Also note that testing how long it will take using a Share count is not consistent.
On the same setup, 1000 Shares can take varying amounts of time.
A share is random, just like a block (it is a difficulty "1" block)
Utility will vary also.
Only the actual average hash rate should be somewhat consistent.
Topic: OFFICIAL CGMINER mining software thread for linux/win/osx/mips/arm/r-pi 4.11.0 - page 772. (Read 5805728 times)
September 20, 2011, 03:40:46 AM
September 19, 2011, 09:32:05 PM
But going 300 to 150 is arguably negligible. I saved 2w according to my meter. Until i get my script written when i get home i wouldn't bother with it, not worth the time imo.
September 19, 2011, 07:41:33 PM
Well cgminer is reporting 128mhz 
Updated above post with 12hour report, seems 128mhz has no effect on utility or hash rates. I'll have to do a proper test on my single card machine as the one below puts out too much heat and is constantly throttled.
I see, but it's probably more beneficial to play with volts and the core for that. I dont think memory clocks use a big amount of power.
Updated above post with 12hour report, seems 128mhz has no effect on utility or hash rates. I'll have to do a proper test on my single card machine as the one below puts out too much heat and is constantly throttled.
Code:
cgminer version 2.0.3 - Started: [2011-09-19 23:58:37]
--------------------------------------------------------------------------------
(5s):1287.1 (avg):1278.0 Mh/s | Q:6109 A:11210 R:29 HW:0 E:183% U:17.54/m
TQ: 8 ST: 9 SS: 0 DW: 424 NB: 58 LW: 29525 GF: 0 RF: 0 I: 8
Connected to http://us.eclipsemc.com:8337 with LP as user kripz_miner
Block: 000008663d81f4c02b0623cc45a8a72b... Started: [10:07:52]
--------------------------------------------------------------------------------
[P]ool management [G]PU management ettings [D]isplay options [Q]uit
GPU 0: 65.0C 2969RPM | 338.6/339.0Mh/s | A:2981 R:6 HW:0 U:4.66/m
GPU 1: 87.0C 3083RPM | 299.2/296.0Mh/s | A:2546 R:12 HW:0 U:3.98/m
GPU 2: 72.0C 2999RPM | 340.4/340.3Mh/s | A:3067 R:9 HW:0 U:4.80/m
GPU 3: 87.0C 3164RPM | 305.7/302.8Mh/s | A:2617 R:2 HW:0 U:4.09/m
--------------------------------------------------------------------------------
Quote
Right, what I'm saying is supplemental to that -- a third factor if you will. What I'm saying is that even though you might get slightly higher hash rates at, say 300 with that worksize as a sweet spot, that depending on your electricity costs, you might benefit more from setting the memory clock as low as it will go, say to 150, even though your hash rate will drop slightly.
I see, but it's probably more beneficial to play with volts and the core for that. I dont think memory clocks use a big amount of power.
September 19, 2011, 07:40:33 PM
I'd compare the differences in wattage at those different memory clock speeds. It might still net you more bitcoins to run the memory clock as low as possible, after accounting for electricity costs.
Not sure if you're interpreting the graph correctly. The point of my post is that according to the graph, if you have the mem clocks at 300 (which most miners do), it's best to run VECTORS2 with worksize of 128 or 256 (depending on the bus width of your graphics card).
Right, what I'm saying is supplemental to that -- a third factor if you will. What I'm saying is that even though you might get slightly higher hash rates at, say 300 with that worksize as a sweet spot, that depending on your electricity costs, you might benefit more from setting the memory clock as low as it will go, say to 150, even though your hash rate will drop slightly.
September 19, 2011, 06:56:08 PM
I better check that then, aticonfig i assume.
This is why cgminer reports back the actual speeds to you after you set them on the fly in the menu, as the driver can lie, saying it set the speed fine but the card just refuses it. September 19, 2011, 06:53:46 PM
I better check that then, aticonfig i assume.
Had a good idea last night, will attempt to write a quick bashscript to tell cgminer to mine 1000 shares and then get the script to log the final output to a file and repeat with different settings.
pseudocode:
Had a good idea last night, will attempt to write a quick bashscript to tell cgminer to mine 1000 shares and then get the script to log the final output to a file and repeat with different settings.
pseudocode:
Code:
Array speedstoTest = 0,100,150,200,250,300,350
for i = 0; i < speedstoTest.length; i++
{
if(speedstoTest[i] == 0)
// run normal cgminer no downclocks
else
cgminer -m `speedstoTest[i]` --shares 1000 > log
}
September 19, 2011, 12:19:30 PM
I completely missed the 2.0.3 announcement.
