Im wait windows x64 on Nvidia 
.
Me too on comp 5.0 and 5.2. . 
Topic: SILENTARMY v5: Zcash miner, 115 sol/s on R9 Nano, 70 sol/s on GTX 1070 - page 90. (Read 209286 times)
November 03, 2016, 03:25:18 PM
November 03, 2016, 01:29:47 PM
Im wait windows x64 on Nvidia .
. 
November 03, 2016, 11:37:42 AM
Marc, have you considered making NR_ROWS_LOG fixed at 20, and clearing cnt for each round? Then each row would have only collisions at each round, avoiding the need to search for collisions. Then NR_SLOTS/OVERHEAD could be significantly reduced.
To avoid time penalty for clearing the cnt values, instead of calling kernel_init_ht at each round, you could zero the count after checking the row. When the DRAM page is already open to check the cnt value, the time cost of doing a write back to the same page (in fact, same 64-byte cache line) is minimal.
To avoid time penalty for clearing the cnt values, instead of calling kernel_init_ht at each round, you could zero the count after checking the row. When the DRAM page is already open to check the cnt value, the time cost of doing a write back to the same page (in fact, same 64-byte cache line) is minimal.
November 03, 2016, 10:59:25 AM
Either those numbers are low, or the bursts at the end of a kernel run are over 1x/Gen1 250MB/s bus capacity. I tested with 2 R9 380s at the same clock speeds (950/1600) and the card in a 16x slot gets ~25H/s vs 21 for the card on a 1x riser.
Thanks for this data. This perf impact of x1 vs x16 is a bit more than what I thought it would be. The transfers are bursty at the end of an equihash run. Basically with a very fast GPU every ~40 msec a 4MB blob is sent over the PCIe link.
It wouldn't be difficult to make the code more efficient and have the kernel filter more of the invalid solutions (duplicate inputs) in order to reduce this 4MB down to a few dozen kB.
November 03, 2016, 10:28:34 AM
CPU speed should not affect Sol/s by more than ±10%. The silentarmy kernel continuously streams 10-80 MB/s (depending on your GPU speed) of potential solution data back to the CPU for analysis, but this is easily handled even by weak CPUs.
Either those numbers are low, or the bursts at the end of a kernel run are over 1x/Gen1 250MB/s bus capacity. I tested with 2 R9 380s at the same clock speeds (950/1600) and the card in a 16x slot gets ~25H/s vs 21 for the card on a 1x riser.
November 03, 2016, 02:12:24 AM
CPU speed should not affect Sol/s by more than ±10%. The silentarmy kernel continuously streams 10-80 MB/s (depending on your GPU speed) of potential solution data back to the CPU for analysis, but this is easily handled even by weak CPUs.
Thanks for clearing that up. I'm assuming that's 10-80MB/s per GPU? So 70-560MB/s for a 7 GPU rig?
Worst case scenario is PCIE 1.0 1x which is 250MB/s. PCIE 2.0 1x and PCIE 3.0 1x are 500 and 1000MB/s respectively, which is plenty of bandwidth in all cases.
November 03, 2016, 02:08:02 AM
CPU speed should not affect Sol/s by more than ±10%. The silentarmy kernel continuously streams 10-80 MB/s (depending on your GPU speed) of potential solution data back to the CPU for analysis, but this is easily handled even by weak CPUs.
November 02, 2016, 11:11:22 PM
Hello,
I wrote SILENTARMY which is a Zcash Equihash solver. I initially posted a thread on the Zcash forums, but I figured I might as well start a thread here.
The latest news is that I just released v2 which now supports GCN 1 hardware. In fact all AMD GPUs should now be supported, but keep in mind very little to no testing has been done on TeraScale (pre-GCN) GPUs as well as Nvidia GPUs. But I need people with Nvidia to test if possible. Now that v2 is released, my roadmap is:
GitHub: https://github.com/mbevand/silentarmy
Zcash forum thread: https://forum.z.cash/t/silentarmy-gcn-1-now-supported-45-7-sol-s-with-one-r9-nano/3580
Does anyone compare cpu speed affect sols? I wrote SILENTARMY which is a Zcash Equihash solver. I initially posted a thread on the Zcash forums, but I figured I might as well start a thread here.
