Not I - 382.53 because everything I've tried from NVidia to date that is newer has been crashy/buggy/unstable as heck - except the LINUX 384.98 they released to support the 1070 ti.
Topic: [ANN] dstm's ZCash / Equihash Nvidia Miner v0.6.2 (Linux / Windows) - page 82. (Read 224961 times)
December 21, 2017, 12:16:37 AM
December 20, 2017, 11:05:21 PM
Anyone test the miner on 388.71 drivers from NVIDIA ?
December 20, 2017, 07:16:30 PM
Is this miner Zcash specific, or can I use it for any equihash mining?
Can I for example, point it at Miningpoolhub's equihash multicoin equihash port?
Can I for example, point it at Miningpoolhub's equihash multicoin equihash port?
equihash is equihash
You can use it for any coin with same algo.Thank you DSTM For now I could not realize a big difference with my 1060's but my 3GB's are a few percent faster now without to change anything.
Before 300@89W
Now 310@89W
and maybe a few invalid shares less. but for that its maybe to early to say. Will update soon.
December 20, 2017, 07:05:31 PM
Cuda 9.1 is out, any chance you will utilize for more performance Sir DSTM?
December 20, 2017, 03:34:39 PM
Hello i switched over from EWBF i dont get much of a boost my 1070 OC Ti's are the same i get about 10 more Sol's wiht my 1070 OC, i have heard this miner was better i was just wondering if anyone got similar cards could share what there using for a boost trying to get solid clock with the miner. because on EWBF i can --no fee and i dont think im making more then 2% of fee instead 0 on the EWBF but just looking for a little help consering everyone seems to be moving to this, just looking for a decent clock for 1070's on this miner and can work from there really apprecaite your time.
the hard cold reality is EWBF is still the better option for 1080TI's and possibly for 1070 TI's because for high end cards the performance difference between the 2 miners is negligible at best (and not even noticeable for some users), so EWBF wins decidedly because you can reduce its fees to 0 or 0.1... Granted I personally tested this DSTM miner about a month ago, so I'm open to any feedback if recently there has been any significant change in the 1080TI performance with DSTM...
It's definitely a case of TEST IT YOURSELF - the results are close enough that it could easily come down to specific settings on specific cards.
I still prefer the EBWF "set your own fee rate" option though - this whole 2% fixed fee stuff is a RIPOFF that I blame Claymore for, and have tended to AVOID using his miners as a result since they rarely offer enough better performance to make up for the ripoff-level fee.
December 20, 2017, 03:32:37 PM
I get more then 1% rejected shares because of: "GPU0 rejected share: [21,"Stale job!"]"
Pool is eu1-zcash.flypool.org. <------------------------------------------------------------------------ (arrow not in original)
Can i do something to fix or mitigate ?
greetz, ezfox
Pool is eu1-zcash.flypool.org. <------------------------------------------------------------------------ (arrow not in original)
Can i do something to fix or mitigate ?
greetz, ezfox
More than 1% - that's a significant value. After what time do you get this results/stats? - it shouldn't be that high after 24h. What pool are you using?
Did you bother reading, they SAID they were using flypool.
December 20, 2017, 12:29:41 PM
Is this miner Zcash specific, or can I use it for any equihash mining?
Can I for example, point it at Miningpoolhub's equihash multicoin equihash port?
Can I for example, point it at Miningpoolhub's equihash multicoin equihash port?
December 20, 2017, 11:38:29 AM
I've tried both and saw a 6% increase in hashrate with DSTM vs EWBF, so even with the 2% devfee I'm still "on top".
This was on 1070's. With the 1070ti comparison is much closer and the ability to go --fee 0 on EWBF gives it a slight edge.
This was on 1070's. With the 1070ti comparison is much closer and the ability to go --fee 0 on EWBF gives it a slight edge.
