Bigadv base bonus points slashed

Fair points, well made, but don't you think they can do better? They have already said that they will continue to provide advanced notice of the termination of a project type. That would pretty much be the most difficult decision to predict given what you describe above.

From what I understand, the speed of the decision in this case has come from donor complaints and a sudden realisation that the various projects are imbalanced. These where not scientific imponderables that couldn't be managed. They didn't look at the data until it was flagged to them even though they were launching a new work type with silly points. I think they can do better.

I'm not suggesting that they can predict where they'll be in 6 months time but just that they should provide some insight into where they are now and where they think they're going (again, appropriately caveatted) in the near future. In other words, planning followed by transparency.

And ranting would be just fine Christopher, I've not been blameless myself.

 

I think Stanford could certainly do better at their PR, or should that be DR for Donor Relations. For a project that depends on donors, the amount of explanation and feedback is incredibly poor.

For something like a major reduction in points, which predictably causes cries of outrage, a terse announcement with immediate effect is simply not good enough. I think Vijay should not only mention it in his blog (before it happens), but also put up a video explaining why the change is necessary, and what is expected as a result. This isn't going to require a huge effort and use of resources. And the same should be done as projects approach their end, for whatever reason.

I do understand that the PG can not respond to every message claiming that points are too low or too high. I can see that using a reference machine is a good idea, and gives a certain amount of objectivity to the points system. But recent events have made me wonder just how much the PG are sticking to their own rules.

I'm also not clear how much the moderators on the Stanford Forums can feed back to the PG group, when the Donor community start to grumble.

 

Well I've had it with F@H now. After 4.5 days crunching my last 6904 (downloaded 6 hours before they posted their points change) it was rejected by the server and dumped - what a waste.
God knows why it didn't like it - perhaps because it was going to earn too many points?

Anyway, as PG don't want so many 6903/4's crunched I'm off on another tack for a while until they sort out what direction they want us all to row in - bye bye and thanks for all the bananas.

I made it over 20M points in the last 7 years so that's something I suppose.
Let's hope it helped towards something.

BOINC, boinc, boinc......

P.S. I've got a shelf full of GPUs that PG now say are useless for folding (ATI 4890's) but surprise surprise there are several other DC projects that quite like them - and they won't cost me a penny.

 

I've only once had a "server reports problem with unit" (if that's what you had) I remember I stopped/started the client several times and put it down to that. I've noticed these p6904s work the CPU more than the normal big advs, I've had 2 crashes so far, one last night, and the temps are hotter. Maybe you've a bad OC? or maybe just a bad WU.
I tried seti@home a few weeks back and to be honest it didn't inspire me, but good luck in whatever you decide to do.

 

Well I am donating 50% of my cpu time to Rosetta which is another protine-folding program but from a different direction to F@H, so I am still trying to help humanity cure itself.
I fell out with SETI about 8 years ago, after donating to it for almost 3 years. It was one of those never-ending projects and I couldn't see a real purpose to it (the chances of actually finding aliens that want to talk to us rather than eat or swat us is less than being hit by lightening or winning the lottery multiplied together).

The client had only been restarted once during the processing and it resumed from checkpoint OK. I've never had one be rejected by the server before and it did not say it was corrupted, just that it didn't like it. I am surmising that the WU is tagged with how many points it's worth and it didn't match what the new calculation is. Perhaps it takes the start time of the unit as the last time it started (after the checkpoint) and that was after PG had changed the points. Who knows. I don't really care any more as there is nothing I can do about it as they dumped it. It's not a bad OC - I've been running that rig like that for ages and have had no similar problems - it's water cooled all over and the temps were well within limits. It did not freeze up or anything. Ho hum.

All the best.

Dave.

 

I still feel folding should be judged on WU's rather than points. I guess pushing GPU's is better since you can fit several into a rig and they tend to have a faster return than the bigadvs. I donno why bigadvs take so long or whats going on with the "folding" in each project but if more WU's completed produces faster results i'll be happy. I would also be very very happy if AMD cards or APU's could fold..........

 

I'll be bringing the dual hex core rigs back online next weekend hopefully now I've sorted some additional cooling, they were getting far too hot for my liking once the warm weather hit.

I was also in the process of building a 4p X7560 based system but that is on hold at the moment for two reasons firstly the points reduction has made me think twice and I haven't seen anyone get a 64 thread system running F@H correctly and I don't want to build one to find the points aren't much better than a couple of dual hex systems which would be cheaper to build.

 

You're right, I've seen a number of mixed messages about the success of 64 threads.

A shame though, that would've been a beast.

 

Yeah it might still happen I just need to consider my options, would love to give it a go as it would be a real challenge to get it folding right, but an expensive gamble if it doesn't work out.

Just got one of the dual hex cores back online and guess what was the first WU it picked up, yep 2684!

Was hoping for one of the juicy 6903/6904's.

 

It's a lottery with those darn 2684s plus there seems to have been a glut of them yesterday. I hope you had better luck today.

 

Doing a bit better today, got a 2689, 6900 & 6904, seem to be running about 30-40k down per machine on the last time I ran them which seems about right with the points drop I think?

 

I'm afraid that sounds about right Dave. I'm seeing roughly a 20% drop.

 

They are changing some bigadv scoring again.

From http://foldingforum.org/viewtopic.php?f=24&t=19180 Wed Jul 20:

This will certainly reduce my ppd.

 

It shouldn't make any difference to you, I can make those deadlines easily on my single hex core machines, so it will be no problem for your SR2. Remember it's just the preferred deadline, meaning anyone returning those WU after that deadline won't receive bonus points, returning those WU before the deadline will result in exactly the same points as before.

Sadly it won't

 

I've got a couple of i7 980's (without the 'X') clocked at 4GHz, and they won't make the new preferred deadlines - estimated at a couple of hours too long over the five day period, so that the PPD drops to about 6k

 

That really suprises me, do those rigs run graphics cards as well? what TPF are you getting for the p6903/4?

I currently doing both a p6903 & p6904:



The closest deadline is the p6904 but I still have well over a day to spare. I know mine is at 4.2 but at 4ghz I know I'd still make it.
I use Ubuntu 10.10 (ck patched) and the kraken, what distro are you using? The ck patched kernel in Ubuntu makes a hell of a difference, also if you're using another distro or kernel (non ck) the kraken on its own makes a hell of a difference too.

 

Surprises me too but I'm not particularly familiar the 980. Shouldn't they be as fast as the 980X, just without the ability to increase the multiplier above 25? The preferred deadline for the 6904 is now 5.6 days so with just over 5 you should still be OK. The 6903 new deadline of 5 days is easier to achieve.

On a side note, but related to the events covered in this thread, I've moved over all but one client to BOINC projects for the time being. Although I still think the cause is extremely worthy, my heart's just not in fah at the moment. The straw that broke the camel's back was the silence from PG and the ff.org admin indirectly telling me to go to my room and think about what I did. He's got a tough job and I smile about it now, but still.

It's a shame that CPC bit-tech doesn't have an established BOINC team.

 

Thanks for the reply.
No GPU's Folding on these machines. The TPF was about 81.5mins - about 1min too long. I'm using Ubuntu 10.04 LTS, but haven't done any mods to it.
I've not heard of the "ck patch" - what does it do, and how would one go about implementing it (please note my Linux skills are fairly rudimentary!!)?

@phoenicis - I think the other difference with the 980 is that the QPI is 4.8GT/s, rather than 6.4 - perhaps this has a big effect?

 

it's a patched kernel that is better for folding than the generic one [how/why I do not know], it based on a slightly older kernel than the one in the Lucid repositories [2.6.32-33.33 -v- 2.6.32-33.39]

to add it... bring up the terminal... and do the following...

Code:

sudo -s
add-apt-repository ppa:chogydan/ppa
apt-get update
apt-get install linux-generic-ck
exit [to leave 'root']

then during startup choose the kernel ending in 'generic-ck' [i forget the GRUB fudging needed to auto select the -ck kernel]

not currently on ubuntu so i may have missed something... i think that's everything

 

It uses the BFS (banning material acronyn) scheduler versus the standard CFS (Completely Fair Scheduler). This improves performance on dedicated machines with upto 24 threads.

You can compile kernels and apply the ck patch right upto 2.6.39. Doc, let me know if you want to do this and I'll dig out the link.