urubu
Member

Offline
Activity: 87
Merit: 10
|
 |
April 03, 2014, 10:24:53 PM |
|
How are 290x improvements coming along? Eta?
I want to buy this for a couple of my cards, but if 290x improvements are coming soon, I would prefer to run it on those instead.
|
|
|
|
primeGPU (OP)
|
 |
April 03, 2014, 10:49:48 PM |
|
How are 290x improvements coming along? Eta? I want to buy this for a couple of my cards, but if 290x improvements are coming soon, I would prefer to run it on those instead.
We're planning to make a 290x-optimized version in 1 month, but it can take a bit more time.
|
|
|
|
trogdorjw73
|
 |
April 03, 2014, 10:56:59 PM |
|
So testing on my 3x7950 rig, I finally (FINALLY!) found a block:  Great! Except, where are the coins stored? I have the client running on another system, and here's what I see: new height:473347 new height:473348 new height:473349 ==== BLOCK FOUND ==== (see debug.log for details). new height:473350 new height:473351 new height:473352 ...SNIP!... new height:473486 new height:473487 new height:473488 ==== CHAIN FOUND ==== (see debug.log for details). new height:473489 new height:473490 new height:473491
Checking out debug.log: ERROR: PrimecoinMiner : generated block is stale Son of a.... So I've now "found" 1 block and four chains, but no coins to date. Been running about two days now. Blarg. Some people get lucky... some people write blogs. 
|
|
|
|
trogdorjw73
|
 |
April 03, 2014, 10:59:21 PM |
|
How are 290x improvements coming along? Eta? I want to buy this for a couple of my cards, but if 290x improvements are coming soon, I would prefer to run it on those instead.
We're planning to make a 290x-optimized version in 1 month, but it can take a bit more time. Moooo! Milk it for all it's worth!
|
|
|
|
primeGPU (OP)
|
 |
April 03, 2014, 11:07:17 PM |
|
Son of a.... So I've now "found" 1 block and four chains, but no coins to date. Been running about two days now. Blarg. Some people get lucky... some people write blogs.  Ouch... On the bright side, your 'time' value is pretty good for 7950.
|
|
|
|
merc84
|
 |
April 03, 2014, 11:52:04 PM |
|
Does the wallet need be open for this to work correctly seems the miner is running but it gave error when the wallet was open?
|
|
|
|
Primorial
Newbie
Offline
Activity: 7
Merit: 0
|
 |
April 03, 2014, 11:57:12 PM |
|
It seems to me that member mtrlt was on the right track but needed some help with an efficient multiple-precision arithmetic implementation. https://bitcointalk.org/index.php?topic=258540.0Although I have almost no experience with OpenCL, it will not take too much time to supercharge mtrlt's code with an efficient implementation of some functions. I will have some playtime in May to mess around with the code. You are right, his code makes perfect sense but contains several bugs, some of them are obvious, some are hard to even detect. Making a functional prototype out of it seems to be moderately hard (don't ask me how do I know), getting competitive performance even on GCN is much harder, as usual, though. Modern AMD's OpenCL compiler is unstable, unpredictable and buggy, it is no fun to play with. Do you have any suggestions for the best OpenCL compiler version to use? ISA Assembler? Thanks in advance All the versions supporting R9 290 (since Catalyst 13.11 Beta 6) are essentially sub-par, and it appears benefitial to try different releases for each moderately complicated kernel individually. Miscompiles are frequent for fragments with intensive integer computations, some versions are unable to compile heavily unrolled kernels at all. There is no official AMD GCN ISA assembler available to the public. This http://openwall.info/wiki/john/development/GCN-ISA unofficial implementation appears to be most promising, but needs work.
|
|
|
|
primeGPU (OP)
|
 |
April 04, 2014, 12:02:04 AM |
|
All the versions supporting R9 290 (since Catalyst 13.11 Beta 6) are sub-par, hence each moderately complicated kernel benefits from choosing its own particular version. Miscompiles are frequent for fragments with intensive integer computations, some versions are unable to compile heavily unrolled kernels at all. There is no official AMD GCN ISA assembler available to the public. This http://openwall.info/wiki/john/development/GCN-ISA unofficial implementation appears to be most promising, but needs work. Could you please stop derailing the thread, thank you.
|
|
|
|
Primorial
Newbie
Offline
Activity: 7
Merit: 0
|
 |
April 04, 2014, 12:05:30 AM |
|
Could you please stop derailing the thread, thank you.