December 20, 2017, 11:10:37 AM
Hello i switched over from EWBF i dont get much of a boost my 1070 OC Ti's are the same i get about 10 more Sol's wiht my 1070 OC, i have heard this miner was better i was just wondering if anyone got similar cards could share what there using for a boost trying to get solid clock with the miner. because on EWBF i can --no fee and i dont think im making more then 2% of fee instead 0 on the EWBF but just looking for a little help consering everyone seems to be moving to this, just looking for a decent clock for 1070's on this miner and can work from there really apprecaite your time.
the hard cold reality is EWBF is still the better option for 1080TI's and possibly for 1070 TI's because for high end cards the performance difference between the 2 miners is negligible at best (and not even noticeable for some users), so EWBF wins decidedly because you can reduce its fees to 0 or 0.1... Granted I personally tested this DSTM miner about a month ago, so I'm open to any feedback if recently there has been any significant change in the 1080TI performance with DSTM...
December 20, 2017, 09:40:11 AM
Awesome, thanks!
December 20, 2017, 09:27:49 AM
So I've got a question. The telemetry HTTP interface lists the card type but the JSON-RPC does not - however, the card order in the miner does not match nvidia-smi or PCI ordering. How does the miner order the cards?
Thx for pointing this out.
ZM uses CUDA_DEVICE_ORDER. Nvidia specifies it's default behaviour like this:
FASTEST_FIRST causes CUDA to guess which device is fastest using a simple heuristic, and make that device 0, leaving the order of the rest of the devices unspecified.
The default behaviour can be changed by setting the environment variable 'CUDA_DEVICE_ORDER' to 'PCI_BUS_ID' which causes CUDA to order the devices by '... PCI bus ID in ascending order'.
The documentation for this is located at: http://docs.nvidia.com/cuda/cuda-c-programming-guide/index.html#env-vars
Quote
It'd be great if in the telemetry JSON-RPC you included the uuid for the card -- for larger deployments it'd let us identify specifically which card the miner was referring to. Or the PCI slot ID or anything. 
That's a good point, I'll include both: PCI-Bus-ID and UUID.
Quote
If this information is already there or there's a logical method that the miner organizes the cards please let me know. For example, in one of my rigs I have the following
1060 PCI bus: 0:01:00.0
1060 PCI bus: 0:02:00.0
1070 PCI bus: 0:03:00.0
1070ti PCI bus: 0:05:00.0
1070ti PCI bus: 0:06:00.0
That's also how they are ordered in nvidia SMI and nvidia-settings (all linux obviously). However, the miner reports them as follows:
0 GeForce GTX 1070 Ti
1 GeForce GTX 1060
2 GeForce GTX 1060
3 GeForce GTX 1070
4 GeForce GTX 1070 Ti
The 1060 in pci bus 01 is the "active" interface carrying xorg for reference.
1060 PCI bus: 0:01:00.0
1060 PCI bus: 0:02:00.0
1070 PCI bus: 0:03:00.0
1070ti PCI bus: 0:05:00.0
1070ti PCI bus: 0:06:00.0
That's also how they are ordered in nvidia SMI and nvidia-settings (all linux obviously). However, the miner reports them as follows:
0 GeForce GTX 1070 Ti
1 GeForce GTX 1060
2 GeForce GTX 1060
3 GeForce GTX 1070
4 GeForce GTX 1070 Ti
The 1060 in pci bus 01 is the "active" interface carrying xorg for reference.
December 20, 2017, 09:23:20 AM
What are the best settings of core/mem for a 1070?
December 20, 2017, 09:13:19 AM
I wish it were that simple, unfortunately no - lspci also follows the nvidia-smi output (which is expected since that is orders in PCI bus order):
So that's the two 1060's, then the 1070 followed by the 1070ti's as expected.
