Slightly offtopic: Kept trying to figure out why I never got any notifications from this forum thread and finally found just this thread being binned as spam by google mail. Something about the topic makes it unhappy, probably the word earnings. Carry on.
Bad Google omen? I hope not since I am contracted by Coinlab. |
|
|
|
|
Slightly offtopic: Kept trying to figure out why I never got any notifications from this forum thread and finally found just this thread being binned as spam by google mail. Something about the topic makes it unhappy, probably the word earnings. Carry on.
|
|
|
|
Hi, sorry for being stupid – What do I put if I don’t want to touch a rather finicky GPU?
Like 5850/5830/5770 with 5830 being faulty – 750,???,900 core, 250,???,250 memory 100,???,100 fan?
I cannot touch any of those without a forced restart. Setting defaults does not work either. Auto gpu/fan restarts as well.
The card works okay with no modification on any kernel/vectors/aggression setting, the only thing is, I cannot overclock the stable cards on the same machine. Or can i (without afterburner's/etc's manual set up)?
TIA.
Just don't give it any gpu parameters and it won't try to set anything. Or if you want to disable all the gpu monitoring entirely, start it with --no-adl |
|
|
|
On network delays 2.7.4 seems to not bleed shares to my backup pool. I am showing 600Mhahs on pool 0 with pool not providing work fast enough. Using a BFL single with 864 bitstream. Not throttling and no HW errors. If the pool it is on is able to handle the load then it will hit 860 and stay at greater then or equal to 859. What can I do to help track it down as an issue with eclipsemc seems to be causing it and it will likely get fixed fast enough?
This is confirmed not working in 2.7.4 now. It should be fixed in git. |
|
|
|
I'm trying to compile cgminer on my mac. My card is a 6750M so I don't expect much performance. I'm just doing this to see if I can.
I have MacOS 10.8. I installed XCode, XCode command line tools, MacPorts and the following from MacPorts: git yasm curl ncurses automake autoconf autogen.
I copied adl_defines.h, adl_sdk.h and adl_structures.h from one of my linux machines into cgminer/ADL_SDK/. I had to modify adl_sdk.h to remove the #if defined (LINUX) condition.
Why did you do that then? There is no ADL support on macos. So is cgminer on MacOS possible? If not I will give up. DONT DO THAT. |
|
|
|
I'm trying to compile cgminer on my mac. My card is a 6750M so I don't expect much performance. I'm just doing this to see if I can.
I have MacOS 10.8. I installed XCode, XCode command line tools, MacPorts and the following from MacPorts: git yasm curl ncurses automake autoconf autogen.
I copied adl_defines.h, adl_sdk.h and adl_structures.h from one of my linux machines into cgminer/ADL_SDK/. I had to modify adl_sdk.h to remove the #if defined (LINUX) condition.
Why did you do that then? There is no ADL support on macos. So is cgminer on MacOS possible? If not I will give up. It is, without the adl support... |
|
|
|
I'm trying to compile cgminer on my mac. My card is a 6750M so I don't expect much performance. I'm just doing this to see if I can.
I have MacOS 10.8. I installed XCode, XCode command line tools, MacPorts and the following from MacPorts: git yasm curl ncurses automake autoconf autogen.
I copied adl_defines.h, adl_sdk.h and adl_structures.h from one of my linux machines into cgminer/ADL_SDK/. I had to modify adl_sdk.h to remove the #if defined (LINUX) condition.
Why did you do that then? There is no ADL support on macos. |
|
|
|
|
Note that cgminer does this already intrinsically and adding this proxy with it serves no real purpose. I can't speak for all other mining software though.
|
|
|
|
Hello
I really like the new version, it is giving better hashrate a little bit on my 7970, now its ~ 685.
Sad thing is that with this new version i starting to see error about driver recovery, when using mozilla firefox with many tabs opened.
Any suggestions? Thanks
I think ckolivas has said many times that the latest iterations are more aggressive/efficient and hence the GPU is getting even less idling time. This means any basic rendering it has to do even for web browsing or flash gets squeezed even more. Either back down your clock rate while surfing or turn down intensity. Thanks for the answer, i had suspicions about that by myself. Curently using intensity of 9. What is tipycal/best intensity for 7970 while doing something on PC (just web basicly, flash hardware acceleration is turned off) and what would be the best intensity if nothing else on computer is done? Tried dynamic mode? |
|
|
|
On network delays 2.7.4 seems to not bleed shares to my backup pool. I am showing 600Mhahs on pool 0 with pool not providing work fast enough. Using a BFL single with 864 bitstream. Not throttling and no HW errors. If the pool it is on is able to handle the load then it will hit 860 and stay at greater then or equal to 859. What can I do to help track it down as an issue with eclipsemc seems to be causing it and it will likely get fixed fast enough?
Edit: stupid question restarted with debug on. Sadly display debug would have likely shown me sooner.
Edt: Edit: Output shows http error couldnt rewind something. errors and rpc errors. Wish I would have grabbed a screenshot as it may be more help.
Do you have --failover-only enabled? If so, don't, if you want shares to leak. |
|
|
|
2 Bitcoins sent!
089df03143be8503228442753641ad65d037f7394826e32bd9f8b3c3efb3cf14
Thank you my friend!!
Cheers!
Thiago
Noted in the wallet already, thanks!  |
|
|
|
....
vectors 1, worksize 64, kernel poclbm - these are the defaults it will set if you set nothing and are best for 7970
gpu-memdiff -150, , intensity 9 (11 only on linux)
You cannot set memory to 250 on a 7970 with cgminer, only unofficial overclocking tools on windows can do this.
Thank you CK! I'm a Linux-only guy... Best! Thiago In that case do this first: export GPU_USE_SYNC_OBJECTS=1 and then you can set intensity to 11 as well  |
|
|
|
Hi! Is this the best for 7970's: "intensity" : "11",
"vectors" : "2",
"worksize" : "128",
"kernel" : "diablo",
"gpu-engine" : "550-1000",
"gpu-fan" : "0-75",
"gpu-memclock" : "150",
"gpu-memdiff" : "0",
"gpu-powertune" : "0",
"gpu-vddc" : "0.000",
"temp-cutoff" : "95",
"temp-overheat" : "85",
"temp-target" : "75",
"api-allow" : "W:192.168.X.0/24",
"api-listen" : true,
"api-network" : true,
"api-port" : "4028",
"auto-fan" : true,
"auto-gpu" : true,
"expiry" : "120",
"failover-only" : true,
"gpu-dyninterval" : "7",
"gpu-platform" : "0",
"gpu-threads" : "2",
"log" : "5",
"queue" : "1",
"retry-pause" : "5",
"scan-time" : "60",
"temp-hysteresis" : "3",
"shares" : "0",
"kernel-path" : "/usr/bin"
? Or should I use MEM @ 250 and a different worksize, or is it fine? Thanks! Thiago vectors 1, worksize 64, kernel poclbm - these are the defaults it will set if you set nothing and are best for 7970 gpu-memdiff -150, , intensity 9 (11 only on linux) You cannot set memory to 250 on a 7970 with cgminer, only unofficial overclocking tools on windows can do this. |
|
|
|
Intensity is the amount of work in one go the GPU has to do before it can return its results. It needs to be large enough to keep the device fully utilised yet not so overloaded that the device becomes unresponsive. The faster the GPU, the higher the sweet spot is. Ram is of no use to mining whatsoever. They could even mine fine with only a few MB of ram.
pretty sure most people will run into a roadblock with their CPU before their GPU. you could keep increasing it and get some infinitesimally higher results except your cpu will hit 100% usage i have 4 5870's on a single core sempron that i run at agg 7 in phoenix, because at agg 8 the cpu is 100% That is a driver problem. The CPU usage required to set higher intensities intrinsically in cgminer is actually less than at lower intensities. |
|
|
|
|
Upgrading 2.7.4 to stable status.
|
|
|
|
Pool rotate is broken in 2.7.4 (don't know about 2.7.2/3 was working in 2.7.1)
Have it set up to rotate every 20 mins. All pools are shown as alive, but it hasn't moved off pool 0 for 1h 10m.
Yeah looks like it's off. The new "balance" strategy should have a similar effect in the meantime. |
|
|
|
New version: 2.7.4, 23rd August 2012
I knowh you have scrypt... but 274 on scrypt on p2pool still not counting "Q:" correctly (stuck on 1 no matter of longppols and new blocks). Using your latest binary form #1 post. Win7 32. Yep, looks like the counter is broken just in scrypt mode. Update: not stuck, just counting... something ;] E: is hitting the roof ;] cgminer version 2.7.4 - Started: [2012-08-24 14:04:38]
--------------------------------------------------------------------------------
(5s):13.5 (avg):14.3 Kh/s | Q:10 A:1639 R:83 HW:0 E:16390% U:21.0/m
TQ: 0 ST: 1 SS: 0 DW: 10 NB: 34 LW: 1118 GF: 0 RF: 0 WU: 22.1
Connected to http://rav3n.dtdns.net:9327 with LP as user toy.gpu
Block: ae4d022c87f97081ff8fc269ec1706a1... Started: [15:17:59]
--------------------------------------------------------------------------------
[P]ool management [G]PU management [S]ettings [D]isplay options [Q]uit
GPU 0: 61.0C 85% | 13.6/ 14.3Kh/s | A:1641 R:83 HW:0 U:20.99/m I:10
--------------------------------------------------------------------------------
[2012-08-24 15:22:28] LONGPOLL from pool 0 requested work restart
[2012-08-24 15:22:34] Accepted 2c42840c.5e4a4c84 GPU 0
[2012-08-24 15:22:42] LONGPOLL from pool 0 requested work restart
[2012-08-24 15:22:45] Accepted d02a3fcb.9a5425c1 GPU 0
[2012-08-24 15:22:48] LONGPOLL from pool 0 requested work restart
[2012-08-24 15:22:49] Accepted 71a67c93.7953be2d GPU 0
[2012-08-24 15:22:51] Accepted 30c09893.8920e169 GPU 0
[2012-08-24 15:22:53] Accepted 084d5cc0.98c6ddca GPU 0
[2012-08-24 15:22:55] Accepted 324867e9.ea605510 GPU 0
[2012-08-24 15:22:58] Accepted 0a0fe324.669c0c9b GPU 0
[2012-08-24 15:22:59] LONGPOLL from pool 0 requested work restart
[2012-08-24 15:23:00] Accepted 9a71534b.fa2044d5 GPU 0
[2012-08-24 15:23:00] Accepted 9b28aa94.0af3ebee GPU 0
In that case then, it's not a bug, but more a sign of the combination of extended rolltime support and mining on a low difficulty pool. |
|
|
|
- Pick worksize 256 with Cypress if none is specified.
I wondered why recent git versions have been slower on my 5870, about 414 vs. 421 Mhash/s. Using worksize 128 brought it back to the old speed. (Using fglrx 11.12 and SDK 2.5) Yeah sorry, it seemed that most people using 58x0 were using very low memory speeds and 2.1 sdk which does better with worksize 256. |
|
|
|
New version: 2.7.4, 23rd August 2012
I knowh you have scrypt... but 274 on scrypt on p2pool still not counting "Q:" correctly (stuck on 1 no matter of longppols and new blocks). Using your latest binary form #1 post. Win7 32. Yep, looks like the counter is broken just in scrypt mode. |
|
|
|
|
Why on earth would AMD put an inefficient CPU usage bug based on openCL global worksizes in there on purpose for BFL to release ASICs?
|
|
|
|
|
Show Posts
