[Dev] NeoScrypt GPU Miner - Public Beta Test
-
xIntensity works by being a shader multiplier for the thread intensity. It gives you finer control over thread intensity than just using regular Intensity settings of I1-20 which use a standard 2^x = thread intensity (where x is the intensity setting). You need to specify the number of shaders your card has in your config file to get this, or raw intensity to work properly because, again, xIntensity is a shader multiplier. I’ve used both xIntensity and rawintensity settings before, and let me just say that they are great features, but don’t expect a miraculous 100% increase in hash rate. Using these settings may or may not give you better hash rate, and to be honest, you will be spending a lot more time tweaking your config if you use xIntensity or rawintensity. rawintensity isn’t a multiplier. It’s a method of directly controlling the thread intensity. So going off of the traditional intensity 2^x = thread intensity (where x is intensity setting), you are simply entering the total number of thread intensity rather than using an equation or a shader multiplier. xIntensity and rawintensity settings also require extensive tweaking of the thread-concurrency as well, so traditional values of TC that worked while using standard intensity I1-20 will most likely require different TC settings for every single adjustment you make to xIntensity or rawintensity. So, HW errors are being caused because majsta hasn’t posted the number of shaders, nor have they posted the TC setting, which again is very important for using xIntensity and rawintensity. Here’s a good place to start for learning about xIntensity in particular:
https://github.com/sgminer-dev/sgminer/commit/7aeae40af22e6108aab8b68a229eea25a639d650
xIntensity Example: r9 280x has 2048 shaders
I15 = 2^15 = 32768 thread intensity
I16 = 2^16 = 65536 thread intensity
xIntensity setting 16 = 16 * 2048 = 32768 thread intensity
xIntensity setting 17 = 17 * 2048 = 34816 thread intensity
So using regular I1-20, there’s a huge, exponential jump in the number of thread intensity even at increment +1 (from I15-I16 it’s about a 2x increase in thread intensity). Using xIntensity, there’s a very small increase in thread intensity from xintensity16 to xintensity17, which again, may or may not give you better, or worse, hash rates.
no new info for me…
still something i am missing, hw only what ever i do when using xi
xi nor working in any way for neo?? why?
-
no new info for me…
still something i am missing, hw only what ever i do when using xi
xi nor working in any way for neo?? why?
Both xintensity and rawintensity work properly under sgminer5-dev and with neoscrypt. I’ve used xintensity on my 290x with neoscrypt and it works fine, albeit with tiny less hash rate with the settings I used. It’s probably the thread concurrency you need to adjust here, and what magical TC is majsta using? I have no idea :(
Maybe try using thread-concurrency32768 or thread-concurrency16384. I know without entering TC while using xintensity in sgminer will cause HW errors and other weirdness.
-
Well, weird… I rebooted my rig, and the 3% gained from the 112 worksize turned into about a 5% loss in hashrate over the 96 settings, but running the 96 settings was right where it’s been at…
So, I decided to do the old driver swaparoo with all the OCL stuff posted here (and wipe the bin file during swaps)
With 13/96/2 and 1100/1450 the results are 290/290x hashrate:
The 13.9 gets 161k145k
The 14.2 gets 298/294k
The 14.4 gets 255/242k
The 14.6 (rc2) gets 298/294k
The 14.7 (rc1) gets 298/294k
Tried running the worksize 112 again after all that, and it shows the 3% gain again, and lower hashrate on the 290x.
-
PPL like I said there is nothing special in my settings, it works every time on anything I just need to setup --xintensity 3. I m using 14.6 RC2 drivers and all other settings are the same like for any other miner. Here is complete line.
sgminer.exe -k neoscrypt --lookup-gap 2 --xintensity 3 --worksize 256 -g 2 -o xxxx:xxxx -u x -p x --gpu-engine “1000,1000,1000” --thread-concurrency “8192,8192,8192” --gpu-memclock “1500,1500,1500”
One more thing, I published this 5 days ago on my local forum and there is about 20 miners who use this settings for last 5 days or so.
-
Both xintensity and rawintensity work properly under sgminer5-dev and with neoscrypt. I’ve used xintensity on my 290x with neoscrypt and it works fine, albeit with tiny less hash rate with the settings I used. It’s probably the thread concurrency you need to adjust here, and what magical TC is majsta using? I have no idea :(
miner compiled by wolf? 4.2.2 so called 5.xx??
im on win 7 64 bit . xi no use… i have no idea why
-
Well, weird… I rebooted my rig, and the 3% gained from the 112 worksize turned into about a 5% loss in hashrate over the 96 settings, but running the 96 settings was right where it’s been at…
So, I decided to do the old driver swaparoo with all the OCL stuff posted here (and wipe the bin file during swaps)
With 13/96/2 and 1100/1450 the results are 290/290x hashrate:
The 13.9 gets 161k145k
The 14.2 gets 298/294k
The 14.4 gets 255/242k
The 14.6 (rc2) gets 298/294k
The 14.7 (rc1) gets 298/294k
Tried running the worksize 112 again after all that, and it shows the 3% gain again, and lower hashrate on the 290x.
try my dllls from previous post and update your results??
edit: my best test
getting best results so far using dlls from 14.201.1009( very little effect):
-
PPL like I said there is nothing special in my settings, it works every time on anything I just need to setup --xintensity 3. I m using 14.6 RC2 drivers and all other settings are the same like for any other miner. Here is complete line.
sgminer.exe -k neoscrypt --lookup-gap 2 --xintensity 3 --worksize 256 -g 2 -o xxxx:xxxx -u x -p x --gpu-engine “1000,1000,1000” --thread-concurrency “8192,8192,8192” --gpu-memclock “1500,1500,1500”
One more thing, I published this 5 days ago on my local forum and there is about 20 miners who use this settings for last 5 days or so.
equal to mine, why it wont work for me???
-
And definitely going to look into the x-intensity, I tried it at the very beginning of running neoscrypt, and it didn’t work well because I had no idea where to tune the cards at.
Now I can work backward and find a starting point to try.
-
miner compiled by wolf? 4.2.2 so called 5.xx??
im on win 7 64 bit . xi no use… i have no idea why
Make sure you are removing regular intensity setting from your config when you run xintensity.
-
No need since adding xintensity overrides intensity settings.
-
Make sure you are removing regular intensity setting from your config when you run xintensity.
yep, im not so stupid. still no idea what this is…
we been fooled…
-
try my dllls from previous post and update your results??
edit: my best test
getting best results so far using dlls from 14.201.1009( very little effect):
The 14.2 is what you posted. Exact same results as 14.6 and 14.7 for me.
-
yep, im not so stupid. still no idea what this is…
we been fooled…
Have you seen the screenshots from Wolf0 where he shows off his new kernel speeds? Notice how he uses just regular intensity? There’s a reason for that: xintensity and rawintensity are not necessarily the holy grail of mining. It’s a bitch to set up properly, and while it may work for one person, it may not work for another. This isn’t the first time I have seen someone make these type of claims about using xintensity or rawintensity, and it won’t be the last. I’ve spent countless hours tweaking settings to get xintensity and rawintensity configs for miners, and while sometimes it delivers, most times I just go back to regular intensity configs. I wouldn’t get worked up over claims of crazy high hash rates using xintensity or rawintensity.
-
The 14.2 is what you posted. Exact same results as 14.6 and 14.7 for me.
I believe they are going off the actual DLL versions of the files, and not using the Catalyst version like most of us do. So when they say 14.2 it is probably Catalyst 14.6 RC2 or Catalyst 14.7 RC3.
-
miner compiled by wolf? 4.2.2 so called 5.xx??
im on win 7 64 bit . xi no use… i have no idea why
No, I am not using the miner compiled by wolf0. I am using the sgminer-5.1.0-dev builds that are built by Nicehash.
Use the sgminer I posted here to use the same one as me:
-
^^ same build of SGminer that I have…
-
i made some notepad work, we should get 50% of original scrypt hashrate when kernel is optimal
-
Guys i use this miner
I have included a sgminer-5.1.0-dev 7z file with all of the changes listed above here:
https://drive.google.com/file/d/0B06N0FYvyFgDeHBnTVEtR3BVN2c/view?usp=sharing
and @majsta settings in my .bat file
sgminer.exe -k neoscrypt --xintensity 3 --worksize 256 -g 2 -o xxxx:xxxx -u x -p x --thread-concurrency 8192 -d 0,1
(280x+7950…3rd is 5850 and its off). 14.4 drivers, W7x64, 1000(gpu)/1500(mem) for both cards
and you can see results
edit:
.bat file without --thread-concurrency got only HW…
-
Well, after spending the afternoon screwing around with the xintensity, the best I could get was a bump up from 325k to 327k for the 290’s and a loss from 308k to 298k for the 290x’s.
Back to straight intensities for me… (for now)
-
Glad to see some are finally take note about xIntensity I posted about way back in this thread. I have never use sgminer much at all
and had pretty bad luck with it when I did test with it way back and since this went to neoscrypt so I didn’t put much time in trying to
find a working version of sgminer that has xIntensity. I still have my old tweaked settings for all my AMD cards and will test those out
soon and if there good to go I’ll post those to help others with the type of AMD cards I have on hand.
I did try some testing with cgminer back a few weeks ago but didn’t have the neoscrypt kernel we have now to test with. ;-)