with the help of Stephan: win7 x64 professional sp1 /// 5*hd7950 (1050/1250) ///8Gb ram+12GB page geth + ethminer 124000000 h/s - max 120000000 h/s - typical
120 MHs / 5 cards --> that would be 24 MHs per single 7950 card . Well, this is very high number. I see others (mostly on Linux) getting approx 14 Mhs per 7950 card. Is there any Windows/Linux trick with ethminer? Which Catalyst drivers (13.x, 14.x, 15.x) are best?
I was asking myself the same, seems like those are overclocked no ? 5*hd7950 (1050/1250)
mine clocks at 850 MHz, dunno if that 1050 is the clock
I did some tests myself with clocks 1100/1500 but couldn't get more then 14 MHs on Ubuntu 14.04, drivers fglrx 15.20.2 (latest from fglrx-updates, equal to Catalyst 15.7 on Windows: http://www2.ati.com/rss/catalyst.rss). So, any hints on how to get those 24 MHs?
Interesting, stilt's back in play. So your 7950s on stilt make 24mh/s each but are your 290s on stilt's? You lowered the 290 memclock to 1000 and they do 26mh/s but that doesn't seem like much of an improvement.
Windows 7 / 15.7/i5-2500k/8GB RAM ASUS R9 280X DirectCU II TOP 3 GB @1150gpu/1625mem pulling 27.6MH/s ~230Watt on a silver PSU. @1070gpu/1600mem pulling 26MH/s ~200Watt on a silver PSU. more test coming as I will get more gpu' s home.
with the help of Stephan: win7 x64 professional sp1 /// 5*hd7950 (1050/1250) ///8Gb ram+12GB page geth + ethminer 124000000 h/s - max 120000000 h/s - typical
120 MHs / 5 cards --> that would be 24 MHs per single 7950 card . Well, this is very high number. I see others (mostly on Linux) getting approx 14 Mhs per 7950 card. Is there any Windows/Linux trick with ethminer? Which Catalyst drivers (13.x, 14.x, 15.x) are best?
I was asking myself the same, seems like those are overclocked no ? 5*hd7950 (1050/1250)
mine clocks at 850 MHz, dunno if that 1050 is the clock
I did some tests myself with clocks 1100/1500 but couldn't get more then 14 MHs on Ubuntu 14.04, drivers fglrx 15.20.2 (latest from fglrx-updates, equal to Catalyst 15.7 on Windows: http://www2.ati.com/rss/catalyst.rss). So, any hints on how to get those 24 MHs?
OK, did some more testing. Switched to Ubuntu 14.10, drivers fglrx 14.20.7 (equal to Catalyst 14.4) and I now get 24 MHs on 7950 and 26 MHs on 7970. That's pretty decent.
Btw: I have a total of apprx. 150 MHs ... is it worth to solo-mine or pool mining is actually only option with the current Network hashrate?
11x HD7970 3gb 22 mh 242mh total 3x R9 280x 3gb 25 mh 75mh total 2x R9 290 4gb 23.5 mh 47mh total 1x HD7950 3gb 20 mh 20mh total
What motherboard are you using to get 11 GPUs going? Are you using PCIe splitter? I am looking for mobo setup with 8+ cards.
I use 3mobo (asrocks 970extr) with 5 gpu and 1 mobo with 2gpu i use psu lepa1600w with 10/2x pci-e connections, and i use setup with 2x psu750w with splitter. All mounted in a aluminium rack connected with pci-e risers
i wouldn't want to use more than 5 gpu per mobo but its possible bilding a rig.
@taylor_luk I suspect this may have something to do with the mining thread still working on the previous block, while the thread that validates the solution is already on the next block. But it could also be a GPU glitch. Anyway, as long as you only get these sporadically, it isn't a big deal.
However I see that the load of the GPU (AMD 7950) is at 100%.
How can I make sure that the GPU is actually mining and is there any option to show the "miner ..." lines? I even tried adding verbosity parameter "-v 9" but nothing changed.
Anybody with similar issues?
edit:
hmm, it looks that the ethminer from ubuntu repository is built with less verbosity ... I complied my own version and I see full "miner ..." lines now ...
with the help of Stephan: win7 x64 professional sp1 /// 5*hd7950 (1050/1250) ///8Gb ram+12GB page geth + ethminer 124000000 h/s - max 120000000 h/s - typical
120 MHs / 5 cards --> that would be 24 MHs per single 7950 card . Well, this is very high number. I see others (mostly on Linux) getting approx 14 Mhs per 7950 card. Is there any Windows/Linux trick with ethminer? Which Catalyst drivers (13.x, 14.x, 15.x) are best?
I was asking myself the same, seems like those are overclocked no ? 5*hd7950 (1050/1250)
mine clocks at 850 MHz, dunno if that 1050 is the clock
I did some tests myself with clocks 1100/1500 but couldn't get more then 14 MHs on Ubuntu 14.04, drivers fglrx 15.20.2 (latest from fglrx-updates, equal to Catalyst 15.7 on Windows: http://www2.ati.com/rss/catalyst.rss). So, any hints on how to get those 24 MHs?
I run standard clocks and get around 23.5 mhs/s on a 7 gpu rig i see hash rates as high as 166 mh/s and as low as 155 , when it gets cooler i will try a mild overclock and try and squeeze out another 1-2 mh/s per card
if there are some Ethminer or Geth API that exposes individual miners Hash rate like how they do it on nanopool?
Taylor
Actually if you do " geth attach" on your geth instance and then type "eth"
you will see the curretn hash rate is displayed there its some big number not sure what unit of mesaure they use but it should be trivail to create a dashboard that pulls this number in realtime and displays it in mh/s
Actually if you do " geth attach" on your geth instance and then type "eth"
you will see the curretn hash rate is displayed there its some big number not sure what unit of mesaure they use but it should be trivail to create a dashboard that pulls this number in realtime and displays it in mh/s
Hmm, I see this number:
hashrate: 72955810727690180,
But this number does not make any sense at all. I'm mining with three rigs 165 Mhs + 25 Mhs + 15 Mhs therefore the total hashrate should be approx. 200 Mhs
Moreover, after a minute this number changes a lot:
hashrate: 5973435984647768000
Now, what exactly is this number, what does it represent? Some number of accepted shares?
Do you want to share the individual hash rate, core/mem clock and volt for these cars?
I have a 7990 does about 50MH undervolt at 950/1250mhz, that 390 must be close to 40MH.
The 7990 is running at 1175 core and 1500 mem (cannot down clock mem on this card...) The R9390 is running at 1175 core and 1175mem(on this specific card syncing the mem and core clocks gives me the best result and keeps temps lower) Combined they run best at global work size of 16384 and 8192 with not much effective difference given average block time of 17s.
The 7990's is running at 59mh/s and the R9390 at 32.5 mh/s
Note these are at stock voltages. The 7990 can go to 1200core stable without voltage tweaking but temps do rise here...runs at 1.2v The R9390 can go 1225core stable with a slight voltage tweak to 1.192v but again temps start rising rapidly here so I prefer to run them a bit cooler at 1175.
Do you want to share the individual hash rate, core/mem clock and volt for these cars?
I have a 7990 does about 50MH undervolt at 950/1250mhz, that 390 must be close to 40MH.
The 7990 is running at 1175 core and 1500 mem (cannot down clock mem on this card...) The R9390 is running at 1175 core and 1175mem(on this specific card syncing the mem and core clocks gives me the best result and keeps temps lower) Combined they run best at global work size of 16384 and 8192 with not much effective difference given average block time of 17s.
The 7990's is running at 59mh/s and the R9390 at 32.5 mh/s
Just so you know, the ethereum hash algo is quite sensitive to VRAM speed / clocks. I found that when I was underclocking the VRAM on my R9 280Xs to 1100 MHz, I was giving up a pretty big percentage of my hashrate vs running them at the stock 1500 MHz.
I do of course undervolt the gpu core, but found that the kw/hash ratio more or less remained the same with the VRAM @ 1100 vs 1500 with heat being the only difference. So, when underclocking VRAM, you're really only slowing down your hashing and saving a little due to less heat -- and that savings due to less heat is probably negligible at best.
Do you want to share the individual hash rate, core/mem clock and volt for these cars?
I have a 7990 does about 50MH undervolt at 950/1250mhz, that 390 must be close to 40MH.
The 7990 is running at 1175 core and 1500 mem (cannot down clock mem on this card...) The R9390 is running at 1175 core and 1175mem(on this specific card syncing the mem and core clocks gives me the best result and keeps temps lower) Combined they run best at global work size of 16384 and 8192 with not much effective difference given average block time of 17s.
The 7990's is running at 59mh/s and the R9390 at 32.5 mh/s
Just so you know, the ethereum hash algo is quite sensitive to VRAM speed / clocks. I found that when I was underclocking the VRAM on my R9 280Xs to 1100 MHz, I was giving up a pretty big percentage of my hashrate vs running them at the stock 1500 MHz.
I do of course undervolt the gpu core, but found that the kw/hash ratio more or less remained the same with the VRAM @ 1100 vs 1500 with heat being the only difference. So, when underclocking VRAM, you're really only slowing down your hashing and saving a little due to less heat -- and that savings due to less heat is probably negligible at best.
For the R9 390's I get the best rate if I sync the mem clock with the core clock. Increasing the mem clock to stock simply degrades my Mh/s rate. Not by very much but around 300kh/s. So no point in wasting electricity and adding heat. Cards likely differ. With the 7990 it is exactly the opposite. If I underclock the mem on that it gives me a lot less Mh/s.
Comments
I have lowered to @950/1250 and it pulls @120mh/s now.
all the 7950gpus are stilt-biosed
PS 5*r9 290 /// @950gpu/1000mem /// Win7 x64 /// Cat 14.12
132mh/s
3xMSI R9-390 8GB RAM
75 MH/s
ASUS R9 280X DirectCU II TOP 3 GB
@1150gpu/1625mem pulling 27.6MH/s ~230Watt on a silver PSU.
@1070gpu/1600mem pulling 26MH/s ~200Watt on a silver PSU.
more test coming as I will get more gpu' s home.
Btw: I have a total of apprx. 150 MHs ... is it worth to solo-mine or pool mining is actually only option with the current Network hashrate?
Sapphire R9 290 Trixx
CORE +38 / CLOCK 1200 / MEM 1400
min/mean/max: 30650414/30736763/3087628
inner mean: 30719040 H/s
Phoning home to find world ranking...
if there are some Ethminer or Geth API that exposes individual miners Hash rate like how they do it on nanopool?
Taylor
11x HD7970 3gb 22 mh 242mh total
3x R9 280x 3gb 25 mh 75mh total
2x R9 290 4gb 23.5 mh 47mh total
1x HD7950 3gb 20 mh 20mh total
2*R9 390 8GB 32mh 64mh total
All mounted in a aluminium rack connected with pci-e risers
i wouldn't want to use more than 5 gpu per mobo but its possible bilding a rig.
global-work @8192
local-work @128 I have seen it solved 3 blocks but reports GPU gave incorrect result.
Are these params too aggressive that it returns incorrect result?
✘ 18:03:19| ethminer FAILURE: GPU gave incorrect result!
Miner2 - 3 R9 280x Toxics + 1 7970 sapphire ~ 107Mh/s
Miner3 - 4 7970 Sapphires ~83Mh/s
Miner4 - 1 R9 270x + 3 7970 sapphire ~97Mh/s
Ave Range Total (depending how flaky my router and internet speed is): 375Mh/s - 500Mh/s
All cards are used AMD Sapphire cards bought off of ebay.
Board used for all is the Arock H81 BTC (horrible MOBO) but $59.
ℹ 21:01:35|gpuminer0 workLoop 1 #6d29f6dd… #6d29f6dd…
ℹ 21:01:36|gpuminer0 workLoop 1 #6d29f6dd… #6d29f6dd…
...
I don't see any lines "miner .... " lines
However I see that the load of the GPU (AMD 7950) is at 100%.
How can I make sure that the GPU is actually mining and is there any option to show the "miner ..." lines? I even tried adding verbosity parameter "-v 9" but nothing changed.
Anybody with similar issues?
edit:
hmm, it looks that the ethminer from ubuntu repository is built with less verbosity ... I complied my own version and I see full "miner ..." lines now ...
On win 7X64.
on a 7 gpu rig i see hash rates as high as 166 mh/s and as low as 155 , when it gets cooler i will try a mild overclock and try and squeeze out another 1-2 mh/s per card
you will see the curretn hash rate is displayed there its some big number not sure what unit of mesaure they use but it should be trivail to create a dashboard that pulls this number in realtime and displays it in mh/s
hashrate: 72955810727690180,
But this number does not make any sense at all. I'm mining with three rigs 165 Mhs + 25 Mhs + 15 Mhs therefore the total hashrate should be approx. 200 Mhs
Moreover, after a minute this number changes a lot:
hashrate: 5973435984647768000
Now, what exactly is this number, what does it represent? Some number of accepted shares?
I have a 7990 does about 50MH undervolt at 950/1250mhz, that 390 must be close to 40MH.
The R9390 is running at 1175 core and 1175mem(on this specific card syncing the mem and core clocks gives me the best result and keeps temps lower)
Combined they run best at global work size of 16384 and 8192 with not much effective difference given average block time of 17s.
The 7990's is running at 59mh/s and the R9390 at 32.5 mh/s
The 7990 can go to 1200core stable without voltage tweaking but temps do rise here...runs at 1.2v
The R9390 can go 1225core stable with a slight voltage tweak to 1.192v but again temps start rising rapidly here so I prefer to run them a bit cooler at 1175.
I do of course undervolt the gpu core, but found that the kw/hash ratio more or less remained the same with the VRAM @ 1100 vs 1500 with heat being the only difference. So, when underclocking VRAM, you're really only slowing down your hashing and saving a little due to less heat -- and that savings due to less heat is probably negligible at best.