SILENTARMY v5: Zcash miner, 115 sol/s on R9 Nano, 70 sol/s on GTX 1070

[zawawa]

NR_ROWS_LOG=14 turned out to be less efficient than NR_ROWS_LOG=15.
Oh well. I m getting 147 sol/s with RX 480 with quite a bit of room for further optimization left, so that should be good enough.
I will create a new public GitHub repo and start a new topic for my fork shortly.

[Amph]

I will, I will. No worries. I will release my fork in a day or two.
Although it is not as fast as Claymore's yet, it should be much faster than SA v5 on AMD cards.
I already sold most of my NVIDIA cards, so I have no idea how it would compare on NVIDIA cards.
If I had received a lot of donations, I could have justified purchasing GTX 1060 to my wife, but I could only dream...

why you have sold them? they perform better than amd, just look at nice hash result 370 sol on a 1070, you couldn't achieve such speed?

He probably sold before the recent Nvidia speed boosts, which made them somewhat in the ballpark cost/perfwise.  370 is on a highly overclocked card, realistically 1070 gets ~330@150w, and also pool locked. So sure they peform ~50% better currently, but cost more than twice as much.

216 sol a 1060 that is cheap and with less than 90 watt, better than a 470, 1070 is still not there yet i agree, but nvidia isn't only 1070 you know, there is 1050ti/1060 etc...

1060 3GB is a decent card for mining.  Similar speed to a a 470 (which does ~200 with stock clocks), but also costs more. Can get a 470 for $175, 1060 is $200+ ($200 is for garbage cooling). 216 sol is heavy OC, I have moderate OC on a 1060 3GB, pulling 200, my other 2 are only doing 180, so I highly doubt you using 90w.  The other issue is currently Zec profitability is pretty crap, I can make $30/month Eth only with a 470 easy (no memory OC), pulling very little watts, or $22 on Zec.  Nvidia missed the party for Zec it seems.

1070 can do 310 sol at 100 watt so i can't see how the 1060 is not doing 200 sol and less than 90 watt, it should be 70-80, nvidia is currently more efficent than amd that is certain

talking about the nicehash last miner of course

It seems that the 1070 is better than the R9 390 or even the Nano. The nano can do 300 at 140W.

Looking at my card in GPU-Z again, it may be possible, the shown power % of TDP on it yesterday was ~80%, now its ~50-60% for some reason.  Maybe was looking at different brands.....  Don't feel like moving my Killawatt.  

Anyways, currently a 1070 would make ~$30/month, and have ~14 month ROI, and a 1060 3GB would have ~11 month ROI, add in system costs, and it is doesn't look any better.  470 on Ethereum would have ~7 month ROI. So like I said, Nvidia totally missed the party.  At this point, being slightly more efficient for ZEC isn't really a win, unless there is some major pump happening on ZEC, because Ethereum is so much more profitable.

how is etheruem more profitable? dunno where you are taking those numbers, but ethereum is not more profitable than zcash, not even for amd, they are there

a 1070 is making 1.2 net per day currently at 10 cent electricity, that is $36 per month, which is 12 months roi, if you get it used($300 second hand) or founder edition it's even 10 months roi or less

a 470 will roi on 7 months, but you need to roi more than one 470 card to equal one 1070, because the 1070 is faster even on etheruem 32-33MH vs 27mh(i don't remember the exact hahs of a 470 there)?

Profitability depends on card, though right now Zec is particularly bad.  A 470 doin 27.7MH Eth (strap mod, no OC) make $30 (according to whattomine, and 0.068/kwh), doing Zec it will do 200H, and get ~$20.

A 1060 will make ~$20 Zec, or ~20 Eth.

A 1070 would make ~$32 Zec, or ~$34 Eth (with crazy high memory OC)

7000/200/300/Fury cards, I think are similar Zec and Eth profit since Zec is in the toilet right now, before they were better off doing Zec.

So for Polaris cards, Eth is hands down better (~50% more profit), and for other cards its similar, with Eth sometimes having a small lead.

If you are talking about used cards, you can get a 470 for $145, giving it a 5 month ROI.

No matter how you slice it, even with recent Nvidia boost to Zec, it just isn't as good. Sure a highly OC 1070 rig doing Eth will make slightly more profit than a 470 stock clocks, but it costs twice as much, so not a win. 1060 does maybe 19MH eth with +200 mem, and 1070 memory bus is only 33% wider, so 33MH must be some crazy insane torture for a 1070. An abused 470 will do ~31MH.

