 480s which it seems really should have had an 8 pin.

[citronick]

... that means powered risers only will not be enough if you OC? Likely need molex powered too, in case of H81 mobo.

I am guessing probably that's why my rig has no issues because the risers are powered via SATA from PSU. And the ASrock H81 Pro BTC board has also molex powered drawing from PSU. With both of them powered, it will give enough juice to 480 when OC.

Meanwhile, with same mobo setup, I have 5 x 390s OC to 1050Mhz/1125MHz giving me 30.5MHs per card, and no issues with that too. Both risers and molex are powered. The 390s draw much much more power than the 480s.

The reason the 390s are fine is because they have sufficient power from the pci-e connectors on the card.

You need the Asrock H81 pro btc boards, they use two extra molex connectors on the mother board, and do not use the USB riser which does not connect to the PCIE slot +12V.

How to tell if the riser u buy connects to the pcie slot +12V? I believe these btc boards were designed when unpowered ribbon risers were the norm, these additional molex connections should be redundant when using powered risers.

I also have a rig with 6 x Nanos OC to 1040/300 with Heliox ROM Mod hashing 175MHs, with same H81 mobo setup. Both are molex and risers powered from PSU. Nanos have same 6-pin connector like the 480s. This Nano rig is also functioning well and hashing for more than a month.

The Nano's are fine with 6pin pcie because thats all they need, unlike the 480s which it seems really should have had an 8 pin.

According to the H81 Pro BTC manual, the powered molex are used for SLI/XFIRE and there 2 connectors on the mobo.

If I power up the mobo without the molex connected onto the mobo, upon bootup there will be a BIOS screen flash to remind you to connect the molex.

[Q_R_V]

The Nano's are fine with 6pin pcie because thats all they need, unlike the 480s which it seems really should have had an 8 pin.

But Nano's are reference design 8pin pcie. Are there boards with 6 pin only?

[citronick]

The Nano's are fine with 6pin pcie because thats all they need, unlike the 480s which it seems really should have had an 8 pin.

But Nano's are reference design 8pin pcie. Are there boards with 6 pin only?

my bad, Nanos are 8pins, 390s use 8+6pins

[BtcCrazy1]

I've had mine running for 3 days now maybe I should get powered risers just in case I get 85mhs on average with 2 480's and 1 390 without a problem but if I get the powered risers I should be extra safe.

[Gotumoot]

I've had mine running for 3 days now maybe I should get powered risers just in case I get 85mhs on average with 2 480's and 1 390 without a problem but if I get the powered risers I should be extra safe.

If you use the motherboards such as Asrock H81/61 Pro Btc, which has extra onboard molex connectors, you are fine.

[fredeq]

It doesnt make any sense that 6PIN cant deliver more than 75W.
Suddenly 2 extra cables give 75W more?
Also it seems the 2 extra cables are ground only, check the image:


For me it rather makes sense that some hardware on 480 is routing too much power from PCI-E, while it should just get much more from the 6PIN.

Honestly I have never mined with powered risers... I am using MSI boards with EPS12V (8PIN), maybe that saves my boards
The usual molex ones with basic resistor look so crappy to me that I am afraid to use them
The usb ones look decent, but you say they are not that good?
Any good boards besides Asrock H81/61 Pro Btc that you can recommend? (to handle extra PCI-E power)

[Fanpant]

6 Pin PCIE power connectors are rated at 75W, but they can hold much more power, up to 150W, widely used in bitcoin mining.

[molitar]

OK two things the person with the burned mobo due to RX 480 stop cross threading!  This thread is about the Claymore Dual Miner and the RX 480 discussion is for that thread!  Now what I want to know and which is of most importance is this latest released affected by the DOS attack that is affecting mining speeds?  Right now they are telling those to go back to the geth 1.4.7 right now until a new version is released that fixes the DOS vulnerability.   Someone really seems to be trying to kill off Ethereum.

[fr4nkthetank]

Its hard to stay on topic   We'll just have to wait until Claymore brings us that new sacrifice for the gods (new dual coin to mine).  Other than that, i'd love to mine on etherminer, but the damn 4444 port is not working.  anybody have any ideas for me, i didnt really try anything special yet.   etherpool works fine, but I'd rather have ethermine.

[d57heinz]

It doesnt make any sense that 6PIN cant deliver more than 75W.
Suddenly 2 extra cables give 75W more?
Also it seems the 2 extra cables are ground only, check the image:


For me it rather makes sense that some hardware on 480 is routing too much power from PCI-E, while it should just get much more from the 6PIN.

