Topic: Sapphire R9 290X Tri-X OC Hashrate? - page 2. (Read 36578 times)

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.

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

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,

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"

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.

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!

Post your config.

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 

^I would appreciate that! Thanks for the post.

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'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?

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

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?

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

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

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!

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?

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 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)

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?