just a moment where etheruem is better but it will not last, but those price are off in my country, for example i have the 470 and 1060 at the same price, and a 1060 is better at zec than a 470 same hash but lower wattage

let's talk once again about real numbers, assuming an investment of 3k, with 1070 i would buy 6 x 1070, 1 full rig, which will do 1920 sol at 650watt(320 sol/90watt per card, check nicehash thread for reference)

with the 470 i would buy 11 gpu because you will lose some money on the second rig due to additional components, these 2 rig will hash at 2200/1100watt(200 sol at 90 watt per gpu)

now take a look at whattomine, we are basically doing the same net profit per day at 0.1 cent electricity, so again amd being better it's a myth

actually a 1070 can hash at 370 sol/140 watt, at least some model can, so you have better hash(2220) and lower wattage(900), on zcash nvidia win, on etheruem not, but etheruem will be dead soon and i don't care

[Eliovp]

just a moment where etheruem is better but it will not last, but those price are off in my country, for example i have the 470 and 1060 at the same price, and a 1060 is better at zec than a 470 same hash but lower wattage

let's talk once again about real number, assuming an investment of 3k, with 1070 i would buy 6 x 1070, 1 full rig, which will do 1920 sol at 650watt

with the 470 i would buy 11 gpu because you will lose some money on the second rig due to additional component, these 2 rig will hash at 2200/1100watt

now take a look at whattomine, we are basically doing the same net profit per day at 0.1 cent electricity, so again amd being better it's a myth

actually a 1070 can hash at 370 sol/140 watt, at least some model can, so you have better hash(2220) and lower wattage(900), on zcash nvidia win, on etheruem not, but etheruem will be dead soon and i don't care

The only thing happening here is the fact that the AMD VS NVidia "war" is yet again filling up a thread with nonsense not related to the thread itself.

We can discuss this subject endlessly, there will simply always be people who prefer AMD and vice versa.
It's like the endless Iphone(ios) vs Samsung(android) discussion...


@zawawa

I'm following your progress up close.

Some extra options for us Linux users is always very welcome.

Greetings

[laik2]

@zawawa
With new AMD linux drivers I'm getting average 2/3S/s more than with 16.40 version.
I have r9 390 for testing purpuses and it gets almost equal hash speed as rx480, but 390s are 512bit and should overperform 480s with at least 30%.
If you don't have Hawaii or older GPUs - PM me and I will try to bring some vps up for testing.

[zawawa]

Peeps, the wait is OVER!

https://github.com/zawawawa/gatelessgate

150 sol/s on RX 480 with a modded BIOS. Enjoy!

[zawawa]

@zawawa
With new AMD linux drivers I'm getting average 2/3S/s more than with 16.40 version.
I have r9 390 for testing purpuses and it gets almost equal hash speed as rx480, but 390s are 512bit and should overperform 480s with at least 30%.
If you don't have Hawaii or older GPUs - PM me and I will try to bring some vps up for testing.

I will definitely contact you once the dust settles a bit.
Unfortunately, I sold my good old 290X's, too...

[laik2]

@zawawa
With new AMD linux drivers I'm getting average 2/3S/s more than with 16.40 version.
I have r9 390 for testing purpuses and it gets almost equal hash speed as rx480, but 390s are 512bit and should overperform 480s with at least 30%.
If you don't have Hawaii or older GPUs - PM me and I will try to bring some vps up for testing.

I will definitely contact you once the dust settles a bit.
Unfortunately, I sold my good old 290X's, too...

In an hour I will bring up Linux with R9 390 to test your new fork. Unfortunately main.c has some windows code that cannot be compiled under linux platform. I will try to port it for Ubuntu 16.04 later.

[zawawa]

I also started a new topic for my fork:

https://bitcointalk.org/index.php?topic=1716584

I would really appreciate feedback as my testing environments are quite limited.

[zawawa]

@zawawa
With new AMD linux drivers I'm getting average 2/3S/s more than with 16.40 version.
I have r9 390 for testing purpuses and it gets almost equal hash speed as rx480, but 390s are 512bit and should overperform 480s with at least 30%.e
If you don't have Hawaii or older GPUs - PM me and I will try to bring some vps up for testing.

I will definitely contact you once the dust settles a bit.
Unfortunately, I sold my good old 290X's, too...

In an hour I will bring up Linux with R9 390 to test your new fork. Unfortunately main.c has some windows code that cannot be compiled under linux platform. I will try to port it for Ubuntu 16.04 later.

Oops, my bad... Can you show me the errors? It should be quite easy to fix them.

[Eliovp]

@zawawa, compiling on ubuntu 14.04 and getting some errors here mate. (to many to list to be honest :p)

Edit: just checked your thread, i see you're on it

[laik2]

@zawawa
With new AMD linux drivers I'm getting average 2/3S/s more than with 16.40 version.
I have r9 390 for testing purpuses and it gets almost equal hash speed as rx480, but 390s are 512bit and should overperform 480s with at least 30%.e
If you don't have Hawaii or older GPUs - PM me and I will try to bring some vps up for testing.

I will definitely contact you once the dust settles a bit.
Unfortunately, I sold my good old 290X's, too...

