It could be:
OS related - memory management / scheduling / general efficiency / better throughput / less overhead / etc.
Compiler related - better optimizations / better vectorization / better use of newer instructions / etc.
It could be any combination of the above AND maybe some things we haven't considered yet.
TL;DR - we don't know for sure. We can only guess.
well, the driver crashed again overnight. The network light was still on so I'll be watching the ppd to see if boinc is still running.
Is there anyway to skip the GUI altogether and just get to the console? If boinc starts even before login in then I could avoid some headaches that way.
1. In bios, set the system to halt on "no errors"
2. take out the vidya card
That's it.
Also, it's advisable to set up SSH beforehand, for maintenance / checking up. Also, if You set up remote access in boinc, You can connect with a boinc manager from another computer to check on boinc when needed.
Currently I'm only doing DSFL, as it's one of my less-crunched projects (at least compared to CEP2). I'll be switching over to all FAAH at some point. But it is just faster as well
An OC'ed hex may be capable of doing a good 13-14k. I'm getting 8.5-9k or so with the 2600k, so 50% more cores running ever so slightly higher could do amazingly
Good plan! I'm setting up the i3-2100+HD7950 as my gaming system again so that I can dedicate my best to Linux for WCG, and I intend to put Linux on the 2700k & 3930k at some point (by mid-August at the latest). Maybe the 3770k as well, but we'll see. I want to keep around a Windows system for FAH.
1. In bios, set the system to halt on "no errors"
2. take out the vidya card
That's it.
Also, it's advisable to set up SSH beforehand, for maintenance / checking up. Also, if You set up remote access in boinc, You can connect with a boinc manager from another computer to check on boinc when needed.
It could be:
OS related - memory management / scheduling / general efficiency / better throughput / less overhead / etc.
Compiler related - better optimizations / better vectorization / better use of newer instructions / etc.
It could be any combination of the above AND maybe some things we haven't considered yet.
TL;DR - we don't know for sure. We can only guess.
ok, edited the grub file to start directly into terminal and installed htop to keep track of boinc. What should I do to enable SSH? I had set it up for VNC previously, is that enough? Will I be able to log into it with TightVNC or should I use telnet?
Hadn't restarted it yet.
edit: can't connect with telnet or VNC. Gotta find a program.
edit2: sorry, I wasn't thinking straight. Now that there's no GUI what I should do is setup boinc to control the clients on the linux crunchers. I was using VNC before to report the completed tasks remotely... in fact if I could setup the clients to report tasks immediately (a la windows) I wouldn't even need to do that.
ok, edited the grub file to start directly into terminal and installed htop to keep track of boinc. What should I do to enable SSH? I had set it up for VNC previously, is that enough? Will I be able to log into it with TightVNC or should I use telnet?
You need the GUI running in onder to use VNC, as it lets You take over the existing one. AFAIK.
Well, since You said You did not restart it yet, I suppose You still have access.
To set up ssh: 1. run this in the terminal
Code:
sudo apt-get install ssh
2. edit Your /etc/ssh/sshd_config file to suit Your needs [1] 3. run
Code:
sudo service ssh reload
4. all set
----
To manage BOINC from a remote manager:
On the machine You want to manage, go to the boinc directory [2] (can't remember what the default was. [Ion], help me out here - check what directory Your Linux boinc is using) and create a file remote_hosts.cfg there. In it, add IPs of the machines You want to allow access from. Then look at the contents of the gui_rpc_auth.cfg file - that's the remote access password for it. You can change the file contents to set Your own password. It has to be at least 6 characters long, IIRC.
Then restart boinc by doing [3]:
Code:
sudo service boinc-client restart
Then, in the computer You want to access that boinc from, with advanced view enabled:
Click Advanced -> Select Computer, then in the "Host name" field, enter the IP of the machine with the boinc You want to manage and in the password field - the contents of the gui_rpc_auth.cfg file I mentioned before. Click OK. It then will connect to that computer and You will see its boinc status & be able to change settings as normal.
To show the "local" boinc there again either restart the manager OR in Advanced -> Select Computer enter "localhost" as hostname and click OK (it fills in the password automatically for localhost)
----
[1] if the computer is not directly visible to the Great Wilds - if it's in Your own LAN, only has an "inside" IP, but not an IP which would make it directly accessible from the internet, You don't need to change much. Just make sure PermitRootLogin yes and PasswordAuthentication yes are set and not commented out. And You should be able to SSH into Your machine for administration.