After updating I noticed that the fans spun up to full speed and stayed there.
Also I got the "Failed to init GPU thread 0, disabling device 0" that I got with 2.0.1 and 2.0.2. So I copied the .bin files that 2.0.0 generated into the directory and it started up correctly.
Pressing Q to close the program still causes it to crash on Windoze XP.
When the program starts the fans go to full speed and settle down to a reasonable speed and sound level, I guess its supposed to do that?
Thanks for the update,
Sam
After updating I noticed that the fans spun up to full speed and stayed there.
Also I got the "Failed to init GPU thread 0, disabling device 0" that I got with 2.0.1 and 2.0.2. So I copied the .bin files that 2.0.0 generated into the directory and it started up correctly.
Pressing Q to close the program still causes it to crash on Windoze XP.
When the program starts the fans go to full speed and settle down to a reasonable speed and sound level, I guess its supposed to do that?
Thanks for the update,
Sam
September 19, 2011, 09:10:59 AM
Well i just tried 128mhz (just because its a power of 2). Power usage is down 2 watts. Keep in mind cgminer is throttling my core speeds due to heat so take with a grain of salt.
300mhz
128mhz
I'm wondering how low i can go before i have problems. So far there doesnt seem to be a disadvantage to such low memory clocks (dedicated nix miner with no desktop installed so i dont notice any problems), i think on windows the desktop will be unusable.
Wish i had the time to play around with more settings.
300mhz
Quote
cgminer version 2.0.3 - Started: [2011-09-18 20:31:54]
--------------------------------------------------------------------------------
(5s):1258.0 (avg):1302.4 Mh/s | Q:15835 A:29526 R:42 HW:0 E:186% U:17.94/m
TQ: 8 ST: 8 SS: 0 DW: 1045 NB: 150 LW: 68082 GF: 0 RF: 2 I: 8
Connected to http://us.eclipsemc.com:8337 with LP as user kripz_miner
Block: 000004b8e281c024a40ffc939f4f4c87... Started: [23:28:17]
--------------------------------------------------------------------------------
[P]ool management [G]PU managementettings [D]isplay options [Q]uit
GPU 0: 67.5C 2985RPM | 343.0/343.3Mh/s | A:7795 R:7 HW:0 U:4.74/m
GPU 1: 88.0C 3087RPM | 295.7/305.2Mh/s | A:6835 R:12 HW:0 U:4.15/m
GPU 2: 74.5C 3001RPM | 342.9/342.9Mh/s | A:7807 R:16 HW:0 U:4.74/m
GPU 3: 86.5C 3166RPM | 286.7/311.1Mh/s | A:7092 R:7 HW:0 U:4.31/m
--------------------------------------------------------------------------------
--------------------------------------------------------------------------------
(5s):1258.0 (avg):1302.4 Mh/s | Q:15835 A:29526 R:42 HW:0 E:186% U:17.94/m
TQ: 8 ST: 8 SS: 0 DW: 1045 NB: 150 LW: 68082 GF: 0 RF: 2 I: 8
Connected to http://us.eclipsemc.com:8337 with LP as user kripz_miner
Block: 000004b8e281c024a40ffc939f4f4c87... Started: [23:28:17]
--------------------------------------------------------------------------------
[P]ool management [G]PU management
GPU 0: 67.5C 2985RPM | 343.0/343.3Mh/s | A:7795 R:7 HW:0 U:4.74/m
GPU 1: 88.0C 3087RPM | 295.7/305.2Mh/s | A:6835 R:12 HW:0 U:4.15/m
GPU 2: 74.5C 3001RPM | 342.9/342.9Mh/s | A:7807 R:16 HW:0 U:4.74/m
GPU 3: 86.5C 3166RPM | 286.7/311.1Mh/s | A:7092 R:7 HW:0 U:4.31/m
--------------------------------------------------------------------------------
128mhz
Quote
cgminer version 2.0.3 - Started: [2011-09-19 23:58:37]
--------------------------------------------------------------------------------
(5s):1287.1 (avg):1278.0 Mh/s | Q:6109 A:11210 R:29 HW:0 E:183% U:17.54/m
TQ: 8 ST: 9 SS: 0 DW: 424 NB: 58 LW: 29525 GF: 0 RF: 0 I: 8
Connected to http://us.eclipsemc.com:8337 with LP as user kripz_miner
Block: 000008663d81f4c02b0623cc45a8a72b... Started: [10:07:52]
--------------------------------------------------------------------------------
