 Show Posts
|
|
Pages: « 1 2 [3] 4 »
|
what method are you guys using to have a fixed core clock ?
i set it up for let's say 1400mhz for vega56 using overdriventool but when watching it with HWiNFO is not the same it's about 1350Mhz
any hint appreciated
from yesterday I experience same problem with Vega 56 / 64 cards. Have you tried disabling all lower p-states in OverdriveNTool, i.e. click the little text "P2", "P3" etc left of the MHz column so they are greyed out? Only using the highest state has helped me sometimes to force the driver to use the specified clocks. Yes kerney, that is my standard settings in OvedriveNtool. Just did some more test and found, when applying overdrive and look in the HWiNFO64 i see that GPUP7 clocks is set on higher values of set in overdrive.ini ? And when i start miner values go down? Down in profiles of ovedrivetool.ini i added actual clock values read by HWiNFO before and after start miner: Run_OverdriveNTool.bat (OverdriveNTool.exe -r1 -r2 -r3 -r4 -p1Sapphire -p2Nitro_RX570 -p3Asus_Vega64 -p4Gigabyte) GPU_P6=1212;905;0 before start miner after start miner GPU_P7=1408;955 1437;955 1378 and 1381 (changeing) GPU_P7=1490;930 1535;930 1463 and 1466 and 1467 (changeing) GPU_P7=1408;950 1444;950 1381 and 1382 and 1385 (changeing) This is standard behavior of Vegas, due to ACG/AVFS. The lower your voltages, the more clock throttling under load. This has nothing to do with the miner, other than the fact that it puts a certain signature load on the GPU. Thank you pbfarmer that makes sense. You force me to learn more of Advanced Clock Generator(ACG)/Advanced Voltage Frequency Scaling(AVFS) :-) Never noticed that behavior before in HWiNFO. Found some interesting stuff here: https://www.reddit.com/r/Amd/comments/7dzc9j/vega_56_w_64_bios_tweak_result_good_chip_or/ |
|
|
|
V1.7.0
- Speed increase on BitTube/Haven/Heavy/Italo algos up to ~10% (depends of gpu)
+ Performance boost for BitTube/Haven/Heavy/Italo algos up to ~10%, some cards gained 1% some 10%, i don't have exact numbers, you should try it out and see.
Vegas gained the most, also i hope that i fixed the Ellesmere 8Gb cards problem, not reaching max hashing speed. If you had 1k+ on some earlier versions, you should be able to reach it now again but you MUST set fragments to 0 if you are setting it by hand, otherwise it probably won't reach max speeds.
And no fee increase because of the extra performance, it stays at ~0.85%
Nice work "Dok" on heavy Algo (Bit Tube 2), on same settings Vega cards got impressive H/s about 10% Rx570 4GB got 3% :-) But... stress on the cards is increased so must tuning again :-( I have stable rig in bittube2 for days, but now restarted once on hour... That leads me to this: can you please consider my recommendation:1. put mining fee to your adress all the time of mining like TRM do? So, if miner restart miner's will not lose profit for additional fees. Extra bonus for miners and dev is that we will never disconnect and connect to pool (my test with supportxmr.com pool show benefit in income, because always connected to pool) 2. Increase fee to 1,00% because you will loose some fee now earning because restarting miner and mining fee for first 15 minutes. Please... |
|
|
|
what method are you guys using to have a fixed core clock ?
i set it up for let's say 1400mhz for vega56 using overdriventool but when watching it with HWiNFO is not the same it's about 1350Mhz
any hint appreciated
from yesterday I experience same problem with Vega 56 / 64 cards. Have you tried disabling all lower p-states in OverdriveNTool, i.e. click the little text "P2", "P3" etc left of the MHz column so they are greyed out? Only using the highest state has helped me sometimes to force the driver to use the specified clocks. Yes kerney, that is my standard settings in OvedriveNtool. Just did some more test and found, when applying overdrive and look in the HWiNFO64 i see that GPUP7 clocks is set on higher values of set in overdrive.ini ? And when i start miner values go down? Down in profiles of ovedrivetool.ini i added actual clock values read by HWiNFO before and after start miner: Run_OverdriveNTool.bat (OverdriveNTool.exe -r1 -r2 -r3 -r4 -p1Sapphire -p2Nitro_RX570 -p3Asus_Vega64 -p4Gigabyte) [General] ShowBusNumber=1 ShowPnpString=0 ShowAdapterIndex=0 ShowFriendlyName=0 ShowRegistryKey=1 MainWindowLeft=984 MainWindowTop=244 [Profile_0] Name=Sapphire GPU_P0=852;900;0 GPU_P1=991;900;0 GPU_P2=1084;900;0 GPU_P3=1138;900;0 GPU_P4=1150;900;0 GPU_P5=1202;900;0 GPU_P6=1212;905;0 before start miner after start miner GPU_P7=1408;955 1437;955 1378 and 1381 (changeing) Mem_P0=167;900;0 Mem_P1=500;900;0 Mem_P2=800;900;0 Mem_P3=1088;945 Fan_Min=3500 Fan_Max=4900 Fan_Target=75 Fan_Acoustic=2400 Power_Temp=85 Power_Target=0 [Profile_1] Name=Nitro_RX570 GPU_P0=300;750;0 GPU_P1=588;765;0 GPU_P2=952;943;0 GPU_P3=1041;1012;0 GPU_P4=1106;1075;0 GPU_P5=1168;1143;0 GPU_P6=1209;1150;0 GPU_P7=1170;850 Mem_P0=300;750;0 Mem_P1=1000;800;0 Mem_P2=1970;900 Fan_Min=576 Fan_Max=2150 Fan_Target=50 Fan_Acoustic=1340 Power_Temp=85 Power_Target=0 [Profile_2] Name=Asus_Vega64 GPU_P0=852;900;0 GPU_P1=991;900;0 GPU_P2=1084;900;0 GPU_P3=1138;900;0 GPU_P4=1150;900;0 GPU_P5=1202;900;0 GPU_P6=1212;905;0 GPU_P7=1490;930 1535;930 1463 and 1466 and 1467 (changeing) Mem_P0=167;900;0 Mem_P1=500;900;0 Mem_P2=800;900;0 Mem_P3=1100;910 Fan_Min=2300 Fan_Max=3500 Fan_Target=50 Fan_Acoustic=1600 Power_Temp=85 Power_Target=0 [Profile_3] Name=Gigabyte GPU_P0=852;900;0 GPU_P1=991;900;0 GPU_P2=1084;900;0 GPU_P3=1138;900;0 GPU_P4=1150;900;0 GPU_P5=1202;900;0 GPU_P6=1212;905;0 GPU_P7=1408;950 1444;950 1381 and 1382 and 1385 (changeing) Mem_P0=167;900;0 Mem_P1=500;900;0 Mem_P2=800;900;0 Mem_P3=1070;950 Fan_Min=3350 Fan_Max=4900 Fan_Target=75 Fan_Acoustic=2400 Power_Temp=85 Power_Target=0 |
|
|
|
what method are you guys using to have a fixed core clock ?
i set it up for let's say 1400mhz for vega56 using overdriventool but when watching it with HWiNFO is not the same it's about 1350Mhz
any hint appreciated
from yesterday I experience same problem with Vega 56 / 64 cards. |
|
|
|
these guys with rx 570/580 hashing below 1kh/s i think are doing it wrong. check your straps and oc.
all of my rx 570 hashes above atleast 1kh be it 4gb or 8gb. the only problem i have is micron, which i think is the worst in cn variants.
Agree, matches what we found as well. Haven't really found any card stuck below 1kh/s, except for Micron mem where we've been stuck around 860 h/s on some cards whatever we try for straps and clocks. Samsung by far the easiest, Hynix often also just a copy of lower-to-higher straps. And, I'm really not that good with straps, I should add. Great job TeamRed! Read every post here, testing rig for about 8 hours without crash. Now trying to fine tune. Here is mixed rig (1 x rx570 4GB Elpida, 2 x Vega 56 referent + 1 Vega 64 Asus Strix) Vega 56 hashing about 2000 H/s, Vega 64 2240 H/s). Vega 56 have low clock and memory 1408/1088 because on SRB is overheating and crashing on more clocks. Vega 64 is on 1520/1100. Funny thing about RX570 is that hashrate of RX570 somehow depends of configuration on Vega cards. When i first start miner (without --cn_config) RX570 Nitro + (Elpida) hash 920 H/s with timings 1200/1980. Than I experiment and put on RX570 7+7 (default was 8+7) and lower clock and memory to 1225/1940 (as user Dragonmike recommended). Playing with OverdriveNtool when hashing, been able to get 942 H/s with 1200/1920 but i think that time i put Vega 56 to 15+15. When i restarted rig, RX 570 hash just 880 H/s? Try appy OverdriveNtool with same clocks as before 1200/1920 but still 880 H/s. Than i apply 1250/1980 and still 880 H/s. mV for clock and memory 900 mV. What am i missing and is it possible and where to delete kernel files to make automatic initialization of the cards again? That sounds a little weird. I don’t know if you restarted the miner or rebooted the rig in between runs? One issue we have (which I’m a little ashamed of tbh) is our unclean shutdown. It has led to me losing perf between restarts at times. Can you reproduce your test, i.e. whenever you run the 570 only after a clean reboot you hit ~940 h/s, but 570+Vegas after a clean reboot and it’s lower for the 570? Yes, something with RX570 4GB is weird. Tested a little, two times rebooted the whole rig and whatever clocks applied max hash is just 880 H/s. Than rebooted third time, and my Task Scheduler automatic run SRB Miner.bat and apply Overdrive profile and my RX570 hash 910 H/s in V8 which is normal to me in SRB Miner for this card. And than weird again :-) start TRM and RX570 hash with same 910 H/s like in SRB :-) Bump from 880 to 910.I leave it that way from last night almost 24 hours stable running: [2018-11-06 21:28:13] Stats GPU 0 - cnv8: 910.2 h/s, avg 906.8 h/s, pool 924.0 h/s 386/0 [2018-11-06 21:28:13] Stats GPU 1 - cnv8: 1.998kh/s, avg 1.996kh/s, pool 1.895kh/s 787/0 [2018-11-06 21:28:13] Stats GPU 2 - cnv8: 2.202kh/s, avg 2.242kh/s, pool 2.232kh/s 926/0 [2018-11-06 21:28:13] Stats GPU 3 - cnv8: 1.997kh/s, avg 2.002kh/s, pool 2.002kh/s 833/0 [2018-11-06 21:28:13] Stats Total - cnv8: 7.107kh/s, avg 7.148kh/s, pool 7.053kh/s Here is this (for me) stable Setup of mixed rig with 730W from the wall (-50W and + 300 H/s): Win 10, Adrenalin 18.6.1 GPU 0 (RX 570 4GB Elpida Nitro+) 7+7 core 1180/850 mV memory: 1980/900 mV. One click strap. GPU 1 (Vega 56 ref. with 64 bios ) 16+14 core 1408/950 mV memory: 1070/950 mV GPU 2 (Vega 64 Asus Strix) 16+14 core 1520/920 mV memory: 1100/910 mV GPU 3 (Vega 56 ref. with 64 bios ) 16+14 core 1408/955 mV memory: 1088/945 mV As you can see, Vega 56 ref with Samsung mem can in your rig hash much better with more clock on core and mem, i must lower it because rig is in small space and overheating. Waiting for the winter to come :-) I tested intensity and found that Vega 56 Reference even on this lower clocks still more like 16+14, than 15+15. |
|
|
|
Dok... JCE and Teamredminer are really upping their game with Heavy and CNv8.
