Pages:
Author

Topic: python OpenCL bitcoin miner - page 20. (Read 1239060 times)

newbie
Activity: 15
Merit: 0
March 30, 2011, 08:29:21 PM
try sdk 2.2 instead

Using SDK 2.2 instead of 2.3 or 2.4, I still get 2 cores put to 100% load using a HD 5870 and HD 5670 on Win7.

For the meanwhile anyone with this problem can set CPU affinity to only 1 (same for all) core for all poclbm processes so only one core is fully loaded, that way less electricity is used and less heat is made until a proper solution arises.. it doesn't seem to effect hash rate at all.
full member
Activity: 171
Merit: 127
March 29, 2011, 02:43:09 PM
Thanks nelisky! I'll add this in next version.
legendary
Activity: 1540
Merit: 1002
March 29, 2011, 12:30:01 PM
Not sure if this was discussed before, I didn't find anything on my 5 seconds search. I needed to use ssl for remote miners and while bitcoind now supports that out of the box, poclbm didn't seem to, at least not the version I was using (and I'm too lazy to try and upgrade a perfectly running system) so here's a patch to allow for that. Patch is for BitcoinMiner.py and you just supply the host part with 'https://' prefix to make it use ssl.

Code:
158,159c158,162
<
< self.host = '%s:%s' % (host.replace('http://', ''), port)
---
> if host.find('https:') >= 0:
> self.connector = httplib.HTTPSConnection
> else:
> self.connector = httplib.HTTPConnection
> self.host = '%s:%s' % (host.replace('http://', '').replace('https://', ''), port)
220c223
< self.connection = httplib.HTTPConnection(self.host, strict=True, timeout=5)
---
> self.connection = self.connector(self.host, strict=True, timeout=5)
full member
Activity: 170
Merit: 100
March 29, 2011, 06:38:30 AM
Version 20110325 is a good 5%-10% slower on my 2nd and 3rd card compared to version 20110311.

i have a drop of ~2000 khash/s on 6950 and ~1000 khash/s on 5970 and 5870, nothing major, not even 1%
member
Activity: 84
Merit: 10
March 27, 2011, 03:43:50 PM
Oh thanks for the prompt reply, will overclocking failure reflect rejected?
full member
Activity: 263
Merit: 100
YGOLD is a Defi platform
March 27, 2011, 03:42:16 PM
Hi does anyone know if the accepted shares should be shown?

I am running solo with the latest poclbm client and unlike mining in pool the command prompt doesn't show any accepted shares or rejected shares. Is this normal?

Cheers!
There are no shares when mining solo, only found blocks.
member
Activity: 84
Merit: 10
March 27, 2011, 03:39:10 PM
Hi does anyone know if the accepted shares should be shown?

I am running solo with the latest poclbm client and unlike mining in pool the command prompt doesn't show any accepted shares or rejected shares. Is this normal?

Cheers!
full member
Activity: 226
Merit: 100
March 27, 2011, 06:41:27 AM
try sdk 2.2 instead
full member
Activity: 238
Merit: 100
March 27, 2011, 02:29:30 AM
On the contrary, on my 2x5970 Linux mining rig, no difference in hashspeed compared to older versions. Everything works fine.

Crossfire disabled?

Yes, disabled.
qed
full member
Activity: 196
Merit: 100
March 27, 2011, 01:14:13 AM
Version 20110325 is a good 5%-10% slower on my 2nd and 3rd card compared to version 20110311.

On the contrary, on my 2x5970 Linux mining rig, no difference in hashspeed compared to older versions. Everything works fine.

Crossfire disabled?
member
Activity: 60
Merit: 10
March 27, 2011, 12:42:37 AM
Has anyone found some remedies I can get rid of the 100% core usage of poclbm?  I have 2 miners running with my 6970s and they each load up a core on my system.

The system seems to be tied to SDK 2.3 + 69xx cards.  I have two 5870s running on SDK 2.3 also and their instances of poclbm do not suck up much cpu time at all.

