Topic: Claymore's Dual Ethereum AMD+NVIDIA GPU Miner v15.0 (Windows/Linux) - page 1248. (Read 6590757 times)

Also noticed on this rig when using -allcoins 1 flag overclocking does not work on 3/5 gpu notice this from logfile.


Cards available: 5
11:21:52:182   8b8   GPU #0: name: Tahiti, 3072 MB available, 28 compute units
11:21:52:182   8b8   GPU #1: name: Tahiti, 3072 MB available, 28 compute units
11:21:52:182   8b8   GPU #2: name: Tahiti, 3072 MB available, 28 compute units
11:21:52:182   8b8   GPU #3: name: Tonga, 4096 MB available, 28 compute units
11:21:52:182   8b8   GPU #4: name: Tonga, 4096 MB available, 28 compute units
11:21:52:182   8b8   Total cards: 5
11:21:52:182   8b8   Initializing...

11:21:52:182   8b8   GPU #3 recognized as Radeon 285/380
11:21:52:182   8b8   GPU #4 recognized as Radeon 285/380

With -allcoins 1 overclocking only works on cards listed as "recongnized" also on this rig i have issue of "GPU 0, OpenCL error -4 - cannot create DAG on GPU" it is always on cards 0-2 never on cards 3-4, it eventually loads fine after random number of watchdog resets.

Compare this log to my 280x rig

08:55:48:967   44c   
Cards available: 5
08:55:48:967   44c   GPU #0: name: Tahiti, 3072 MB available, 32 compute units
08:55:48:972   44c   GPU #1: name: Tahiti, 3072 MB available, 32 compute units
08:55:48:972   44c   GPU #2: name: Tahiti, 3072 MB available, 32 compute units
08:55:48:972   44c   GPU #3: name: Tahiti, 3072 MB available, 32 compute units
08:55:48:972   44c   GPU #4: name: Tahiti, 3072 MB available, 32 compute units
08:55:48:972   44c   Total cards: 5
08:55:48:972   44c   Initializing...

08:55:48:972   44c   GPU #0 recognized as Radeon 280X/380X
08:55:48:972   44c   GPU #1 recognized as Radeon 280X/380X
08:55:48:972   44c   GPU #2 recognized as Radeon 280X/380X
08:55:48:972   44c   GPU #3 recognized as Radeon 280X/380X
08:55:48:972   44c   GPU #4 recognized as Radeon 280X/380X

All cards recognized and all clocks set correctly.

I used the driver on the CD and did not install any of the bloatware.

I have strange issue when mining Expanse:
ETH - Total Speed: 116.244 Mh/s, Total Shares: 57, Rejected: 0, Time: 00:18
ETH: GPU0 25.200 Mh/s, GPU1 25.221 Mh/s, GPU2 25.482 Mh/s, GPU3 20.171 Mh/s, GPU
4 20.170 Mh/s
Gpu 3 and 4 are r9 380 remaining gpu's are 7950's why would the r9 380 be so much slower given the lower difficulty of exp?

Thank you for clearing that up for me. I appreciate you taking the time to provide a thorough explanation.  I would have to agree with you after reading up on it.  This brings to light a serious issue in pcie configuration does it not.  Providing three positive lines while only providing 2 grounds.  75 watts gets split by a third over three but only half over the ground.  Thus most likely causing the overheating.   Thanks again for stearing me straight.

Best regards
D57heinz

24 MH/s with only 106 watt is a good balance 

my 330watt x51r2 with XFX RX480 8GB no OC, been running for 24hr+ @24Mh/s

LOL... no you cannot. You need sgminer (AMD), ccminer (Nvida) or those X11 ASICs miners like Baikal, iBelink...

However, Claymore is adding a 2nd coin to mine with ETH... let's wait and see patiently

I have a feeling it's going to be something good... like his other Cryptonight miner.

hi,
i'm not linux specialist, however, i know and successfully use claymore miner, i see missing -epsw x command,
and "GPU #0 recognized as Radeon 290X", however, a far as i know, 380x is simplified version of 290x

i would try this way

add -epsw x
disable dcr mining and try eth only mining
if not successfull check gpu core clock and if low add a command setting target gpu core clock speed in Mhz
it is important to disable dcr mining, as you are interested in gpu clock speed for eth mining
gl
mxclene

This miner is failing on my 7870 no matter what I do. I have tried the latest drivers, I tried 15.12 as suggested in the first post, and I tried 14.9 as mentioned in another person's response who was dealing with the same issue. I have configured virtual memory to 16GB, I have set GPU_MAX_ALLOC_PERCENT 100 and GPU_SINGLE_ALLOC_PERCENT 100, I have tried running it without the Decred miner.

Regardless of everything, I continue to encounter the issue with the DAG buffer. The machine is Windows 10 x64, 12GB RAM, everything should be perfect but this just fails hard. Anyone have any advice?