Honestly I have never mined with powered risers... I am using MSI boards with EPS12V (8PIN), maybe that saves my boards
The usual molex ones with basic resistor look so crappy to me that I am afraid to use them
The usb ones look decent, but you say they are not that good?
Any good boards besides Asrock H81/61 Pro Btc that you can recommend? (to handle extra PCI-E power)

Ya we stray a bit ot from time to time.  On the pcie. Reason the 8 pin can deliver more is the extra ground.  If you have had a pcie burn it's usually the ground side.  Since one wire is used to sense voltage and the other two are ground.  That coupled with the fact most psu come with 3 hot wires in the 6 pin pcie. So it can technically feed more thru the 12+ then the ground can return.  Yes there is some burnt off as heat but most comes back down the return. And if that isn't the same as the 12+  then you have meltdown.  Seen this many times now and that's the best I came up with as to why the 8 pin can deliver so much more and doesnt add anything more other than two more grounds and another voltage sensor.

Best regards
d57heinz

[Caemyr]

Well there are a lot of reports of PCI slots crapping out, I came home to this in my rig that had 3 RX-480s in it.  They all went back for a refund today.

http://www.holylands.com/mb.jpg

Now, please excuse me if this was already mentioned, as I just did a quick look at last few pages, but there is something I don't understand about this incident.

The pins burned on that pic are +3.3V/Ground (ATX 24pin, pins 2,3). How could RX 480 possibly draw so much 3.3V from PCIe to burn the ATX pin? The biggest draw measured on that line was 5W (PCper), with others pointing even lower values (2.5W during benchmark).

EDIT: Also https://youtu.be/ZjAlrGzHAkI?t=773

[madmartyk]

Its hard to stay on topic   We'll just have to wait until Claymore brings us that new sacrifice for the gods (new dual coin to mine).  Other than that, i'd love to mine on etherminer, but the damn 4444 port is not working.  anybody have any ideas for me, i didnt really try anything special yet.   etherpool works fine, but I'd rather have ethermine.

Sorry!!

[fredeq]

Ya we stray a bit ot from time to time.  On the pcie. Reason the 8 pin can deliver more is the extra ground.  If you have had a pcie burn it's usually the ground side.  Since one wire is used to sense voltage and the other two are ground.  That coupled with the fact most psu come with 3 hot wires in the 6 pin pcie. So it can technically feed more thru the 12+ then the ground can return.  Yes there is some burnt off as heat but most comes back down the return. And if that isn't the same as the 12+  then you have meltdown.  Seen this many times now and that's the best I came up with as to why the 8 pin can deliver so much more and doesnt add anything more other than two more grounds and another voltage sensor.

Best regards
d57heinz

Interesting argument for the extra ground, though shouldnt it be 3/2 * 75W = 112.5W for the 8PIN?

Btw I have just finished new calc section for this merge mine: http://www.whattomine.com/eth_dcr
Will add the new mystery coin as well when released.

[Overdud]

]

Now, please excuse me if this was already mentioned, as I just did a quick look at last few pages, but there is something I don't understand about this incident.

The pins burned on that pic are +3.3V/Ground (ATX 24pin, pins 2,3). How could RX 480 possibly draw so much 3.3V from PCIe to burn the ATX pin? The biggest draw measured on that line was 5W (PCper), with others pointing even lower values (2.5W during benchmark).

EDIT: Also https://youtu.be/ZjAlrGzHAkI?t=773

+3v and +3v will not short.  The issue is not the power, it's drawing that much power thru connectors, traces, pad, etc that were designed for less current.  It's important to keep paths in the circuit insulated from one another.  These paths were designed and rated, intended for the load.  This is why there is a max specified load, the components, pathways, connectors, traces, pads, etc all have an intend range.

A short circuit occurs when positive finds a straight path to ground.  If there were no other paths or traces nearby, a larger than intend load could only act as a fuse, zip and the connection is gone. 
In reality, there are plenty of pathways on a circuit board. They are really really close to each other.  So, what happens... the extra load causes excessive heat, lowering resistance until an actual short occurs between positive and negative. Positive straight path to ground.

The picture, you see the middle pin above the two +3v pins?  It is the -12v.  Usually, the ground is far away from the positive in a lot of plugs.  Had the -12v been by the +12 in this case, there would have been a much larger arc blast.  See, the -12v on this plug was the heat that lowered resistance, melting and the the +3v lanes had a path straight to ground. 

+3v to +3v would do nothing.  +3v straight to negative....you see the result