In an hour I will bring up Linux with R9 390 to test your new fork. Unfortunately main.c has some windows code that cannot be compiled under linux platform. I will try to port it for Ubuntu 16.04 later.

Oops, my bad... Can you show me the errors? It should be quite easy to fix them.

I will message you login credentials to see for yourself.

[xeridea]

just a moment where etheruem is better but it will not last, but those price are off in my country, for example i have the 470 and 1060 at the same price, and a 1060 is better at zec than a 470 same hash but lower wattage

let's talk once again about real numbers, assuming an investment of 3k, with 1070 i would buy 6 x 1070, 1 full rig, which will do 1920 sol at 650watt(320 sol/90watt per card, check nicehash thread for reference)

with the 470 i would buy 11 gpu because you will lose some money on the second rig due to additional components, these 2 rig will hash at 2200/1100watt(200 sol at 90 watt per gpu)

now take a look at whattomine, we are basically doing the same net profit per day at 0.1 cent electricity, so again amd being better it's a myth

actually a 1070 can hash at 370 sol/140 watt, at least some model can, so you have better hash(2220) and lower wattage(900), on zcash nvidia win, on etheruem not, but etheruem will be dead soon and i don't care

Issues with your calculations:

Arbitrary cash limit, gimping the 470 setup.

470 in US costs $170, not $200.  1060 3GB are all 200+, looking at ~$220 unless you want garbage cooler.

So your arbitrary rig would cost $330 less, or have 2 more cards.

If you are OCing the crap out of a 1070, it will not last a long time, you could also OC the 470.

Ethereum won't be dead soon, PoS is a while out, probably will be delayed due to recent happenings. So you could take the 13 470s and mine Eth for a clear win.

Of course prices can always change, but it seems Zec is overmined, due to pre-polaris cards being good at it, and now some 1070 owners jumping on the cool train, that they already missed.

I have my 1060 3GB cards doing Zec, just because they aren't very good at Eth, but I wouldn't go and buy new ones to mine, and definitely not 1070. I go where the money is, I would buy whatever card had best ROI.  It just happens that whenever I have mined, AMD has had better ROI.

[nerdralph]

Alright, I'm hitting 140 sol/s with RX 480, and I haven't even tried NR_ROWS_LOG=14 yet.
Good stuff. I think I can go well over 200 sol/s with a GCN assembler now.

CLRX is my favorite.

I love it, too! That's the way to go, methinks.

Re-writing everything in asm looks like a lot of work.  You might be able to take the .isa output from the OpenCL code and modify it rather than starting from scratch.
I've been looking at the llvm AMDGPU back-end to see if I can use a couple asm functions with the rest of the code in OpenCL.

[Amph]

just a moment where etheruem is better but it will not last, but those price are off in my country, for example i have the 470 and 1060 at the same price, and a 1060 is better at zec than a 470 same hash but lower wattage

let's talk once again about real numbers, assuming an investment of 3k, with 1070 i would buy 6 x 1070, 1 full rig, which will do 1920 sol at 650watt(320 sol/90watt per card, check nicehash thread for reference)

with the 470 i would buy 11 gpu because you will lose some money on the second rig due to additional components, these 2 rig will hash at 2200/1100watt(200 sol at 90 watt per gpu)

now take a look at whattomine, we are basically doing the same net profit per day at 0.1 cent electricity, so again amd being better it's a myth

actually a 1070 can hash at 370 sol/140 watt, at least some model can, so you have better hash(2220) and lower wattage(900), on zcash nvidia win, on etheruem not, but etheruem will be dead soon and i don't care

Issues with your calculations:

Arbitrary cash limit, gimping the 470 setup.

470 in US costs $170, not $200.  1060 3GB are all 200+, looking at ~$220 unless you want garbage cooler.

So your arbitrary rig would cost $330 less, or have 2 more cards.

If you are OCing the crap out of a 1070, it will not last a long time, you could also OC the 470.

Ethereum won't be dead soon, PoS is a while out, probably will be delayed due to recent happenings. So you could take the 13 470s and mine Eth for a clear win.

Of course prices can always change, but it seems Zec is overmined, due to pre-polaris cards being good at it, and now some 1070 owners jumping on the cool train, that they already missed.

