-ck (OP)
Legendary
Offline
Activity: 4242
Merit: 1644
Ruu \o/
|
|
January 12, 2012, 11:01:56 AM |
|
@conman I suggest taking a look at this https://github.com/ius/radeonvolt and see if the license permits, to use it. The software is from what I see, able to detect VRM temps on a 58XX card(only, but still better than nothing). It also appears to enumare the number of GPUs in a better manner than the way you do it(by querying the opencl devices). People need to know the VRM temps. My 5850's core can be 65C at max load, when my VRM temps skyrocket to 110C. VERY bad/dangerous idea, see https://bitcointalk.org/index.php?topic=44495.0 as for why. Thanks. I thought the code looked very quirky and operating system dependant anyway (which others had already noted) so I wasn't planning on it.
|
Developer/maintainer for cgminer, ckpool/ckproxy, and the -ck kernel 2% Fee Solo mining at solo.ckpool.org -ck
|
|
|
Remember remember the 5th of November
Legendary
Offline
Activity: 1862
Merit: 1011
Reverse engineer from time to time
|
|
January 12, 2012, 12:10:38 PM |
|
@conman I suggest taking a look at this https://github.com/ius/radeonvolt and see if the license permits, to use it. The software is from what I see, able to detect VRM temps on a 58XX card(only, but still better than nothing). It also appears to enumare the number of GPUs in a better manner than the way you do it(by querying the opencl devices). People need to know the VRM temps. My 5850's core can be 65C at max load, when my VRM temps skyrocket to 110C. VERY bad/dangerous idea, see https://bitcointalk.org/index.php?topic=44495.0 as for why. I've looked at that. But who said we are going to run a thousand APPS to monitor the GPU sensors? And the threads says that it's monitoring voltage that causes this. ADL already provides info about that.
|
BTC:1AiCRMxgf1ptVQwx6hDuKMu4f7F27QmJC2
|
|
|
dlasher
|
|
January 12, 2012, 06:38:16 PM |
|
AMD/ATI have been a bunch of fsking morons with their drivers and the driver returns a different set of values to the ATI display library and all I can do is enumerate them in the order they appear in each library and hope they're one and the same. It usually correlates with PCI slot but even then, mobo manufacturers don't enumerate them always in order either.... sigh
The 6990 / dual GPU cards also are special cases in point and while the ADL library does give tiny hints about how they're related, I can't poke and prod around them to find what the values mean because I don't own one.
Thx for the reply. Any chance we can get what you CAN get, added into the GPU section, based on the results of the "--list-adapters" output? Instead of: GPU 0: 429.2 / 391.7 Mh/s | A:50595 R:120 HW:3 U:5.30/m I:9 76.0 C F: 38% (1566 RPM) E: 960 MHz M: 300 Mhz V: 1.088V A: 99% P: 0%
how about something like this? GPU 0: 429.2 / 391.7 Mh/s | A:50595 R:120 HW:3 U:5.30/m I:9 76.0 C F: 38% (1566 RPM) E: 960 MHz M: 300 Mhz V: 1.088V A: 99% P: 0% CT: AMD Radeon HD 6900 Series X-ID: 0. (08:00.0)
|
|
|
|
-ck (OP)
Legendary
Offline
Activity: 4242
Merit: 1644
Ruu \o/
|
|
January 12, 2012, 09:26:52 PM |
|
going back to the problem I had o the last page.... I installed/re-installed sdk 2.5 then deleted and re-installed 2.1.2 - got my 20MH/s back I noticed that deleting the .bin (which I assume are the engines?) caused the cant find opencl error, and for some reason wouldn't re-generate them - only after re-installing 2.1.2 did they re-appear. I guess the .bin files generated before I installed the last catalyst were still being used, hence i didn't see the performance drop some have. after deleting the .bins and they had to be generated they used the slower 2.6 sdk libs. With catalyst 11.12 (including sdk) I get reduced hashrate with cgminer 2.1.2. If I use cgminer 2.1.0, I not only get my performance back, I get a slight boost over older drivers. I've replicated this on 4 machines so far. I didn't bother checking 2.1.1 because I am lazy. The binary kernel that is generated occurs only ONCE when you first run cgminer. That means that if you run cgminer with sdk 2.5 and then install sdk 2.6, it won't make any difference. However if you install sdk 2.6 first and then run cgminer for the first time, you'll get a binary kernel from the 2.6 sdk (i.e. the slower one). Since none of that code was changed in cgminer between those versions, I suspect that's what you're seeing. The newer *driver* may be faster, but the newer *sdk* is what causes the problem. Try deleting the *.bin files and start cgminer again and you'll see.
|
Developer/maintainer for cgminer, ckpool/ckproxy, and the -ck kernel 2% Fee Solo mining at solo.ckpool.org -ck
|
|
|
swapper
Newbie
Offline
Activity: 17
Merit: 0
|
|
January 12, 2012, 10:21:52 PM |
|
Hi. Do you also get that BSOD if you change the video card clocks back with Afterburner after quitting cgminer? cgminer changes the clocks back when you quit, so I wonder if changing them back causes the problem or if it is more likely related to cgminer directly.
No, I can change frequencies in Afterburner after quitting cgminer, and while cgminer is executing, without BSOD. Thanks.
|
|
|
|
swapper
Newbie
Offline
Activity: 17
Merit: 0
|
|
January 12, 2012, 11:42:27 PM |
|
Hi. I believe the problem is MSI Afterburner usually automatically sets Cayman based cards to a voltage of 1.175, as it does with my 6970. There is an option to select stock voltage setting, which is 1.15. Check what voltage it has it set for in your MSI Afterburner and what your default card voltage is, maybe you are giving your card a slight voltage boost unintentionally.
Try adding this argument in cgminer to see if the BSOD problem is fixed, --gpu-vddc 1.175, Also if you add --gpu-powertune 10, it should help stabilize it, but you should know best what powertune setting to use.
I have checked with GPUZ the voltage of my 6990 when cgminer is running is 1.175v, I think it's a correct value when the card switch is in overclock position. If I execute "cgminer.exe -o http://de.btcguild.com:8332 -u xxx -p xxx --donation 1.0 --temp-cutoff 85 --temp-overheat 80 --temp-target 60 -I 5 --gpu-vddc 1.175 " or "cgminer.exe -o http://de.btcguild.com:8332 -u xxx -p xxx --donation 1.0 --temp-cutoff 85 --temp-overheat 80 --temp-target 60 -I 5 --gpu-vddc 1.175 --gpu-powertune 10", after quitting cgminer I receive a BSOD also, when MSI Afterburner is running, and when Afterburner is not started (and the card clocks are in the default 880/1250 Mhz) The only strange thing I can see is the fan speed: http://img560.imageshack.us/img560/498/cgminerfan.png . My 6990 is watercooled and without fan, but cgminer shows a GPU0 fan of 960 RPM and GPU1 fan of a strange number. Thanks.
|
|
|
|
Brunic
|
|
January 13, 2012, 01:22:02 AM |
|
Hi guys,
I don't why or how, but using cgminer on my miners seems to overload my network or something like that.
If I use cgminer on one or two workers, everything is fine. But usually, as soon as I use it for 4 workers or more, my network jam completely. In cgminer, I see "Lost connection to pool" a bunch of others message linked to network failure. Everything fails so much that even computers on the network that don't mine (like my work computer) get an horrible latency while browsing the web, getting easily "Could not find this website blablabla".
Cgminer 2.1.1 provoke this phenomenon, and 2.1.2 didn't seem to correct it either.
I've used Phoenix for more than 6 months and I never had a similar problem. When I got this problem, I've reverted all my miners back to Phoenix, and everything went fine. All my miners and my computer are connected into a hub, and that hub is connected to the home DSL modem.
I'm really interested in using cgminer, with the new RPC interface. If you guys could help me diagnose the problem, or suggest me things to try, I'll appreciate it!
|
|
|
|
DeathAndTaxes
Donator
Legendary
Offline
Activity: 1218
Merit: 1079
Gerald Davis
|
|
January 13, 2012, 01:31:14 AM Last edit: January 13, 2012, 01:51:00 AM by DeathAndTaxes |
|
Brunic do you have a router? You say the hub (switch?) is connected directly to the DSL modem.
Some routers especially older crappier ones can't sustain a large number of simultaneous connections.
The problem you describe I have seen described with users having issues w/ bittorrent. I don't know how many connections cgminer creates but the problem sounds the same.
The other thing is you said you use 4 workers? You aren't running multiple copies of cgminer on 1 rig are you? That is a completely wasteful way to run. 1 rig = 1cgminer instance = 1 worker.