In case the computer IS visible from the Great Wild, set Port to something other than 22, preferably in the 5000+ range. That will prevent most crackers from detecting SSH on it and thus prevent attacks. In such situations it's also good to disable Password authentication and use private key authentication. But I won't go into detail unless You ask me about it; it's a bit complex.
WHAT!? Only two active projects left!?
Really, at this rate not only we won't get any GPU work, but we might even run out of CPU work!!! I don't like this one bit. And am very pissed. (DSFL went intermittent just now)
Yes, I know it's a double post, but this is completely different kind of info compared to my previous post.
FAAH still has about a year of work left, although with no other project competing for CPU time (CEP2 doesn't really count since the boinc client only processes 1 WU at a time, unless you change the stock settings) that estimate should go down. Maybe 200 days left?
WHAT!? Only two active projects left!?
Really, at this rate not only we won't get any GPU work, but we might even run out of CPU work!!! I don't like this one bit. And am very pissed. (DSFL went intermittent just now)
FAAH still has about a year of work left, although with no other project competing for CPU time (CEP2 doesn't really count since the boinc client only processes 1 WU at a time, unless you change the stock settings) that estimate should go down. Maybe 200 days left?
I wish my rig was running as cool. I just made a little change to sig rig.
The result..
4GHz on air!
k10temp says I'm at 40.6c. Not sure which one to believe. I can touch the base of the heatsink without burning myself, but that could just be poor thermal transfer to the heatsink (something I suspect). Seems stable, so hey!
Code:
Mon 08 Jul 2013 08:05:05 PM PDT | | Benchmark results:
Mon 08 Jul 2013 08:05:05 PM PDT | | Number of CPUs: 6
Mon 08 Jul 2013 08:05:05 PM PDT | | 3014 floating point MIPS (Whetstone) per CPU
Mon 08 Jul 2013 08:05:05 PM PDT | | 19065 integer MIPS (Dhrystone) per CPU
Temps are running around 45C for the rig but the A880GZ board doesn't really read temps all that well. The case is well ventilated w/plenty of fans and it's running a Xig Gaia on the chip. In this weather, I would only run the 1045T rigs here.
@manofthem- yeah it gets pretty hot/humid here but it doesn't stay like it does where you are.... at the most we get this weather for 2-3 week spells and 6 months from now it could be -5F with a -30F wind chill BRRRrrrr!!!!
@xvi- I have a set of the Xig adapters to turn that cooler to blow out the back of the case if your interested
*Note- you will need the socket 1366/775 mounts for the cooler to use them
Switched some GPUs around today to make them run cooler and whatnot, and diagnose why the 2600K rig likes to lock up. My mATX 3770K has seemingly memory related issues as well, as I remote in and WCG and F@H "have stopped working" so I pulled the 7770 and GTX 660Ti out of it and put in the GTX 470. Perhaps the 530w PSU was overworked, but we'll see.
I think that the power saving features is what is crashing my Mint crunchers. When crashed I've found them either with the screen turned off or with a drm_panic.
I think that the driver sometime fails to turn off/on the screen and crashes right there.
Is there anyway to make setterm stick? I've found some ways but they are for other distros and Mint doesn't seem to be setup the same way (I can't find the files they refer to).
@xvi- I have a set of the Xig adapters to turn that cooler to blow out the back of the case if your interested
*Note- you will need the socket 1366/775 mounts for the cooler to use them
Xigmatek Crossbow? Got it already, but thanks! Originally had it pointed towards the back in my old case, but moved it back since the Phantom has a large exhaust fan up top. Wasn't sure if the orientation made any difference.
I've been curious to know how a FX 8350/8150 at average overclock would stack up against my X6. I've been looking for people who seem to have both a Phenom II and a FX crunching (TRWOV, Norton, Mad_Shot) and trying to find PPD via hosts (Norton's, for example). Norton's sig says an FX 8350 4.4GHz for 7 Day Average of 5,346. That certainly beats my X6's 4,180 by ~27.8%. So what's the general consensus on FX processors for crunching? Good value? How hard was it to get those speeds, Norton?
![[Ion]](https://tpucdn.com/forums/data/avatars/l/75/75675.jpg?1443792369){kind=avatar}
That is odd... the admins on the WCG forum keep stating they have work for us but I guess we'll have to see.
