maybe, but did u test it by urself?.. I often got a sick or dead gpu, even upped the vgpu, sometimes just a hard reset after 12hours.
It works for me so yes maybe coincidence..
I don't have and never had setx commands in the batch. As to why is it coincidence, setx sets environment variables, which 'stick'. So if you did it once and rebooted - no need to set them over and over again. Those particular settings MAX_ALLOC and SYNC are used by OCL runtime. It either finds them in the environment or not. When you removed setx commands from the batch they are set anyway and OCL runtime still uses them. So nothing was changed when you removed them from the batch. But you say something changed - coincidence. If you want you can read more about setx, tho it's a lengthy and confusing time to time ; ) |
|
|
|
Finally it is stable for 24hours..I think I know why it is stable now.
I deleted
setx GPU_MAX_ALLOC_PERCENT 100
setx GPU_USE_SYNC_OBJECTS 1
Coincidence, these commands are not needed in the batch and have no effect if present. |
|
|
|
290x with The Stilt BIOS, Windows, 13.12 drivers
Scrypt: 905kh
N-Scrypt: 453kh
X11: 3.6 Mh
X13: 2.6 Mh
290, Stilt, W2k8 r2, 13.12 mod v3 850/1250 scrypt, arebyp, 32768: 840khs 850/1250 scryptn, bufius, 28159: 366khs 950/1250 x11, x11mod(darkcoin-mod), 8192: 3.4mhs 950/1250 x13, x13mod(marucoin-mod), 8192: 2.6mhs I 20 |
|
|
|
|
Any bounty for bug free multi-algo miner?
|
|
|
|
Anyone know how to enable logging in the batch file that doesn't overwrite itself each time it starts? I want to be able to check the log after a crash, but sometimes I don't catch it in time, and the evidence is lost.
@echo off
for /F "usebackq tokens=1,2 delims==" %%i in (`wmic os get LocalDateTime /VALUE 2^>NUL`) do if '.%%i.'=='.LocalDateTime.' set ldt=%%j
rem set ldt=%ldt:~0,4%-%ldt:~4,2%-%ldt:~6,2% %ldt:~8,2%:%ldt:~10,2%:%ldt:~12,6%
set ldt=%ldt:~0,4%-%ldt:~4,2%-%ldt:~6,2%_%ldt:~8,2%-%ldt:~10,2%
cgminer 2>"c:\log\cgminer_jane_%ldt%.log"
|
|
|
|
Multi-algo ports all seem down.
+1 |
|
|
|
lbr, are there any inner-kernel settings to change with regards to SPH_BLAKE or any others? That could be different for Hawaiis?
Nothing I know of for x11/x13/keccak. I'm using different kernel, arebyp, for scrypt tho. And for scryptn, some changes do increase hashrate, but not significantly enough. |
|
|
|
since it's the same IP address...
Same IP yes, but different services on different ports. Lets say you spam incorrect work on scrypt pool -it may ban you for that, however x11 pool won't. At least it may make some sense if anti-DDOS/hacking is implemented and nicehash does have something implemented. |
|
|
|
|
Ok, I've just tested 3333, 3335, 3336, 3337 and 3338 - all of them are ALIVE.
|
|
|
|
3336 is ALIVE for me right now..
well, then I have absolutely no idea what's going on here... I have also tried to remove the pool and re-add it but it didn't help If it's alive for me and dead for you, either nicehash banned all and whitelisted me or it banned you or you have connection issues to nicehash. In either way nothing to be done.. maybe ask nicehash? Ah, maybe you specified incorrect BTC address? It does go DEAD if it is not valid. |
|
|
|
|
3336 is ALIVE for me right now..
|
|
|
|
I have set stratum+tcp://stratum.nicehash.com:4333, stratum+tcp://stratum.nicehash.com:4335, stratum+tcp://stratum.nicehash.com:4336, stratum+tcp://stratum.nicehash.com:4337 and stratum+tcp://stratum.nicehash.com:4338 as pools, with no backup because I just wanted to check if it works but it looks like there's no work on neither one of these pools. But this is strange. Are all of them DEAD? Right now.. my rigs switched to x11 from x13.. ~10 mins ago. |
|
|
|
So is it possible that for example stratum+tcp://stratum.nicehash.com:3333 which is a port for scrypt algo works, but 4333 does not work?
Yes of course, 3xxx pools are always online. But 4xxx pools are only online when they are most profitable, so multi-algo configuration must include(ideally) all 4xxx pools. |
|
|
|
I was just trying to use the auto-switching but it doesn't seem to work for me. When I set pool ports to 33xx, everything works perfectly, I can manually switch pools and all pools work, however when I switch it over to 43xx and start the sgminer, it says that all pools are dead after a while. Any suggestions?
Perfectly normal; in order to 'switch' only one algorithm can be 'selected' at a time. All of NH ports that start with 43xx are multi-algo ports, which will shut, so when 1 is selected, the other 4 algos go dead. If you have no failovers, it is possible you may get no work served if there are not enough orders up for that algo. I understand that, but there surely IS some work at least for scrypt, since when I go manual, it works. So it's unlikely that there's no work for any algo... There is no 'at least'. And there are no multi-algo pools. There are pools with different algos which go offline/dead if they are not profitable. So you must configure all 4xxx pools for nicehash in order to get this to work. Tho, check nicehash thread, apparently they ban by IP for 'hacking'.. |
|
|
|
"hamsi-expand-big" : "1", Default value is 4, but 1 works better for many cards (for example Sapphire R9 280X Dual-X). For 290 '1' is better also. Just add --kernel setting back.. what would be the point of putting --kernel back in when the kernels are all set with --algorithm? The point is that instead of modifying existing kernel I would create new kernel and set it with --kernel option. That's why --kernel option was implemented in the first place. To be clear, --algo looks if specified kernel exist in the current sgminer build and if not it loads ckolivas. So, I can't use ckolivas_my_pimped_kernel.cl And adding hamsi-expand-big to config.. f.e. scrypt does not need that option. Lets add some.. dunno.. SPH_ECHO_64 or SPH_COMPACT_BLAKE_64 or any other defines in ocl kernels then.. |
|
|
|
"hamsi-expand-big" : "1", Default value is 4, but 1 works better for many cards (for example Sapphire R9 280X Dual-X). For 290 '1' is better also. Just add --kernel setting back.. |
|
|
|
--gpu-reorder makes sgminer go crazy ; ( 6x290 sgminer-5.0-pre-release-1-win32.zip or sgminer-5.0-pre-release-2014-06-06_01-win32.zip or sgminer_v5_0_06062014.zip Error -46: Creating ExtraKernel #0 from program. (clCreateKernel)
|
|
|
|
[19:48:43] Building binary marucoin-modHawaiiglg2tc12800nf10w256l8.bin
[19:48:43] Error -11: Building Program (clBuildProgram)
[19:48:43] "/tmp/OCL13381T8.cl", line 883: error: identifier "buf" is undefined
INPUT_BIG;
marucoin works perfectly, but marucoin-mod gives me this. I've tried deleting all the .bin files.
Edit: It was a simple solution (it usually is). I just had to clear my tmp folder. There must've been a bunch of old junk in there screwing things up. So to any linux newbs having problems try:
cd /tmp/
sudo rm -r *
you call other linux newbs and you delete everything in /tmp including your unix sockets you NEED (or you have to restart you wm)?! He does not call anyone anything.. So to any linux newbs having problems try: This means - "if you feel that you are a *nix n00b try:" Btw, interesting, OCL compiler should have overwritten .cl in /tmp/ before compilation. Maybe there is a bug with OCL runtime on *nix, on win it does replace .cl files everytime. |
|
|
|
Yep I get -4 Errors on Hawaii chipsets whenever I try and build a scrypt or scrypt-N bin with this miner. For the time being I'm running it as X11/X13/Keccak combo-flip.
For me it's only the switch between scryptn and scrypt is not working, each algo separately runs fine with the same miner(sgminer-5.0-pre-release-1-win32.zip or sgminer_v5_0_06062014.zip). If it does not work for you when only scrypt or scryptn is specified in config - something is jsut wrong in the config(or driver). |
|
|
|
|
Show Posts