Code:
greg@pallas:~$ lspci -tv
-[0000:00]-+-00.0 Intel Corporation Device 590f
+-01.0-[01]--+-00.0 NVIDIA Corporation Device 1c03
| \-00.1 NVIDIA Corporation Device 10f1
+-08.0 Intel Corporation Sky Lake Gaussian Mixture Model
+-14.0 Intel Corporation Device a2af
+-16.0 Intel Corporation Device a2ba
+-17.0 Intel Corporation Device a282
+-1b.0-[02]--+-00.0 NVIDIA Corporation Device 1c03
| \-00.1 NVIDIA Corporation Device 10f1
+-1b.5-[03]--+-00.0 NVIDIA Corporation Device 1b81
| \-00.1 NVIDIA Corporation Device 10f0
+-1c.0-[04]----00.0 Realtek Semiconductor Co., Ltd. RTL8111/8168/8411 PCI Express Gigabit Ethernet Controller
+-1c.5-[05]--+-00.0 NVIDIA Corporation Device 1b82
| \-00.1 NVIDIA Corporation Device 10f0
+-1c.7-[06]--+-00.0 NVIDIA Corporation Device 1b82
| \-00.1 NVIDIA Corporation Device 10f0
+-1f.0 Intel Corporation Device a2c8
+-1f.2 Intel Corporation Device a2a1
+-1f.3 Intel Corporation Device a2f0
\-1f.4 Intel Corporation Device a2a3
-[0000:00]-+-00.0 Intel Corporation Device 590f
+-01.0-[01]--+-00.0 NVIDIA Corporation Device 1c03
| \-00.1 NVIDIA Corporation Device 10f1
+-08.0 Intel Corporation Sky Lake Gaussian Mixture Model
+-14.0 Intel Corporation Device a2af
+-16.0 Intel Corporation Device a2ba
+-17.0 Intel Corporation Device a282
+-1b.0-[02]--+-00.0 NVIDIA Corporation Device 1c03
| \-00.1 NVIDIA Corporation Device 10f1
+-1b.5-[03]--+-00.0 NVIDIA Corporation Device 1b81
| \-00.1 NVIDIA Corporation Device 10f0
+-1c.0-[04]----00.0 Realtek Semiconductor Co., Ltd. RTL8111/8168/8411 PCI Express Gigabit Ethernet Controller
+-1c.5-[05]--+-00.0 NVIDIA Corporation Device 1b82
| \-00.1 NVIDIA Corporation Device 10f0
+-1c.7-[06]--+-00.0 NVIDIA Corporation Device 1b82
| \-00.1 NVIDIA Corporation Device 10f0
+-1f.0 Intel Corporation Device a2c8
+-1f.2 Intel Corporation Device a2a1
+-1f.3 Intel Corporation Device a2f0
\-1f.4 Intel Corporation Device a2a3
So that's the two 1060's, then the 1070 followed by the 1070ti's as expected.
December 20, 2017, 05:48:39 AM
So I've got a question. The telemetry HTTP interface lists the card type but the JSON-RPC does not - however, the card order in the miner does not match nvidia-smi or PCI ordering. How does the miner order the cards?
It'd be great if in the telemetry JSON-RPC you included the uuid for the card -- for larger deployments it'd let us identify specifically which card the miner was referring to. Or the PCI slot ID or anything.
If this information is already there or there's a logical method that the miner organizes the cards please let me know. For example, in one of my rigs I have the following
1060 PCI bus: 0:01:00.0
1060 PCI bus: 0:02:00.0
1070 PCI bus: 0:03:00.0
1070ti PCI bus: 0:05:00.0
1070ti PCI bus: 0:06:00.0
That's also how they are ordered in nvidia SMI and nvidia-settings (all linux obviously). However, the miner reports them as follows:
0 GeForce GTX 1070 Ti
1 GeForce GTX 1060
2 GeForce GTX 1060
3 GeForce GTX 1070
4 GeForce GTX 1070 Ti
The 1060 in pci bus 01 is the "active" interface carrying xorg for reference.
It'd be great if in the telemetry JSON-RPC you included the uuid for the card -- for larger deployments it'd let us identify specifically which card the miner was referring to. Or the PCI slot ID or anything.
If this information is already there or there's a logical method that the miner organizes the cards please let me know. For example, in one of my rigs I have the following
1060 PCI bus: 0:01:00.0
1060 PCI bus: 0:02:00.0
1070 PCI bus: 0:03:00.0
1070ti PCI bus: 0:05:00.0
1070ti PCI bus: 0:06:00.0
That's also how they are ordered in nvidia SMI and nvidia-settings (all linux obviously). However, the miner reports them as follows:
0 GeForce GTX 1070 Ti
1 GeForce GTX 1060
2 GeForce GTX 1060
3 GeForce GTX 1070
4 GeForce GTX 1070 Ti
The 1060 in pci bus 01 is the "active" interface carrying xorg for reference.
I'm not currently running 'dstm', but try 'lspci -t' to see if it follows the PCI-tree.
December 20, 2017, 03:07:34 AM
I just wanted to add that I did a little experiment with drivers with the latest release.
0.5.7 is unstable with NVidia 388.xx drivers. It instant crashes for me and never restarts. I want to add that new games like PUBG did the same.
So I tried older drivers, I found that 384.94 works flawless with 0.5.7 and earlier release + I got increased hashrate compared to 388.xx drivers.
This was tested for my 3 rigs with 1080ti and 1070 on windows 10 with latest updates.
0.5.7 is unstable with NVidia 388.xx drivers. It instant crashes for me and never restarts. I want to add that new games like PUBG did the same.
So I tried older drivers, I found that 384.94 works flawless with 0.5.7 and earlier release + I got increased hashrate compared to 388.xx drivers.
This was tested for my 3 rigs with 1080ti and 1070 on windows 10 with latest updates.
I'm using the latest 388.xx drivers on windows 10 with a mix of 1070 and 1070ti and it's been running for days without a hitch.
Weird that different people have such different outcomes.
I forgot to mention that my 1080ti are overclocked with +200 mem and core. 1070 are overclocked with +100 mem and core. My post was more directed towards users that overclock and uses DTSM miner.
My 1070ti's are overclocked with +200 core and +700 mem and my 1070's are overclocked with +120 core and +650 mem. I am using DSTM 5.7 and have found it slightly better in every way than the excellent 5.6
I run a total of 17 cards. 8 1070ti and 8 1070 in rigs, and one 1070 in a laptop. All running DSTM 24/7 and so far have not had any crashes or stability issues with DSTM.
Whats your powerlimit? Mines 78%
65%
Oh I misread your earlier posts. You got 1070s, dtsm miner is optimized for 1060 and 1070s. I wonder when he starts with the 1080/1080 optimization. Still thats great its stable for you.
December 20, 2017, 01:47:43 AM
Hi ppl, help pls,
i have evga 1070 2k frequency, driver 388.59, dstm miner, flyypool
I launch all ok works as it is necessary
after 30 min-1hour GPUz view 1,6k frequency and temperature 40 c,
dstm miner view temperature 59 c and higter sols, but in fact they are smaller
sorry bad eng
ps 408 avg sols norm?
i have evga 1070 2k frequency, driver 388.59, dstm miner, flyypool
I launch all ok works as it is necessary
after 30 min-1hour GPUz view 1,6k frequency and temperature 40 c,
dstm miner view temperature 59 c and higter sols, but in fact they are smaller
sorry bad eng
ps 408 avg sols norm?
December 20, 2017, 12:19:02 AM
Hello i switched over from EWBF i dont get much of a boost my 1070 OC Ti's are the same i get about 10 more Sol's wiht my 1070 OC, i have heard this miner was better i was just wondering if anyone got similar cards could share what there using for a boost trying to get solid clock with the miner. because on EWBF i can --no fee and i dont think im making more then 2% of fee instead 0 on the EWBF but just looking for a little help consering everyone seems to be moving to this, just looking for a decent clock for 1070's on this miner and can work from there really apprecaite your time.
December 19, 2017, 11:24:12 PM
Have anyone tested dstm's miner ver.0.5.7 with GTX 1060 3G? Also, please share your setting?
I could not purchase more P106-100, but there're some MSI GTX 1060 3G are available, so I'm still considering.
Thanks!
I could not purchase more P106-100, but there're some MSI GTX 1060 3G are available, so I'm still considering.
Thanks!
