Saturday 03/08/02

The Top 20 has been updated.

I made a few adjustments to the way I check the Folding Club numbers so hopefully it should be more up to date from now on.

Sunday 04/08/02

The Top 500 is up, thanks xylus and Team Weekly has also been updated.

Thanks to looktall for bringing this to our attention. Apparently Vijay has found a serious bug in the stats system and a lot of us have been getting no credit for some of our returned work. They are going to sort it out and give everyone there missing points back. The thread is here at the Folding-Community

ok, I found a pretty major bug in the stats. This bug was giving credit, but the credit value was wrong (set to 0). It's clear how this would be bad.

First thing (definitely on Monday, hopefully tomorrow) is we'll get this fixed for new WUs. Next, since all this is in the logs, we can credit these WUs appropriately.

So, thanks for bearing with us. It will take us a few days to get this in (I want to do it carefully as to not to cause any more problems), but once in, new WUs won't suffer from this problem and there will be a lot of old points coming back.

Thanks for all your help!
 

Friday 16/08/02

Finally we are back up and running at 21:00 WST.

The Folding Clubs have been updated to the last Statsman update, everyone should be in there right clubs but if not you know the drill.

Try the new Milestone clubs and see what you think.

Also give the new Members page a whirl to.

With these gromac proteins flying around these days it may pay to try out TheWeatherman's EM3 program to monitor them. Download it from here its a great program and is updated regularly.

Saturday 17/08/02

The Top 20 has been updated with numbers for two weeks not the usual one.

Sunday 18/08/02

The xylus 500 is up.

Team Weekly has been updated.

News from Stanford about the Gromac work units and some problems some users with AMD K6 cpu's.

8/17/2002 Issues with the Gromacs WUs

There have been some issues with the Gromacs WUs. Most notably for users with AMD K6's as well as users who restart the cores frequently. Guha has figured out fixes for each of the problems and new Gro cores should roll out Monday or Tuesday.

8/16/2002 New servers online

We've added some new servers (.108, .109, .119). They should help the network load. We've also moved our servers to be directly off of Stanford's main backbone. That should help as well. This is good timing, since F@H continues to grow, and has broken the 40,000 active CPU point recently.

Monday 19/08/02

Trivia time, did you know that Overclockers Australia hit the No1 position for the first time on October 29th last year with a team score of 7842 points with 187 members, 12 days after FAH2 started.

xylus a man of many talents has started a thread in the Forums so that we can all post our Folding achievements and Milestones in. Instead of having numerous single threads we now have a common place to brag and boast of our Folding prowess.

I thought I would take a look at the Folding poll that has been running for a while in the forums and it seems that Queensland and NSW were neck and neck on 142 votes each and vying to be the top Folding state in Australia. Since then NSW have taken the lead by one, so if you haven't voted yet hop to it and do your state or country proud.

There has been a bit of concern over the latest proteins that Stanford have released namely project 180 (p180_enghomeoGB) and project 181 (p181_enghomeoPB0.5), they are worth 5.6 and 8 points respectively and only consist of 10 frames. The concern is about checkpointing and the length of time it takes to do a frame on these monsters. A lot of time is going to be lost on the any machine and in particularly the slower ones where it can take over 3 hrs or more to complete a single frame, if computer has to be shut down or folding is shut down mid frame. Vijay is aware of this and all future big proteins like these will be back to the normal 100 frames. Here is a couple of snips from the Folding-community forums concerning this ...

I've increased the # of checkpoint frames to try to make the checkpointing work out better.

This will affect new WUs, but not old ones. We decreased this in order to try to minimize net bandwidth, but it clearly does create issues for you guys in checkpointing.

Due to the internal nature of Tinker, the number of checkpoints is directly related to the number of frames we get back.

We've talked about decoupling this, but this would take some re-hashing of Tinker and since we're pushing to move to Gromacs right now, what work was proposed internally, but never assigned.

Bojan got these new runs up and wanted to decrease the net bandwidth, but forgot about the issue of how decreasing the # of frames back would impact on the checkpointing.

When I noticed the posts here, I set it back to the normal manner.

Sorry for the confusion,

Vijay

Its time to update your EM3 program again to version 1.9.3.

Tuesday 20/08/02

As you can see I have been playing again and reorganized the stats bar. I have included a Current Projects section and the other sections have been modified a little. The top section has been spread out a bit and should be easier to read.

Queensland have hit the lead by one. Vote here and get behind your state.

The p180_enghomeoGB and p181_enghomeoPB0.5 proteins are now 100 framers and we all can breathe easy now and not worry to much about losing folding time. TheWeatherman has got a new emprotx file out so EM3 recognises the proteins with 100 frames.

Wednesday 21/08/02

A note from mashee that the Gromac core (FahCore_78) uses a lot less virtual memory than the tinker core (FahCore_65). That should please some of the multiple instance people.

Gromac

FahCore_78: 5792 KB memory, 6716 KB Virtual Memory.

Tinker

FahCore_65: 5344KB memory, 57140KB Virtual Memory.

This one is courtesy of looktall, if you have to make adjustments to your config don't answer "y" instead of "yes" to the questions because the client only accepts "yes", otherwise you end up with a username = y and team = y. Also don't edit the client.cfg with Wordpad because that completely stuffs up the information in the config, notepad is better but still dangerous if you mess with the formatting. Do it the proper way by using the -config switch (for console users that is) and right clicking the FAH icon and selecting "Configure" for graphical users.

Thursday 22/08/02

Here is a treat for you Linux users out there, a faster core, apparently the current core was a bit slow.

People have noticed that the Linux core_65 recently got slower (http://forum.folding-community.org/viewtopic.php?t=1287&highlight=). Guha has put a new one up.

If you'd like to use it, just delete your current core.

Thanks,

Vijay

Stanford have a new page up relating to their servers and what's hot and what's not. Very interesting and a very popular page it will be no doubt. I have added the page to the Stanford section on the left side of this page, right under Current Projects.

8/21/2002 Server status page

We've put up a new server status page at http://folding.stanford.edu/serverstat.html It checks several statistics to tell if a server is down. It updates every 10 minutes and the page should auto refresh.

Friday 23/08/02

Congratulations to the [H]ardOCP team on being the first team to reach the magic one million points mark, well done dudes we will be there to in a few days.

Today we have a new beta client for windows and Linux users. We are up to version 3.14 and its available from here.

We've posted beta clients (v. 3.14) for Windows and Linux at folding.stanford.edu/beta3 . It doesn't have all the things listed in the To Do list some of you have prepared (we'll get to that), but just the following:
- Configuration option added of disabling assembly optimizations in core (applicable only to Core_78).
- Bug fixed where if changed certain Configuration options, core saw as an improper shutdown.
- Send/autosend conflict resolved.
- If core outdated error arises for a work unit, and still remains after a new core has been downloaded, that work unit is deleted.

The Gromac proteins are back...

We finished changes to the Gromacs cores. We haven't started assigning to the Gromacs server widely again yet -- we'll let you know when we do. Here are the key changes:
- One issue in the old core was that for certain processors (particularly the K6), Gromacs detected wrongly the presence of 3DNow support. Now, in general, if there is a bad exit, on the restart, assembly loops will be disabled, and if this succeeds, then they are disabled permanently for future restarts too.
- On certain processors that did have support for SSE or 3DNow, we noticed that these instructions actually do not perform correctly consistently -- sometimes resulting up as instability on a WU that we know to be stable. The core now does a periodic test calculation and if the answer is incorrect, restarts with assembly optimizations disabled.
- There is also support for a client configuration option of manually disabling assembly loops (see soon-to-appear post on beta client 3.14).

One other thing is that we're going to start maintaining a page listing the changes from core version to core version. The beginnings of it are at folding.stanford.edu/corenotes.html .
 

Sunday 25/08/02

The Top 500 is up, thanks xylus. Also Team Weekly has been updated to.

I added a bit of colour to the stats bar and a few more bits of info.

A little diversion from the usual Folding News regarding Project Dolphin which is a project that counts keystrokes yeah that's right keystrokes. We now have a team thanks to jaani the founder and Pyro for introducing it to us.

Project Dolphin is a project established for a small circle of contacts on the internet, in order to track how many keys were being typed by them. After finding some rather fortunate publicity on the internet and on television over the last couple of months, the project has taken on several thousand users and has grown way past its initial goals. It is now striving to mature to the needs and requests of a larger user base than had ever been anticipated.

The program doesn't compete with your folding for CPU cycles and it only counts your keystrokes, so its not a threat there. There is no security issues at all because it only sends the number of keystroke back to the server. When you install the program it puts a small icon in your systray this is called the Pulse which does all the counting.

Pulse is a small 32-bit Windows program that remains in your computer's memory while it is on. Its job is to keep a running total of the number of keys you've typed since it last "pulsed" this server.

A "pulse" is simply a notification sent to this server, informing it of the number of keys you've hit.