The latest news is that I just released v2 which now supports GCN 1 hardware. In fact all AMD GPUs should now be supported, but keep in mind very little to no testing has been done on TeraScale (pre-GCN) GPUs as well as Nvidia GPUs. But I need people with Nvidia to test if possible. Now that v2 is released, my roadmap is:
- v2:
support GCN 1→ done - v3: make a full miner supporting stratum
- v4: Nvidia support
- v5: Windows support
GitHub: https://github.com/mbevand/silentarmy
Zcash forum thread: https://forum.z.cash/t/silentarmy-gcn-1-now-supported-45-7-sol-s-with-one-r9-nano/3580
November 02, 2016, 04:56:07 AM
Is anyone here mining with R9 Nanos and reaches the proclaimed "45.7 Sol/s with one R9 Nano" of Github/Silentarmy page?
With a single mining task, every of my Nano mines fine with ~23-24 Sol. But with second task, hashrates droppes to (2x) 15-16 Sol per card.
Any other experiences?
Update: with new CoinsForAll silentarmy miner 0.2.3 I reach now over 44 Sol per Nano. Very cool!
With a single mining task, every of my Nano mines fine with ~23-24 Sol. But with second task, hashrates droppes to (2x) 15-16 Sol per card.
Any other experiences?
Update: with new CoinsForAll silentarmy miner 0.2.3 I reach now over 44 Sol per Nano. Very cool!
November 01, 2016, 06:59:52 PM
i find it working very bad, and i'm not alone with this.
Genoil's may not be the problem, overclocking and pool instability don't help. I got mine stable over 7 hours after downclocking memory and change of pool. And with the same setup zogminer is now stable too.
A lot of people just switched their rigs from eth to zec , it may not be that simple for some reason.
November 01, 2016, 05:16:45 PM
i find it working very bad, and i'm not alone with this.
October 31, 2016, 03:54:29 PM
Excellent work. I'm finding Genoil's Windows implementation of this to be working great.
October 31, 2016, 02:29:26 PM
I have a fresh install of Ubuntu 16.04. So install the AMD Drivers first using instructions on your github page and then for the miner install? I am fairly new to Ubuntu trying to get my feet wet with it. Would you mind giving instructions on how to install and run the miner? I have 5x 280x and can't wait to get those running. Thank you.
October 31, 2016, 12:21:58 PM
silent army kernels is awesome and is part of genoil-zec miner now and it did wonders to my R9 280 and 2 R7 370 .
it work with pitcairn and tahiti ,I'm using it now with a tahiti and 2 pitcaim cards .
it work with pitcairn and tahiti ,I'm using it now with a tahiti and 2 pitcaim cards .
October 31, 2016, 11:29:10 AM
Just did a git pull to refresh your repo, and confirm pitcairn works. ~20sol/s for a 370 clocked at 1125/1500.
October 31, 2016, 11:26:39 AM
Please put for Linux only. You do not say what OS it's for until people waste there time clicking on the page.
Someone running Windoze that doesn't like wasting their time? LOL! October 31, 2016, 11:00:59 AM
I updated the title ("Linux only").
Nvidia support will come in v4, and Windows support probably in v5. I updated the initial post with a roadmap.
Nvidia support will come in v4, and Windows support probably in v5. I updated the initial post with a roadmap.
October 31, 2016, 10:06:25 AM
good work!
I will try it on some of my older rigs with Tahiti
I will try it on some of my older rigs with Tahiti
October 31, 2016, 08:21:19 AM
@ mrb
are you thinking of adding nvidia support?
can you make a windows version ?
If you make a windows release I would gladly give it a try
are you thinking of adding nvidia support?
can you make a windows version ?
If you make a windows release I would gladly give it a try
October 30, 2016, 11:40:26 PM
Please put for Linux only. You do not say what OS it's for until people waste there time clicking on the page.
Wonderfull!!! Agree totally!
Jump to: