16x r9 280x mining rig arrived today. big problems. help will get rewarded.

[chip99]

so my 16 gpus arrived. im testing two currently
setup
2x Sapphire Vapor-X Radeon R9 280X
Gigabyte G1.Sniper
AMD A88X AMD A8 Series A8-6600K 4x 3.90GHz
900 Watt Inter-Tech Nitrox Nobility Modular
80+ 64GB SanDisk Solid State Disk 2.5"
win 8.1
amd 3.11 beta driver
amd sdk 2.9

config


important notes:
i have no psyhical access to the hardware its in my brothers house
he has only one display
he is doing the hardware im doing the software
im managing this with teamviewer

so far this is the best i could get

setx GPU_MAX_ALLOC_PERCENT 100
setx GPU_USE_SYNC_OBJECTS 1
cgminer.exe --scrypt -o stratum+tcp://doge.scryptpools.com:3333 -O chip99.1:1 --thread-concurrency 15000 -g 2 -I 13 -w 256

after hours of hours of testing im at this now.

http://imgur.com/rTov2Fa

note the big difference in temperature and one gpu is always having some trouble and dropping to 600 khash or even lower


im losing alot currently cuz i cant mine with my 16 gpus properly 

GPUs that i need to config are

5x     Club 3D Radeon R9 280X royalQueen, 3GB GDDR5  CGAX-R928X7
3x     Sapphire Vapor-X Radeon R9 280X, 3GB GDDR5 11221-02-40G 
8x MSI Radeon R9 280X Gaming 3 GB DDR5 PCIe V277-053R




BTC whoever can help me to get 720-760khash stable will get rewarded with 1 LTC

[1714780459]

1714780459

[1714780459]

1714780459

[Lexomatico]

I'm no expert myself on 280x' but this guide here looks pretty in-depth and could help you out finding the right settings.

http://rumorscity.com/2013/12/14/radeon-r9-280x-litecoin-mining-and-sweet-spot-for-700khashsec/

If it helped a small tip is appreciated

LKsS2E9mVm3WTuqRK3zLvv6Lsjh9Cp9Hq7

[diac21]

skip the tread concurrency and switch to: --shader 4096

use pci-e 16x riser cards to better separate the cards. blow a fan directly onto the existing fan for extra cooling.

btw. i didn't realize this at first. but when you are using teamviewer, or any other remote connect app. hash rates WILL drop. if you disconnect it WILL rise. I think nothing is wrong with your setup. let him hook a display on it and he can check for you.

Even though it won't rise too 740. Your first card will also get 715 khash. second card will stay 715 hash. if you think this was worth: LRwuHV1yE6BTUNfSdAcnb8qzn5NHXtJs8A

[Mirsad]

Best thread ever!
Buying 16 cards and have no clue about multi-GPU mining. 

Are you going to run only 2 cards per mainboard? That's a pretty waste.
4 or 5 would be good. But you'll need pcie risers and better psus.

[GBattaglia]

I prefer 3x GPU builds myself.  They are convenient and quite cheap.  PSUs that are quality and capable of powering a 3X rig are easy to come by (at a decent price), and configuration is much easier.  You can even use a standard cheap H61 style motherboard for a 3x build. 

[traderman]

how many mobo's you got? how much ram per mobo?

[Mirsad]

Reward 1LTC is a joke...
With 16 280x, you should be able to afford at least 0.5 BTC.

Otherwise, just search the forums and do the work yourself.

[chip99]

how many mobo's you got? how much ram per mobo?

8 mobos 4-8gb .ram will get upgraded soon.

currently i flashed one gpu with this http://www.techpowerup.com/vgabios/150288/sapphire-r9280x-3072-131127.html
because of this comment

ALL 7970/280X VAPOR-X MODELS HAVE LOOSENED RAM TIMING SO THEY CAN CLOCK HIGHER
LOOSE RAM TIMINGS HURT SCRYPT PERFORMANCE
FLASH THE NON-VAPORX BIOS FOR YOUR CARD (in this case reference should work.)

after reboot started miner and the flashed gpu was showing 0 khash and just wrong everything. then a freeze occured.
so i cant connect now and nobody can currently restart the computer there so i have to wait.
dont think its really broken i also saved original bios and there is a switch for bios1 and bios2 on the gpu. i just cant do anything for now . have to wait.



riser cards are vital ? every guide had them under "optional"

also 2 gpus per mainboard was decision of my bro he said 3-4 wont fit on a mobo without problems.



not sure what to do tomorrow. guess i will have to flash back. someone just said that vapor-x bios have issues on higher gpu clock and it seemed true i had 700khash with 1020mhz and 600-500 with 1080mhz...
also the fan kicking into other gpu might as well play a big role. i guess its a combination of many issues.
and the fact that i dont have physical access to the miners doesnt hep at all.
this shit is hard as fuck . i imagined it would be alot easier.

well you could say i should be happy with 650-700khash. but im not accepting that . everyone is getting atleast 700khash with this gpus. and i will too sooner or later  . it also have some inpact with 16 gpus every khash is important.

[traderman]

The r9 280 has proximately 2048 stream processing units.

Thread concurrency should be @ ~ 20,000 - 21000 and intensity should be @ 18-19. that will solve your speed issues. Make sure the gpu have space between them and a fan or something else blowing on them to get good air circulation. And get rid of the -g command.

[samesstee]

you need to run another instance of cgminer, infact one for each type of GPU you have as they all require different settings (I'm guessing you know that right??)

then use the -d x,x,x,x in cgminer with , to select which gpu(s) to execute

example
cgminer version 1 -
'-d 0,1'
cgminer version 2 -
'-d 2,3'

[Kreativekrypto]

Alright Chip my ltc address is in sig I have a couple of the cards you just got r9 280x are beast

Step 1: Here we go Download AMD Catalyst 13.1 Software Suite Vista/7 64-bit AMD AND AMD APP SDK v2.8 most these are performing best for those cards right now.

Step 2: use my config file (i havent given this to anyone before but its time to share.)

Step 3 740+ khash

Step 4 Im jealous you have that many gpu's

"pools" : [
   {
      "url" : "yourpool",
      "user" : "workername",
      "pass" : "password"
   }
]
,
"intensity" : "13",
"vectors" : "1",
"worksize" : "256",
"kernel" : "scrypt",
"lookup-gap" : "2",
"thread-concurrency" : "0",
"shaders" : "2048",
"gpu-engine" : "1080-1080",
"gpu-fan" : "0-85",
"gpu-memclock" : "1500",
"gpu-memdiff" : "0",
"gpu-powertune" : "0",
"gpu-vddc" : "1.200",
"temp-overheat" : "85",
"api-mcast-port" : "4028",
"api-port" : "4028",
"auto-fan" : true,
"expiry" : "120",
"expiry-lp" : "3600",
"gpu-dyninterval" : "7",
"gpu-platform" : "0",
"gpu-threads" : "2",
"log" : "5",
"no-pool-disable" : true,
"no-show-processors" : true,
"no-show-procs" : true,
"no-unicode" : true,
"queue" : "3",
"scan-time" : "60",
"scrypt" : true,
"skip-security-checks" : "0",
"submit-stale" : true,
"temp-hysteresis" : "3",
"shares" : "0",
"kernel-path" : "/usr/local/bin"
}


btw if I helped anyone else please let me know with contribution or +1's are nice

[revvv]

 -I 13 -g 2 -w 256 --thread-concurrency 8192 --gpu-engine 1070 --gpu-memclock 1500  --auto-fan

I get 740kh/s each card but the temp is a big issue. Keep them cool if the temp reaches 80c you'll see a drop in hashrate.

Let me know

[Chrophedra]

I have 3x Sapphire Vapor-X R9 280x running at 1080/1500 undervolted with SAPPHIRE TRIXX to 1137

intensity 13, thread-concurrency 8192, GPU-THREADS 2, shaders 2048, -w 256 --auto-fan 0-95

750khs each


Asrock motherboard and amd cpu. I was having problems until I disabled the onboard GPU in UEFI/BIOS. You might want to check that out. (you could check without disabling in bios by starting cgminer with -d 0,1 (gpu 0 and 1, if those are your R9's)

And define heat problems? they are supposed to run like 75 degrees. I use 2 boards to run 3 cards (I use two cases too, duh. Cooler master Haf X and a Storm Trooper)

My fans are at 1500/2500rpm and temps stay at 72-75degress , I use autofan

PS

the other card is run off an asus intel motherboard with an i5 CPU , same settings and same hashrate @ 750khs

Oh and running Windows 7 ultimate with latest amd drivers 13.9 and latest gpu capable cgminer 3.7.2

[tichwich85]

Same as Chrop, no issues on those settings and i got it even undervolted to 1.112 and stable temp around 71-73 doing 740-750 khs

[farkedup]

do this: https://app.box.com/s/kslalufsu1rdxhxi2cup

[nonny12]

I spent way too much time optimizing my Sapphire R9 280x Vapor-X. The hashrate would drop after a while for no apparent reason. I noticed that the GPU usage would bounce around between 100% and 60% after running for about one minute (see Afterburner screenshot below).

https://i.imgur.com/aLlaY91.png

The cause of this was a really HIGH VRM1 temperature around 110 deg celcius. Check this temperature using GPU-Z - it is shown in the bottom of the sensors tab. Undervolting the card and turning up the fan speed helped with this issue. You can use Sapphire TRIXX to undervolt. Using '--gpu-powertune -20' will also decrease VRM temps and the cards will use less power with no negative effect on hashrate. GFX card core temp isn't the issue (VRM temp is), so set a sort of high fan speed (mine is 65%) instead of using auto fan.

Finally, I ended up lowering the memory clock rate, even though everything I had read indicated that higher memclock rates would increase hashrate. I ended up at 1500, like the posters above. The optimal engine clock speed varies from card to card. While the people above had the most success at engine clock 1080, mine is the fastest at 1030. Start somewhere around 1020 and increase in steps of 5. If hashrate drops after increasing by a step of 5, then lower the rate 2 or 3.

For thread concurrency and shaders flags, use one or the other, not both. My optimal thread concurrency is 11200, but you should also try 8192 or 8191. Also, you will want to close GPU-Z, Afterburner, Trixx or any other GPU monitoring software before you start cgminer. My hashrate increases by about 5% when these programs are not running.

My final startup.bat is as follows

Code:

setx GPU_MAX_ALLOC_PERCENT 100
setx GPU_USE_SYNC_OBJECTS 1
cgminer.exe --scrypt -o stratum+tcp://xxx.com:3333 -u xxx.xxx -p xxx --thread-concurrency 11200 -g 2 -w 256 --gpu-engine 1032 --gpu-memclock 1500 --gpu-fan 65 -I 13 --gpu-powertune -20

LTC Address: LbGjCkLc6ZPYSV6fGqsoSt38Zj2o37j3Bf

[chip99]

thanks for the replies i will try all this stuff tomorrow. well there is a hardware build issue aswell i guess. he also said he is putting the miners into case now so just like regular computers because of some issues . well i will have to talk to him tomorrow.
also realized that i forgot to buy resistors for dummy plugs . he have to connect his one display to every new miner now.


well the major problem is that one of the 2 gpus is never staying stable and just keep going down lower and lower. will have to find a reason for this.

my top 3 guess are
gpu model issue (only have 3 of the vapor x, couldnt test mining with the other models yet)
missing dummy plug resistors
hardware build issues

so sounds like it must be hardware. i tried to optimize alot tried lots of stuff whole day today with software but the core problem that one gpu is behaving bad is not going away.
so the problem is on his side end i guess.

the mining rig will be outside on the balcony. so cooling might be not a issue. its winter afterall.

[STT]

Use the shaders switch instead of TC
At least to start off with, once you have stable decent hash config then you can try to improve but as you say you are losing mining time messing with it meanwhile

powertune -20 will keep the cards cooler if you are having heat problems but it'll restrict the hash possibly.  Are you paying for power ?   hash per watt can better and easier to get with less power.  -10 drops the volts enough on mine to run 'cool'

Final thing is get your poor brother an infrared temp gauge.   He doesnt have to touch anything or use software, just point the reader at a hotspot and it'll show temperature and with this much high end hardware getting worked hard, ventilation & heat buildup could be a problem on card components even when CG says GPU is not overheated

Quick google - http://www.ebay.co.uk/sch/sis.html?_nkw=Digital%20Infrared%20Temp%20Gun%20Gauge%20Thermometer

[Eliavres]

what clocks are your x280 currently at? that cpu you got, its an APU right? you could mine with the embedded gpu in them too I think, would probably get you at least 80khash each

getting different models of 280x wasnt exactly a good idea, different models from different brands have different asic percentages, which will influence the max OC you can do on them and the voltage needed for it. if you really want to get the most out of those cards, you cant just set things up like "one size fits all", each card can reach its own max and it takes time to figure it out.

[chip99]

I have 3x Sapphire Vapor-X R9 280x running at 1080/1500 undervolted with SAPPHIRE TRIXX to 1137

intensity 13, thread-concurrency 8192, GPU-THREADS 2, shaders 2048, -w 256 --auto-fan 0-95

750khs each


Asrock motherboard and amd cpu. I was having problems until I disabled the onboard GPU in UEFI/BIOS. You might want to check that out. (you could check without disabling in bios by starting cgminer with -d 0,1 (gpu 0 and 1, if those are your R9's)

And define heat problems? they are supposed to run like 75 degrees. I use 2 boards to run 3 cards (I use two cases too, duh. Cooler master Haf X and a Storm Trooper)

My fans are at 1500/2500rpm and temps stay at 72-75degress , I use autofan

PS

the other card is run off an asus intel motherboard with an i5 CPU , same settings and same hashrate @ 750khs

Oh and running Windows 7 ultimate with latest amd drivers 13.9 and latest gpu capable cgminer 3.7.2

there is always atleast 10-15c temp difference between gpu0 and gpu1
so one of them get too hot and get throttled i guess. not sure.
will look into the onboard gpu in bios. dont know if my brother disabled it.

@Eliavres i ordered what was available. everything is out of stock.

anyone tried the new amd driver 13.12?