Bitcoin Forum

Hey guys, can anybody who has a R9 290X Tri-X post their hashrate? I'm in the market for new hardware and I'm trying to see whether I should buy 290Xs or 290s. Thanks.

around 800

You need to learn how to tune...

I have 4x Sapphire R9 290 (NON-X) Tri-X OC cards running at 945KH/s and 900KH/s (for a card that won't take the memory OC).

Rig Specs:
Gigabyte Z87X-UD3H
4GB DDR3 RAM (cheap stuff)
Cheapest LGA1155 CPU Available
4x Sapphire R9 290 Tri-X OC

SGMiner Config:

Key points: if your card won't overclock to 1499,1500, then downclock it to 1250 to see some performance gain.

Every tri-x OC card should be able to do 900KH/s+ straight out of the box, with no burn in.

My Gigabyte R9 290X is getting 863Kh/s with WU:780/m and 79C

setx GPU_MAX_ALLOC_PERCENT 100
cgminer --scrypt -o stratum+tcp://scrypt.megamultipool.com:80 -u user.user -p password --intensity 20 -g 1 -w 256 --lookup-gap 2 --thread-concurrency 24550 --gpu-engine 920 --auto-fan --gpu-memclock 1500 --gpu-powertune 20 --temp-target 80 --temp-cutoff 95 --temp-overheat 92

Why do you have GPU engine so low?  And stop using intensity - switch to xintensity and you will see several percent hashrate increase.

I've tried about a dozen other higher GPU engine values higher and they all resulted in a lower hash by about 100-150Kh/s lower.
What GPU engine value do you recommend I try? I've never heard of xintensity, I will definitely look into it.  As a side note. I've never noticed any difference at all in performance with the following included:
setx GPU_MAX_ALLOC_PERCENT 100
setx GPU_USE_SYNC_OBJECTS 1
Am I doing something wrong with the above code?

You only need to run the setx command _once_ on a windows machine, and the variable will be fixed forever.  If you don't do GPU_MAX_ALLOC_PERCENT, cgminer simply won't work with higher threadconcurrencies.

GPU engine depends massively on BIOS.  Some 290 (and X) BIOSes will drop in performance past 1000, others drop past 1020, and still others keep climbing all the way up to and beyond 1080.  Typically 1000 seems to work on everything tho, if it doesn't, try another BIOS (e.g. a different manufacture like Sapphire).

My favorite BIOSes on the 290 at the moment are:
Powercolor AXR9 290 4GBD5-MDH/O  015.042.000.000.000000  (does 930KH/s at 1499/1040 mem/engine)
Sapphire R9 290 4096 MB 015.041.000.000.000000 (does 899KH/s at 1499/1000 mem/engine - DO NOT USE THE 015.041.000.001.003747 BIOS, you will lose 20-30KH/s at the same clocks)
and
Sapphire R9 290 4096 MB Tri-X OC 015.042.000.000.000000 (does 961KH/s at 1499/1080 and 954KH/s at 1499/1070 mem/engine)

You will have to experiment a bit to find the best BIOS for your 290X.

I've now got it set up with a .conf file and running kalroth's 3.7.2 cgminer and can't get as high of a hash as I was getting before without the xintensity.  I think my card has 2816 shaders so the xintensity would be 372?

Try bumping it up a good chunk, say double.  Personally I didn't have much luck with kalroth's miner, even though it has the same xintensity code as sgminer.  372 makes sense by my math.  I use 1100 on my 290 (non-X).

Also set your tc to a multiple of 2816, and then add or subtract 1 to see which gives more efficiency (usually adding 1 will give you a bump).  Say, 28161 or something.  xintensity means you don't have to push TC so high to get the same performance, because you are aligning the number of threads launched with the number of threads you can actually run at one time.

It's better to check Watt rate and temperature also.

A have 2 of 290(not X) Tri-X and get:

863 KH/s each with 632 Watt and temperature 77`C (965KHz Core, 1250 Ram)
905 KH/s each with 677 Watt and temperature 88`C (1000KHz Core, 1500 Ram)

It's 4.8% rate gain and 7.1% Watt gain. Also I don't like 88`C and possible card crash.

So it mine already 6 days 24/7 without problems. I use sgminer 4.0.0  xI:400

PS: I use TC: 20481 don't know where I found it, but it give me better rate, than higher values
PS: xI don't make much difference. I get almost same rate from 200 to 600

Why?  Power consumption is almost totally irrelevant at current scrypt profitability.  You could use double the power for 10% more hashrate and probably come out ahead right now.

This said, my 945KH/s+ cards run at a cool 65 degrees C max (lowest is 57 degrees).


Would be shaders * 8 + 1, which makes sense.  I find shaders * 10 + 1 to be solid on these cards, but this also won't make a huge difference.


If by "almost the same rate" you mean several percent, sure.  But yes, the key is really that using xI over I makes a decent difference.

Looks like you don't use standard Sapphire Tri-X card. It's quite hot. 74'C is minimum for me :(

Of course I use standard Sapphire Tri-X cards...  Sapphire R9 290 Tri-X OC, no mods.  I just have plenty of experience running rigs.  

I have banks of reference Sapphire R9 290s running at <60 degrees too:
https://i.imgur.com/XBZjAej.png

Stock unmodified reference cards.  899KH/s stable as a rock too.

whats your vcore's for these cards?
got 5 pieces.

one is going dead on stock (gpu0) trying on lower gpu speed (now 900 , stock 1000 ) at 1000 is going dead after few seconds - probably I dl RMA.
Mem clock lowered 1300->1250


cooling is same for all - but temps are weird high.
Seems like they had been assembled in rush!

http://i61.tinypic.com/wj80h3.jpg

Stock.  I never use windows, so can't touch the voltages on Hawaii chips yet.

Not exactly uncommon - you could RMA it, but is it worth a 10% hashrate boost to lose time waiting for a replacement?  Have you tried 947 engine?

You can re-thermal paste the cards if you don't mind losing your warranty, and it might help even things out while shedding a few degrees.  I have also observed some cards that read at higher temperatures but don't actually seem to be (e.g. the measurement seems skewed).  As long as they are within range, I wouldn't be too concerned.

I do agree they seem like they were assembled in a bit of a rush.  But not as bad as the new MSI 290 cards... those are bloody weird, but good hashers anyway.

running 2x 290X.
Gets between 1170-1190k/ each
temp average between 73-75

You set 100% fan. And ambient temperature 10`C? :)
I have tried 100% fan its quite loud and got 69`C. Ambient temperature 21`C.
Set back to 55% fan (2740 RPM). It's 74`C back but quite acceptable.

Outside temp: -1 to 2
Fan in Auto Mode.

but notice 15% invalid shares.

I found these card are undervolted - when high stressed it just cant make it - but in games - ok , yet...

orginal vddc - 1.020 (stress) at 1.070 (stress) should be fine still testing

I have to check every single one

What settings are you using?
Sgminer, karloth or cgminer 3.7.2?

I never got more than 970 from my Sapphire R2 290X

He said - 15% rejects - is this Hrate worth this?

I don't have 290X Tri-X... just Sapphire R9 290 reference. But here's some good settings, and perhaps a good pattern for them that could give good results for the 290X Tri-X :)

https://bitcointalk.org/index.php?topic=460181

Hi Guys,

I try to build a rig with 6 R9 290 Tri-x.

With windows or Linux, The computer just recognize 5 of them.
An idea to put the 6 on the same rig ?


ps : motherboard is Asrock probtc and my pci riser are on usb.



Thanks,

First step is to figure out what the problem spot is - card, cable, or mobo slot.  I would.definitely suggest linux for all testing.

1. Identify the failing card by running 100% fan and noting the one that doesn't spin up.
2. Swap pci-e ports, try again. If a different card fails, it is then pci-e slot. Tell us which one.
3. Swap riser (change both 1X, 16X, usb, and power). If the other card fails, you probably have a faulty riser cable.
4. Finally, if 2/3 had the same card fail consistently, plug it directly into the PCI-e 16x slot, and leave all other cards unplugged. Does it fail to boot? Faulty card

Faulty USB risers are fairly common. They also seem to burn out sometimes if you have too many plugged into a single SATA or Molex cable.

Is it possible to call xintensity inline in a .bat? I'm running a 280x and a 290 on the same rig, and if I completely remove intensity from the config, things start to crawl really slow :-/

--xintensity ...

Of course things slow to a crawl if you remove intensity - it defaults to i 8 then.

Hmm..

So --xIntensity isn't a replacement for --Intensity? I call them in the same config?

Yes, xintensity replaces/overrides intensity.  Are you using a version of cgminer (karoth's), or sgminer that supports.xintensity?

Apparently not! Where can I find a compiled version of SGMiner that can work with this voodoo?

I'm running a 290 (Sapphire Tri-X OC version), but I'm not really able to get above 850ish KH/s (Closer to 860 I guess). I tried a variety of settings (including the ones you posted from your .conf file) to no avail, except the xI, as I was apparently doing it wrong in not actually having a compatible mining software.

My card runs really cool! (at 1040/1500) Never going above 72 degrees (mining next to my 280x, which is hashing away at 760kh/s, super solid!)), so I'm really happy there, but would LOVE to get another 50kh out of 'er...

.bat file as follows:
(Device 0 is the 280x, Device 1 is the 290)

Core clock: 1040
Mem clock: 1500

-I 13,19 -g 2,1 -w 256 --thread-concurrency 8193,20481 --lookup-gap 2

With these settings I get no HW errors, and <1% failure rate on the 290 (less on the 280x). If I increase the TC, or the intensity, I start to get HW errors here and there... Any suggestions for a direction to move? I've tried various clock settings, as well as multiple TC settings. -g 1 and -g 2 don't seem to make much difference either way.

Stock cgminer doesn't support comma deliminated g parameters, so that might be part of your issue.  290s tend not to run very well with 2 threads (save the occational person that reports great results for who knows why...). Being capped around 860KH/s sounds like 2 threads to me - try -g 1 and see if you can get a better hash out of the 290.

As for sgminer binaries, I can't recommend anything myself because I compile everything direct from the git tree, but I believe the reddit announcement thread had binaries posted.  Also, you could try karroth's cgminer fork which I believe he provides binaries for.

You could also just run two copies of cgminer (one for each card), using -d 0 and -d 1 on the two instances. Many people seem to get better results out of heterogenious rigs this way (can't say myself, as I run all homogenius rigs with the same cards, oem, and bios).

