Sounds like a patch is needed ineed.
Guess you have tried to change voltage with forcing `-P 3`, even if not listed by default?
Yeap, no luck. Reflashed with patched BIOS, good results for now, yet have to see how stable they will be: work@miner-1:~/dual-ethereum-decred$ atitweak -s
0. AMD Radeon HD 7900 Series (:0.0)
engine clock 1050MHz, memory clock 1550MHz, core voltage 1.09VDC, performance level 2, utilization 64%
fan speed 49% (2743 RPM) (user-defined)
temperature 65 C
powertune 20%
1. AMD Radeon HD 7900 Series (:0.1)
engine clock 1150MHz, memory clock 1500MHz, core voltage 1.15VDC, performance level 3, utilization 64%
fan speed 100% (3075 RPM) (user-defined)
temperature 70 C
powertune 20%
2. AMD Radeon HD 7900 Series (:0.2)
engine clock 1150MHz, memory clock 1500MHz, core voltage 1.15VDC, performance level 3, utilization 63%
fan speed 95% (2972 RPM) (user-defined)
temperature 63 C
powertune 20%
|
|
|
|
Sure, but that's what I was asked for: work@miner-1:~/dual-ethereum-decred$ atitweak -l
0. AMD Radeon HD 7900 Series (:0.0)
engine clock range is 150 - 1100MHz
memory clock range is 75 - 1575MHz
core voltage range is 0.8 - 1.09VDC
performance level 0: engine clock 300MHz, memory clock 150MHz, core voltage 0.85VDC
performance level 1: engine clock 1050MHz, memory clock 1550MHz, core voltage 1.09VDC
fan speed range: 0 - 100%, 1100 - 3300 RPM
1. AMD Radeon HD 7900 Series (:0.1)
engine clock range is 150 - 1700MHz
memory clock range is 75 - 2000MHz
core voltage range is 0.85 - 1.256VDC
performance level 0: engine clock 300MHz, memory clock 150MHz, core voltage 0.85VDC
performance level 1: engine clock 1150MHz, memory clock 1500MHz, core voltage 1.256VDC
fan speed range: 0 - 100%, 700 - 5900 RPM
2. AMD Radeon HD 7900 Series (:0.2)
engine clock range is 150 - 1700MHz
memory clock range is 75 - 2000MHz
core voltage range is 0.85 - 1.256VDC
performance level 0: engine clock 300MHz, memory clock 150MHz, core voltage 0.85VDC
performance level 1: engine clock 1150MHz, memory clock 1500MHz, core voltage 1.256VDC
fan speed range: 0 - 100%, 700 - 5900 RPM
work@miner-1:~/dual-ethereum-decred$ atitweak -s
0. AMD Radeon HD 7900 Series (:0.0)
engine clock 1050MHz, memory clock 1550MHz, core voltage 1.09VDC, performance level 2, utilization 64%
fan speed 100% (3264 RPM) (user-defined)
temperature 67 C
powertune 20%
1. AMD Radeon HD 7900 Series (:0.1)
engine clock 1150MHz, memory clock 1500MHz, core voltage 1.256VDC, performance level 3, utilization 64%
fan speed 100% (3010 RPM) (user-defined)
temperature 85 C
powertune 20%
2. AMD Radeon HD 7900 Series (:0.2)
engine clock 1150MHz, memory clock 1500MHz, core voltage 1.256VDC, performance level 3, utilization 64%
fan speed 100% (2900 RPM) (user-defined)
temperature 78 C
powertune 20%
That is, it lists only 2 performance levels (0 and 1) and can change settings for them. But cards work actually at perf levels 2 and 3, and there is no way to alter settings. Someone wrote that he patched the tool to use perf levels above 1, but it did not change the voltage. It seems that BIOS mod is the only way to change voltages for 79xx cards (newer cards use version 6 of API and there is another utility for them that should work). |
|
|
|
Undervolt by 100mV and enjoy -10 temps  It works on Windows. But I even asked here if someone managed to undervolt on Linux (w/o BIOS mod) and got no replies. I can reflash BIOS, and even did that. By default these 7970s have 1.256V max, and I patched in 1.100. It worked on Linux, crashed on miner start on Windows (did not use serialised init). With 1.150 they were stable. Still, I wish I find a way to undervolt them on Linux w/o BIOS mod, I don't believe there is no such tool, I just need to find it. |
|
|
|
..and did you measure the power-consumption on each GPU at such clock?
Nope. But I don't stake on those 7970: they are really hot for some reason comparing to any other my cards (7950, 390, 390X). Thermal interfaces were replaced, nothing helped. I bought them used and plan to sell as soon as I replace them by 390X (on order). Previous owner used them for hard gaming with SLI/CrossFire @1220 with original BIOS (they are BE, Boost Edition, and allow such OC). I just run them @ factory boost clock: https://www.msi.com/Graphics-card/R7970-Lightning-BE.html#hero-specificationBoost / Base Core Clock
1070MHz Core (Boost Clock: 1150MHz) |
|
|
|
What hashrate of your card ?