I have my 1060 3GB cards doing Zec, just because they aren't very good at Eth, but I wouldn't go and buy new ones to mine, and definitely not 1070. I go where the money is, I would buy whatever card had best ROI.  It just happens that whenever I have mined, AMD has had better ROI.

it will last much longer than you think, it's not heavy oc just mem oc core is at zero, only on some model it work and can get you 370 for 90 watt

point is 370 at 90 watt is better than 200 at 90 watt, factoring the electricity cost of 10 cent, make the investment equal, there is no amd better ever, roi is the same even with your 13 gpu(which is not the case here, i can't find a 470 for $170) and any case it's 12 gpu not 13

a 1070 cost 400 on newegg, 400 x 6 + 400 for component = 2800 = 170 x 12 + 800(two rig of 470), 370 x 6 at 900watt vs 200 x 12 at 1200watt, same roi, so no amd is not better even on newegg, case closed, here the price is not $170 but 200 so it's far worse...

a 1060 can do 226 sol at 85watt, even if it cost 200 it's a better choice

[Hotmetal]

Would you girls mind creating a new thread and taking it there as this AMD vs Nvidia drama is dragging on.

[tiennou]

@zawawa, compiling on ubuntu 14.04 and getting some errors here mate. (to many to list to be honest :p)

Edit: just checked your thread, i see you're on it

Compiled without any hassle here, with the latest commits.
Getting 115sols/s with a stock rx 470 (for reference, that same card would do ~175sols/s with optiminer - including the fee).

[QuintLeo]

Ethereum won't be dead soon, PoS is a while out, probably will be delayed due to recent happenings. So you could take the 13 470s and mine Eth for a clear win.

Of course prices can always change, but it seems Zec is overmined, due to pre-polaris cards being good at it, and now some 1070 owners jumping on the cool train, that they already missed.

ETH POS move 6 months out give or take per the last timetable I've seen, but as you mention it could be delayed AGAIN.

 That's going to free up a LOT of hashrate when it does happen, at which point profitability for the rest of the current "basket" is going to take a big drop.


 In my case, I jumped on ZEC with my 1070s for the same reason that they were previously doing ETH mining over *most but not all* of the last 6 months - better profitability vs what I originally built the rigs FOR.
 However, once ZEC dumps out far enough, I have  backup plan - go back to the original intent of those rigs, which doesn't involve ETH/ETC/ZEC/ZClassic/XMR or the other "lesser" members of the current "basket of profitable GPU mineable altcoins".



 Is good to see a viable alternative to the "for fee" stuff out of Claymore and Optiminer, but y'all still have a ways to go to match the 200+ sol/s I'm seeing out of my NON-MODDED Sapphire RX 470 4GB cards.

 Keep plugging!

[sp_]

SA v6 (zavasava) 170MHASH@gtx 1070
Eqminer 1.03b 340MHASH@gtx 1070 (+100%)

Same clocks
zavasava's running with 2 threads.
Eqminer 1.03b running with -ca (2 threads)

[zawawa]

SA v6 (zavasava) 170MHASH@gtx 1070
Eqminer 1.03b 340MHASH@gtx 1070 (+100%)

Same clocks
zavasava's running with 2 threads.
Eqminer 1.03b running with -ca (2 threads)

Just to make it very clear, my fork is NOT SA v6.
And I made it very clear that NVIDIA was not my focus. (I don't even own a current-gen NVIDIA card.)
And you intentionally misspelled my name.
You have been so rude to me that I just decided to optimize the hell out of my miner for NVIDIA cards.
My new GTX 1060 is arriving tonight, and I will catch up with Eqminer very soon.

[xeridea]

Ethereum won't be dead soon, PoS is a while out, probably will be delayed due to recent happenings. So you could take the 13 470s and mine Eth for a clear win.

Of course prices can always change, but it seems Zec is overmined, due to pre-polaris cards being good at it, and now some 1070 owners jumping on the cool train, that they already missed.

ETH POS move 6 months out give or take per the last timetable I've seen, but as you mention it could be delayed AGAIN.

 That's going to free up a LOT of hashrate when it does happen, at which point profitability for the rest of the current "basket" is going to take a big drop.


 In my case, I jumped on ZEC with my 1070s for the same reason that they were previously doing ETH mining over *most but not all* of the last 6 months - better profitability vs what I originally built the rigs FOR.
 However, once ZEC dumps out far enough, I have  backup plan - go back to the original intent of those rigs, which doesn't involve ETH/ETC/ZEC/ZClassic/XMR or the other "lesser" members of the current "basket of profitable GPU mineable altcoins".



 Is good to see a viable alternative to the "for fee" stuff out of Claymore and Optiminer, but y'all still have a ways to go to match the 200+ sol/s I'm seeing out of my NON-MODDED Sapphire RX 470 4GB cards.

 Keep plugging!

At current rate, the difficulty bomb would be about 8 months, but judging by the speed of development for Ethereum, it will be delayed a bit.

Point is, if Zec doesn't pick up, you can get ~50% more profit doing Eth until PoS, giving you much quicker ROI, and less risk. If Zec does pick up, Eth profit would also rise, due to people switching to Zec.

If you already have 1070s, sure go ahead and mine Zec, you should always mine what is most profitable. I just wouldn't go out and buy one, and have year+ ROI, which is more risky now due to uncertainty of future of mining.