Ok 
|
|
|
|
merc84
|
 |
April 04, 2014, 12:08:13 AM |
|
cpuusage set to 7 but only using 7-8% of the E3 1230v3; note i used affinity mask for cores 1-7 but only 3 of those cores as showing any use.
|
|
|
|
Starlightbreaker
Legendary
Offline
Activity: 1764
Merit: 1006
|
 |
April 04, 2014, 12:13:39 AM Last edit: September 17, 2016, 06:30:43 PM by Starlightbreaker |
|
Son of a.... So I've now "found" 1 block and four chains, but no coins to date. Been running about two days now. Blarg. Some people get lucky... some people write blogs.  Ouch... On the bright side, your 'time' value is pretty good for 7950. ....wait. so that means my 290 time for 7950 isn't bad at all? now i know. i guess it's better for a 7950 to be mixed with 7970s/280x then?
|
|
|
|
primeGPU (OP)
|
 |
April 04, 2014, 12:21:28 AM |
|
....wait. so that means my 290 time for 7950 isn't bad at all? now i know.
290 is kinda unexpected for me. We should have an edition for 7950/270x really soon. i guess it's better for a 7950 to be mixed with 7970s/280x then?
There shouldn't be any correlation.
|
|
|
|
primeGPU (OP)
|
 |
April 04, 2014, 12:25:37 AM |
|
cpuusage set to 7 but only using 7-8% of the E3 1230v3; note i used affinity mask for cores 1-7 but only 3 of those cores as showing any use.
That's for 1 card, right? We'll probably expand -cpuusage setting in a future version.
|
|
|
|
merc84
|
 |
April 04, 2014, 12:33:29 AM |
|
yes just one card  Time is around 210 and estimated ch/d is 1.36 
|
|
|
|
mustyoshi
|
 |
April 04, 2014, 12:35:15 AM |
|
How do we know this isn't just a clever way to sell a server farm? I'd like to see blockchain embedded proof that these users' GPUs are actually the ones mining these blocks.
|
|
|
|
merc84
|
 |
April 04, 2014, 12:39:23 AM |
|
with such low cpu usage would it be safe to set affinity to just 1 core so i can use the remaining cores for cpu mining?
|
|
|
|
Starlightbreaker
Legendary
Offline
Activity: 1764
Merit: 1006
|
 |
April 04, 2014, 12:44:21 AM Last edit: September 17, 2016, 06:29:56 PM by Starlightbreaker |
|
....wait. so that means my 290 time for 7950 isn't bad at all? now i know.
290 is kinda unexpected for me. We should have an edition for 7950/270x really soon. i guess it's better for a 7950 to be mixed with 7970s/280x then?
There shouldn't be any correlation. not 290 as in r9 290, but 290 as in the time in the miner server for my 7950. 280x now floating around 230-250.
|
|
|
|
Starlightbreaker
Legendary
Offline
Activity: 1764
Merit: 1006
|
 |
April 04, 2014, 12:46:16 AM Last edit: September 17, 2016, 06:28:50 PM by Starlightbreaker |
|
and speaking of luck..  4 blocks in less than an hour.
|
|
|
|
primeGPU (OP)
|
 |
April 04, 2014, 12:49:28 AM |
|
with such low cpu usage would it be safe to set affinity to just 1 core so i can use the remaining cores for cpu mining?
Probably yes.
|
|
|
|
primeGPU (OP)
|
 |
April 04, 2014, 12:51:47 AM |
|
and speaking of luck..
4 blocks in less than an hour.
Khm... Looks like a bug. There is no way it can have a block w/o a 10-chain. Gonna check it right now. Have you by a chance restarted the server w/o restarting the client? (Not saying you shouldn't do this, it'd just helped me in locating the problem)
|
|
|
|
|