09:46:17:669   784   Setting DAG epoch #60 for GPU #0
09:46:17:669   784   Create GPU buffer for GPU #0
09:46:17:669   784   OpenCL error - cannot allocate big buffer for DAG. Check readme.txt for possible solutions.
09:46:20:685   77c   Setting DAG epoch #60 for GPU #0
09:46:20:685   77c   GPU 0, OpenCL error -38 - cannot write buffer for DAG

hi, what settings for miningpoolhub?
doesn't work

Unable to run v4.7 on Arch Linux due to

NO

can this miner be used to mine x11? dash? tnx

The burned pins are both +12V of ATX connector. I think you forget that pinouts are of the plug, not socket.
Why only +12V burned? Because there are many more pins for ground, so each had less current.

Kirchhoff's first law disagrees with this. Current incoming is always equal to current outgoing.

6 pin can deliver much more power in reality. I know, I use it to power some devices from computer PSU. It's just about limiting how much power should be drained, probably because many PSU are crap.

I would also like to state that -12V is almost not used today. It may by used by audio circuit on motherboard, but it's absolutely not used with +12V to power anything heavy. See how small current is allowed (specification of the PSU). Voltage is too small to create any arc. You make some bizarre explanation to explain bizarre situation, but it was just the +12V pins that burned because of too much load.

There are people that already asked to end this off topic discussion, so I don't want to keep it going and going.

I will try to be clear so we can move on....
 as others noted in the thread-
 6pin PCIe to 8 pin, its added ground path. Look at the pin out on those connectors. 6pin = three +12V and -12v, rated for 75watt. This are typically heavier gauge than the motherboard 24pin.  If you look closely at the mother board connector pin out, there are two +12v wires yet only one single -12v to connect the loop. You have to realize, it should be obvious, there is only one negative path from the motherboard plug to the psu.  Remember now, the 75watt rated 6 pin PCIe has 3 pins for -12v which are more robust on top of that.  Am 8pin is not much different than the 6, it's the same three +12v with the extra pins being added -12v to accommodate the extra load.
The motherboard, it has one single -12v pin. Should be obvious, it's not made robust. Significantly less than the PCIe 6pin.

The -12v would have to carry the entire 12v load, on this connector, it is the weakest point.  If there is an excessive 12v load, on this plug the sole -12v pin would be the weak point, most likely place of failure. Heating up metal, plastic, air, and gases in and around that particular pin, lowering resistance and creating the perfect condition for an arc.  Since current is returning thru the negative pathway, it is (-) remember this. The short occurred not thru the +3v but from the +3v which had found a path straight to the ground (-12v pin).  The 3v was straight from the psu and is supposed to enter the board but instead had found a path straight back thru the -12v beside it.  

You would not expect the -12v pin to explode, it will carry all the load the metal will allow, the heat melts plastic before the copper and metal pins, the heated gases ionizing, then the 3v finds its destiny shorting thru the -12v.  

I hope you get it, I don't want to be redundant.  Too high current heats up and melts things, if it's not way past the specified load then this will not be sudden or instantaneous. a short to ground will create  a much mote dramatic blast. a sudden explosion.  The -12v pin would not explode, current wouldn't flow out of it to other pins as it don't work like that.  The +3v found its way to ground.  

Could you post claymore JSON-RPC command list and the meaning of the returned value ?

Have you tried the Stilt BIOS mod for 290s?

yeah i know it works fine, but for some reason port 4444 is a no go for me.  some people have said I should ask my isp or something, but idk, i'm kinda lazy.  maybe i'll ask next time I call for a price reduction.  I do that every year, call, ask for customer retention, just be nice and ask for a better deal.

Still, -12V is pretty much intact, i can only see melted plastic from +3.3V connector. The two pins that got fried were 2 and 3 - +3.3V and Ground. If it would really be -12V shorting with +3.3V, wouldn't we see burns on both and pin3 with no damage?

It worked!

But one problem... For some reason, after running the software for a few hours, it closed and wouldn't open again. I did get the 4.7 to run. But then after a while, one gpu stopped running, and some other weird stuff. And I guess my hashrate is generally low with 14.9 drivers, so I tried to update to 15.12. I got some "catastrophic error" on a dll file. Something like "amdocl" or something like that, and it told me there were no cards in the list. So I ended up just reverting back. I set power limit to -10% and gave a bit of an overclock to the card and it seems alright for now... It occured to me, that perhaps my both cards 1050/1450 was pulling too much power from the mobo, which is why I lowered the power limit. If that even does much.

Anyway, hopefully the current setup will run for a whole day straight without issue. Then I'll figure out how to update drivers properly and whatnot. If anything, the next thing I'll probably try are the most recent drivers to see if that helps anything... I was informed that my 2 R9 270's should be able to break 30MH/s without an OC, so about 23MH/s right now kinda sucks.