there's a new competitor in town: https://github.com/mbevand/silentarmyextract: "Performance
45.7 Sol/s with one R9 Nano
39.6 Sol/s with one RX 480
Note: the code is currently very poorly optimized; it makes zero attempts to keep the queue of OpenCL commands full, therefore one needs to run 2 instances of SILENTARMY in parallel on the same GPU to keep it well utilized." |
|
|
|
Any indications on what difference, if any, there are on RX-series cards running ETH optimized BIOS (1500 straps) vs running stock. Is there a better route to go with BIOS mods for this algo?
See my results above in the thread, I'm running 1375 strap. I think it improves ~10% compared to stock results I saw earlier. Any HW errors running the 1375 strap? If you mean in general, never under 2000 memclock. Mostly 2010 is ok too. SGminer give me errors at that speed though. Over 2010 I get errors, but same with other straps. Running zcash miners, no errors for the little while I mined, it does not stress memory as much as ETH it seems. Very similar to XMR mining in how the hashrate increases/decreases with both coreclock and memclock. |
|
|
|
Any indications on what difference, if any, there are on RX-series cards running ETH optimized BIOS (1500 straps) vs running stock. Is there a better route to go with BIOS mods for this algo?
See my results above in the thread, I'm running 1375 strap. I think it improves ~10% compared to stock results I saw earlier. |
|
|
|
|
RX 470 4g with modified bios at 1275/2010:
[GPU 0] T=-1C A=-1% sols=19.950 with ~100W at the wall
at a more reasonable 1150/2000 I still get 19.2 with 85W
|
|
|
|
|
unless you have a big house and free electricity, you cannot.
|
|
|
|
so six cards per board is expensive and thats only what? maybe $80 per week? sorry I dont think any one has thousands of PCs mining in their house.
No of course they don't mine in their house, they have factories. see : http://www.coindesk.com/bitcoin-miners-ethereum/ |
|
|
|
295x2 is from the wall, I don't care about manufacturer tdp.
If I mod that bios for xmr mining it will most probably do more, those numbers are for dagger optimized card.
Wonder how much wolf0 gets out of hawaii?
Same here, sapphire rx 470 4g ref @778H/s, TDP 110 is from the wall : system Iddles at 75W (including ~10W for the car) and full operates at 175 => 110W BTW with wolf's miner I get even 5W less. Nota: modded bios and undervolting of course. |
|
|
|
I did test under windows with ETH, and yes I have memory errors (hwinfo) although there's is no direct correlation between increasing HW errors and memory errors. With wolf's xmr miner and claymore's ETH or Cryptonight I don't have memory errors at the same clock (2010 and 2000),though I will recheck that later
I downclocked to 1990 to see what happens that way.
I don't have memory errors anymore, but still 12 HW errors/h After two hours, I stoped the miner and started it with lower intensity, It crashed after 5 minutes,computer reboot. Strange. |
|
|
|
I did test under windows with ETH, and yes I have memory errors (hwinfo) although there's is no direct correlation between increasing HW errors and memory errors. With wolf's xmr miner and claymore's ETH or Cryptonight I don't have memory errors at the same clock (2010 and 2000),though I will recheck that later
I downclocked to 1990 to see what happens that way.
I don't have memory errors anymore, but still 12 HW errors/h |
|
|
|
The hardware error is quite large. So I have to reduce the intensity quite a lot. That will reduce the hash rate. [/quote] What card /settings do you use? |
|
|
|
Oh! The (5s) will fluctuate - it's a five second average across all cards. The second line is your current hashrate.
Ok then, sorry. Next I let it run for a while (like an our per run): I run several tests decreasing from 1024/8 (1024/4 was bad). The best mark was 750 H/s with rI=1008. But when I came back after trying 1000 and 992 (and lower...), It didn't go over 737... I tested 512/4 as you asked, very bad hashrate ~500 H/s. After a while in each run I got HW errors and the share count decreased. I will try to reproduce something similar under windows mining eth(unless xmr is running now?) to see if there's a correlation with memory errors. Nota: Wolf's standalone xmr miner give's similar hashrate, but better share count. Awesome, thank you for letting me know! I appreciate it so very, very much! The HW errors don't seem to be normal. We'll sort that out once we have some time. XMR isn't running on Windows still. I'm so sorry. Once we have a moment we'll fix it. We're swamped with preparing for ZCash's launch. I did test under windows with ETH, and yes I have memory errors (hwinfo) although there's is no direct correlation between increasing HW errors and memory errors. With wolf's xmr miner and claymore's ETH or Cryptonight I don't have memory errors at the same clock (2010 and 2000),though I will recheck that later I downclocked to 1990 to see what happens that way. |
|
|
|
Oh! The (5s) will fluctuate - it's a five second average across all cards. The second line is your current hashrate.
Ok then, sorry. Next I let it run for a while (like an our per run): I run several tests decreasing from 1024/8 (1024/4 was bad). The best mark was 750 H/s with rI=1008. But when I came back after trying 1000 and 992 (and lower...), It didn't go over 737... I tested 512/4 as you asked, very bad hashrate ~500 H/s. After a while in each run I got HW errors and the share count decreased. I will try to reproduce something similar under windows mining eth(unless xmr is running now?) to see if there's a correlation with memory errors. Nota: Wolf's standalone xmr miner give's similar hashrate, but better share count. |
|
|
|
Hello I am new to all of this. but I am confused. Even if you dont pay for electricity and already have a computer that can mine at 800 H/S thats only like $13 USD per week. Is that correct? $13 US Dollars per week? whats the point if you have to dedicate a whole computer to do nothing but mine for 24 hours per day just to get $13 per week? Am I missing something?
Pro-miners put six cards per motherboard and have thousands of them. |
|
|
|
hi there,I gave it a go mining xmr (RX 470 1175/2010, rI:1000/WS: 8 )
I wonder if it's expected to have a wide range of fluctuation, with current settings from 600 to 850 H/S ?
tx.
Can you try your rawintensity as 512, and set your worksize to 4, and tell me what you get? Using the formula I posted (which is a reference, but not the rule), you'd be best to work with 1024 - and then, and this applies to everyone - work down in increments of 8, until you stop improving. So many variables come into play here: cache, latency, overall speed, bandwidth. I know it's frustrating to spend so much time tweaking it. But it's the best advice I can offer you. You are currently the only user reporting the fluctuating hash rates, so, if anyone else is having this issue, please, speak up! Just to be sure I'm reading the right output, it's like that: (5s): n (avg:) m H/s GPU0: x / y H/s n is the one fluctuating in the 200 H/s range. The others are quite stable. BTW why is there a difference between the two lines, I would have expected them to be equal with one card only, but they're not. |
|
|
|
hi there,I gave it a go mining xmr (RX 470 1175/2010, rI:1000/WS: 8 )
I wonder if it's expected to have a wide range of fluctuation, with current settings from 600 to 850 H/S ?
tx.
MY HASH RATES WERE STEADY-- The hash rates on my four card R9 280X rig stayed close to 530H/s per card during the half-hour mining run reported above. --scryptr tx, then I guess I should change my parameters, I already tried lower intensity and lower mem clock, no real change a part for lower hashrate. Any clue anybody? |
|
|
|
|
hi there,I gave it a go mining xmr (RX 470 1175/2010, rI:1000/WS: 8 )
I wonder if it's expected to have a wide range of fluctuation, with current settings from 600 to 850 H/S ?
tx.
|
|
|
|
|
Hi there,
got it working for my Sapphire RX 470 4Gb ref under ubuntu 16.04:
cclock 1150/ mclock 2010 custom kernel: 750mhs with rawintensity=1000, worksize=8 for best results.
Claymore gives 778mhs with the same clocks on windows.
the interesting part is the power consumption, it's lower in linux with stock voltage (if someone nows how to undervolt in linux...) as in windows I was highly undervolted.
Nota: it seems to be 10mhs lower when window manager is started.
|
|
|
|
Why amd and dont nvidia ( gtx 1060/1070 ) ?
Nvidia is the best for mining low wattage and high power hash , high initial price it is repaid with a lower electricity bill .
For the Ethereum mining, the nVidia 1070 has the same efficiency as the AMD RX 480. There is no difference. A part from the card's price.... |
|
|
|
I'm stuck at 778H/s too. my mem is not stable over 2010... that's why I use 1375 strap.
You might use higher memory strap to get more stable mining of the Monero. If it is too high, it is not so stable. It's very stable at that mark, run it for a week flawlessly with expected revenue. |
|
|
|
Two things:
First 470's seem to have a hardlock on 2100 Mhz RAM -my Nitro cards behave like they could blast through that limit with ease.
Does anyone know how to remove the 2100 Mhz limit?
Well, there is a 2100 limit in the bios that can be changed with PBE for the cards that do not have Elpida memory. |
|
|
|
|
Show Posts