My MSI r9 280x has got 19.2->19.6Mhash/s as 7950, driver 15.12 core 1030 -mem clock 1250
If you asked me, then under Linux ETH - Total Speed: 61.890 Mh/s, Total Shares: 3663, Rejected: 0, Time: 33:30
ETH: GPU0 19.493 Mh/s, GPU1 21.205 Mh/s, GPU2 21.192 Mh/s
DCR - Total Speed: 1237.804 Mh/s, Total Shares: 5537, Rejected: 66
DCR: GPU0 389.854 Mh/s, GPU1 424.104 Mh/s, GPU2 423.846 Mh/s
GPU0 - HD7950 @ 1050/1550, -dcri 40 (ASUS, somehow lesser than @Windows where similar card gives 20MH/s) GPU1 - HD7970 @ 1150/1500, -dcri 40 (MSI Lightning BE) GPU2 - HD7970 @ 1150/1500, -dcri 40 (MSI Lightning BE) |
|
|
|
@Claymore: after 24h+ runtime the Linux version of miner stopped sending stats  Miner itself works but no monitoring. There were two EthMon apps connected (one on my PC and another on Windows rig for web stats). None of them report now (web or app). Miner shown in red. Check miner, may be it was restarted and after restarting stopped stats? Ot miner shows 24h+ so it was not restarted? PM me the log file then. ETH - Total Speed: 61.891 Mh/s, Total Shares: 3525, Rejected: 0, Time: 32:16 I will PM logs |
|
|
|
@Claymore: after 24h+ runtime the Linux version of miner stopped sending stats Miner itself works but no monitoring. There were two EthMon apps connected (one on my PC and another on Windows rig for web stats). None of them report now (web or app). Miner shown in red. |
|
|
|
GenEthOS stock OS
dont oc with CDM commands, use MSI afterburner instead that way You wont have problems it was mensioned couple times in thread I think he'd be glad to use MSI AB as soon as you or somebody else ports it to Linux (see above, he runs Linux OS). BTW, similar question: who and how downvoltages HD79xx/R92xx AMD cards on Linux? I tried atitweak but it shows performance levels 0 and 1 only, but cards run at 2 and 3. No way to change voltage. od6config is for newer cards. Any other ways apart from modding the BIOS? Any tool for Linux? |
|
|
|
Btw.. what about the monitoring tool that should be available?
Where can I get it?
The 1st page, the 1st message of this thread... |
|
|
|
I've read all, and all I found was that both PSU must be grounded together. But for me that is obvious: if no connection between grounds - you may definitely fry the motherboard and/or something else. Also one should use powered risers or connect both extra on-board Molex connectors to a PSU to provide more power to 6 GPUs. To be on topic: @Claymore, any progress with stability of stats fetching on Linux? 4.3 works well for me but even pressing "s" once gave me assertion failure. Also automatic refresh wia EthMan causes this often. I set the miner as offline and update stats only when necessary. |
|
|
|
fyi, asrock h81 btc pro doesnt work with 2 psu (another problem miner argh)
Can you explain this in details, I am going to use one server PSU for cards and one PC PSU for H81 BTC Pro just bought. Any issues? |
|
|
|
I have been mining on ethermine.org but its a set payout. It doesnt seem like luck is involved at all. A pool has huge hashrate comparing to one miner, the law of large numbers works here ( https://en.wikipedia.org/wiki/Law_of_large_numbers). That means less variance in a number of shares found by pool. It finds them so many per hour so you don't see any luck. But mining solo, if your hashrate is little, you may find many blocks in a day and them find nothing in a week. Also, say, DwarfPool uses HBPPS (hour-based pay per share). It gets some Ethers for an hour, then calcluates number of shares submitted in this hour, and distributes Ethers proportionally to shares of each miner. Payment depends on (1) percent of shares submitted by you (based on total shares per hour and your shares) and (2) number of blocks found (and payments for them). Ethermine uses PPLNS (pay per last N shares). It calculates moving average of your shares in last hour (60 minutes) and uses this average to distribute Ethers from every block. Since it uses average, your payments are instant (calculated after every block, not in the end of an hour), so you see a smooth stream of payments after the 1st block found since you started. Again, it works for large numbers (hashrates), for small ones you will see how the luck is involved. |
|
|
|
Over time, If you solo mine 100 blocks, 1 block will be missed due to the dev fee on average. If your on the pool your hashrate will be shown 1% less on average.
Yes, and if one has higher hashrate for the remaining 99%, then on the long run one will have more blocks found, say, not 100 but 105. Minus 1.05 lost blocks due to DevFee (1%), and still 103.95 blocks found on average. |
|
|
|
|
I think that the probability to get a block in 1-2% of time is compensated by increased probability to find it in the remaining 98-99% of time due to higher hash rate. So it still win. If you miss a block in this 1-2% of time, you just get a new hash to calculate, and the probability to find this one will be exactly the same. So as soon as your average hash rate (in 100% of time) is higher, you win.
|
|
|
|
I would think I should still see accepted shares as rarely are they blocks on a pool. I avg 450 accepted shares per hour on a pool like Nano, so I was expecting to see similar results solo mining.
You should read the theory of pool operation to understand the whole thing. A miner calculates a "checksum" of data in the block using some hashing algorithm. The result is a very long number. To solve a block this number must be lesser than some value called target. Target depends on the speed of finding blocks by the whole network and adapts so the average rate of new blocks is constant regardless of number of miners in the network. That is, more miners - more hashpower - lesser target - harder to find a block solution. When you mine with a pool, pool sets target for miners much higher than actual target for the block solution. So miners with low hash rates can find at least something in reasonable time frame. Those solutions are called "shares" because they do NOT solve a block. But pool can estimate miner hashrate by the rate of submitted shares. Finally some share from some miner will be smaller than a block solution. This share will solve the block, but usually miner does not know that it was its share (some pools like suprnova and pool.mn report block founders, though, for stats). Now compare the typical target from a pool: 0x0000000112e0be82 (4611686018) with your target 0x000000000006f3f2 (455666). 4611686018/455666 ~= 10120. This means that in solo mode only 1 of 10120 such pool shares will solve a block. If your miner finds 100 shares per hour, you need 101 hour to find a solution in solo mode. Until that you will not see any shares in solo mode. You see only block solution. |
|
|
|
Connects, looks like it starts mining and Geth console shows the connection. But not seeing any shares? only temps and new jobs. On pool, i'm seeing accepted shares almost instantly. Doing something wrong here?
ETH: 05/19/16-17:44:09 - New job (target 0x000000000006f68d) from localhost:8545
I didn't try solo mining. But I guess that with solo mode your "share" means to be a block solution. With such difficulty difference (compare yours and pool ones) and your hashrate you should mine few days, maybe a week on average, to find a block. I think so, but might be wrong. |
|
|
|
Great. I hope someone writes a nice webapp for mobiles that uses the same API to (only) get stats. ATM I use internal server but it is Windows only, and I wish to go Linux completely.
Try to use wine - what problem? This makes using Linux for mining meaningless. I use diskless machines with minimum resources with no GUI at all and control them via ssh. I also see no reason to run Wine just to expose few lines of stats :-) I could make an application for the "Android". But we need to API from the author. Also, access to the API must be via a password. API was actually explained few posts above, it's just JSON via TCP. But I would prefer not an application for Android but simple web app usable everywhere (iOS, Android, anything else). Actually what is exposed is enough, but the author writes miner, not UI. So maybe someone else... BTW, I've just got the idea to add few lines to my OpenHAB smarthome configuration. Then I can fetch stats by JSON and expose them in OpenHAB UI. It will be the simplest way for me. It also eliminates the need for password. But for the case you mentioned there is an option to provide stats only (using negative port number). I don't mind if someone will see some stats of some miner. But if he will not be able to alter config, all will be OK. is this purple +_+ temp/fan line detachable?
IMHO the miner should concentrate on mining and not adopt Afterburner's job.
I find it much better if there would be a status line wich informs how many shares each card found the last hour and/or overall. I found the -tt switch very useful on Linux. So if that is free for us - then easier to use less programs if possible. Anyway, it is already implemented. The log details that's another issue. |
|
|
|
|
Great. I hope someone writes a nice webapp for mobiles that uses the same API to (only) get stats. ATM I use internal server but it is Windows only, and I wish to go Linux completely.
|
|
|
|
Anybody has issues with 4.3 monitoring on Linux? @Claymore: Tried few 4.3 beta (as soon as they were updated). One of them dumped some error and segfaulted. Updated to the latest, the same issue. Restarted - it worked some time with monitoring utility, then turned red and no more updates. The miner itself worked. Restart helped to recover, worked few seconds and gave me the same error: ETH - Total Speed: 39.123 Mh/s, Total Shares: 0, Rejected: 0, Time: 00:00
ETH: GPU0 19.556 Mh/s, GPU1 19.567 Mh/s
DCR - Total Speed: 782.459 Mh/s, Total Shares: 1, Rejected: 0
DCR: GPU0 391.114 Mh/s, GPU1 391.345 Mh/s
GPU0 t=64C fan=26%, GPU1 t=70C fan=95%
DCR: 05/19/16-11:26:09 - SHARE FOUND - (GPU 0)
DCR: Share accepted (82 ms)!
DCR: 05/19/16-11:26:10 - SHARE FOUND - (GPU 1)
DCR: Share accepted (83 ms)!
[xcb] Unknown request in queue while dequeuing
[xcb] Most likely this is a multi-threaded client and XInitThreads has not been called
[xcb] Aborting, sorry about that.
ethdcrminer64: ../../src/xcb_io.c:179: dequeue_pending_request: Assertion `!xcb_xlib_unknown_req_in_deq' failed.
./EthDcrMiner.sh: line 71: 2366 Aborted I think that is some problem with monitoring in the miner, found on Linux only. |
|
|
|
|
Show Posts