[P]ool management [G]PU managementettings [D]isplay options [Q]uit
GPU 0: 65.0C 2969RPM | 338.6/339.0Mh/s | A:2981 R:6 HW:0 U:4.66/m
GPU 1: 87.0C 3083RPM | 299.2/296.0Mh/s | A:2546 R:12 HW:0 U:3.98/m
GPU 2: 72.0C 2999RPM | 340.4/340.3Mh/s | A:3067 R:9 HW:0 U:4.80/m
GPU 3: 87.0C 3164RPM | 305.7/302.8Mh/s | A:2617 R:2 HW:0 U:4.09/m
--------------------------------------------------------------------------------
--------------------------------------------------------------------------------
(5s):1287.1 (avg):1278.0 Mh/s | Q:6109 A:11210 R:29 HW:0 E:183% U:17.54/m
TQ: 8 ST: 9 SS: 0 DW: 424 NB: 58 LW: 29525 GF: 0 RF: 0 I: 8
Connected to http://us.eclipsemc.com:8337 with LP as user kripz_miner
Block: 000008663d81f4c02b0623cc45a8a72b... Started: [10:07:52]
--------------------------------------------------------------------------------
[P]ool management [G]PU management
GPU 0: 65.0C 2969RPM | 338.6/339.0Mh/s | A:2981 R:6 HW:0 U:4.66/m
GPU 1: 87.0C 3083RPM | 299.2/296.0Mh/s | A:2546 R:12 HW:0 U:3.98/m
GPU 2: 72.0C 2999RPM | 340.4/340.3Mh/s | A:3067 R:9 HW:0 U:4.80/m
GPU 3: 87.0C 3164RPM | 305.7/302.8Mh/s | A:2617 R:2 HW:0 U:4.09/m
--------------------------------------------------------------------------------
I'm wondering how low i can go before i have problems. So far there doesnt seem to be a disadvantage to such low memory clocks (dedicated nix miner with no desktop installed so i dont notice any problems), i think on windows the desktop will be unusable.
Wish i had the time to play around with more settings.
September 19, 2011, 08:59:42 AM
Actually the 1/3rd rule has proven to be sheer coincidence. There is no obvious relationship between the GPU and memory clock speeds that equates to any 1:3 equation. It just happens that there's some kind of performance peak if you lower memory speed (on some, not all) ATI cards to ~35-40% of the speed, not to mention the heat advantages.
September 19, 2011, 08:55:54 AM
Probably because nobody bothered to benchmark and compare.
I'll try your suggestions and see how it goes.
Diablo who writes another mining program says mem should be core mhz divide 3 otherwise it's not optimal because of sync/timing or something
I'll try your suggestions and see how it goes.
Diablo who writes another mining program says mem should be core mhz divide 3 otherwise it's not optimal because of sync/timing or something
September 19, 2011, 05:48:14 AM
I have thoroughly tested all 11 of my GPU's with all different settings and conditions. I used to set memclock to 300, not sure why this has become the go-to number, but I have found 180 does not lower speed on any cards, increases speed on some cards (compared to 300) and of course uses less electricity and creates less heat.
I have been mining for about 6 months now with these cards, and these are the optimum settings I have found. Hope this helps some of you as a reference.
(GPU) | (Core/Mem/FanSpeed) | (Mhash/Temp)
(4) Saphire Vapor-X 5870 | 970/180/75% | 445-470/60s(c)
(2) AMD Reference 6870 | 1012/180/75% | 320-340/60s(c)
(2) Saphire 6870 1012/180/75% | 310-320/60s(c)
(1) Saphire 2gb 5870 Eyefinity Six Edition | 970/180/75% | 445-470/60s(c)
(1) AMD Reference 5850 | 875/180/75% | 350-370/60s(c)
(1) XFX 5830 | 980/180/75% | 305-315/60s(c)
(note: the temps are low because I have very good cooling setup, including a/c and multiple fan exhaust system - you could never run this many cards in the same room as each other without air conditioning)
and of course these speeds are all using CGMINER which is the best miner in the world.
I have been mining for about 6 months now with these cards, and these are the optimum settings I have found. Hope this helps some of you as a reference.
