the result vega56samsung(v64bios) srb 1.8.6 slower 5% & higher power usage 6% compared teamred, hope next iteration will better
Srb is 100% faster on Litev7, Tube, Haven, Gpu, Hycon, Mox, Wownero, than teamred. Just because TRM not supporting them But it's still true. |
|
|
|
the result vega56samsung(v64bios) srb 1.8.6 slower 5% & higher power usage 6% compared teamred, hope next iteration will better
Srb is 100% faster on Litev7, Tube, Haven, Gpu, Hycon, Mox, Wownero, than teamred. |
|
|
|
When I use min_rig_speed option with 6000 H/s with this config the option says : Your average Hashrate is lower then 6000 (2100 H/s sometime 4200 H/s) but in SRB my average hashrate is 6400 H/s. That's normal ?
Nope, thats not normal, can you try to set 'min_rig_speed_duration' to for example 1800, that is 30 minutes. Im curious if the same thing will happen. [2019-04-29 12:49:55] GPU0 : 2216 H/s [T: 59c, RPM: 1737, CC: 1502 MHz, MC: 1100 MHz][BUS:10] [2019-04-29 12:49:55] GPU1 : 2218 H/s [T: 60c, RPM: 1843, CC: 1502 MHz, MC: 1100 MHz][BUS:7] [2019-04-29 12:49:55] GPU2 : 2220 H/s [T: 61c, RPM: 1905, CC: 1502 MHz, MC: 1100 MHz][BUS:3] [2019-04-29 12:49:55] Total: 6654 H/s [2019-04-29 12:51:34] Average hashing speed (1106 H/S) is lower than the defined minimum (6600 H/S) speed! @doktor83 Ok, i will check it out edit: this is happening only on CN/R because of kernel precompilation. It messes up the average speed. I will think of a way to work this out. |
|
|
|
I used timing = 0 (auto) option from drivers with srbminer.
When I applied custom timings shown above I used timing = 1 driver mode.
Don't know if it makes any difference
Probably not. Thanks for helping.  upx2 fork support with auto mode in 1.8.6? --enablecoinforking Yes, i tested on testpool and it was ok. |
|
|
|
I used timing = 0 (auto) option from drivers with srbminer.
When I applied custom timings shown above I used timing = 1 driver mode.
Don't know if it makes any difference
Probably not. Thanks for helping. |
|
|
|
Any reason off the top of your head why the tweak profile would be completely unresponsive? It mines the same without it and does not add a line (like in your video) "Applied Tweak Profile 5...." etc
Rig of vega 64's
{
"cryptonight_type" : "normalv4",
"intensity" : 0,
"tweak_profile" : 5,
"double_threads" : true,
}
My first guess would be that you are not running the miner with admin privileges. But you can see in the log what is exactly the problem. |
|
|
|
my timings are stable up to 1090mhz hbm clock. Yours feel fine with 1107 but I didn't test for long.
Mine are 2350h/s cn-r (teamred) vs yours 2200h/s (teamred) with 1090mhz clock. Claymore's is 49.5 vs 48 mhs eth.
With higher mem clock (1107mhz) with your timings results are getting better but didn't catch mine at 1090mhz.
Thanks for the stock values. That is a huge difference in CN/R and it shouldn't be, im wondering what settings did you get for 'my' timings. Can you share them ? |
|
|
|
When I use min_rig_speed option with 6000 H/s with this config the option says : Your average Hashrate is lower then 6000 (2100 H/s sometime 4200 H/s) but in SRB my average hashrate is 6400 H/s. That's normal ?
Nope, thats not normal, can you try to set 'min_rig_speed_duration' to for example 1800, that is 30 minutes. Im curious if the same thing will happen. |
|
|
|
Hi guys !
I'm really lost with those config.. My rig isn't stable. I got all the time crash, or bug and I don't know why.
I'm using Windows 10 1809 with 18.6.1 drivers and I got 3 Sapphire RX Vegas 64 Nitro+ and 1 Sapphire Radeon HD 7950
I'm using SRB Miner 1.8.6 with my 3 Vegas and i'm using XMR Stak with my Radeon HD 7950.
This is my config in SRB Miner :
{
"cryptonight_type" : "normalv4",
"intensity" : 0,
"double_threads" : true,
"gpu_conf" :
[
{ "id" : 1, "intensity" : 122, "worksize" : 16, "threads" : 2, "tweak_profile" : 0},
{ "id" : 2, "intensity" : 122, "worksize" : 16, "threads" : 2, "tweak_profile" : 0},
{ "id" : 3, "intensity" : 122, "worksize" : 16, "threads" : 2, "tweak_profile" : 0}
]
}
-----------------------------------------------------------------------------------------------------
SRBMiner-CN.exe --config Config\config-normalv4.txt --pools pools.txt --logfile %LOGTIME%
----------------------------------------------------------------------------------------------------
Sometimes I got GPU crash then SRB try to shutdown my PC and it freezes.
This is my Wattman config :
GPU Clock : -8%
Memory Clock : 1105 Mhz
My Hashrate with tweak profil to 3 :
GPU0: 2149 H/s [T:68c][BUS:10]
GPU1: 2155 H/s [T:68c][BUS:7]
GPU2: 2157 H/s [T:67c][BUS:3]
hashrate: Total: 6461 H/s
That is really low with Vegas 64 Nitro+.. Doktor83, with your 56 you have 2100 Hashrate with each cards with tweak profil 5.. (In your video)
----------------------------------------------------------------------------------------------------
Something to change ?
Doktor83 I have try to use "min_rig_speed" option but there is a bug with this one in SRB 1.8.6. This is not the good average Hashrate. If someone can try and confirm that is a bug..
I don't know why you have crashes, but your vega temperatures are too high (thats +10c on the vram). 1.8.6 has 10 tweak profiles, so you have 7 more to try Also you can disable the watchdog so it wont restart the miner on gpu crash, use --disablegpuwatchdog. I would first try to lower memory clock, get temps down to at least 60c. min_rig_speed gets the average hashrate for 'min_rig_speed_duration' time, that is 5 minutes if not set by the user. So please describe what you think is a bug. |
|
|
|
Hi, Dok!
Just examined your timings for profile 10.
They are good for hynix HBM2 but for samsung (both low voltage 800-940 mhz vega56 and high voltage 945-1100 mhz vega64 I have much better (more stable and fast) mem timings)
I can share to you if you want ))
Ok, so now i have examined your V64 samsung timings and i can't see why they would be much better, as you say. Can you run profile 10 on V64 with samsung mem? If you can, what's the hashrate with profile 10 vs your timing on the same algo? |
|
|
|
Ok, this is for Vega 64 LC
This if for reference vega56, stock bios
win10, 19.4.1 driver
What is the default value for RFC on the samsung V56 ? <values cl="20" ras="33" rcdrd="16" rcdwr="10" rcab="47" rcpb="47" rpab="14" rppb="14" rrds="4" rrdl="6" rtp="5" faw="16" cwl="7" wtrs="4" wtrl="9" wr="16" rrefd="8" rdrddd="3" rdrdsd="2" rdrdsc="1" rdrdscl="3" wrwrdd="3" wrwrsd="2" wrwrsc="1" wrwrscl="3" wrrd="1" rdwr="19" ref="3900" mrd="8" mod="15" xs="270" xsmrs="8" pd="8" cksre="10" cksrx="10" rfcpb="160" stag="7" xp="8" cpded="2" cke="8" rddata="19" wrlat="5" rdlat="14" wrdata="1" ckestag="2" rfc="260"></values>
vega64LE(samsung):<values cl="20" ras="33" rcdrd="16" rcdwr="10" rcab="47" rcpb="47" rpab="14" rppb="14" rrds="4" rrdl="6" rtp="5" faw="16" cwl="7" wtrs="4" wtrl="9" wr="16" rrefd="8" rdrddd="3" rdrdsd="2" rdrdsc="1" rdrdscl="3" wrwrdd="3" wrwrsd="2" wrwrsc="1" wrwrscl="3" wrrd="1" rdwr="19" ref="3900" mrd="8" mod="15" xs="270" xsmrs="8" pd="8" cksre="10" cksrx="10" rfcpb="160" stag="7" xp="8" cpded="2" cke="8" rddata="19" wrlat="5" rdlat="14" wrdata="1" ckestag="2" rfc="260"></values> vega56 nitro+(hynix)stock bios:DEFAULT - GPU 5: Vega 10 XL/XT [Radeon RX Vega 56/64] Memory state: 0x118 (800MHz) Timing 1 CL: 18 RAS: 24 RCDRD: 22 RCDWR: 10 Timing 2 RCAb (RC): 37 RCPb (RC): 37 RPAb (RP): 13 RPPb (RP): 13 Timing 3 RRDS: 3 RRDL: 3 RTP: 6 Timing 4 FAW: 13 Timing 5 CWL: 6 WTRS: 4 WTRL: 4 Timing 6 WR: 12 Timing 7 RREFD: 7 Timing 8 RDRDDD: 3 RDRDSD: 2 RDRDSC: 1 RDRDSCL: 3 Timing 9 WRWRDD: 3 WRWRSD: 2 WRWRSC: 1 WRWRSCL: 3 Timing 10 WRRD: 1 RDWR: 18 Timing 12 REF: 3200 Timing 13 MRD: 4 MOD: 12 Timing 14 XS: 216 Timing 16 XSMRS: 216 Timing 17 PD: 7 CKSRE: 8 CKSRX: 8 Timing 20 RFCPB: 96 STAG: 7 Timing 21 XP: 7 CPDED: 2 CKE: 7 Timing 22 RDDATA: 17 WRLAT: 4 RDLAT: 14 WRDATA: 1 Timing 23 CKESTAG: 2 RFC Timing RFC: 176 Thanks but i still miss the V56 Samsung default RFC ) Not V56 flashed to V64, because that uses the V64 timings. |
|
|
|
Ok, this is for Vega 64 LC
This if for reference vega56, stock bios
win10, 19.4.1 driver
What is the default value for RFC on the samsung V56 ? |
|
|
|
Hi, Dok!
Just examined your timings for profile 10.
They are good for hynix HBM2 but for samsung (both low voltage 800-940 mhz vega56 and high voltage 945-1100 mhz vega64 I have much better (more stable and fast) mem timings)
I can share to you if you want ))
You can share to help the folks here, im not using predefined straps in SRB. They are calculated based on existing settings. |
|
|
|
Hey. On one rig srb issues a message "Job timeout detected, reconnecting the pool". This condition can last for hours and the reconnect does not pass. In this case, the cards are under 100% load , srb writes that the shares, but in fact the pool does not receive them.
Windows 10, SRB 1.8.6, moneroocean
You are right, after the big networking code rewrite, i forgot to test every option, and this one isn't working as its supposed to. Going to fix in next release, thanks for reporting. As a workaround increase "job_timeout" for example to 1 hour (3600 sec) |
|
|
|
i test it with 18.6.1 it's not submit any shear. (uninstall with ddu )
i use with this --disabletweaking but no shear find.
hash rate is fine but after 30m no shear find .
Try maybe changing the value for fragments. What does your config looks like? Try renaming the cache file from v182 that works for you to 186 like : g1CNV4w168f4hm3_ 182.srb to g1CNV4w168f4hm3_ 186.srb I just tried and 182 kernel works on 186 |
|
|
|
@doktor83
Tweaking seems still disabled for VII. Tried from both the config file and the command line, with different values for tweak_profile, but it always says it's disabled.
Paste here the first 10 sec of your log |
|
|
|
|
not really except DDU and reinstall 18.6.1 drivers
|
|
|
|
Hello doktor83!! i just wanna make a suggestion, why don't you just make a simple miner just like "TRM" where are we can make timing with other program..
TRM is everything but not simple You can 'make timing' with other program just use --disabletweaking parameter. it's all written on the first page and in the readme. |
|
|
|
vega 56-hynix
i use this setup
--rp 14 --rc 42 --rfc 156 --rrds 4 --rrdl 5 --rrd 3 --rcdrd 18 --rcdwr 3 --REF 15600
work 10-12 hour with 2-3 error and block all comp and dont go tu restart:(
what i must chage? what step i must use?
etc: rp14 to rp15
rfc 156 chg to rfc 157 ? or i must use 168 or higher?
what step use ? 1? or nubmer /with 8 or 16?
--rp 14 --rc 42 --rfc 156 --rrds 4 --rrdl 5 --rrd 3 --rcdrd 18 --rcdwr 3 --REF 15600 - work but i must little downclock this, but i dont know how and what...
i try down mem 1100 to 1090mhz but again have error
i try give more power. now work with 870mv and i try with 920mv but again have error
pls any tutotrial what formula work with rp rc rfc ref etc....
read back a little for formulas if your hynix memclk is really 1100 then and it's stable at stock timings, then rfc and rc and rcdwr are too low imho. i read all page but not see any forumula..(i go read again) my mistake- 1100 for samsung. 1 samsung card always have error 1100 or 1090 . |
|
|
|
|
Warning:
Someone created a page on Github (github.com/SRBMiner), be careful because it wasn't created by me, so don't download from there.
|
|
|
|
|
Show Posts