Vega 56 1750h/s on heavy
Vega 64 2200h/s on CNv8
Are you planning any performance increases with SRB soon?
Hmm, come on mk111, you expect to Doktor do miracle (once again), and expect it today? Think what U can do to improve miner, lets start with reading Doktor post on this same page: are you going to reduce power usee like team red miner did, it uses 10 to 15 percent power same hash
too unstable for me tho compared to srb
Well that miner gave me homework to do, so i am working on something since then. It will definitely increase the hashrate, but don't know about the power consumption. we will see. |
|
|
|
these guys with rx 570/580 hashing below 1kh/s i think are doing it wrong. check your straps and oc.
all of my rx 570 hashes above atleast 1kh be it 4gb or 8gb. the only problem i have is micron, which i think is the worst in cn variants.
Agree, matches what we found as well. Haven't really found any card stuck below 1kh/s, except for Micron mem where we've been stuck around 860 h/s on some cards whatever we try for straps and clocks. Samsung by far the easiest, Hynix often also just a copy of lower-to-higher straps. And, I'm really not that good with straps, I should add. Great job TeamRed! Read every post here, testing rig for about 8 hours without crash. Now trying to fine tune. Here is mixed rig (1 x rx570 4GB Elpida, 2 x Vega 56 referent + 1 Vega 64 Asus Strix) Vega 56 hashing about 2000 H/s, Vega 64 2240 H/s). Vega 56 have low clock and memory 1408/1088 because on SRB is overheating and crashing on more clocks. Vega 64 is on 1520/1100. Funny thing about RX570 is that hashrate of RX570 somehow depends of configuration on Vega cards. When i first start miner (without --cn_config) RX570 Nitro + (Elpida) hash 920 H/s with timings 1200/1980. Than I experiment and put on RX570 7+7 (default was 8+7) and lower clock and memory to 1225/1940 (as user Dragonmike recommended). Playing with OverdriveNtool when hashing, been able to get 942 H/s with 1200/1920 but i think that time i put Vega 56 to 15+15. When i restarted rig, RX 570 hash just 880 H/s? Try appy OverdriveNtool with same clocks as before 1200/1920 but still 880 H/s. Than i apply 1250/1980 and still 880 H/s. mV for clock and memory 900 mV. What am i missing and is it possible and where to delete kernel files to make automatic initialization of the cards again? |
|
|
|
|
Doktore after 12 to 24 hours mining V8 algo, Miner just ends (program closed) without noted error in log file.
I want to try disable gpu watchdog and use "reboot_script_gpu_watchdog" to try restart miner (not reboot rig). "reboot_script_gpu_watchdog" : filename to a batch file in miner directory, if set it turns off built in miner reset procedure on gpu failure, and instead runs this script
Is this possible and where i can find working example of script for restarting srbminer?
Txs!
|
|
|
|
I'm just struggling with the Vegas.
Doesn't seem to matter whether I increase voltage or lower clocks, after a few minutes mining Monerov8, all GPUs go boom, i.e. SRBminer says hashrate is zero for every single card... and it will fail automatic re-initialisation.
Using 120/16/2. They're all 56's flashed to 64.
Maybe I'm just at the limit of what the PSU is able to cope with... Power draw is between 1400 and 1435 for a EVGA 1500 G2...
