DiabloMiner GPU Miner - page 37. | Bitcointalksearch.org

DiabloD3

legendary

Activity: 1162

Merit: 1000

DiabloMiner author

Quote from: ssateneth on January 12, 2012, 04:06:03 PM

Quote from: DiabloD3 on January 12, 2012, 11:55:42 AM

Apparently, the Chinese love DiabloMiner

http://item.taobao.com/item.htm?id=14062839037

Scroll down for screenshot

Never mind the miner screenshot, i'm more interested in the PCI-E splitter they used! What IS that abomination? ... I want one!

Thats the product page for that thing. It even seems to have two pci-e 6 pin plugs for powering the extra slots for larger cards.

ssateneth

legendary

Activity: 1344

Merit: 1004

Quote from: DiabloD3 on January 12, 2012, 11:55:42 AM

Apparently, the Chinese love DiabloMiner

http://item.taobao.com/item.htm?id=14062839037

Scroll down for screenshot

Never mind the miner screenshot, i'm more interested in the PCI-E splitter they used! What IS that abomination? ... I want one!

DiabloD3

legendary

Activity: 1162

Merit: 1000

DiabloMiner author

Update: Add -dt argument for debug timer, quits after 1 min

DiabloD3

legendary

Activity: 1162

Merit: 1000

DiabloMiner author

Apparently, the Chinese love DiabloMiner

http://item.taobao.com/item.htm?id=14062839037

Scroll down for screenshot

DiabloD3

legendary

Activity: 1162

Merit: 1000

DiabloMiner author

Update: Switch back to 3 threads, helps double DMA chips like 69xx and 79xx

DiabloD3

legendary

Activity: 1162

Merit: 1000

DiabloMiner author

Update: Added 1onevvolf's idea for shuffle() + vstore() instead of branchy array output. Use -vs to use it.

Benchmarks using it on SDK 2.6, with my 5850 at 765/1000 stock, without on the left, with on the right, uint3 does not work with shuffle+vstore.

Code:

-w 64      
-v 1       245 245
-v 2       280 280
-v 3       214 xxx
-v 4       245 275
-v 8       222 163
-v 1,1     255 255
-v 2,2     233 233
-v 3,3     180 xxx
-v 4,4     221 163
-v 1,1,1   219 219
-v 2,2,2   260 230
-v 3,3,3   136 xxx
-v 4,4,4   163 109
-v 1,1,1,1 211 211
-v 2,2,2,2 211 261

-w 128
-v 1       245 245
-v 2       285 283
-v 3       222 xxx
-v 4       251 233
-v 1,1     258 258
-v 2,2     239 241
-v 3,3     180 xxx
-v 4,4     221 120
-v 1,1,1   247 247
-v 2,2,2   221 228
-v 3,3,3    99 xxx
-v 4,4,4   120 107
-v 1,1,1,1 221 221
-v 2,2,2,2 210 221

-w 192
-v 1       244 244
-v 2       281 279
-v 3       234 xxx
-v 4       263 184
-v 1,1     255 255
-v 2,2     252 252
-v 3,3     142 xxx
-v 4,4     174 161
-v 1,1,1   245 245
-v 2,2,2   175 236

-w 256
-v 1       245 245
-v 2       278 276
-v 3       206 xxx
-v 4       238 229
-v 1,1     255 255
-v 2,2     225 226
-v 3,3     180 xxx
-v 4,4     219 211
-v 1,1,1   241 241
-v 2,2,2   218 205

-v 2 -w 128 without -a and -vs is still the fastest on SDK 2.6 at 285 mhash. In comparison, the same on SDK 2.1 does 314, and the same on SDK 2.5 does 305.

Goddamnit 2.6.

DiabloD3

legendary

Activity: 1162

Merit: 1000

DiabloMiner author

Quote from: 1onevvolf on January 12, 2012, 06:34:34 AM

Quote from: ?? on ??

What's the recommended -v argument for GCN?

I found that -v 1 works best, with the rest of the settings at default.

@DiabloD3: I like the vector changes! Much cleaner than what I cooked up. By the way, I found that -v 16 isn't working with the new changes. The nonce.sN components aren't correctly declared in hex when N > 9. Here's a fix:

Code:

              if(vectors[y] > 1) {
                for(int i = 0; i < vectors[y]; i++)
-                  end += replace.replace(lastVar, lastVar + ".s" + i).replaceAll("nonce", "nonce.s" + i) + "\n";
+                  end += replace.replace(lastVar, lastVar + ".s" + Integer.toHexString(i)).replaceAll("nonce", "nonce.s" + Integer.toHexString(i)) + "\n";
              } else {
                end += replace;
              }

BTW, I figured out how to make your shuffle() trick actually work. I'll commit it in a bit. It doesn't improve it the way you hoped.

1onevvolf

newbie

Activity: 43

Merit: 0

Quote from: ?? on ??

What's the recommended -v argument for GCN?

I found that -v 1 works best, with the rest of the settings at default.

@DiabloD3: I like the vector changes! Much cleaner than what I cooked up. By the way, I found that -v 16 isn't working with the new changes. The nonce.sN components aren't correctly declared in hex when N > 9. Here's a fix:

Code:

              if(vectors[y] > 1) {
                for(int i = 0; i < vectors[y]; i++)
-                  end += replace.replace(lastVar, lastVar + ".s" + i).replaceAll("nonce", "nonce.s" + i) + "\n";
+                  end += replace.replace(lastVar, lastVar + ".s" + Integer.toHexString(i)).replaceAll("nonce", "nonce.s" + Integer.toHexString(i)) + "\n";
              } else {
                end += replace;
              }

despoiler

member

Activity: 94

Merit: 10

Just got my Sapphire 7970 installed. Been doing some testing. Confirming what others have been seeing. Stock clocks ~550mh/s. 1125/1375 OC is giving me ~671mh/s peak with ~655mh/s average. Default miner settings. 11.12 drivers from the below address. 2.6 SDK of course.

http://support.amd.com/us/kbarticles/Pages/radeon-7900-series.aspx

DiabloD3

legendary

Activity: 1162

Merit: 1000

DiabloMiner author

Quote from: ?? on ??

What's the recommended -v argument for GCN?

No one knows yet. 1onevvolf has a thread going that is exploring this until I get mine.

https://bitcointalksearch.org/topic/my-initial-radeon-hd-7970-mining-benchmarks-56630

DiabloD3

legendary

Activity: 1162

Merit: 1000

DiabloMiner author

Update: I have changed how the vector argument works to better serve new and upcoming GPUs. It now uses comma separated values, and vectors up to 16 in length. It also now supports 3 length vectors as well.

If you used this -> switch to this...

-v 1 -> -v 1
-v 2 -> -v 2
-v 3 -> -v 2,1
-v 4 - > -v 2,2
-v 18 -> -v 1,1
-v 19 -> -v 1,1,1
-v 36 -> -v 4
-v 40 -> -v 4,4

freshzive

sr. member

Activity: 447

Merit: 250

-f 120 seems to have done the trick. hashrate drops by about 5-10mhash when I'm using the computer, but average goes back up to where it was before when I'm afk

thanks for the help Grin

DiabloD3

legendary

Activity: 1162

Merit: 1000

DiabloMiner author

Quote from: Queelis on January 10, 2012, 08:05:04 AM

Is there any benefit to having multiples of 60?

Yeah. There are some things hardwired in the driver to work around 60hz (the refresh rate of virtually every LCD monitor on DVI/HDMI/Displayport), instead of just being triggered either after the last iteration is done, or on the actual refresh rate. So, multiples and divisors of 60 seem to give higher hash rates and/or give better desktop latency.

Queelis

newbie

Activity: 36

Merit: 0

Is there any benefit to having multiples of 60?

DiabloD3

legendary

Activity: 1162

Merit: 1000

DiabloMiner author

Quote from: freshzive on January 09, 2012, 10:46:48 PM

Is there anyway to reduce the amount of graphics lag this causes in OS X? The latest version works perfectly on my iMac w/ a Radeon 4850 (getting 50mhash), but it makes everything so laggy i can only run it when afk.

Currently the only flag I'm using is -w 64 as that's recommended in the OP.

Add -f 60 or a higher number (-f should always be a divisor or multiple of 60)

Quote from: Queelis on January 10, 2012, 03:30:08 AM

If that's anywhere close to -f settings in GUIminer, -f 60 lets me watch GPU-accelerated HD videos without a problem. Even -f 55 is ok.

Same goes for you, 55 isn't a divisor or multiple of 60. -f 60, 120, etc are nicer.

Queelis

newbie

Activity: 36

Merit: 0

If that's anywhere close to -f settings in GUIminer, -f 60 lets me watch GPU-accelerated HD videos without a problem. Even -f 55 is ok.

Eveofwar

sr. member

Activity: 406

Merit: 250

Quote from: freshzive on January 09, 2012, 10:46:48 PM

Is there anyway to reduce the amount of graphics lag this causes in OS X? The latest version works perfectly on my iMac w/ a Radeon 4850 (getting 50mhash), but it makes everything so laggy i can only run it when afk.

Currently the only flag I'm using is -w 64 as that's recommended in the OP.

Try adding -f 120

Tweak the number up/down...higher you go = smoother desktop experience, but slower mining

freshzive

sr. member

Activity: 447

Merit: 250

Is there anyway to reduce the amount of graphics lag this causes in OS X? The latest version works perfectly on my iMac w/ a Radeon 4850 (getting 50mhash), but it makes everything so laggy i can only run it when afk.

Currently the only flag I'm using is -w 64 as that's recommended in the OP.

DiabloD3

legendary

Activity: 1162

Merit: 1000

DiabloMiner author

BTW, anyone using the donate function, update to newest. That multipool bug apparently screws up donating too, it hops to the wrong pool to donate on.

Gabi

legendary

Activity: 1148

Merit: 1008

If you want to walk on water, get out of the boat

Yup i know. I just reported it for the lulz. Fail defender when it report a safe thing as dangerous

Topic: DiabloMiner GPU Miner - page 37. (Read 866596 times)