Folding Forum

Dr.G wrote: Am thinking of trying a Linux system, but will take time... any advice appreciated, thanks!

This is off-topic so the mods may want to move it.

Based on my own experience only, you may have more stability problems with Linux, and there's no equivalent to Afterburner or its ilk. Just the NVSettings utility.

WUs run faster- I get about 20-30% higher PPD on smp WUs (at cpu:3) with Linux. I can't give an equivalent figure for GPU folding because it won't fold under Linux at the same clock as I use for Windows, but as an example:

Windows (Vista), GTX 780 Ti, factory o/c'd to 1020MHz plus another 32MHz by me- PPD on a P10467 approx 250k
Linux, same PC, same card but underclocked from factory setting by 100MHz, same project, same PPD.

And no Core_15/16's

No Core_18's either of course, but hopefully they'll come in time.

New AS sends out older WUs to ALL types of hardware, not just your newer hardware. Having newer hardware doesn't make your computer any more entitled to one type of work unit or another.

And why would the newer hardware be especially susceptible to stability issues? Your new fancy hardware should rip that stuff up, not choke on it. Maybe it wasn't so high end after all.

@Dr.G

Just to be clear, I'm not asking why so many Core 16's are being sent out. There are other threads that are though as you have linked to. What I'm asking is why can't users exclude a Core in the client so when we run into situations such as you and I have, we can avoid having to dump a bunch of WU's that we have no way of completing.

As 7IM has pointed out, it isn't as easy as it sounds and people will attempt to exploit it which I agree would happen. However, people will only exploit something if they think there is an incentive to do it. Remove the incentive and the exploitation will go away. So I don't consider possible exploitation to be a reason not to provide users an option. Some Core's are clearly going to give some users serious issues on their hardware. Users may or may not be able to find a work around. If it is a driver issue then those users are stuck between a rock and a hard place. I just don't see how improving the user experience even if it is a small number of users overall is a bad thing.

This isn't just an issue with Core 16's. Surely this isn't the first time this sort of thing has happened and it won't be the last.

I would actually like to see a response from PandeGroup on this because I don't think the request is unreasonable.

Update on status:

I've had very few issues since my post a day ago. I did have a few Core 16 WU's that wouldn't get to 1% which I had to dump. I've had mostly Core 17 WU's since which has lead to essentially a stable environment. For the most part it appears my machines are getting Core 17 WU's though.

ChasingTheDream wrote:I would actually like to see a response from PandeGroup on this because I don't think the request is unreasonable.

Is it more reasonable for every donor to take responsibility for knowing which projects cannot be run on their system and managing them 24 hrs per day or is it more reasonable for the Pande Group to assign only projects that your system is able to run?

In either case, mistakes do happen, and based on today's posts by Dr. Vijay Pande, himself, in another topic he seems to feel it's a mistake and it's his team's responsibility to fix those problems.

Even if every donor used the capability you're asking for responsibly, which is doubtful, adding that option to the client would take weeks, and even then, not everybody would install the new client, whereas fixing it on the server should be possible in a day or two.

@gwildperson

If the answer is to have each donor babysit their folding machines 24x7 then I suspect I'm in the wrong place. lol That is a level of dedication that I don't even put into my full time job.

It may take months to put what I'm suggesting into the client. What I'm suggesting wouldn't be implemented in time to deal with the Core 16 issues I was having but I would bet money that this isn't the first time people have had issues with a specific core whether it be an old core or a core that is new and just released. I believe donors should have a way of dealing with a problem such as we just ran into besides dumping WU and waiting to see if we can get the next WU through on whatever core is giving us a problem.

People will upgrade their client when they need to. If I ran into the problem I just ran into and saw that a newer client was available that allowed me to exclude a core then I think it is fair to say I would have an incentive to do the upgrade.

Just a quick update on the problems I was having with my machines and Core 16 WU's. It's been three days since any issues have come up and I've seen at least a dozen Core 16's run on the various machines. I've made no changes to the machines what-so-ever. So unless I'm now getting some better behaved Core 16 WU's as Bruce had suggested before I have no other explanation as to why the machines are now running and before they simply wouldn't.

My rig has also settled down, with no crashes or faulty WU's etc. Twin R2 290x's only receiving core 0x17 now and all running smoothly (touch wood)... phew no more core 0x16's. In my humble opinion, the settings on AS seem to be much better. Thanks Vijay.

Dr.G wrote:... phew no more core 0x16's. In my humble opinion, the settings on AS seem to be much better. Thanks Vijay.

Yes, I am not getting Core 16 now either. They caused one of my HD 7870s to hang up for a day, and I don't even dare use my HD 7790s when they are around, because they always crash.