So at the moment we have 27 members and rising rapidly and we have accumulated 2,831,558 keystrokes, which puts into 423rd position. Guess who is just up the road a bit in position 311. So if you would like to have a bit of harmless fun and help along another OCAU team join up and after you have become a member log in and join the Overclockers Australia team.

A little problem with the Gromacs proteins and SSE being disabled after a lock up, this was brought to out attention by Cooties and Falls. Here is Cooties quote from the Forums...

Earlier today I got my first gromacs protein (hooray). When I first got it it took 8 minutes (approx) to do a "frame" on my AXP @ 1600 mhz. However later on in the day my computer crashed (as it occasionally does) and when it restarted the protein was taking 20+ minutes to complete. I just left it thinking it was just some sort of wierdness with the new protein. However I found in the log this line :

Read that assembly optimizations off.

Does that mean that its turned off SSE (or whatever the AMD equivalent is). I read elsewhere that these gromacs are alot quicker with SSE Cpu's.

Any help or ideas as to whats causing this slowness and how to fix it ??

thanks

Apparently a file called "machinedependent.dat" is created in your folding directory after a lock up which tells the core to not use SSE optimisations when processing the protein. To fix this up delete the "machinedependent.dat" file and all should be well again.

There is also a thread in the Folding_Community forums regarding this subject.

Wednesday 28/08/02

An issue this time with the FAH client doing Gromac proteins when the client is set up as a service (W2K Pro/XP) and the user logs out. This causes the client to close down folding until someone else logs back in, this has affected coops and DGROMS and many others from other teams. Stanford know about it and have had communication with DG and a fix was made to the client and tested. DG reported the new client fixed the problem and hopefully it will be released for generally use soon.

A few of us me included were effected by the noname proteins handed out be server 109 yesterday. The core didn't like the protein and came up with a "Client-core communications error: ERROR 0x1" and the client proceeded to download a new core thinking that it was at fault, this is repeated over and over until the client was stopped. But really the faulty protein was the culprit, if you had the latest FAH client 3.11 this wouldn't have been a problem because of the changes made to it over the previous client releases. This is a snip from a post by Vijay...

In this particular case, an old feature of the client caused a problem. If the client can't reach the AS (eg if it's down), then it will go back to the last server that was ok. This worked great when we only had 1 core and when it was easy to keep all the platforms up to date. Now, this can cause a real problem -- clients can go to servers they should not go to.

Once we saw what the issue was, Guha has made the fix in the client within hours and the next builds will have this problem removed.
 

So make sure your clients are the latest available to avoid being caught out by this. There is also a few beta clients up to version 3.14 available here to try. A relevant thread can be found here at the Folding Community forums.

Congratulations goes out to badg3er for getting the team some publicity in the latest issue of the Bulletin magazine out now, the article can also be read here online. The story was well written and explains a fair bit about what Folding@home is about and the culture of the teams that participate in it and how we are all dedicated to the cause and so competitive with other teams. It also includes a few quotes from other members that submitted their own story about how folding got them hooked, well done Stephen.

Thursday 29/08/02

Another tweak to the stats bar with the inclusion of member production section which shows how many people are producing points equal to the different folding clubs. Also I have doubled up the first frame to include the PPW and the PPD numbers, to view the PPD click the link in the bar above the frame.

The article on the Bulletin's web page yesterday didn't include a link to Overclockers Australia web site or to here the folding page. So we need to write a quick message to the editor extolling the virtues of folding and a few comments about the article including a URL to the folding page and/or to the main OCAU page. Here is a post from the forums by the reporter that wrote the article...

Greetings - I'm actually the author of the Bulletin article on folding. My apologies for omitting the Overclockers URL - a bad oversight I confess.

Could I suggest - albeit with no promises - that a short letter to the editor in response to the article that includes the URL might help to rectify things and attract some more folders. (Of course the more mail, the more chance one or two get published.)

The e-mail address is bulletinletters@acp.com.au.

Thanks to all of you who assisted me, and keep up the good work.

Regards,

Paul Syvret

Project Dolphin has taken off like a rocket with 106 members madly typing away. Our rank has gone from 423 to 242 in just 4 days. Pyro is keeping track of our progress in this thread.

The Milestone Club main page has had a make over with a spiff new Milestone selector.

With the Folding Club main page I have put the badges next to the club numbers and if you hover the curser over the badge it displays the correct forum signature, colour and hex number. Two badges are missing ATM (375 and 750), I have to get on the Darkness to whip a couple up for us.