Yes, I am running 5 EVGA GTX 1060 3G SC, and since Energy Cost is not a factor for me I am running using Afterburn (yes, the software for MSI cards because the EVGA software would not work with more than 4 GPUs) so that my limit is 90°C, and no overclock (the gain is minimal vs crash frequency).
In summary/average they are running at 293 Sol/s, at 87°C, with 120 W consumption.
Since I bought them for US$200, I think it is a great value.
Finally, as stated above, I have not seen an increase in the last 2-3 versions in hash rate.
Hope this helps.
December 19, 2017, 09:47:02 PM
So I've got a question. The telemetry HTTP interface lists the card type but the JSON-RPC does not - however, the card order in the miner does not match nvidia-smi or PCI ordering. How does the miner order the cards?
It'd be great if in the telemetry JSON-RPC you included the uuid for the card -- for larger deployments it'd let us identify specifically which card the miner was referring to. Or the PCI slot ID or anything.
If this information is already there or there's a logical method that the miner organizes the cards please let me know. For example, in one of my rigs I have the following
1060 PCI bus: 0:01:00.0
1060 PCI bus: 0:02:00.0
1070 PCI bus: 0:03:00.0
1070ti PCI bus: 0:05:00.0
1070ti PCI bus: 0:06:00.0
That's also how they are ordered in nvidia SMI and nvidia-settings (all linux obviously). However, the miner reports them as follows:
0 GeForce GTX 1070 Ti
1 GeForce GTX 1060
2 GeForce GTX 1060
3 GeForce GTX 1070
4 GeForce GTX 1070 Ti
The 1060 in pci bus 01 is the "active" interface carrying xorg for reference.
It'd be great if in the telemetry JSON-RPC you included the uuid for the card -- for larger deployments it'd let us identify specifically which card the miner was referring to. Or the PCI slot ID or anything.
If this information is already there or there's a logical method that the miner organizes the cards please let me know. For example, in one of my rigs I have the following
1060 PCI bus: 0:01:00.0
1060 PCI bus: 0:02:00.0
1070 PCI bus: 0:03:00.0
1070ti PCI bus: 0:05:00.0
1070ti PCI bus: 0:06:00.0
That's also how they are ordered in nvidia SMI and nvidia-settings (all linux obviously). However, the miner reports them as follows:
0 GeForce GTX 1070 Ti
1 GeForce GTX 1060
2 GeForce GTX 1060
3 GeForce GTX 1070
4 GeForce GTX 1070 Ti
The 1060 in pci bus 01 is the "active" interface carrying xorg for reference.
December 19, 2017, 08:55:00 PM
I just wanted to add that I did a little experiment with drivers with the latest release.
0.5.7 is unstable with NVidia 388.xx drivers. It instant crashes for me and never restarts. I want to add that new games like PUBG did the same.
So I tried older drivers, I found that 384.94 works flawless with 0.5.7 and earlier release + I got increased hashrate compared to 388.xx drivers.
This was tested for my 3 rigs with 1080ti and 1070 on windows 10 with latest updates.
0.5.7 is unstable with NVidia 388.xx drivers. It instant crashes for me and never restarts. I want to add that new games like PUBG did the same.
So I tried older drivers, I found that 384.94 works flawless with 0.5.7 and earlier release + I got increased hashrate compared to 388.xx drivers.
This was tested for my 3 rigs with 1080ti and 1070 on windows 10 with latest updates.
I'm using the latest 388.xx drivers on windows 10 with a mix of 1070 and 1070ti and it's been running for days without a hitch.
Weird that different people have such different outcomes.
I forgot to mention that my 1080ti are overclocked with +200 mem and core. 1070 are overclocked with +100 mem and core. My post was more directed towards users that overclock and uses DTSM miner.
My 1070ti's are overclocked with +200 core and +700 mem and my 1070's are overclocked with +120 core and +650 mem. I am using DSTM 5.7 and have found it slightly better in every way than the excellent 5.6
I run a total of 17 cards. 8 1070ti and 8 1070 in rigs, and one 1070 in a laptop. All running DSTM 24/7 and so far have not had any crashes or stability issues with DSTM.
Whats your powerlimit? Mines 78%
65%
Jump to: