Bitcoin Forum
February 24, 2019, 02:22:24 AM *
News: Latest Bitcoin Core release: 0.17.1 [Torrent]
 
   Home   Help Search Login Register More  
Pages: « 1 ... 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 [193] 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 ... 1136 »
  Print  
Author Topic: [ANN] cudaMiner & ccMiner CUDA based mining applications [Windows/Linux/MacOSX]  (Read 3403853 times)
relm9
Hero Member
*****
Offline Offline

Activity: 840
Merit: 1000



View Profile
February 06, 2014, 01:54:31 AM
 #3841

Currently you can test keccak in github. use the K kernel (just imagine K stands for Keccak).

The specified blocks/warps config is used for keccak. A good guess would be a larger multiple of your
GPU's SMX with 32 warps.  Fermi CPUs can only run 16 warps.

Autotune is definitely NOT working.

the keccak256 code for maxcoin is currently compiled against compute_10 which means it runs on any GPU.
Performance isn't stellar yet.

Code:
cudaminer --algo=keccak -d gtx780 -L 16 -l K192x32 --benchmark

some 20 MHash/s already...  Beats my CPU! Wink   Only 40% TDP. There's headroom!

More work to be done tomorrow. I need to get rid of the huge scrypt scratchpad buffers. They are not needed for keccak.
Also we may want to have some autotune. I currently use the -L parameter to artificially make the scratchpad smaller,
so I can run more blocks.

Uh, PCI express bandwidth is going to be a bottleneck at these MHash rates. Gotta do the hash verification on the GPU!

Christian


That was fast! I'm getting ~40 MHash with my 780. PCI-e 3.0 x8 here.

Wonder why the big difference, you on PCI-e 2.0?
1550974944
Hero Member
*
Offline Offline

Posts: 1550974944

View Profile Personal Message (Offline)

Ignore
1550974944
Reply with quote  #2

1550974944
Report to moderator
1550974944
Hero Member
*
Offline Offline

Posts: 1550974944

View Profile Personal Message (Offline)

Ignore
1550974944
Reply with quote  #2

1550974944
Report to moderator
1550974944
Hero Member
*
Offline Offline

Posts: 1550974944

View Profile Personal Message (Offline)

Ignore
1550974944
Reply with quote  #2

1550974944
Report to moderator
Your Bitcoin transactions
The Ultimate Bitcoin mixer
made truly anonymous.
with an advanced technology.
Mix coins
Advertised sites are not endorsed by the Bitcoin Forum. They may be unsafe, untrustworthy, or illegal in your jurisdiction. Advertise here.
1550974944
Hero Member
*
Offline Offline

Posts: 1550974944

View Profile Personal Message (Offline)

Ignore
1550974944
Reply with quote  #2

1550974944
Report to moderator
1550974944
Hero Member
*
Offline Offline

Posts: 1550974944

View Profile Personal Message (Offline)

Ignore
1550974944
Reply with quote  #2

1550974944
Report to moderator
1550974944
Hero Member
*
Offline Offline

Posts: 1550974944

View Profile Personal Message (Offline)

Ignore
1550974944
Reply with quote  #2

1550974944
Report to moderator
cbuchner1
Hero Member
*****
Offline Offline

Activity: 756
Merit: 500


View Profile
February 06, 2014, 01:57:48 AM
 #3842

i lost hashrate in a 550gtx updating to the new version

yes Fermi didn't get a speed boost, but a slight performance degradation due to added flexibility for VertCoin and scrypt-jane coins.
cbuchner1
Hero Member
*****
Offline Offline

Activity: 756
Merit: 500


View Profile
February 06, 2014, 01:59:06 AM
 #3843


That was fast! I'm getting ~40 MHash with my 780. PCI-e 3.0 x8 here.

yes, I am on PCI express 2.0 (MSI K9A2 Platinum V2 mainboard)


insert this at the end of do_keccak256 function

    checkCudaErrors(cudaStreamSynchronize(context_streams[0][thr_id]));

remove this from scanhash_keccak function

    cuda_scrypt_sync(thr_id, 0);

This will possibly double your hash rates. Wink
cbuchner1
Hero Member
*****
Offline Offline

Activity: 756
Merit: 500


View Profile
February 06, 2014, 02:00:47 AM
 #3844

looks like it everyone who went and bought cpu farms won't have the advantage anymore! Cheesy

we will OUTFARM these guys.
bigjme
Sr. Member
****
Offline Offline

Activity: 350
Merit: 250


View Profile
February 06, 2014, 02:03:49 AM
 #3845

Im running full pci 3 16x so I shouldnt be getting any bottle necks for my 780. How are you testing these?

Just download the latest, set the algo and run benchmark?

Owner of: cudamining.co.uk
manofcolombia
Member
**
Offline Offline

Activity: 84
Merit: 10

SizzleBits


View Profile WWW
February 06, 2014, 02:09:52 AM
 #3846

I went and looked at the maxcoin thread but I don't fully understand. Do we really need to get a cloud server? I don't understand this whole cloud mining aspect.

cbuchner1
Hero Member
*****
Offline Offline

Activity: 756
Merit: 500


View Profile
February 06, 2014, 02:10:23 AM
 #3847

Im running full pci 3 16x so I shouldnt be getting any bottle necks for my 780. How are you testing these?

Just download the latest, set the algo and run benchmark?

copy my command line I'd say...
cbuchner1
Hero Member
*****
Offline Offline

Activity: 756
Merit: 500


View Profile
February 06, 2014, 02:11:16 AM
 #3848

I went and looked at the maxcoin thread but I don't fully understand. Do we really need to get a cloud server? I don't understand this whole cloud mining aspect.

search for the maxcoin announcement thread instead...
relm9
Hero Member
*****
Offline Offline

Activity: 840
Merit: 1000



View Profile
February 06, 2014, 02:11:52 AM
 #3849


That was fast! I'm getting ~40 MHash with my 780. PCI-e 3.0 x8 here.

yes, I am on PCI express 2.0 (MSI K9A2 Platinum V2 mainboard)


insert this at the end of do_keccak256 function

    checkCudaErrors(cudaStreamSynchronize(context_streams[0][thr_id]));

remove this from scanhash_keccak function

    cuda_scrypt_sync(thr_id, 0);

This will possibly double your hash rates. Wink


Up to ~46 MH now so it's an improvement, maybe I'm hitting the limits of 3.0 x8?

GPU utilization at 60%, btw.
bigjme
Sr. Member
****
Offline Offline

Activity: 350
Merit: 250


View Profile
February 06, 2014, 02:13:52 AM
Last edit: February 06, 2014, 02:42:44 AM by bigjme
 #3850

34008-34160khash/s using christians code
unsure how to do what you did relm9 else i would tell you if its an 8x limit or not

Owner of: cudamining.co.uk
manofcolombia
Member
**
Offline Offline

Activity: 84
Merit: 10

SizzleBits


View Profile WWW
February 06, 2014, 02:17:44 AM
 #3851

I went and looked at the maxcoin thread but I don't fully understand. Do we really need to get a cloud server? I don't understand this whole cloud mining aspect.

search for the maxcoin announcement thread instead...

Ok so we don't need a cloud server...
I went and read the announcement. So its just like any other coin...except the retarget rate is much faster and its a different algorithm?

bigjme
Sr. Member
****
Offline Offline

Activity: 350
Merit: 250


View Profile
February 06, 2014, 02:20:17 AM
 #3852

2:30 am here so if you want anything testing please let me know relatively soon, current config is
cudaminer --algo=keccak -d gtx780 -L 16 -l K192x32 --benchmark

currently doing i believe it is 34Mhash's
still lower then bathrobes, even though im on PCI express 3.0 x16

Owner of: cudamining.co.uk
relm9
Hero Member
*****
Offline Offline

Activity: 840
Merit: 1000



View Profile
February 06, 2014, 02:25:39 AM
 #3853

2:30 am here so if you want anything testing please let me know relatively soon, current config is
cudaminer --algo=keccak -d gtx780 -L 16 -l K192x32 --benchmark

currently doing i believe it is 34Mhash's
still lower then bathrobes, even though im on PCI express 3.0 x16

Are you testing in Linux? Not sure if it'd make a difference but I'm in Windows right now
bigjme
Sr. Member
****
Offline Offline

Activity: 350
Merit: 250


View Profile
February 06, 2014, 02:30:41 AM
 #3854

yes on linux. the bandwidth could still be a huge factor, as our systems are running pci express 3, and Christian is only on 2

ok so here is a little speed breakdown, fastest to slowest

PCI Express 3.0 x16
PCI Express 3.0 x8   === PCI Express 2.0 x16
PCI Express 2.0 x8

Relm9 you would be on the same speed a PCI Express 2.0 x16, so what speed PCI Express 2.0 are you running at Christian?

Owner of: cudamining.co.uk
djm34
Legendary
*
Offline Offline

Activity: 1288
Merit: 1040


View Profile WWW
February 06, 2014, 02:53:59 AM
 #3855

a rapid test without the change in the code and K192x32
my gtx780ti get 55,3Mhash/sec (but it is running 70% TDP... wow it might save the planet too...)
How much does a cpu exactly ?
edit: autotune gives line in Giga Hash/s  Shocked

djm34 facebook page
BTC: 1NENYmxwZGHsKFmyjTc5WferTn5VTFb7Ze
Pledge for neoscrypt ccminer to that address: 16UoC4DmTz2pvhFvcfTQrzkPTrXkWijzXw
bacon_wrangler
Newbie
*
Offline Offline

Activity: 26
Merit: 0


View Profile
February 06, 2014, 03:17:42 AM
 #3856

Hey folks, new miner here. I think I may be missing something in my configuration, as I'm not seeing any improvement with the new version of cudaMiner. I have a GTX 650 Ti BOOST, and am getting 187 Kh/s in both 2-2 and 2-4 using -l K4x32 -i 0 -H 1

Any ideas how I can squeeze a bit more out of this card? I'm running the default settings aside from running the fan at 70, and using MSI Afterburner to monitor the card. (I also use CUDAmanager in case something happens while I'm at work and things overheat, so the miner will be shut down.)
paul09
Newbie
*
Offline Offline

Activity: 56
Merit: 0


View Profile
February 06, 2014, 03:31:09 AM
 #3857

Hmm I have lost around 120kh/s per card on the new version, using 2x GTX770

Was previously ~360kh/s per card now down to like 240 each.

My settings:
cudaminer -r 10 -R 30 -T 30 -H 1 -i 0 -l K16x16 -C 0
relm9
Hero Member
*****
Offline Offline

Activity: 840
Merit: 1000



View Profile
February 06, 2014, 03:38:07 AM
 #3858

yes on linux. the bandwidth could still be a huge factor, as our systems are running pci express 3, and Christian is only on 2

ok so here is a little speed breakdown, fastest to slowest

PCI Express 3.0 x16
PCI Express 3.0 x8   === PCI Express 2.0 x16
PCI Express 2.0 x8

Relm9 you would be on the same speed a PCI Express 2.0 x16, so what speed PCI Express 2.0 are you running at Christian?

Yeah you are right, the first slot on my mobo is actually x16 but since I have two GPUs plugged in both slots are brought down to x8.
djm34
Legendary
*
Offline Offline

Activity: 1288
Merit: 1040


View Profile WWW
February 06, 2014, 03:47:54 AM
 #3859

a rapid test without the change in the code and K192x32
my gtx780ti get 55,3Mhash/sec (but it is running 70% TDP... wow it might save the planet too...)
How much does a cpu exactly ?
edit: autotune gives line in Giga Hash/s  Shocked

I found that the result with the change in the code gives something 1 Mega more.
Beside that, the autotuning gave me this value: K1024x4 where I get 57,8MHash/s
Everything is overclocked... however it uses only 72% of the TDP and 60% of GPU usage.

I also tested the gtx 660oem, I obtained at K215x19 41,9MHash/s
(the card was also overclocked, however, the programm was running at  vely low clock value (same for tdp and gpu usage)

djm34 facebook page
BTC: 1NENYmxwZGHsKFmyjTc5WferTn5VTFb7Ze
Pledge for neoscrypt ccminer to that address: 16UoC4DmTz2pvhFvcfTQrzkPTrXkWijzXw
fcode
Full Member
***
Offline Offline

Activity: 168
Merit: 100


View Profile
February 06, 2014, 03:54:57 AM
 #3860

The new keccak code seems faster than a CPU I think.

Q9550 @3.4 Ghz (one core)
minerd-maxcoin --algo=keccak --benchmark
1151 khash/s.

i3-4130T @2.90GHz (one core)
minerd-maxcoin --algo=keccak --benchmark
599.59 khash/s.  (Huh?? must be a fluke, lots of stuff running on this machine)

Celeron G1630 @ 2.80GHz (one core)
minerd-maxcoin --algo=keccak --benchmark
1020 khash/s

GTX560-448
cudaminer --algo=keccak -d 0 --benchmark -L16 -l F14x16
28468 Kh/s

I could not get cudaminer to run on my Quadro 295NVS. Not sure of the kernel parameters, it wouldn't start (not that I would want to run on this card anyway).
Pages: « 1 ... 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 [193] 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 ... 1136 »
  Print  
 
Jump to:  

Bitcointalk.org is not available or authorized for sale. Do not believe any fake listings.
Sponsored by , a Bitcoin-accepting VPN.
Powered by MySQL Powered by PHP Powered by SMF 1.1.19 | SMF © 2006-2009, Simple Machines Valid XHTML 1.0! Valid CSS!