search

cbuchner1 / CudaMiner

a CUDA accelerated litecoin mining application based on pooler's CPU miner
Other
692 stars 304 forks source link

GeForce GTX 780 Ti result does not validate on CPU #43

Closed phobeus81 closed 10 years ago

phobeus81 commented 10 years ago

Upgraded today from a GTX560 to a GTX 780 TI. While the old card worked flawless, the new version seems to having issues with the card. Whenever running, just a handful of requests seems to be valid. Most requests are failing with the message "GeForce GTX 780 Ti result does not validate on CPU!".

This system is running on a Fedora 19 (64 bit system) with NVIDIA 331.20 and cuda 5.5.22. This issue occurs with the version from the 20th November. Unable to use a more recent version due to another issue reported.

It seems not to make a difference if setting up auto tuning or a profile and results in both cases in lots of fails. When running in debug mode, it looks like nearly all failed requests are having the same value.

Any ideas?

./cudaminer [...] -i 0 -H 1 -lT30x16 -D * CudaMiner for nVidia GPUs by Christian Buchner * This is version 2013-11-20 (alpha) based on pooler-cpuminer 2.3.2 (c) 2010 Jeff Garzik, 2012 pooler Cuda additions Copyright 2013 Christian Buchner My donation address: LKS1WDKGED647msBQfLBHV3Ls8sveGncnm

[2013-12-05 22:16:40] 1 miner threads started, using 'scrypt' algorithm. [...] [2013-12-05 22:16:41] Failed to get Stratum session id [2013-12-05 22:16:41] Stratum difficulty set to 16 [2013-12-05 22:16:41] DEBUG: job_id='7d1' extranonce2=00000000 ntime=52a0ed34 [2013-12-05 22:16:41] Stratum detected new block [2013-12-05 22:16:42] GPU #0: GeForce GTX 780 Ti with compute capability 3.5 [2013-12-05 22:16:42] GPU #0: interactive: 0, tex-cache: 0 , single-alloc: 0 [2013-12-05 22:16:42] GPU #0: Given launch config 'T30x16' requires too much memory. [2013-12-05 22:16:42] GPU #0: using launch configuration T30x16 [2013-12-05 22:16:42] GPU #0: GeForce GTX 780 Ti, 15360 hashes, 74.57 khash/s [2013-12-05 22:16:42] DEBUG: hash <= target Hash: 00000735d19f6cbda52f6bc8f9c71ea3583927ec408d8fff1226a5e2acc3ddf3 Target: 00000ffff0000000000000000000000000000000000000000000000000000000 [2013-12-05 22:16:42] GPU #0: GeForce GTX 780 Ti result does not validate on CPU! [2013-12-05 22:16:43] DEBUG: hash <= target Hash: 00000d3f00c065f2eeed8caea3d7518030f5e4593bdeff382f8a23dbee233f43 Target: 00000ffff0000000000000000000000000000000000000000000000000000000 [2013-12-05 22:16:43] GPU #0: GeForce GTX 780 Ti result does not validate on CPU! [2013-12-05 22:16:43] DEBUG: hash <= target Hash: 0000042049ea7495aca88147fef1924e1305f8d9fe40e6bf4fbac21312e4bafc Target: 00000ffff0000000000000000000000000000000000000000000000000000000 [2013-12-05 22:16:43] GPU #0: GeForce GTX 780 Ti result does not validate on CPU! [2013-12-05 22:16:45] DEBUG: hash <= target Hash: 000003312bf1aba279d41c92dd969496448fc9a54ee4c87421ae64ef2d854fd2 Target: 00000ffff0000000000000000000000000000000000000000000000000000000 [2013-12-05 22:16:45] GPU #0: GeForce GTX 780 Ti result does not validate on CPU!

cbuchner1 commented 10 years ago

[2013-12-05 22:16:42] GPU #0: Given launch config 'T30x16' requires too much memory.

weird! this card has 3 GB of it and I also run T30x16

cbuchner1 commented 10 years ago

I fixed a bug in the memory allocation routine in the 2013-12-01 version. you seem to be hitting it

phobeus81 commented 10 years ago

Allright, confirming. I just build the master. The auto detection might be this time a bit to optimistic: 461.59 khash/s with configuration T105x4 But having a rate arround 320khash/s and all blocks being valid. So I think you can consider this issue being fixed already in the master. Nice work.

cbuchner1 commented 10 years ago

You passed the -H 1 flag? Otherwise the CPU is a bottleneck. I get 450 per card, non overclocked

Von meinem iPhone gesendet

Am 07.12.2013 um 14:32 schrieb phobeus81 notifications@github.com:

Allright, confirming. I just build the master. The auto detection might be this time a bit to optimistic: 461.59 khash/s with configuration T105x4 But having a rate arround 320khash/s and all blocks being valid. So I think you can consider this issue being fixed already in the master. Nice work.

— Reply to this email directly or view it on GitHub.

phobeus81 commented 10 years ago

You're right. The -H1 flag was missing. Adding it gives me a nice boost arround 380, but doesn't seems to come over 400 here (cuda 5.5)