same here. Try 112/16/2, and Core to 1408, works for me... but maybe it is power related... |
|
|
|
Can someone help me to create a batch for set gpu and mem voltages?  + "startup_script" can be used to run a script (batch file) before miner starts. You can for example set clocks, voltages, etc. specific to that algo I'm also interested in how to set GPU and memory clocks and voltages with SRBMiner... Is it possible? Can't find any examples... First in start.bat ad this: --startupscript Run_OverdriveNTool.bat In Run_OverdriveNTool.bat you must set path in OverdriveNtool.ini with proper setting for your card and specific algo. Example: OverdriveNTool.exe -r1 -r2 -r3 -r4 -p1Sapphire_BT -p2Nitro_RX570_BT -p3Asus_Vega64_BT -p4Gigabyte_BT <-- on that way i set up miner for mining BitTube algo. Now set up OverdriveNtool.ini for specific card and algo (same name from OverdriveNtool.bat) Core and memory clocks are set with OverdriveNTool: https://forums.guru3d.com/threads/overdriventool-tool-for-amd-gpus.416116/From first page then open table to see best results: ) Here is a table maintained and edited by users with their settings for various cards : https://docs.google.com/spreadsheets/d/1pH326fwnYWC5Jef2FdF4KnJ4CLj0cwDDtGjrlgekk1A/edit#gid=514064850 |
|
|
|
what cooling you have for rig, what is the temperature core and HBM of the cards, and RPM?
Air cooling /open rig/ , 55 core temperature and 65 HBM temperature of the cards, and vga fan - auto /~3500 RPM/
txs. That explain everything. My rig are in small space, running hotter with 66 core temp. and over 77 HBM temp. For now i testing rig with lowering clocks and more mV. Here is 2 Vega 56 and one Vega Strix 64. Strix is better cooled and work without problems: GPU0 : 1877 H/s [T: 63c, RPM: 3577, CC: 1407 MHz, MC: 1050 MHz][BUS:3] - "more problematic Vega 56" GPU3 : 1940 H/s [T: 66c, RPM: 3516, CC: 1407 MHz, MC: 1095 MHz][BUS:14] - "other Vega 56" GPU2 : 2029 H/s [T: 56c, RPM: 3473, CC: 1442 MHz, MC: 1100 MHz][BUS:11] - Vega 64 Strix |
|
|
|
I am using latest version 1.6.8 and ati drivers 18.6.1. Mining rig is 4 x Vega 56 (flashed with Vega 64 bios). Core clock 1450@900mv memory clock 1100@900mv . /Hellae's powerplay table mod/ v7 - 8200 h/s @ 700 W at the wall. v8 - 7800 h/s @ 800 W at the wall. config_normalv8.txt { "cryptonight_type" : "normalv8", "intensity" : 0, "double_threads" : true, "target_temperature" : 55, "shutdown_temperature" : 80, "persistent_memory" : true, /*"gpu_conf" : [ { "id" : 0, "intensity" : 120, "worksize" : 16, "threads" : 2}, { "id" : 1, "intensity" : 120, "worksize" : 16, "threads" : 2}, { "id" : 3, "intensity" : 120, "worksize" : 16, "threads" : 2}, { "id" : 4, "intensity" : 120, "worksize" : 16, "threads" : 2} ]*/ } Check https://docs.google.com/spreadsheets/d/1pH326fwnYWC5Jef2FdF4KnJ4CLj0cwDDtGjrlgekk1A/edit#gid=514064850logicus that is great result  Please tell me, what cooling you have for rig, what is the temperature core and HBM of the cards, and RPM? |
|
|
|
Did you flash to 64 bios?
yes, of course. |
|
|
|
Vega owners, check your hotspot temps. My Vega didnt survive the night mining v8 even when I dropped the core to 1250@865mV
And a big fat F U to Monero devs
Over the night SRBMiner reset every cca 2 hours, ONE Vega 56 (gpu 0) drop hash to 0 and reset. And that is after I changed power and intensity. Try more mV on core and memory to see the result. Now testing OverdriveNtool: Problematic card Vega 56 reference: [Profile_0] Name=Sapphire GPU_P0=852;900 GPU_P1=991;900;0 GPU_P2=1084;900;0 GPU_P3=1138;900;0 GPU_P4=1150;900;0 GPU_P5=1202;900;0 GPU_P6=1212;905;0 GPU_P7=1407;950 Mem_P0=167;900;0 Mem_P1=500;900;0 Mem_P2=800;900;0 Mem_P3=1050;950 Fan_Min=3350 Fan_Max=4900 Fan_Target=75 Fan_Acoustic=2400 Power_Temp=85 Power_Target=0 Other Vega 56 reference: [Profile_3] Name=Gigabyte GPU_P0=852;900 GPU_P1=991;900;0 GPU_P2=1084;900;0 GPU_P3=1138;900;0 GPU_P4=1150;900;0 GPU_P5=1202;900;0 GPU_P6=1212;905;0 GPU_P7=1407;950 Mem_P0=167;900;0 Mem_P1=500;900;0 Mem_P2=800;900;0 Mem_P3=1095;950 Fan_Min=3350 Fan_Max=4900 Fan_Target=75 Fan_Acoustic=2400 Power_Temp=85 Power_Target=0 |
|
|
|
Wow, I am impressed with this miner ....
2 Vega 56 and 2 Vega 64 (all are reference cards), from 8300 on V7 drop a little at 7800, using default config.
yes, that is normal drop 100 H/s cca per Vega. But are 56 stable or not? Please post OverdriveNtool.ini file for cards, and what soft power play table you are using if rig is stable? Temperature core of the cards when mining for couple of hours? |
|
|
|
Do i put the gpu_conf settings in the same config-normalv8.txt ?
yes |
|
|
|
So people, V8 is here and are pain in the a..  Vega 56 reference is max affected with change because new algo is max memory intensive and produce even more heat on poor Vega 56 reference design. See first test of my "rig" with two Vega 56 ref, one Vega 64 Strix, and one RX570 4GB: Power consumption for rig is 15% more, from 680W to 770W Hasrate of the rig is down with same configuration: V7: 7170 H/s ; V8: 6870 H/s But... temperature for Vega 56 is increased from 60 C to 67 C ;-) my Vegas 56 are throttling again and in V7 is super stable... For now I tried to lower CC from 1442 to 1407 and in configuration of Vega 56 lower intensity from 120 / 16 / 2 to 112 / 16 / 2. Result is lower H/s for about 40 H/s for vega 56, but stability is improved for now: GPU:4 Threads:8 H/s CC MC °C RPM
Radeon RX Vega Bus:3 1938 1407 1095 63 3459
Radeon RX 580 Series Bus:6 898 1150 1980 51 1739
Radeon RX Vega Bus:11 2032 1442 1100 58 3484
Radeon RX Vega Bus:14 1936 1407 1095 66 3532 |
So lets all test our Vega-s, post results and we will find best solution for all... And yes, Rx 570 4GB Nitro, drop for just 4 H/s from 902 to 898, so developers of Monero have older cards and make new algo optimized for Polaris  |
|
|
|
Talking about power... CNv2 (Monerov8) does use quite a bit more. Getting very close to the acceptable limit on my 1500W PSU on that 6x Vega rig. I need to undervolt a tad more.
Bad decision from the Monero team, this new algo draws a lot more power, and gives less hashrate than the previous V7. Monero's price should go up if they want people to continue mining their coin. You are not quite right. The hashrate will fall for everyone in proportion, which means that all miners will receive the same amount of coins as before the fork. That is not true. Hashrate will NOT fall for everyone in proportion, if you have Vega card your lose will be higher than lose on RX cards. |
|
|
|
Doktore, txs for constant development Just quick test new option: "bralock": 1-256, parameter to tune, it *may* add some extra hash, it can differ between algos On V7 algo tested with RX570 4GB, Vega 64, Vega 56. Tried several parameter like: 1, 10, 33, 100, 150, 200, 256. Don't see any benefits for V7 algo so far, but maybe i must wait longer before switching to another parameter in config file... |
|
|
|
OK, so after testing with 18.3.4 and V8 here is the conclusion :
Don't use 18.3.4 if you want to mine V8!
It's something in the V8 code that driver doesn't like, and probably not something i can fix.
I will put this info on the first page.
Txs Doktor! That is my problem with small H/s on Vega and 18.3.4 So, you tested V8 with 18.6.1 and 18.5.1, for Vega 56 is one of these two drivers make better H/s? |
|
|
|
|