(GPU) | (Core/Mem/FanSpeed) | (Mhash/Temp)
(4) Saphire Vapor-X 5870 | 970/180/75% | 445-470/60s(c)
(2) AMD Reference 6870 | 1012/180/75% | 320-340/60s(c)
(2) Saphire 6870 1012/180/75% | 310-320/60s(c)
(1) Saphire 2gb 5870 Eyefinity Six Edition | 970/180/75% | 445-470/60s(c)
(1) AMD Reference 5850 | 875/180/75% | 350-370/60s(c)
(1) XFX 5830 | 980/180/75% | 305-315/60s(c)
(note: the temps are low because I have very good cooling setup, including a/c and multiple fan exhaust system - you could never run this many cards in the same room as each other without air conditioning)
and of course these speeds are all using CGMINER which is the best miner in the world.
September 19, 2011, 02:45:12 AM
So i saw this graph:
Now my 5850 is a 256bit card and cgminer detects the worksize as 128. I manually set it to 256 and now getting 20mh/s more per card (4x 5850).
Mem at 300mhz. Results may vary. Possibly placebo.
Now my 5850 is a 256bit card and cgminer detects the worksize as 128. I manually set it to 256 and now getting 20mh/s more per card (4x 5850).
Mem at 300mhz. Results may vary. Possibly placebo.
I'd compare the differences in wattage at those different memory clock speeds. It might still net you more bitcoins to run the memory clock as low as possible, after accounting for electricity costs.
Not sure if you're interpreting the graph correctly. The point of my post is that according to the graph, if you have the mem clocks at 300 (which most miners do), it's best to run VECTORS2 with worksize of 128 or 256 (depending on the bus width of your graphics card).
September 19, 2011, 01:16:01 AM
Or, an even less elegant solution is just scheduling a 'kill task' and 'start task' every 30 minutes regardless of the state of things. Nothing fancy, but it works.
Yeah, at one point I was even rebooting my computer every 3 hours. I like to see the accumulated stats, though, so I don't do anything like that now.
September 19, 2011, 01:13:35 AM
So i saw this graph:
Now my 5850 is a 256bit card and cgminer detects the worksize as 128. I manually set it to 256 and now getting 20mh/s more per card (4x 5850).
Mem at 300mhz. Results may vary. Possibly placebo.
Now my 5850 is a 256bit card and cgminer detects the worksize as 128. I manually set it to 256 and now getting 20mh/s more per card (4x 5850).
Mem at 300mhz. Results may vary. Possibly placebo.
I'd compare the differences in wattage at those different memory clock speeds. It might still net you more bitcoins to run the memory clock as low as possible, after accounting for electricity costs.
September 19, 2011, 12:46:46 AM
in windows7, how do u guys restart your cgiminer every 30 mins? in case there is card that crash, at least it restarts
I'm using a scheduled task that runs a batch file using grep for Windows to check the log every so often for the word DEAD. If it finds a line containing that, I have another set of batch files that then run to make a backup copy of the log for later analysis, clearing the log, and restarting cgminer. Not the most elegant solution, but it seems to do the job.
September 19, 2011, 12:10:14 AM
Try clearing out your drivers entirely with some kind of driver removal program whatever they're called on windows, and install the 11.6 driver instead.
September 19, 2011, 12:08:35 AM
When the fan is set to auto adjust speed in cgminer ...whenever it issues the fan change speed windows UI becomes extremely laggy, is there any way to prevent this? Thanks
running it on windows here too (xp, vista and 7) and no increase in screen lag when it changes fan speed.
what card and settings?
HD 6950 890mhz / 150mhz stock voltage.
Vista 32 , 11.8 drivers.
Thats all I can think of...
September 18, 2011, 05:05:04 PM
When the fan is set to auto adjust speed in cgminer ...whenever it issues the fan change speed windows UI becomes extremely laggy, is there any way to prevent this? Thanks
running it on windows here too (xp, vista and 7) and no increase in screen lag when it changes fan speed.
what card and settings?
September 18, 2011, 01:09:20 PM
When the fan is set to auto adjust speed in cgminer ...whenever it issues the fan change speed windows UI becomes extremely laggy, is there any way to prevent this? Thanks
September 18, 2011, 06:23:14 AM
From my point of view there is no doubt, that CGMINER become the best available miner on the market (at least for now), and it is a pity that it does not work with P2POOL (hopefully designer of P2POOL will fix that)
Jump to: