December 7th, 2013 
CryptoBadger Now that the excellent Radeon 7950 video cards have essentially become unobtainable, I’m getting a lot of questions from people about how to get the most out of the new Radeon R9 280X in terms of mining. I managed to get my hands on a few Sapphire 280X cards this past week, and have done a little bit of experimenting.
I’ll get around to updating my guide for the 280X at some point in the next couple weeks, but if you’re looking for a good starting point as far as cgminer and undervolting settings go, read on.
Undervolting and Power Consumption
First, the 280X is essentially a re-branded 7970, which means they can be undervolted using VBE7 and the method I previously outlined for the 7950. My Sapphire cards run completely stable at 1137 mV, so if you’re following my undervolting guide, simply use 1137 in place of the 1081 that I recommended for the 7950. I’ll test lower voltages at some point soon, but the 7970/280X doesn’t have a stellar reputation as far as undervolting goes, so I wouldn’t expect too much more.
Running at 1137 mV, you can expect a 3 GPU rig to use between 850-900 watts at the wall, assuming you’re using the rest of the hardware that I recommend in my 7950 guide. You will, of course, need to upgrade the power supply to something a bit larger to accommodate the 280X’s larger power draw—I recommend a 1000+ watt gold or platinum rated PSU. The best choice for those of you without access to very cheap electricity is probably something like this platinum-rated Seasonic PSU. This gold-rated Corsair is a bit less expensive and also a good choice. If you’re looking to run four GPUs, this eVGA PSU has enough PCIe connectors without dealing with splitters.
Cgminer Settings and Performance
These are early test results, but they should get most of you up and running with acceptable performance. My tests were done on a platform with all of the hardware outlined in my guide, the only exception being that the PSU was substituted for a larger 1000 watt unit, and of course the 7950 GPUs were replaced with 280X cards.
For my tests, I was running Xubuntu 12.04 with the latest beta AMD Catalyst drivers installed, AMD SDK 2.7, and cgminer 3.7.2. I’m also using a 280X-optimized scrypt kernal in cgminer. I’ll be looking at Xubuntu 13.x and other driver versions (and Windows, too!) in the coming weeks.
The best results I saw were with the following cgminer settings:
-I 13 -g 2 -w 256 --lookup-gap 2 --thread-concurrency 8192 --gpu-engine 1000 --gpu-memclock 1500 --temp-target 70 --auto-fan
This gets me roughly 710 Kh/sec per GPU, completely stable. Oddly enough, setting the core clock speed any higher than 1000 mhz results in lower cgminer performance.
I’ve read about some people getting speeds up to 750 Kh/sec with other brands and/or other configurations, but in my limited testing that was the best that I was able to achieve without compromising stability.
If you have another brand of 280X, you may want to try 1080 mhz or 1060 mhz for the core clock speed (--gpu-engine 1080), as many people report the best success in that range.
One interesting note is that the 280X seems to be pretty sensitive to heat, and will self-throttle its performance down whenever GPU temperatures get over 70C or so (at least, my Sapphires do). Setting cgminer’s auto-fan temp at 70C prevents this, although I run my rigs in a fairly cold room—some of you may need to point a box fan at your rigs to get the best performance out of your 280Xs.
I’ll be trying to get my hands on other brands of 280X cards, especially the ones by Gigabyte and MSI (although many are reporting good results with the Asus and PowerColor models, too)—although I’m sure most of you have noticed that they’re fairly difficult to come by right now.
If you’re currently running a 280X-based rig, please feel free to leave a comment below with your settings and results!
Posted in 
I have in my house 2 R9 280x TOXIC but i cant mount in in the rig because i dnt have raisers, someone tested the toxic?
i got 3 toxic, they really look mad!
I have currently problems with them, they crashed after 1 hr mining and now I can only run one at a time. When I put 2 or more, it freezes instantly or one card crashes, saying Temp 511°C & Fan speed 0 RPM.
2 of them running fine at the moment, but still need to undervolt them.
I get 750 kh/s with 280x Toxics:
“intensity” : “13”,
“auto-gpu” : true,
“gpu-fan” : “0-55”,
“temp-cutoff” : “85”,
“worksize” : “256”,
“kernel” : “scrypt”,
“lookup-gap” : “2”,
“thread-concurrency” : “8192”,
“expiry” : “120”,
“gpu-dyninterval” : “7”,
“gpu-platform” : “0”,
“gpu-threads” : “2”,
“gpu-engine” : “1080”,
“gpu-memclock” : “1500”,
“gpu-powertune” : “0”,
“log” : “5”,
“no-pool-disable” : true,
“queue” : “0”,
“scan-time” : “30”,
“scrypt” : true,
“failover” : true,
“shares” : “0”,
“kernel-path” : “/usr/local/bin”
Hope this helps
Funny, I’ve been using the exact same settings after much much experimentation over the last week. 750.1KH/s under Win7/64bit.
I can get 2KH/s more with 1081MHz core, but then I seem to get 1-2 HW errors per day, so I set it at 1080 for now. Haven’t found a good setting for high tc/one thread mining.
Very annoying that the scrypt kernel doesn’t respond well to higher core clocks, games run perfect on my card up to 1200MHz.
did you undervolt your toxics? How?
I just flashed the bios with another version and got it to 1.2V, but I think its still locked…
After DAYS of tinkering, I was able to get all 3 of my Sapphire Toxic R9 280X’s undervolted to 1100 mV and getting a stable 745 Kh/s on each card – peaking up to 750 Kh/s. 3 more cards are being added soon! Here are my settings:
Using cgminer v3.7.0 on Windows 7 Pro 64-bit, Intel Celeron, MSI Z77A-GD65 motherboard, 8GB DDR3, 64GB SSD
setx GPU_MAX_ALLOC_PERCENT 100
cgminer –scrypt -o stratum+tcp://URL -u USER -p PASS -I 13 -g 2 -w 256 –thread-concurrency 8192 –shaders 2048 –gpu-engine 1080 –gpu-memclock 1500 –temp-target 70 –auto-fan
How did you manage to undervolt them pdxmicro?
I undervolted them exactly how it is laid out on this website, but for the numbers, I used: 1100. They have been running great and using around 250W per card.
Veit, have you figured out how to undervolt the toxics?
At the moment Im happy that they are running. I installed afterburner, but I dont know if its changing the V, cuz the miner displays an other value.
I got 550Kh/ card and im figuring out if its stable through the night.
The toxic is a sapphire card, no?
if so just use sapphires oc tool, trixx. It let’s you play with gpu/mem clock, over or undervolt and set custom ram speeds
It is not possible to adjust the voltage in neither trixx or afterburner for this card.
I am gladly to share my stats of the 280x from gigabyte i am currently at 1020 mhz and 1500 on the ram and i am geting about 710, didnt messed with the voltage aldow but i am having a temperature of 78 dg with a 6gpu config. Do you think that the teps are ok ??
I hear the gigabytes are voltage locked so you probably wont be able to undervolt them.
78 seems a bit high, but every situation is different.
What are you gpu fan speeds set at?
are you giving cgminer commands to temp and fan regulate?
what is your cooling setup for your rig?
remember voltage doesnt directly affect hash speeds. It is all about mem and gpu clock. Once you find your sweet spot with clocks then you play with voltage to see how low you can get while remaining stable.
you’ll know when the voltage is too low as it will crash.lol.kind of the same way as when you oc the memory too high
Hi,
could you send your config? I’m struggling with my Gigabyte R9 280X at ~630kH/s with this settings:
–scrypt -I 13 –thread-concurrency 8192 –lookup-gap 2 -g 2 –auto-fan -w 256
while my Gigabyte 7950 happily hashes at ~665kH/s..
holy crap… finally I realized the only difference: the GPU freq. Set 1020 from 1000 and it went up to 710…
But the power usage is still relatively high, powertune param does not have any influence on the VDDC (it lowered at my 7950), so to achieve a better wattage I’m afraid I cannot avoid the BIOS reflash…
Thanks alot for the info, very helpful so far.
Q: So for the voltage setting is it only for the sake of running more GPU together and saving power or will it also lower heat?
Evening all
Just got my mb, do I need to update the BIOS to the latest version for best results or leave as is?
Cheers
Brandon,
Any chance you could set out the commands to update software packages. I downloaded BAMT 1.2. But my Sapphire R9 280x Dual-X OCs still only get about 540Khs each. Not sure but I don’t think bee updated everything you did, or at least not using the same packages. I’m rock solid stuck, and know enough about this to know exact versions make a difference but don’t know how to install the version specific updates. Thanks.
Sorry Badger, I may have assumed you were the same guy on the litecoin forums… But the above message is for you.
I had same problem on Asus R9 780x matrix edition, then realised I only had 1 gpu thread, you need to ensure you have 2 gpu threads for this model. Mine hashe at 720 now, and I have not fine tuned them yet.
Thanks! Don’t know how I missed it but now hashisng over 600 as high as 650-75.
hey, try out using thread-concurrency 8191 (with a 1) and use 2 threads. Set your core clock somewhere between 1020 and 1030 and you should be good to go. It finally got my Dual-X up to 710-730khash/s
Thanks Seth,
Your 8191 suggestion is really helpful
Just installed my Gigabyte 280X tonight, and I’m hitting speeds of about 660 kh/s at a temp. of 73 degrees.
Core clock: 1000
Mem clock: 1500
Cooling setup: Terrible. I have a single case fan that is crap, so I’ve just taken the side of the case off to allow more cool air in (I live in a basement apartment and it stays relatively cool).
Hello CryptoBadger and Team.
Just recently competing the build on my machine. The past 3 days been working hard on getting the miner optimized and running smoothly. I noticed that when I installed the R9 280 GPUs that after running the fglrx commands that the GPU were labeled as HD 7900 and not the R9 280X models. So, I believe this is because the drivers are not installed.
At the moment I am setting up a fresh install of Xubuntu 12.04 with the latest beta AMD Catalyst drivers installed, AMD SDK 2.7, and cgminer 3.7.2. The AMD Catalyst drivers for linux call for the installation of the following for optimized performance:
> gimp-help-en
> gimp-help-common
> XFree86-Mesa-libGL
> libstdc++
> libgcc
> XFree86-libs
> fontconfig
> freetype
> zlib
> gcc
Are these already included? Do I need to find each and install it?
This is my first rig build and really would like to get it right. The past few weeks I have learned a lot but still not have a stable rig up and running. Hoping the drivers will help.
Also, is the “sudo apt-get install fglrx-updates fglrx-amdcccle-updates fglrx-updates-dev” command needed if I am installing the updated catalyst drivers?
Should AMD SDK be downloaded separately?
After this build is setup, I am hoping that the overheated GPU and low hash rate mining will be resolved.
Thank you all for your time and any feedback.
Cheers!
R9 280x are re branded hd 7970 so maybe software is detecting them as such?
As stated somewhere else please run (as root)
Prerequisites: Idk the status on 12.04 for I used 13.10, so maybe not all dependencies will be installed.
Try apt-get install *PACKET NAME* for each packet from your list.
After that:
apt-get -y remove fglrx-updates fglrx-amdcccle-updates fglrx-updates-dev
cd ~
wget http://www2.ati.com/drivers/beta/amd-catalyst-13.11-beta-v9.4-linux-x86.x86_64.run.zip
unzip amd-catalyst-13.11-beta-v9.4-linux-x86.x86_64.run.zip
rm -f amd-catalyst-13.11-beta-v9.4-linux-x86.x86_64.run.zip
./amd-catalyst*.run
aticonfig –lsa
PS: As AMD states, the SDK 2.9 is already included in the bundle !
Thank you N8c! Great information. I was pulled away from the build for a few days but have just had time to implement the discussed topics. I can see all three GPUs labeled as R9 200 Series versus HD7900.
On to the next steps. Cheers!
Does anyone know how to unlock the voltage on Gigabyte R9 280x’s? I am still pretty new to mining and haven’t figured out this step.
Gigabytes are voltage locked. I think only ones that were unlocked were the review samples.
so none of the nomral oc tools will allow you to do so. That said im sure u could try flashing other card bios but I have no experience doing so and have no idea if it will help or harm ur card.
have I researched what other ppl have done/tried so far?
This was very useful thanks. I am at least up and running I will try to optimize and get back to the group. I am at A:704.5Kh/s with a Gigabyte GPU… I think… as I am a Nooob.
I found a cheap supply of r9 280x TOXIC. I want to put 6 of them on a rig with a 1600W PSU. Can I undervolt the cards and fit them on? I could possibly underclock the card as well to fit them on.
Any ideas?
Also, I currently have a r9 270x (Sapphire Dual-X) that I am using to experiment before I invest. I can’t seem to change the voltage settings on this card thru Trixx (I also tried afterburner and the ASUS tool). Any help on this as well please?
Thanks in advance.
When you’re done buying how about sharing that location? it’ll be sold out in 5 min 🙂
As for 6 x 1600 you’ll be pushing it, cards alone are over 1600W at full load, could try undervolting but no guarantees, why not do dual 1000-1200W PSUs, probably cheaper also?
Thanks for replying Todd!
I’m in Australia and buying in bulk and getting decent discounts for it, however they don’t have stock till first week of 2014 but will honour the price for me. If any Australians are interested, maybe we can negotiate an even better price by buying together.
Any ideas how I could put the dual PSU together? 1000W + 750W joined with an Add2PSU connector?
Will that be enough power for the 6 of them or should i get two 850W and join them together, might be better to have the same wattage from each or does it not matter? I have heard the Add2PSU connector has issues but I may just get this cheap thing from ebay which does the same thing anyway http://www.ebay.com.au/itm/161172899312
On another note, I don’t want to put the 270X to waste. Can I put a 270X and 5 280X on a 1500W or should I just be safe and put it on a 1600W?
Thanks again =)
I might try and run a 1250W and a 650W thermaltake power express together. Gives me 1900W which should be plenty for a 6 r9 280X or 5 r9 280x +270x.
>6 of them on a rig with a 1600W PSU
Don’t do that. They will overheat. In my limited experience 3 7970s are max that can run in a crate and stay under 80^C. I would expect similar issues with 280Xs. Better two setups with two mobos.
Not even if I spread out six of the 280X with risers and power them with two 1000W PSUs?
So I feel like an idiot that I have to ask this but how do you know what GPU is what? I have 3 Sapphire 280xs and 2 Gigabyte 280xs. 1 of the 5 is hashing at 600k, the rest are all at 705k. I am fairly confused about why that would be and which one of the cards it is 🙂 I’ve tried all the above settings and I can’t find anything that gives me more than the ~705 mark on any of the cards 🙁 Was really hoping to hit 750.
It’s fairly easy:
1) I _suppose_ 0 is the 1st PCI-e slot, 1 the 2nd and so on. (starting on the left / in the middle of your board)
2) Only feed work to all cards minus the one you want to know where it sits. Wait 1 min, check heat emission 😉
Ok thanks I have to figure out how to do them one at a time 🙂 The 1x slots seem to be in a different order than the 16x slots based on what I saw as I added more and more cards.
You can do this with -d 0 for example.
The readme states that it doesn’t work with -d 0,2,3,4 like the other parameters, but that you have to call it via -d 0 -d 2 -d 3 -d 4.
Hf testing 🙂
With 800 khs toxic con 71c of temp, if i change the engine to 1080 and mem to 1500 i get 740 khz with 65c of temp
“intensity” : “13”,
“vectors” : “1”,
“worksize” : “256”,
“kernel” : “scrypt”,
“lookup-gap” : “2”,
“api-port” : “4028”,
“expiry” : “120”,
“gpu-dyninterval” : “7”,
“gpu-platform” : “0”,
“gpu-threads” : “2”,
“gpu-engine” : “1141”,
“gpu-memclock” : “1875”,
“gpu-powertune” : “0”,
“thread-concurrency” : “11136”,
“log” : “5”,
“no-pool-disable” : true,
“queue” : “1”,
“scan-time” : “60”,
“scrypt” : true,
“shares” : “0”,
“kernel-path” : “/usr/local/bin”
I get 790KH/s with this setting, however it draws 70W more (!) at the socket than the 1080/1500 setting I’ve been running at 750KH/s. One card only.
I have 2 rigs each with 3 x r9 280x MSI, got problems beacouse I cant run 3 of them, It like 2 are about 700k/s and one is 20k/s. I was able to run them like 2-3 times, but had to restart rig, and still have the same issue. I use 1150 Wat power supply. Any idea?
R9 280x TOXIC, with the config 800kh/s and 69c of temp>
“intensity” : “13”,
“vectors” : “1”,
“worksize” : “256”,
“kernel” : “scrypt”,
“lookup-gap” : “2”,
“api-port” : “4028”,
“expiry” : “120”,
“gpu-dyninterval” : “7”,
“gpu-platform” : “0”,
“gpu-threads” : “2”,
“gpu-engine” : “1141”,
“gpu-memclock” : “1875”,
“gpu-powertune” : “0”,
“thread-concurrency” : “11136”,
“log” : “5”,
“no-pool-disable” : true,
“queue” : “1”,
“scan-time” : “60”,
“scrypt” : true,
“shares” : “0”,
“kernel-path” : “/usr/local/bin”
With that one 740kh/s with 65c of temp
“intensity” : “13”,
“vectors” : “1”,
“worksize” : “256”,
“kernel” : “scrypt”,
“lookup-gap” : “2”,
“api-port” : “4028”,
“expiry” : “120”,
“gpu-dyninterval” : “7”,
“gpu-platform” : “0”,
“gpu-threads” : “2”,
“gpu-engine” : “1080”,
“gpu-memclock” : “1500”,
“gpu-powertune” : “0”,
“thread-concurrency” : “11136”,
“log” : “5”,
“no-pool-disable” : true,
“queue” : “1”,
“scan-time” : “60”,
“scrypt” : true,
“shares” : “0”,
“kernel-path” : “/usr/local/bin”
A little advice, R9 280x by any manufacturer must go above 730 kh/s..
If you can’t reach at least 730 kh/s stable, than settings are not ok.
My 4x GPU rig:
– 2x Asus R9 280x Top Edition
– 1x Sapphire R9 280x
– 1x Sapphire Dual-X
Here is how it looks: http://img4.imageshack.us/img4/5359/zt6t.png
Here are settings: –scrypt -I 13 -g 2 -w 256 –thread-concurrency 8192 –gpu-engine 1093 –gpu-memclock 1500 –temp-target 73 –auto-fan
The thing i heard is that 280x can go 800 kh/s stable, and this what i am working on and would love to see :))
I posted before one config to get 800kh/s with my toxics, i dont know with other models and i CANT see the VRM GPU temps, but general temp are in 69º stable
on your image, can you explain the kh/s readings?
there are 2 columns. what do they represent?
why is one number larger than the other?
is there a correlation between the distance?
i have a larger gap between my readings, and the column on the left is generally larger than the reading on the right.
there is a supposed setting for 800kh here https://litecointalk.org/index.php?topic=10161.0
thx!
I was running 2 R9 280 x Vapor models perfectly on windows. Then we added 2 more and thats when the problems started we had to change to a 990XFA-GD65 mother board. Then it took time to find slots that powered Riser would detect the cards in only 2 of 4 of the Pci 1x slots detected the cards. Second issue is we moved over to Bamt 1.2. Running the cards at 70-73 degrees for more then an hour at 750khs a card perfectly when the miner suddenly freezes it stays on the hashing power drops to next to nothing which after 2 days is driving me nuts is anyone else having issue with Bamt 1.2 or running 4 sapphire vapor x 280s with out problem what OS system are you using.
Im starting to think its power related Currently running 2 1000w seasonic gold plus with the dual psu adapter could this be the issue.
or is it Bamt 1.2 i have it loaded on HDD drive. Does the program not pay well with sapphire 280x vapors?
my ‘thread-concurrency” : “11136 im going to try 8192 tomorrow see if that helps
any help would be excellent
its always 3 cards running well one just stops then the whole thing stops
Because i CANT see the VRM temp in linux (i use BAMT 1.2 live usb with cgminer 3.7.2) i prefer use that config with 760kh/s and 68º of temp:
“intensity” : “13″,
“vectors” : “1″,
“worksize” : “256″,
“kernel” : “scrypt”,
“lookup-gap” : “2″,
“api-port” : “4028″,
“expiry” : “120″,
“gpu-dyninterval” : “7″,
“gpu-platform” : “0″,
“gpu-threads” : “2″,
“gpu-engine” : “1080″,
“gpu-memclock” : “1500″,
“gpu-powertune” : “0″,
“thread-concurrency” : “11136″,
“log” : “5″,
“no-pool-disable” : true,
“queue” : “1″,
“scan-time” : “60″,
“scrypt” : true,
“shares” : “0″,
“kernel-path” : “/usr/local/bin”
I prefer use that, 800kh7s is so good and only get 2º more, but without see VRM…..
Anyone know any program to see VRM?
Have a bunch of 280x toxics that I’ll setup over the next few days, I’ll let you know. Running alot of HIS 280x turbo boost GPUs at 735-740kh/s at about 71-74C and their VRM readings range from 75-85C (VRM1/VRM2 are always within 5C of each other for HIS cards anyways).
I’ve noticed that the 280x start throttling when VRM temps reach around 88-90C, and these rise quickly if not kept in check.. so I’d suggest to people to keep an eye on these reading more so than the GPU chip temps insofar as stablity and long term performance are concerned. For the HIS cards (which are disturbinly silent lol) there’s about a 10C difference in temps between the GPU chip and it’s VRMs.. so a GPU that hits 80C… yeah not good because it means the VRMs will top 90c and either the card will stop reponding or it will throttle until the temps fall. Either case it’s not good for longterm component life and mining performance.
I have 2 HIS 280x that I can’t get above 685 on Linux. Could you care to share your settings and OS drivers etc?
I’m running 4x HIS Turbo boost cards per system, with the following settings, if you have the 280x Boost (but not turbo.. then the clock is at 850mhz.. the turbos come in at 1050mhz from the factory). Slowly bringing up the clock and mem speeds to 1060-1064mhz clock, 1500mhz mem is the range I got the best results.. so far. Each card is tweaked a bit differently as no two card are identical (hence the variance in the clock speeds). That being said, here are the settings for cgminer:
Intensity 13, lookup gap 2, worksize 256,thread-concurrency 11200 (the txt files for cgminer say 4x#ofshaders for optimal hence with you’ll se alot of people with setting at 8192.. but find that 280x can get pushed a bit more but for 2 threads more than that and kh/s drop. havent tried running off 1 thread, as I’ve seen no concret evidence it performs any better. Insofar as undervolting of powertune.. I’m not a huge fan of it, not enough head room on a 1500w psu to jam a 5th gpu so meh lol though did have to juice the gpu that power the screen as it has a tendency to be finicky (also that card runs slower 1050mhz any faster like the others are its performance drops. If you have a headless rig then that may not apply to you). Hope thhis helps. As I keep tweaking things I’ll add to this.
drivers: The latest AMD drivers and SDK, though its a windows 8.1 cluster.. attempted originally with ubuntu but ran out of patience. Love me some linux, but with windows was fully setup and mining in less than 2 hours, just a personal choice on my part (costvstime analysis type scenario) and in no way indicative of the end results you should be able to get. Hope it helps.
Erebus,
Thanks for the comprehensive reply. I just check my GPU card box and I do not have the turbo version so cards are 850 as you say. I will try do adjust the thread concurrency higher than I have now(8192). I guess without the turbo version of this card I will probably be stuck somewhere in the range(685kh/s) that I am now.
No, the only diff with the boost and turbo boost variant is that the turbo comes factory overclocked (oh and a tiny “turbo” sticker on the card lol) Same card.. You should be able to crank it up to the stated 1060mhz range no prob. Finding that any higher though and vrm temps get dangerously high unless u have AMPLE air flow.. (Build quality not as great as I would like, was expecting more grom HiS) As in box fans and spreadout gpus (say 3 inches using extra long pci-e adapters in an open air rig). If u try this in a regular comp box.. Invest in a fire extinguisher lol, kinding.. They’ll likely just throttle themselve to save from melting and be pretty useless.
Worst case if u cant get to the upper range (due to temps or psu issues etc) Try like 1120-1230 clock and 1450ish mem. Played around with those ranges and got nice temps with khs in the 710-715 range in early experiements a while back. If you can push the cards to higher clock speeds then go for 1500 mem. Thinking the cards should be able to get overclocked to 1150 clock ranges like my toxics but unlike the sapphires their cooling solutions (and vrm designs dont seem to be on the same level). Let me know were you end up. Cheers.
Hmmm, interesting about the differences. I can’t seem to get the engine clock up above 980 on my HISs. Anything above that and the cards just take a nose dive. Played around with different thread-concurrencys but 8192 still seems to be the best. 1050 on my MSI is giving me a steady 735 happy with that. Just gotta keep messing with it I guess.
That is interesting. Perhaps the type of ram used in the non turbo card is slower idk. I’ve heard of some vendors putting cheaper ram in the basic gpus. Tell me, hash rates will fluctuate by alot depending on where the clock/mem speeds are at. There are several “sweet spots”. Typically maintain a .68 to .71 ratio clock to mem speeds. Is the card still stable at 980mhz? Are u bringing up the mem speeds accordingly? From memory the cards sucked a bag of nails depending on where you were at with the clock to mem ratio. If u are going to run it at 980mhz then lower your mem speed correspondingly to 1450 levels. Or again push through and see if indeed when u get to 1060/1500 u get in the low 700s.
Erebus,
My two HIS 280xs are very stable at 980 and 1500. I brought the mem down to 1450 and the kh/s dropped to an unstable ~560. I slowly brought up the mem in 5 kh increments and the whole rig crashed at 1470. I tried leaving the mem at 1500 and increasing the eng from 980 but the kh/s just start dropping.
Just to clarify when u day “they nose dive” are you talking about them crashing/stalling etc or just having generally lower hash rates? If its the crashing then ya.. Prob cheaper parts used to make the cards.. So bring up the clock speeds in the 900s and lower the mem speeds until u find a sweet spot based on the ratios i explained. If its just the hash rate that drops then keep augmenting them and you will see the hash rate come back up dramatically once your in the low 1000s and beyond. Hopefuly for you its the later issue. If not do what u can insofar as get the best stable clock speed and then it becomes a question of playing with the mem speed to get the most out of each card. Last idea.. If they hang or crash.. Maybe juice them up more, increase the powertune (not sure but your cards may be voltage locked as opposed to the turbos with are not, so you might be limited when it comes to what u can squeeze out of them unfortunately) Good luck bud.
Erebus,
Thanks for all of your suggestions. To clarify my nose-dive comments: the KH/s go from the steady 680 to an unstable 580-610 or so. I am running Xubuntu 13.10 so I decided to install the latest amd beta drivers like some others have used but I lost about 10% total kh/s and was not able to tune any better than with the ubuntu fglrx package. Ohh well. Back to the fglrx driver pkg I guess.
Very interesting, from what I’ve read and sourced out (short of contacting HiS directly) is that the boost and turbo cards all supposedly only differentiated by the later being factory overclocked. Its starting to look as though the actual hardware/OS setup of your rig might be the difference maker. In a previous post I stated that I gave up on ubuntu and that was basically the reason why I did so. I’m no linux genious and am not comfortable messing with kernels and custom programming setting etc. in theory a linux should get the same or similar results as windows as if evident by the tesults of other people, but their knowledge of linux and patience far exeeds mine hence why i went with windows. Look around for which linux drivers are more stable for your setup.. There are custom linux drivers out there that are optimised for various linux distros, check up on the various cryptcurrency boards (bitcoin,litecoin etc) The latest ones may not be a good idea as your experimentation is suggesting.
Insofar a info the rest of my setup include Ssd drive, gygatybe 990fxa-ud6 or ud3 boards, 8gb ram per board, sempron chipset (head node has a fx6300 dialed down), enermax maxrevos 1500w for each node. Modular open rig. Each combination of components you have in your rig does play a role in the final outcome, some less than other obviously. This is why whatever people put out there insofar as what is in their cgminer conf or bat file should be guidelines/starting points for each new setup.
-last thing i can think of, always delete the bin file prior to running cgminer (i run 3.7.2) i setup everything in my bat file (so dont use a conf).
Erebus,
Again many thanks for the informative post. I have been using Linux exclusively for 3 year now and I will admit getting this rig going has been the closest I have come to going back to Windows! I decided to leave the The AMD Catalyst™ 13.11 Beta V9.4 driver over-night to see where it was at in the morning. The MSI is doing the same as when I left it however the HIS cards are both up about 30kh/s. I wonder if it is normal for the cards to settle-in to a new config over that period of time? Temperature stability perhaps? My start-up script deletes the .bin file at start-up so I do have that covered.
Just looking at my rig and considering what you said about hardware, I see the the HIS cards are the ones that are directly attached to the Mobo and my MSI is on a riser(more risers on the way). Perhaps that has something to do with it(cooling). It’s minus -17 Celsius right now at my place. I’ll just put the thing outside! 🙂
If your HiS cards are at 30khs its that at some point durin gthe night thr drivers failed and the cards shut off. Unless the system is rebooted they’ll stay there even if u use the disable and restart cammands in cgminer. It can be frustrating indeed. For mining rigs risers should be mandatory the gpu temps might be ok but I’ve seen the VRM temps reach in the 90c and cause throtling or worse faliures. I chalk that up to less than solid construction on the part of HiS unfortunatelly. So these cards (like the MSIs) need a TON of room and cooling. Mine are on risers, hang vertically to maximise heat dissipation and are spread apart a minimum 3inches. The toxic though I can overclock like crazy and keep them in a thighter confirmation (if I choose) and they remain at 71-72 without any external cooling (fans etc). I feel you for the OS, the way I see it it just depends what you wanna get out of it.. Is cryptomining for you a hobby or a business.. If its a hobby then yeah with linux youll get to play in various file inards, learn a ton, and become more involved in that community. But for a business view of things.. Linux distros, special driver tweaks are dependant on individuals in the community.. Were as for windows you pay a corp and expect the best right away type deal. its a personal decision of course and I had to make a hard decision early on, so for me the linux guys and gals are my heroes lol
Your cards have a load temp of 68C, so wouldnt worry about the VRM temps. Thx for the settings you’ve posted, will speed up my tunning and tweaking. I use HWinfo64 but it’s a windows app, sorry ot familiar with programs for linux systems.
Hey Profion, tested the Toxics last night, here are my preliminary findings:
Vrm temps avg about 10c over gpu thermal diode readings .. So if u at 70c the vrm should be at 80-81c so thats good.
Thx for your setting buy try this: therad concurrency 11200, gpu clock 1175, mem 1675. Will bring u up to 819-820kh/s lazer beam solid 🙂 your welcome.
Sorry.. Mem speed at 1875.
Only downside is that its a bit more wattage than your settings, not much though. Seen this card get pushed to clocks of 1200 so will try that tonight with some overvolting.. Thinking 830 is possible and that puts it squarely in 290 gpu territory. Otheriwse amazed at the build quality for the toxic. Its the only cards im buying moving forward. If newwegg gets cleaned out in less than a day when they have stock well its prob because of me lol
setx GPU_MAX_ALLOC_PERCENT 100
setx GPU_USE_SYNC_OBJECTS 0
cgminer.exe –scrypt -u WORKER -p PASSWORD -o stratum+tcp://stratum.wemineftc.com:4444 -I 13 -w 256 –shaders 2048 –thread-concurrency 8192,8192 –gpu-powertune -5,-5 -g 2 –vectors 1 –gpu-threads 2 –gpu-engine 1020,1020
this is the batch file i created that bypasses every other .conf file and i have two gigabyte r9 290x set at 1020 for the clock
i get a stable (as in it hasn’t moved in 9 days, stable) 718.8 per card, running at 75c and 68c
I have a board that can support 3x crossfire so i decided to try a sapphire r9 290x BF4 which is supposed to get anywhere between 880-1000 so i will report back with those results
Thanks for posting your config file Anon, I’m about to build a rig with 4 Radeon r9 290x’s – should I use your same config file for that?
As a baseline yeah but youll be able and should increase the thread concurency and frop the “shaders” line its not relevant (if u add the tC line u dont need the shaders line and vice versa). U have more shaders on a 290x than 280x, scale TC accordingly. Hope u have a solid psu (minimum 1500w) for four 290x. Otherwise intensity might need to be altered and naturally clock and meme speeds is something that you’ll have to play with to get the most out of your gpus. Hope this helps. Let us know how u fare. Ineterestrd to see how much wattage four 290x will run at.
Yesterday I finally got the 1200W Gold PSU & RAM, which without I couldn’t build this thing. Last night I got Win 7 installed, AMD drivers for my R9 290x card, and installed CGminer 3.7.2 using the following conf file:
{
“pools” : [
{
“url” : “stratum+tcp://url-for-wemineltc”,
“user” : “usernameofworker.1”,
“pass” : “password”
}
],
“intensity” : “20”,
“vectors” : “1”,
“worksize” : “256”,
“lookup-gap” : “2”,
“thread-concurrency” : “32765”,
“gpu-engine” : “947”,
“auto-fan” : true,
“gpu-memclock” : “1425”,
“gpu-powertune” : “20”,
“temp-cutoff” : “96”,
“temp-overheat” : “95”,
“temp-target” : “93”,
“api-port” : “4028”,
“expiry” : “120”,
“failover-only” : true,
“gpu-threads” : “1”,
“log” : “5”,
“queue” : “1”,
“scan-time” : “60”,
“temp-hysteresis” : “3”,
“scrypt” : true,
“kernel” : “scrypt”,
“kernel-path” : “/usr/local/bin”
}
Using this conf file, I achieved an average of 850 – 1000 khash for 9 hours (since I let it run last night). At times in my WeMineLTC.com dashboard i’d see an average hash rate of over 1,100 which I thought was crazy, but wasn’t going to argue it 🙂
I do have one question, and I’ve noticed nobody has asked this on here, or any other forum, which I find quite odd considering it’s the only hiccup I’ve run into. How the heck do you run more than one r9 290x card at the same time?
I plugged a second one into the motherboard, and powered it up. And what would happen is Windows would be back to the large icon / no GPU screen and asked to install drivers, which I already installed. So instead I would reboot, and then the screen would be back to normal. I would run CGminer and it would turn into a black command screen without anything running, then the computer would crash and reboot. I tried plugging the 2nd card into other slots, but the same thing would happen.
Sometimes the screen would flicker with a lot of lines across the screen like it was glitching out. Do I need to use dummy plugs? Most people say no with Win 7 and all this new hardware.
I’ve never used more than 1 video card on a PC, so I’m new to this, but I did everything right since I was able to achieve success with one card over night 🙂
I just want to be able to use 2 cards to see how that goes. I ended up buying 4 cards total, crazy….cause that was so much money, and I would love to make my money back making LTC. So any help would be great on how to hook up 2 to 4 cards to 1 motherboard.
And Erebus, since my PSU is only 1200W I guess I should only run 3 290x cards and return one of them back to amazon?
“thread-concurrency” : “32765″,
that looks, really, really high.
how did you come to be using that number?
The thread concurrency is from a config file that I found on another forum if you scroll down:
https://forum.give-me-coins.com/index.php?p=/discussion/381/cgminer-hardware-errors-r9-290-sapphire-battlefield-4-edition/p1
That’s the only config file that works for me, and I’m at a constant 900 – 1000 or more khash rate.
My only issue is not being able to plug in more than one card. I bought four R9 290x’s but am only using one sadly 🙁
I’ve tried everything. Going to take my rig to a local computer shop where the owner says he can hook it all up with Crossfire or something.
so i tried windows wont install right this system cant turn off with the dual psu addapter cant say i like this dual psu system anyone have a single psu that works for 5-6 cards
This monster is powering my 5 280xs no sweat:
http://www.amazon.com/dp/B007W5TV6K/ref=as_li_ss_til?tag=onsimobicomps-20&camp=0&creative=0&linkCode=as4&creativeASIN=B007W5TV6K&adid=0TSSEEV0RC81HPVBKQD4
Hey Paul, did u have to undervolt ir underclock them to acheive stability? 4 280x have me tapped at about 1400w but i have the enermax maxrevos psus (1500w top end). What’s total wattage are u using with 5 cards?
Great set of articles, really got me going.
I originally build myself a rig with 2 HD7950’s. I added a R9 280x now. However due to the fact that the HD7950’s prefer gpu-thread to be 1 and 280x likes 2 I have a problem setting up nice config files. I now ended up using 2 instances of cgminer with command line parameters to get this done. Should you run into this too, I’d like to know if there is an elegant solution as I don’t like editing long command line scripts.
Keep up the good work!
Had the same issue and looked into it. The cgminer guy didn’t believe gpu-threads needed to be a local variable and so it’s global. Guess he didn’t try and mix cards that need it. I have 280x’s (needs 2) and 1 5870 (needs 1). Only option I found was to use the -d/device options to separate them and run two instances. This doesn’t really work with cryptobadger’s screen setup and I kinda gave up. Would probably work if you replicated the cgm in bash.rc and had a cgm1 and cgm2 and ran the two instances in separate screens but I settled for just buying another 280x :). That won’t help you as all your cards are pretty useful. So, no elegant solution that I know of i’m afraid mate.
I found a workaround which isn’t as good as I wanted too but it will do for now. It is ugly in monitoring too, but ok, I need the hashes for the money these cards cost. Also I can’t 100% figure out how cgminer works as I have the impression that below setup gets mixed between the cards, i.e. the HD7950’s get the memory value of the 280x for example.
I setup 2 workers in the pools I mine for. I also launch 2 cgminer processes now, not using a jason stile config, but in command line format. This because I wasn able to address the 2 HD7950’s in 1 config file. The device option didn’t allow for 2 GPU’s.
So I have a script that calls 2 other scripts to start the different cards into screen sessions like below. First line uses “-d 0 -d 1” to call the first two cards being the HD7950’s using worker1. The 2nd line uses -d 2 to call the R9 280X using worker2:
cgminer.sh
find *.bin -delete
sleep 2
# Start HD7950s
screen -dmS cgm0 ./cgminer.hd7950.sh
sleep 2
# Start R9 280X
screen -dmS cgm1 ./cgminer.r9-280x.sh
cgminer.hd7950.sh
export GPU_USE_SYNC_OBJECTS=1
export GPU_MAX_ALLOC_PERCENT=100
export DISPLAY=:0
#
# Start HD7950s
screen -dmS cgm0 ./cgminer –scrypt -o -u worker1 -p xyz –failover-only -o -u worker1 -p 456 -d 0 -d 1 -I 19 -w 256 –thread-concurrency 24000 –gpu-powertune 20 -v 1 –gpu-engine 1110 –gpu-memclock 1575 –auto-fan –auto-gpu –no-submit-stale –scan-time 30 –shaders 1792 –gpu-fan 30-85 –temp-cutoff 90 –temp-overheat 85 –temp-target 80 –api-port 4028 –api-listen –api-network
cgminer.r9-280x.sh
export GPU_USE_SYNC_OBJECTS=1
export GPU_MAX_ALLOC_PERCENT=100
export DISPLAY=:0
#
#Start 280x
screen -dmS cgm1 ./cgminer –scrypt -o -u worker2 -p xyz –failover-only -o -u worker2 -p 456 -d 2 -I 13 -g 2 -w 256 –thread-concurrency 8192 –gpu-powertune 5 -v 1 –gpu-engine 1026 –gpu-memclock 1498 –auto-fan –auto-gpu –no-submit-stale –scan-time 30 –shaders -2048 –gpu-fan 30-85 –temp-cutoff 90 –temp-overheat 85 –temp-target 80 –api-port 4029 –api-listen –api-network
This way I get 610+ Khs for the 2 XFX 7950’s each and 700+ Khs for the XFX R9 280X, for a total of around 1900+ Khs in total. I’m sure there is some hashes to be gained by a litle tweaking from here onward, but ok enough.
——————————————————————————–
[P]ool management [G]PU management [S]ettings [D]isplay options [Q]uit
GPU 0: 79.0C 3322RPM | 626.3K/618.4Kh/s | A:107 R:3 HW:0 U: 4.82/m I:19
GPU 1: 78.0C 70% | 627.5K/618.8Kh/s | A:119 R:4 HW:0 U: 5.36/m I:19
GPU 2: 77.0C 1798RPM | OFF / 0.000h/s | A: 0 R:0 HW:0 U: 0.00/m I:19
——————————————————————————–
——————————————————————————–
[P]ool management [G]PU management [S]ettings [D]isplay options [Q]uit
GPU 0: 79.0C 3324RPM | OFF / 81.00h/s | A: 0 R:0 HW:0 U: 0.00/m I:13
GPU 1: 78.0C 70% | OFF / 0.000h/s | A: 0 R:0 HW:0 U: 0.00/m I:13
GPU 2: 77.0C 1797RPM | 730.4K/710.5Kh/s | A:364 R:1 HW:0 U: 16.67/m I:13
——————————————————————————–
i recently read a thread on the litecoin or bitcoin forums describing running separate instance of cgminer, for different cards on the same rig.
i couldnt come up with a good reason to do that, until now.
i think the r9 series really wants 2 threads, regardless of the readme instructions.
I contemplated going the sell and buy similar cards route until I found the “–remove-disabled” option. Added that to the cgminer startup lines as posted above and now running multi-card setup at max rates I was getting for the cards seperately. Below currently pull just below 1000w from the wall.
For XFX HD7950’s (2x at 650+Khs / 78c / fan hd7950s.log
For XFX R9 280x (1x at 710+Khs / 78c / fan r9.280x.log
Let’s try that again as post got mangled.
I contemplated going the sell and buy similar cards route until I found the “–remove-disabled” option. Added that to the cgminer startup lines as posted above and now running multi-card setup at max rates I was getting for the cards seperately. Below currently pull just below 1000w from the wall.
* For XFX HD7950’s, 2x at 650+Khs 78c fan <80%
./cgminer –scrypt -o -u worker1 -p xyz –failover-only -o -u worker1 -p 456 -d 0 -d 1 –remove-disabled -I 19 -w 256 –thread-concurrency 24000 –gpu-powertune 20 -v 1 –gpu-engine 1110 –gpu-memclock 1575 –gpu-fan 30-90 –auto-fan –auto-gpu –no-submit-stale –scan-time 30 –temp-cutoff 90 –temp-overheat 85 –temp-target 80 –api-port 4028 –api-listen –api-network 2
* For XFX R9 280x, 1x at 710+Khs 78c fan <50%
./cgminer –scrypt -o -u worker2 -p xyz –failover-only -o -u worker2 -p 456 -d 2 –remove-disabled -I 13 -g 2 -w 256 –thread-concurrency 8192 –gpu-powertune 5 -v 1 –gpu-engine 1050 –gpu-memclock 1498 –gpu-fan 30-90 –auto-fan –auto-gpu –no-submit-stale –scan-time 30 –temp-cutoff 90 –temp-overheat 85 –temp-target 80 –api-port 4029 –api-listen –api-network 2
Heterogenous setups are a pain. Advice.. Sell the 7950.. People still want them, youll get good value for them, heck since they are impossible
To find you might be able to sell at premium. Then just buy 280x to fill iuy your rig. Problem solved.
I have 3 Sapphire R9 280xs and 2 Gigabyte R9 280xs and now have it all working well 🙂 I do not have them undervolted yet, that’s the next step but they are all hashing at 740+ on litecoinpool with these settings. Just posting here in case someone else wants to try. Next step is futzing with the individual card settings! Some of these I don’t understand super well so I want to see what happens. I do have 1 card running hot at 79, I think the CPU is exhausting on it which isn’t helping. The others are all steady at 69-70. I do have a box fan pointed at the crate. If anyone has any suggested tweaks let me know!
“intensity” : “13,13,13,13,13”,
“vectors” : “1,1,1,1,1”,
“worksize” : “256,256,256,256,256”,
“kernel” : “scrypt,scrypt,scrypt,scrypt,scrypt”,
“lookup-gap” : “0,0,0,0,0”,
“thread-concurrency” : “8191,8191,8191,8191,8191”,
“shaders” : “2048,2048,2048,2048,2048”,
“gpu-engine” : “1080,1080,1080,1080,1080”,
“gpu-fan” : “0-85,0-85,0-85,0-85,0-85”,
“gpu-memclock” : “1500,1500,1500,1500,1500”,
“gpu-memdiff” : “0,0,0,0,0”,
“gpu-powertune” : “0,0,0,0,0”,
“gpu-vddc” : “0.000,0.000,0.000,0.000,0.00”,
“temp-cutoff” : “90,90,90,90,90”,
“temp-overheat” : “85,85,85,85,85”,
“temp-target” : “72,72,72,72,72”,
“api-port” : “4028”,
“expiry” : “120”,
“failover-only” : true,
“gpu-dyninterval” : “7”,
“gpu-platform” : “0”,
“gpu-threads” : “2”,
“hotplug” : “5”,
“log” : “5”,
“no-pool-disable” : true,
“queue” : “1”,
“scan-time” : “30”,
“scrypt” : true,
“temp-hysteresis” : “3”,
“shares” : “0”,
“kernel-path” : “/usr/local/bin”
keep a fire extinguisher close.
and use a surge protector.
pics?
Hah now you have me nervous!
well, fire is no joke, and this club is pushing the limits of hardware, wiring, and a few misc housing codes.
best to know how may watts you pull at the wall, what that circuit breaker is rated for, and the wire that carries it.
I ran a dedicated 20amp line just for this. Not sure what else I could do. I’m also going to install a smoke detector right over it. Any other safety tips?
sounds like you did the right thing running a dedicated line. as long as your wire is rated, kind of hard to buy something new that isnt, and your breaker is good. i would wonder about your outlet. when i bought they had 15 and 20amp outlets. i went with the 20.
check the amp draw on the label of your psu. mine is 15amp. i didnt want to be using a 15amp breaker, or outlet. but thats just me.
Yep, 20 amp outlet although my PSU is only a 15amp. And 20amp rated wire absolutely. Thankfully it’s in the basement very near my panel so it was an easy job.
I could do pics I guess if I had a place to post them. My setup is sort of ghetto 🙂
Get a clip on fan for example from wallmart to target the 1 gpu thats 79c.. Im assuming its VRM is prob pushing 85-90c or more so get than buttoned up if u want it to have a long life and prevent possible throttling. Otherwise nice job, i havent been able to run more than 4 280x on the 1500w enermax maxrevos.
Hello ! Can you upload the modified kernels you mentioned ? The original thread is closed, and the links were removed. I am interested in it. Thanks !
And please routinely check your PSU cord, plug strips, outlets AND yoru breakers. If anything is getting too warm, it is because it is drawing too much. Reduce the load or move things to a different circuit.
A standard 20 amp household circuit is only rated to be used at 80% of its capacity continuously. That is 16 amps and is 1920 watts at 120 volts. Please do not go over this.
Also most outlets are 15 amps. The assumption is that no single outlet is drawing all the load and is spread around.
Stay safe!
ive seen some comments about apartments, and older houses.
its a worst case scenario, but its easy to assume all the power you want is right there in that little box. its not. there are limits.
they never intended for houses to contain multiple mini-power plants. the basement setup form paul is the best one ive heard yet. new wire, breakers, and boxes, all able to handle the max load, in the basement, preferably unfinished with no drywall or carpet.
clean(relatively) and cool.