I tried the 11.4 RC1 driver that came out a few days ago, no change.  I'm back on 11.2 currently.
full member
Activity: 238
Merit: 100
March 26, 2011, 03:49:53 PM
Version 20110325 is a good 5%-10% slower on my 2nd and 3rd card compared to version 20110311.

On the contrary, on my 2x5970 Linux mining rig, no difference in hashspeed compared to older versions. Everything works fine.
qed
full member
Activity: 196
Merit: 100
March 26, 2011, 11:58:47 AM
Version 20110325 is a good 5%-10% slower on my 2nd and 3rd card compared to version 20110311.

There is a change in the way '-f' works. Simply said, just use lower '-f' now to achieve same performance/desktop lag as before. 5% seems too much though, what's your setup? Is the first card faster?

3x HD 6950@725MHz and 0.950V.
Windows 7 64-bit and catalyst 11.4 RC2.
CrossfireX enabled.

With the 20110311 version everything is fine and i'm getting around 270 Mhash/s on each card. With the 20110325 the crossfire issue is back, random drop at 180-190Mhash/s and an average of 250-260 Mhash/s on the 2nd and 3rd card.

Thanks for your great work.

EDIT: I'm using "-f 60"
full member
Activity: 171
Merit: 127
March 26, 2011, 11:37:02 AM
Version 20110325 is a good 5%-10% slower on my 2nd and 3rd card compared to version 20110311.

There is a change in the way '-f' works. Simply said, just use lower '-f' now to achieve same performance/desktop lag as before. 5% seems too much though, what's your setup? Is the first card faster?
qed
full member
Activity: 196
Merit: 100
March 26, 2011, 08:34:48 AM
Version 20110325 is a good 5%-10% slower on my 2nd and 3rd card compared to version 20110311.
member
Activity: 158
Merit: 10
March 26, 2011, 05:36:55 AM
Eh, do I have to disable crossfire for my 5970(1 card) using this miner guys ?
full member
Activity: 171
Merit: 127
March 26, 2011, 03:27:34 AM

Working much better for me now. Before this update, I had the miner restart if it threw a RPCError in getwork.
Now I see a bunch of "warning: job finished, miner is idle," but it keeps chugging along.
I haven't looked at the new code yet, but would some pre-fetching help?

No. You need as fresh jobs as possible. Try to figure out why it fails to fetch new one in a reasonable time. If not in pool - there is an issue currently which can make getwork() extremely slow, see https://bitcointalksearch.org/topic/createnewblock-too-slow-especially-while-bitcoin-is-under-tx-spam-attack-4853
newbie
Activity: 5
Merit: 0
March 26, 2011, 02:45:13 AM
New version is up. All changes are pool related.

- long polling is slightly better in preventing stales to escape to server;
- miner now supports 'time rolling' whenever there is 'X-Roll-NTime' header in HTTP response;
- improved check for end of current task

Working much better for me now. Before this update, I had the miner restart if it threw a RPCError in getwork.
Now I see a bunch of "warning: job finished, miner is idle," but it keeps chugging along.
I haven't looked at the new code yet, but would some pre-fetching help?
legendary
Activity: 1386
Merit: 1097
March 25, 2011, 09:12:15 PM
With the latest version while mining for myself, I seem to be getting intermittent "problems communicating with bitcoin rpc" errors. They appear and disappear seemingly randomly...

This may be related to the getwork locking in bitcoind, the same trouble as I solved. More people already reported troubles with mining on their own with many pending transactions.

You can a) patch the bitcoind with m0mchil's patch published yesterday b) Join my pool, now you even know how to do that Wink
member
Activity: 73
Merit: 10
March 25, 2011, 08:59:58 PM
With the latest version while mining for myself, I seem to be getting intermittent "problems communicating with bitcoin rpc" errors. They appear and disappear seemingly randomly...
Pages:
Jump to: