Folding Forum

BigAdv would be great to have again! With the new AMD Zen line sometime hopefully next year, it could be like a nice server chip, and I'd hope it to be reasonably priced (something that doesn't exist yet...). I've searched the forum but still do not understand why it was shutdown, but I don't want to focus on the past, let's all just move forward and see if we can make something cool for dual / quad socket CPU users (with high core counts).

How may CPU cores are you thinking about?

Many of the proteins which are amenable to being run with a few hundred parallel threads up to several thousand are configured for GPUs without also being ported to high-core-count CPUs. My guess is that projects that run on over 16 (3d or 4d) threads could be created but there probably wouldn't be enough donors with that class of hardware to make it worthwhile. FAH is targeted at the hardware owned by the "@home" class of donors. That tends to include quite a number of under 16-cpu systems plus a lot of GPUs of various classes.

bruce,

Presently, I run 28 threads. I am now only getting project 7528 for this configuration. I would seriously consider upgrading up to 2 E5-2699V3 for a total of 68 threads (4 threads reserved for the 2 GPUs) if this opens significant benefits to the science.

Lets not even go there, Although I have 3 bigadv capable machines and would personally like to see more projects, the angst it caused amongst donors is not worth the hassle. Whole teams tore themselves apart and many donors quit in the aftermath of the bigadv shut down.

Any new bigadv projects should only be considered once the long rumoured avx core is launched.

@ricky, if you leave the bigadv settings in place, -bigadv -smp on a v6 client, not sure on v7 settings, you will get projects 8106 & 8108, these are the current "bigadv" projects that are available

Nathan_P,

Thanks for the suggestion. It seems to be the case where you can only have one client-type setting. I am already using the client-type setting for beta. I have noted that I have not yet received projects 8106 or 8108 since starting HFM a few months ago.

One note to add to the directions for the current large WU's that use the bigadv folding core, they are only available to systems running Linux. Windows and OS X are not supported.

Nathan_P wrote:Most of the 24+ core WU's were generated by Kasson or one of his associates out of his lab, he has lots of projects listed on the project summary using core A3 and A5 that run great on our multi cpu machines but none of his servers currently appear on the server status page

Does anyone know what has happened to the servers, associated with Kasson, which were issuing the 81xx WU's?

They went down last week, not even on project summary any more - none of his servers are, SMP or BA. Hoping for something soon as 115k PPD from 2 12 core cpu's is painful. On the plus side its only 40 minutes per 9032 WU so lots of work being sent in.

I've switched to a single SMP 24 slot on my 48 thread machine, the other is has a fried mobo and the 32 thread rig is running my gpu's 1with only 1 cpu so I'm not missing out too much whilst the servers are offline

The work servers for Dr. Kasson's lab have not been showing up on the server status page for some time, not just since last week. This dates back to an update to the page, his servers are running an older version of the software and were not picked up for display after the update.

As for the servers themselves, I was able to connect to 3 of the 4 I am aware of being there using a browser. But the WS for bigadv units did not respond. No way of telling how many WU's are there on the servers that were up. I have not received any recent word on the projects his lab is working on.

He may have 3 work servers up but they won't assign any work - His are all using core A3 and no matter what settings I try in terms of core count, OS etc I get nothing from them. Maybe its something to do with the older software he's using?

Or maybe it's the version of software YOU'RE using. Projects will not be assigned unless the client meets certain requirements. Which FAH version are you running?

v6 and v7 using smp 24, smp 32 and smp 48 on linux

So you're saying that V6.xx on Linux is no longer getting 48 or 64 assignments. Right?

How about V7 on Linux?

------------

I've been trying to establish a communications channel with Kasson.

V6 won't pick up anything, v7 will not pick up anything larger than 24 threads

So you're saying that V6 has been obsoleted. I'd like to hear that from PG, but it could be.