How many total rigs do you have and how many total GPUs?
|
|
|
|
-ck (OP)
Legendary
Offline
Activity: 4242
Merit: 1644
Ruu \o/
|
|
January 13, 2012, 01:39:12 AM |
|
You can halve the number of connections potentially by decreasing threads per GPU from 2 to 1
--gpu-threads|-g <arg> Number of threads per GPU (1 - 10) (default: 2)
Try setting it to 1. I know some miners have a little pause between connections to cope (like diablominer) whereas cgminer is absurdly aggressive with trying to keep the GPU busy - this is also why it's more suitable for mining with p2pool. Perhaps yet another option is required...
|
Developer/maintainer for cgminer, ckpool/ckproxy, and the -ck kernel 2% Fee Solo mining at solo.ckpool.org -ck
|
|
|
Brunic
|
|
January 13, 2012, 02:56:11 AM |
|
Brunic do you have a router? You say the hub (switch?) is connected directly to the DSL modem.
Some routers especially older crappier ones can't sustain a large number of simultaneous connections.
The problem you describe I have seen described with users having issues w/ bittorrent. I don't know how many connections cgminer creates but the problem sounds the same.
The DSL modem also act as a router. My modem: http://assistance.sympatico.ca/index.cfm?language=en&method=content.view&category_id=674&content_id=11213My hub: http://www.superwarehouse.com/Linksys_Etherfast_II_10_100_24_Port_Auto-Sensing_Hub/EF2H24/p/44854The hub is connected into the Bell modem. The other thing is you said you use 4 workers? You aren't running multiple copies of cgminer on 1 rig are you? That is a completely wasteful way to run. 1 rig = 1cgminer instance = 1 worker.
How many total rigs do you have and how many total GPUs? No, I'm running only 1 instance of cgminer. When I said 4 workers, I meant 4 different mining rigs. I have a total of 12 mining rigs with 46 GPU total. On my 12 mining rigs, I used Phoenix 1.7.2. Never had any problem related to network with this miner. I've decided to install cgminer. If I test it on 1 or 2 mining rig, I don't have any problems at all. With 3 mining rig, it seems ok, rarely having network problem (but you see them starting to appear). At 4 mining rigs, it really gets bad and at 5 mining rigs with cgminer, the network is completely dead. The network is completely jammed, even on computers not mining. You can halve the number of connections potentially by decreasing threads per GPU from 2 to 1
--gpu-threads|-g <arg> Number of threads per GPU (1 - 10) (default: 2)
Try setting it to 1. I know some miners have a little pause between connections to cope (like diablominer) whereas cgminer is absurdly aggressive with trying to keep the GPU busy - this is also why it's more suitable for mining with p2pool. Perhaps yet another option is required... I'll try that and let you know.
|
|
|
|
DeathAndTaxes
Donator
Legendary
Offline
Activity: 1218
Merit: 1079
Gerald Davis
|
|
January 13, 2012, 03:11:28 AM |
|
You aren't going to like it Brunic BUT it likely is your POS router built in to your modem. The good news is it is possible to get a router and turn off all routing functions in the modem and get a router which has the ability to handle the high number of simultaneous connections. It is called bridging the modem and can be "fun" depending on how crappy the modem is and how poor the documentation is. An example: http://www.no-ip.com/support/guides/routers/bridging_your_dsl_modem.htmlIf it were up to me modems w/ built in routers would be deemed a crime against humanity.
|
|
|
|
-ck (OP)
Legendary
Offline
Activity: 4242
Merit: 1644
Ruu \o/
|
|
January 13, 2012, 03:14:44 AM |
|
It does indeed sound like your network is simply overloaded. cgminer hits the network very hard, and with that many GPUs, especially after a longpoll, they'll all be looking for work at the same time. If it happens at twice as many GPUs when you decrease threads to 1, then that will be diagnostic.
|
Developer/maintainer for cgminer, ckpool/ckproxy, and the -ck kernel 2% Fee Solo mining at solo.ckpool.org -ck
|
|
|
Brunic
|
|
January 13, 2012, 03:48:31 AM |
|
Yeah, even with the threads at 1, it didn't change anything. As soon as I get 4 mining rigs working (or around 16 gpus), apocalypse comes say hello.
I'll look into bridging the modem. I have a couple of router in a box here, I'm going to try that.
Thanks!
|
|
|
|
kano
Legendary
Offline
Activity: 4592
Merit: 1851
Linux since 1997 RedHat 4
|
|
January 13, 2012, 05:53:42 AM |
|
You aren't going to like it Brunic BUT it likely is your POS router built in to your modem. The good news is it is possible to get a router and turn off all routing functions in the modem and get a router which has the ability to handle the high number of simultaneous connections. It is called bridging the modem and can be "fun" depending on how crappy the modem is and how poor the documentation is. An example: http://www.no-ip.com/support/guides/routers/bridging_your_dsl_modem.htmlIf it were up to me modems w/ built in routers would be deemed a crime against humanity. Although I bridge my ADSL modem through my linux 'router' (rp-pppoe of course) I would never do that for any of my clients. As soon as they have internet problems and they call their ISP, having a bridged modem either ends the support conversation, or the support ask them to plug it directly into their computer ... and that's the end of their internet ... until I go visit them - why I stopped doing it. Maybe for someone who knows what they are doing, bridging is good, but for others there can be issues ...
|
|
|
|
jjiimm_64
Legendary
Offline
Activity: 1876
Merit: 1000
|
|
January 13, 2012, 06:32:36 AM |
|
Yeah, even with the threads at 1, it didn't change anything. As soon as I get 4 mining rigs working (or around 16 gpus), apocalypse comes say hello.
I'll look into bridging the modem. I have a couple of router in a box here, I'm going to try that.
Thanks!
I had to buy a new router and install tomato on it before I could use cgminer for more then a couple of rigs.. no problems since.
|
1jimbitm6hAKTjKX4qurCNQubbnk2YsFw
|
|
|
Brunic
|
|
January 13, 2012, 06:34:43 AM |
|
You aren't going to like it Brunic BUT it likely is your POS router built in to your modem. The good news is it is possible to get a router and turn off all routing functions in the modem and get a router which has the ability to handle the high number of simultaneous connections. It is called bridging the modem and can be "fun" depending on how crappy the modem is and how poor the documentation is. An example: http://www.no-ip.com/support/guides/routers/bridging_your_dsl_modem.htmlIf it were up to me modems w/ built in routers would be deemed a crime against humanity. Yeaaaaah! It worked! It was the shitty Bell modem who caused all of this.
|
|
|
|
jjiimm_64
Legendary
Offline
Activity: 1876
Merit: 1000
|
|
January 13, 2012, 06:38:20 AM |
|
Has anyone had problems connecting to the rpc port for cgminer? I continue to get the following error: this is using the api-example.php from a machine on the network with access to the rig "216ERR: socket connect(miner16,4028) failed 'No connection could be made because the target machine actively refused it. ' " I have the latest miner: root@linuxcoin:/home/user/cgminer-2.1.2-x86_64-built# here is my conf file that I start the miner with { "pools" : [ { "url" : "http://us.eclipsemc.com:8337", "user" : "jjiimm_64_miner16", "pass" : "mypass" }, { "url" : "http://192.168.1.100:8332", "user" : "myuser", "pass" : "mypass" } ],
"intensity" : "7", "gpu-engine" : "850", "gpu-fan" : "85", "gpu-memclock" : "300", "api-network" : true, "api-port" : "4028", "auto-fan" : true }
I am using linuxcoin on a stick.. I am sure I have to push or pull some knob in linux, any help would be appreciated.
|
1jimbitm6hAKTjKX4qurCNQubbnk2YsFw
|
|
|
sveetsnelda
|
|
January 13, 2012, 06:55:15 AM |
|
Looks like you forgot "api-listen" : true in there.
|
14u2rp4AqFtN5jkwK944nn741FnfF714m7
|
|
|
jjiimm_64
Legendary
Offline
Activity: 1876
Merit: 1000
|
|
January 13, 2012, 06:58:38 AM |
|
Looks like you forgot "api-listen" : true in there.
Thank you sir. 1btc on its way! edit: i figured if you had api-network it was assumed you wanted it to listen
|
1jimbitm6hAKTjKX4qurCNQubbnk2YsFw
|
|
|
sveetsnelda
|
|
January 13, 2012, 07:19:43 AM |
|
woah... Wasn't excpecting that. Thanks!! First donation. Woo! I'd gotten so used to the trolling and negativity on this forum that I forgot generous and fair people still exist...
|
14u2rp4AqFtN5jkwK944nn741FnfF714m7
|
|
|
|