My tri-x is running quietly at 68c, 840 khs, wu 780, reject rate 2.5% with config below.

I can't get two instance running on sgminer for some reason, even with "device" : "1", and "remove-disabled" : true,

Running my gigabyte 290 (hynix) simultaneously at stock just seems to crash my PC and I don't want to overclock it to 1500 yet, looking for quiet moderate settings to start with. Any advice would be appreciated

"intensity" : "20",
"worksize" : "256",
"lookup-gap" : "2",
"vectors" : "1",
"thread-concurrency" : "25601",
"temp-cutoff" : "88",
"temp-overheat" : "91",
"temp-target" : "69",
"expiry" : "1",
"failover-only" : true,
"temp-hysteresis" : "3",
"api-port" : "4028",
"expiry" : "120",
"log" : "5",
"queue" : "1",
"scan-time" : "60",
"expiry" : "120",
"device" : "1",
"remove-disabled" : true,
"gpu-threads" : "1",
"gpu-powertune" : "20",
"gpu-memclock" : "1250",
"gpu-engine" : "1020",
"gpu-fan" : "50-75",
"auto-fan" : true,
"no-submit-stale" : true

I accidentally posted this as a new thread after I timed out (not spamming intentionally)

975khs each cgminer 3.72+cgwatcher (same as my R9 290x's Club/Sapphire both same)
catalyst 13.12

I have outdoor winter cooling so temps are never over 75-80
2xSapphire R9 290x tri-x

But I broke a fan on one and need to know where to get replacement and how to fix the bugger
I hope can do local and not mail it to Taiwan

Anyone know?

Hi guys

Just bought a brand new Sapphire Tri-X OC R9290X.  I see that everyone is getting at least 850+ hashrates however I'm currently only sitting at 470kH/s.  What on Earth am I doing wrong?  I've even changed the motherboard, CPU, RAM, HDD and PSU to see if that would make a difference and nothing.  I'm running Catalyst 14.1 beta (also tried 13.12).  Using cgminer 2.11.4, also tried other versions as well as sgminer with no luck.  The batch file I'm executing is:
cgminer.exe --scrypt -o stratum+tcp://hk2.wemineltc.com:3336 -u username.1 -p x -I 19 -g 1 -w 256 --thread-concurrency 32765

I've many different variations from many different forums?  I'm started to think I've got a dud card here.  Can somebody help me out please as I'm only new to mining.
My PC is running Asus h81m-e M/B, i5-4670 CPU, 8GB 1333 RAM, 850W PSU (high quality).

Any ideas guys?  I seem to be the only one with this card getting such low hashrates out of the box.  The most I ever saw it get to was 690kH/s but for some reason I haven't been able to replicate this number.

setting problem, what is your clock speed..

GPU clock 1040,  memory clock 1300

I`m running a 290x Sapphire (not the Tri-X, the other one) and the 290x Gigabyte, getting the same hashrate on both, with the same settings.

986 kh/s
< 0.5% rejects
Temp 71-72 (with an extra fan)

(using MSI Afterburner)
+30 Power Limit
1025 Engine Clock
1500 Memory Clock
-44 Core Voltage
Fan Speed 80%

Using SGMiner (4.1.0) with zuikkis kernel on Windows 8.1

-k zuikkis --xintensity 370 -w 256 --thread-concurrency 32765 --queue 0 --scan-time 1 --expiry 30

This being said, I suppose the Tri-X version can easily go over 1 MH/s

Either use 1500 or 1250 memclock.  Start from 1000 GPU clock and go up from there.  Also, what are your temperatures like?  Hot?

Ok I'll give it a try, can you give me an example of the command line I need to execute in my .bat file?  My temperate only gets to about 70 degrees C

by the way, 1500 memclock I've already tried and it made the screen flicker a lot, is something wrong with  my card?

1500 is high for this memory and can cause artifacts... but because of the way timings are setup in the default BIOS, only 1250 or 1500 memclock will yield good performance.

You can try The STILT's 290 BIOS (works on most Tri-X cards) which changes the mem timings so that 1375-1450 will yield the same or better performance then 1500, but I would get your cards stable before flashing to his BIOS.

Basically tho, 1300 will be worse then 1250 on the stock BIOS, possible a lot worse.

I am using The Stilt's BIOS on my Tri-x cards, and am getting over 980KH/s on the best card (1400 mem 1100 engine), and over 920KH/s (1375 mem 1020 engine) on the worst card.

Had anyone had issues in driver crashes with this video card?

When clocking them too high or letting them get too hot, yes.

Have issues even on stock...

Right away, or after running for a while?  What are your temps like before the crash?

After running for a while. 76-78, its well ventilated

Running four cards, waiting on a riser for the fifth. Was unstable due to me not having a couple of meters to see my power draw!  Now I have a 750 and a 1200 watt power supply feeding it. Roughly 300 watts each card and 100 or so for the SSD and motherboard. It gets a little weird when I try to tweak settings but I have left it alone and running 3.64 Mh/s steady with four, hoping for 4.55 Mh/s tonight! Running around 73C with a floor fan on an open air case. Probably need to work on air flow upstairs. It is in the game room. :)

My amateur settings:

-o stratum+tcp://doge.joinaparty.com:22550 -u MYPAYOUTADDRESS -p MYPASSWORD -g 1 -w 256 --lookup-gap 2 --thread-concurrency 24550 -I 20 --gpu-engine 1022 --gpu-memclock 1500 --gpu-powertune 20

I tried the xintensity setting but it doesn't like it. Also trying P2P pool mining. Interesting concept. Connect to a node close to you.  You can browse or connect to test at http://doge.joinaparty.com:22550, no guarantees I will keep it running. Trying to find the best solution for me.

Daz

Have you tried Linux?  I hated the 290s on Windows, drivers were horrible.

Are these 290X or just 290s?  Once you get them stable, try The Stilt's BIOS so you can drop the memclocks down to 1375 but keep the same performance, and gives you more headroom on the gpu-engine.  If all of your cards run at 1500 memclock without crashing, then try pushing the engine up to 1060 - when a card goes sick, decrease only that card's engine by 10.

xintensity will increase your hashrate, period.  No reason not to use it.  What's not to like about a way better way to tune the number of threads launched that actually lines up with your number of shaders?

Can you post a link to the stilt bios for a tri-x r9 290 OC (non-x)?  Thanks in advance!

I tried BAMT, but didn't have much luck with it...

https://litecointalk.org/index.php?topic=12830.0

I saw that earlier but haven't had time to read the whole thread yet. Are there different bios for reference and non-reference cards?

There is only one set of BIOSes by The Stilt for the 290/290X.  And there are only two posts you need to read:
https://litecointalk.org/index.php?topic=12830.msg121827#msg121827
and
https://litecointalk.org/index.php?topic=12830.msg125148#msg125148
(with the first being the important one)

They will work with all revision 1 Tri-X cards.  However, these BIOSes do NOT work with some Tri-X cards that came with the 015.043.000.001.000000 BIOS, because these cards support a new type of Hynix memory that is not supported by the previous BIOS revisions.  Not a big deal however, you won't brick your card by flashing - it just won't be able to detect it's memory and therefore won't work, but can still be re-flashed back to stock (which you can obtain here: http://www.techpowerup.com/vgabios/153454/sapphire-r9290-4096-140115.html)

Thanks!! Very helpful!

Hi

I attempted to flash the STILT bios (MBA_C6710101-100_K2_AGR_STILTMXT_V46.zip (DPM7 - 1.1125V) and got a SubsystemsID's mismatch error?  What am I doing wrong?  I have the 290x Sapphire Tri-x OC

I still can't get my hash rate above 600

Did you use -f (force) ?

e.g.
atiflash -p 0 filename.rom -f

So guys - does anybody run cards that came with the 015.043.000.001.000000 BIOS with speed over 900 KH/sec? I've got 880-890 with CGMiner 3.7.1, Catalyst 13.12 and 1050/1500 clk/mem with MSI AB.

My problem is I start messing with it around 10PM and then run out of patience. :) Messed with it again until 1:30 last night (this morning) and think my fifth card needs another power supply. Pushed I 20 to four cards and then the fifth would cause the dreaded driver failed and was restarted successfully, which is crap, because hash rates never recover without a reboot. Was at 1,080 watts on the 1,200 Corsair. I have another 750 and will try it.  If it is power, then I can mess with the xintensity again. I am hoping it is just frustrating due to the power issue. I just need to understand the math/ratio of threads to shaders. Time to go reading again!

I read through that huge Stilt BIOS thread, wow! I even gave him a post (I think under my real name?) to encourage him to keep coding. Some peeps will drive you crazy with entitlement issues. I have the Sapphire Tri-x OC R9 290 cards. Victim of Newegg price gouging. I like the two BIOS feature so I won't worry about bricking them when I get time to try Stilts' BIOSs. I keep running across stuff to try. Curious to see if the power issue is my problem and then see if the new BIOS will drop the power requirement down to only needing two power supplies. Fun puzzle to work through. If it wasn't for the fast SSD boot times, I would have thrown it out the window. :)  I have it stable at 4.19 Mh/s so I could sleep. Maybe 5 Mh/s if I can get power resolved and run Stilts BIOS? Now to go study your post about which version to try with my cards.

Thanks for the help!  I really try to read and figure stuff out without burdening people. So much info and noise to sort through.

Have a good one!

@DaZuru: You definitely cannot run 5 of these cards on a 1200W... you would be very lucky if you can even run 4 stable for extended periods of time.  I am impressed you have 5 running at all without the PSU shutting off.  Add a 750W or 850W for the last 2 cards.  5MH/s is unlikely - you might manage 980KH/s on a few cards (I do have one that can pull 1.002MH at 1125 engine), but you're more likely aiming for 940 on average or so (every card is different and will take a different engine overclock before becoming unstable).

@vvar10ck: You should be able to manage over 900 with xintensity of 800-1200, 1500 memclock, and whatever engine the card can take (try 1020 and work up; or try 1080 and work down).  The only disadvantage with that BIOS is that you probably can't run STILT's BIOS because you likely have a new model hynix memory onboard.  (on the plus side, you definitely have hynix memory if you are on .043)

Sorry for the confusion, it is running on a 1200 and a 750. I am thinking the 1200 can't handle it and going to test by adding another 750 I have laying around to verify the 1200's ability to provide 1080. 2700 watts should rule out any power problems. I am thinking it is a little weak when they are all at full hash. Even keeping the average I am doing, 4.55 Mh/s should be doable. I'll know in a few hours. :)

Word of advice: Don't just go higher with your GPU clocks, also go lower. There are sweet spots where increasing clocks actually reduces the hashrate. And no, this isn't caused by throttling.

Just go across the entire reasonable spectrum in 10 or 20 MHz steps and try everything.

1200 for 3 cards and a 750 for 2 cards + mobo is more then enough.  Are you using powered risers?

I can run 3 off a 1000W and 2+mobo off an 850W, with heavy overclocks.

This really isn't very true - it depends on the BIOS and memory timings.  There is no 'ratio' that performs better, and a properly tuned BIOS will continue to improve no matter how high you push the engine clock (up until crash of course).

It is true, I suppose, on certainly BIOSes that pushing the engines past a certain point will not provide benefit and may decrease hashrate, but if that's the case you should change the BIOS.

Hi phzi

Thanks for your help the past couple of days, I've finally got my card stable and am ready to give the STILT bios a crack.  Can you please advise which link I need to download as there were several?  I have the Sapphire Tri-X OC 290X with bios version 015.042.000.000.000000 from GPU-Z.  The available BIOS were:

290X:

MBA_C6710101-100_K2_AGR_STILTMXT_V32.zip (DPM7 - 1.2375V)

MBA_C6710101-100_K2_AGR_STILTMXT_V36.zip (DPM7 - 1.2125V)

MBA_C6710101-100_K2_AGR_STILTMXT_V3A.zip (DPM7 - 1.1875V)

MBA_C6710101-100_K2_AGR_STILTMXT_V3E.zip (DPM7 - 1.1625V)

MBA_C6710101-100_K2_AGR_STILTMXT_V42.zip (DPM7 - 1.1375V)

MBA_C6710101-100_K2_AGR_STILTMXT_V46.zip (DPM7 - 1.1125V)

Thanks

Read the whole post I linked to. Each BIOS has a different fixed voltage, instead of variable voltage like the stock BIOS.  I just use the highest (normal) voltage BIOS unless that is unstable, in which case I try the last resort BIOS. If you care about tweaking power consumption, you can try lower voltage versions (maybe work your way up until stable).

You can also grab a multi-meter, or use GPU-z and follow Stilt's info on how to calculate the ideal voltage for each card.

Not a power supply issue, I think. Hooked third one up and even ran extension cord to another circuit. I am not using powered risers. Could that cause the problem? Started digging for the obvious and the card that is giving me trouble was using bios version 015.042 and the other 4 were using bios version 015.043. Flashed it and still a problem going over I 13. Driver flips out. Adjusted GPU and memory but once I hit I 14 it dies. Sitting at 4.18 Mh/s. Time to try Stilts when I get a chance. Running 64 bit Win 8.1. Think 32bit would be better?

I never use Windows for a mining rig - had nothing but stability problems and driver crashes.  Full custom gentoo all the way for me.

Intensity shouldn't play that big of a roll - did you try downclocking the engine to 947? (On the faulting card)

DaZuru I had the same problem when I pushed mine over intensity 13 it would crash.  I solved the problem by putting in more RAM.  12GB works fine for me but 8GB wasn't enough.  I'm not sure why this is the case when others are getting away with far less.

phzi I flashed the STILT BIOS and am now getting 1013kH/s so thank you very much :)

I've got 2 problems with the tri-x's i'm using:

The first is that a 6th card isn't detected for mining. If I plug it in, I can't mine, but "aticonfig --lsa" shows all 6 cards.


The other is that I'm getting a ridiculous amount of invalids, something like 3-5%.


any idea?

What motherboard, what OS, what pool? 3-5% might be perfectly fine depending on where you are mining and what your ping to the stratum server is.

That's just bizarre... I run all my rigs with 4GB of ram - I have never believed the more RAM helps story, because cgminer uses hardly any RAM.  I usually have 90% free RAM with 4GB.

Very nice!

Also interested in RAM question, since I had basically the same issue. Constant driver crashes with 3-4 or more cards.

Scrypt: about 990Kh with very few rejects/no hardware errors. Did some benchmarking and even though there's the occasional reject (less than 1% mining 24 hours), happy with keeping it at 990Kh. Runs at around 78C at 65% fan.

N-Scrypt: about 440Kh with no rejects/no hardware errors. Runs at around 70C at 50% fan.

CPU is 4770k with 16 GB RAM

And oh yeah, I run Ubuntu from a USB stick. But maybe occasionally boot up Win 8.1. Speeds are similar but Linux requires less reboots when restarting CGMiner. Windows tends to do BSOD when CGMiner is closed, probably because of the OC-ed GPU returning to normal

How did you decide to try out upgrading the RAM? Did someone give you any reasoning to do it (because of the driver crashes) or you just decided to spend $200 and try it out?

Did you roll your own or use a Gentoo Mining guide somewhere? I am basically dangerous with Linux, but can follow a recipe. :)

I'm using BAMT 1.5 on a Gigabyte 990FXA-UD7, and I'm using a pool that usually has a very low reject rate (like 0.3% on average). The cards individually can range from somewhere like 3-10%, but usually average 3-5%.

It may be related, but I keep having one+ cards shutting down. I can't seem to get them to work correctly.

Any idea about using 6?

My setup is self-designed.  The whole install fits on a 8GB drive with plenty of free space for compiling other mining software, running coin daemons, etc (the base install is less then 3GB).  I bought a bunch of tiny SSDs (8GB), because I find USB drives to be super flakey as boot drives.  The gentoo install has absolutely bare minimum components, runs a stripped down kernel, is optimized for haswell chips, and includes only the most basic xorg install, xterm, sgminer, and asimple series of startup scripts to launch everything automatically.  Took me a few days to throw together, and I continue to tweak it on occasion.  

The majority is pretty simple straight forward (although it might just be me that thinks so... I have installed gentoo many times over many years):
- follow the gentoo handbook to make a very basic install
- optimize your cflags in make.conf to optimize for your CPU (refer to handbook), and set VIDEO_CARDS="fglrx" in make.conf
- emerge cgminer (which pulls in the most basic x11 install, opencl dependencies, etc)
- emerge xterm and xset (barebones terminal and user preference utility for x11)
- emerge ati-drivers (you might need to pick a specific version e.g. emerge =x11-drivers/ati-drivers-13.12 , or specify the version properly in portage config files - refer to gentoo wiki)
- create a non-root user called mine (or whatever name you desire)
- set the user to auto login (http://wiki.gentoo.org/wiki/Automatic_login_to_virtual_console)
- auto-start x11 by adding "[[ $(tty) = "/dev/tty1" ]] && exec startx" (without quotes) to ~/.bash_profile of your non-root user
- emerge git
- retrieve sgminer source from github, and compile (I pull and install to ~/src/sgminer)
- create a startup script that launches from .bashrc and starts sgminer if needed.  I wrote this script for this purpose:

The above script checks if X11 is running yet, uses a folder as a lockfile (a super handy *nix shell trick), checks if sgminer is already running in a screen, and then launches my mining script if needed.  Since you're calling it from .bashrc, it will launch every time a terminal session is started.  This means that it will launch before X11 starts, and exit because there is no X server; launches again in xterm when xorg starts up (xterm launches 3 times by default, so the lockfile causes 2 instances to exit, and the third to start up mining); and will run when you login via SSH (and display "sgminer is currently running..").

~/mine.sh then contains:

You can then use "screen -r" from ssh to access sgminer, and edit ~/sgminer.conf to change your config.

I should probably throw this together into a decently detailed tutorial in it's own thread - maybe I will do that a bit later today or over the weekend.

^I would appreciate that! Thanks for the post.

I would love to find out how you have your cards set in the bamT.conf. I am running 1 Sapphire R9 290X Tri X (model 11227-00-40G) and i am only getting 671 KH/s.

Here is my setup
990FXA-UD5
AMD Sempron 145
Seasonic X850
Sapphire R9 290X Tri X (Model 11227-00-40G)
4GB memory
All in a Cooler Master HAF X 

Post your config.

I am not sure what is needed from the configuration page so this will be a long post. I apologize in advance for the length of this. In this configuration page i am using GPU 0. I included GPU 1 (which is not in use) because it has different setting than GPU 0. Thanks in advance!

You need to edit cgminer.conf in the same folder as bamf.conf ... bamf.conf doesn't really do much, and the overclocks won't work with a 290.

Im sorry. Is this what i need? I will edit the previous post to delelte that lenghty post!

"intensity" : "20",
"vectors" : "1",
"worksize" : "256",
"kernel" : "scrypt",
"lookup-gap" : "2",
"thread-concurrency" : "40960",
"shaders" : "0",
"api-listen" : true,
"api-port" : "4028",
"api-allow" : "W:127.0.0.1",
"expiry" : "120",
"gpu-dyninterval" : "7",
"gpu-platform" : "0",
"gpu-threads" : "1",
"gpu-engine" : "0",
"gpu-memclock" : "0",
"gpu-powertune" : "0",
"log" : "5",
"no-pool-disable" : true,
"queue" : "1",
"scan-time" : "60",
"scrypt" : true,
"shares" : "0",
"kernel-path" : "/usr/local/bin"

So i was running 2 sapphire R9 280X Tri-x OC's and one of them died after i undervolted it to 1.087mv (the other one was fine but put it back to 1.093 just incase). so since i have to RMA it cause it was a week after the store 15 day direct exchange, it'll be gone for about month+ according to althon. So i started looking for another 280x cause i was gonna run 3 eventually once i got my risers and another pci-e cable from corsair. i couldn't find anything at a decent price that would be worth it so i stopped by the place i got the rest of the rig from and they had an open box sapphire R9 290x Tri-X so i figured why not spend the difference.

so i get everything up and running and the 290x runs about 925kh/s but my remaining 280x dropped down to under 600kh/s. Before the 290x in bamt i was getting 750kh/s from each of the 280x's with like 1-2% returns tops.

i put it back into win8 and cgminer is doing the same thing only about 5% lower on both cards (which is why i was running bamt). config looks like this in win8:

timeout /t 30
setx GPU_MAX_ALLOC_PERCENT 100
setx GPU_USE_SYNC_OBJECTS 1
COLOR 0A
cgminer --scrypt -I 13,20 -g 2 --thread-concurrency 8192,25601  --gpu-engine 1080,950 --gpu-memclock 1500 -o stratum+tcp://us-east.multipool.us:7777 -u xxxxx -p xxxxx

I can run the TC at 32765 in bamt but if i try that in win8 it hangs the card and gets disabled,

got it sorted, running bamt 1.3 getting ~750 from the 280x and ~950 from the 290x 0.085% returns total. :D

I have a Sapphire Tri-X 290x with stock bios, using sgminer 4.1.0.  I have terrible case cooling -- don't ask.

I'm unable to get 950+ at any settings.  I can push 900k but the temps go over 90cGPU/100cVRM.  I tried memclk 1500 and it worked for awhile then started crashing hard.

I found I can drop GPU Engine down to 875/1250mem and still get about 830khs with 75cGPU/80cVRM at 85% fan, but I'm still not sure how to tweak the settings for this.  Shooting 5 variables in the dark at an interation speed of once per 5minutes to let it stabilize, not fun.

So I'm sure these settings are not optimal but I could use help figuring out what to change.  My #1 priority is low temp and 85% fan max.

Only relevant parts quoted - nuff said... fix your cooling situation - VRM and GPU needs to stay cool.

---

That said...

stop using gpu-threads 2... 1 is optimal for 290(X)

Where the heck is that thread-concurrency from?  Try some normal numbers as posted everywhere.

If you are tuning intensity/xintensity, get rid of the shaders line.

Crank your FAN!  Why are you using variable?  Stick it to 90 or 100% if you need to to keep it cooler. temp-* is too low.  cutoff should be 95, overheat something like 97, and target 92.

If you are mining on a pool that has a good reject policy, crank the xintensity up - 1100 is a good value.  Only pool like CM, or solo-mining low-diff coins, will reward a low intensity.

I finally got some good settings for my 290x Tri X OC.  I am running 918KH/s with these settings in BAMT. Mine runs at 73 degrees with these settings. I also have a Cooler Master Haf X. So there are a lot of fans in my case! Thanks for the response rhaikh!

"intensity" : "20",
"vectors" : "1",
"auto-fan" : true,
"temp-target" : "75",
"temp-overheat" : "80",
"temp-cutoff" : "90",
"worksize" : "256",
"kernel" : "scrypt",
"lookup-gap" : "2",
"thread-concurrency" : "32765",
"shaders" : "0",
"api-listen" : true,
"api-port" : "4028",
"api-allow" : "W:127.0.0.1",
"expiry" : "60",
"gpu-dyninterval" : "7",
"gpu-platform" : "0",
"gpu-threads" : "1",
"gpu-engine" : "947",
"gpu-memclock" : "1425",
"gpu-powertune" : "20",
"log" : "5",
"no-pool-disable" : true,
"queue" : "0",
"scan-time" : "2",
"scrypt" : true,
"shares" : "0",
"kernel-path" : "/usr/local/bin"

Thanks for the reply.

The TC is from 2816 * 8 + 1 and just a guess.  Where the heck is 24000 24550 32765 from?

2816 * 5  + 1 = 14081
2816 * 6  + 1 = 16897
2816 * 7  + 1 = 19713
2816 * 8  + 1 = 22529
2816 * 9  + 1 = 25345
2816 * 10  + 1 =  28161
2816 * 11  + 1 = 30977
2816 * 12  + 1 = 33793

It doesn't seem happy using more than 30k thread concurrency so I'm using *8 at the moment.  I can remove the shaders line but I am pretty sure the card has 2816 shaders so I doubt it makes a difference.  I'm willing to entertain other numbers but I'd like to hear a reason.

I want to keep the fan low (hence 60-85) and the temps low to increase overall longevity of the card and avoid a bad time when summer hits.  I'm willing to give up some kh/s to do it but I also want to run an optimal setting.  Giving up 10% so far seems like an OK deal but 5% would be nice.  I'll try a single gpu thread and see where that gets me.  And more trial and error.  I think it was generally better than 2 in my attempts earlier.

Why this #?  People here say try 1000 + x*10 and I guess that makes sense. What speeds under 1000 should be tried?

Intensity 20 on these cards gives me way too many rejects, I found 18 is the best setting (< 1% reject).

Stop using intensity... switch to shader multiple threads - xintensity.  Intensity is completely antiquated now.

Intensity - raw(?) - xintensity
13   8192           2.90909090909091
14   16384   5.81818181818182
15   32768   11.6363636363636
16   65536   23.2727272727273
17   131072   46.5454545454545
18   262144   93.0909090909091
19   524288   186.181818181818
20   1048576   372.363636363636
21   2097152   744.727272727273
22   4194304   1489.45454545455

So you are suggesting an xintensity which corresponds to between 21 and 22 whereas he's saying a value closer to 18 (93) reduces rejects.

In my own setup I tested xintensity keeping other values constant.  Values higher than ~372 (such as 1100 and 702) were causing some throttling issues i'm assuming due to gaps in utilization (temp pretty cool throughout).  Values lower than ~372 (such as 176 and 93) reduced the hash rate with no improved thermal performance.  I don't have any reject % numbers though and I will test that again later.

I'm a bit out of the loop when it comes to newer cgminer versions... I did try an xintensity value of 300 just now and this seems to be working better, rejects are at 4% which is an improvement over the 7% I was getting with intensity 20, and the hash rate is around the same as intensity 20.

My hash rate gain from -I 18 to the new settings only around 4-5% though, so with the rejects factored in it's almost a wash.

I'll have to play around with the settings more... I came across someone's post with 300 xintensity and his rejects were around 1% though, so not sure what's going on. He was running a custom BIOS and I'm still on stock, maybe that has something to do with it.

rhaikh: I don't see where I suggested xi 1100 in that post...

Relm9: rejects are hugely dependant on the pool too. Where are you mining?  BIOS is unlikely to affect your reject rate.

I get ~1% rejects at xi 1100, btw.

Yeah, I did wonder if it was the pool too. ScryptGuild here.

Try WafflePool and see if your rejects are better.  Your high reject rate is very likely due to the way ScryptGuild decides if a share is stale or not.

Hello miners,

I will appreciate if you support me on an issue I am currently experiencing with my RIG.
I will start from scratch trying to simply explain my pain. Thank you in advance for your time! :-)

RIG Specs:

MB  Asrock H81 ProBtc
GPU 4x Saphire 290 Tri-X OC
PSU 1x Corsair i1200ax Platinum
CPU Pentium
RAM 2x2GB 1333mhz

3 GPU's are on risers PCIx1 --> PCIx16
The system is assembled and starting with SMOS BAMT from USB stick. It is working with SMOS config suggested from their site for about 11-36 hours when it suddenly freeze with black screen, no SSH and no web dashboard. The only solution is physically powering off the RIG. I also observe that it wont start mine successfully after reboot with the overclocked cgminer.conf and I have to load a default one from backup to reboot and afterwards to update the overclocked config for another 11-36 hours of successful mining.



So far I have done some investigation and I have the following results:

1) Assuming power consumption issue - I have started the RIG with 3 GPU's where I have the same result. The system works perfectly fine for ~11h and then freeze.
2) Assuming overclocking issue - I have left the RIG with 3/4 GPU's with default cgminer config - the very same result.
3) I have tested also the MEM and can confirm it is fine.
4) GPU temps are 62-69 C with fan's rpm between 55-85%

5)From the logs I can observe the following error:
root@smos-1:~# tail /var/log/messages
Mar  5 20:02:20 smos-1 kernel: [   58.821275] atitweak[3090]: segfault at 2e6e6570 ip b7511cad sp bf85f700 error 6 in libc-2.11.3.so[b74a2000+140000]
Mar  5 20:02:20 smos-1 start_mining[2992]: generating per gpu munin plugins for template gputemp...
Mar  5 20:02:20 smos-1 start_mining[2992]: done generating munin config, starting munin-node
Mar  5 20:02:20 smos-1 kernel: [   59.000845] atitweak[3191]: segfault at 1d ip b75e3cad sp bfd397b0 error 6 in libc-2.11.3.so[b7574000+140000]
Mar  5 20:02:24 smos-1 start_mining[3536]: starting cgminer with cmd: cd /opt/miners/cgminer;/usr/bin/screen -d -m -S cgminer /opt/miners/cgminer/cgminer -c /etc/bamt/cgminer.conf
Mar  5 20:02:24 smos-1 start_mining[3547]: post mining tasks...
Mar  5 20:02:24 smos-1 start_mining[3547]: gathering post-mine stats for offline mode
Mar  5 20:03:01 smos-1 kernel: [   99.895369] atitweak[3905]: segfault at 50a0556 ip b753fcad sp bfd83260 error 6 in libc-2.11.3.so[b74d0000+140000]
Mar  5 20:03:01 smos-1 kernel: [   99.958959] atitweak[3910]: segfault at 64356441 ip b7459cad sp bf862480 error 6 in libc-2.11.3.so[b73ea000+140000]
Mar  5 20:05:01 smos-1 kernel: [  220.020975] atitweak[4060]: segfault at 776f64c2 ip b7590cad sp bff36820 error 6 in libc-2.11.3.so[b7521000+140000]


I have potted that one of the hangs was with GPU 0 DEAD/SICK - How could I confirm on which slot is GPU 0 installed ?
Could you please advice on what should I focus my self to investigate any further the RIG issue ?

Really appreciate your help!

need an intensity that don't make my pc unusable

getting 856 at 72 degrees in an open case.

I have been mining for 8 days straight now so well happy.

the magic TC is "25599"  ;)

My favorite lately is 20479.  Best performance I have found.

Hi all,

Two of My R9 290 Tri-x's just died.  They won't boot up alone directly plugged into an X16 slot.

How do you go about RMA'ing these?  I see everyone talk about RMA, but don't see any info on it.

Also what exactly does RMA stand for ?  Return, Manufacturer .... ???

Cheers,

I realise "You cant polish a turd", but does anyone know of a n alternative/improved BIOS that works with the later 290 Tri-X 043 cards?

eg:  that supports the later Hynix ram.

Best,
-Evil

with what x intensity?

I get about 930khs from my 043 cards in bamt, how much more you expecting?  tc 20481, 1030/1400 i-20