Bigadv base bonus points slashed

Guys i have a question here:

Politics and conspiracy theories aside, as this is a global change for everyone and not just "a team" or "a country" then does the playing field with regards to the acquisition of points stay the same? Am i correct in saying, that even if a WU is worth less points, the actual folding of the protein remains the same, and the scientific result the same?

If this means that you will be getting 50k PPD instead of say 75k PPD then as long as everyone else gets the same, what's the problem?

I can see how this would be annoying but does it really matter that much? This is of course IF there are no corporate influences etc...

The problem is that FAH is growing in popularity and corporate financial influence is inevitable, isn't it? Money (unfortunately) does make the world go round.........no matter how you look at it.

 

Absolutely right mate. My frustration is more about the handling of the change and what it means for the future, than the change itself. It was only a month ago that they introduced the uber bigadv with silly quick return bonus points which I mistook as an affirmation of the importance to the science of the speedy return of bigadv. That would have been a golden opportunity to make a global change and signal that there was too much focus on bigadv by making the points somewhat less outrageous.

Not all teams are structured the same. Some are still quite GPU focused whilst others have built numerous monster 4p rigs. Changing bigadv points in isolation will inevitably favour some teams over others.

But again, it's not really about the competitive element of points so much as being guided by the point structure towards the science that PG feels if most important. PG seem to have flipped their view of this within the space of a month. I don't expect things to remain the same but a modicum of appropriately caveatted discussion would prevent people from purchasing the wrong kit. We may be thinking that we're helping when potentially there is an over demand of a particular work type and therefore we're helping substantially less.

Also, you're right, there will inevitably be corporate influence in all endeavours. We'll typically distance ourselves from it when we find it distasteful.

 

Whilst all research is valuable, they must have priority WU and projects that they wish/need to focus more attention on. My thoughts would be that they already have this with -bigadv but should review and design the WU according to where the folding horse power comes from. I.E. We all know that GPU's can utilise certain mathematical architecture a CPU cant. It is more likely that "Joseph Bloggs" will have a GTX 470 or a 5970 etc.. in his PC he wishes to use for folding than a 990 extreme CPU.

My point is here that they should bolster the modest folders with larger point gains so that they feel more competitive BUT NOT slash into the extreme folders such as yourself who take this project a lot more seriously than the average Joe.

If they embrace the competition that has been created and encourage it (without deviating from the true purpose), then it's a WIN WIN for all parties concerned.

Basically, level the playing field but still allow the true enthusiasts who invest heavily in the project to shine through in the spirit of competition.........

BLOWING UP THE PLAYING FIELD SO THAT NO-ONE WANT TO PLAY ANY MORE HELPS NO-ONE AT ALL!!

 

Glad I saw this thread. I had been toying with the idea of an i7 980 (non extreme version) as they have just come out and look like they would fold the large bigadv.

I gave up folding a while ago because of the cost of electric but the upgrade bug was starting to get me again and the new wu were a good excuse to go for it. They may not be such a good excuse anymore as I can quite happily live with my current rig for everyday use if I don't fold.

I shall put my itchy wallet finger on hold for a while to see how it all pans out but I for one will definately not ever be folding using a GPU

 

Every last single folder is equally important and I would hope that Stanford shows the same disdain for us all without distinction The vast majority of our and most other team's output comes from the regular guy not the 'big guns'. [H] is just about the only exception to this I can think of, with most of their production coming from their 'top 20' producers.

Compared to many who can only run the standard bigadv, the change still leaves me in relatively good shape with a couple of SR2s that can run through the 6903/6904s for more than standard bigadvs were awarded before the change. If the whole thing was handled right, there was a degree of transparency and there was a plan to prevent a recurrence I really don't mind being slashed into

Noooo. I really wouldn't want some venting to discourage anybody from returning to what is still a project with extremely worthy goals.

I'll try to shut up now. I've let off steam, calmed down and life goes on.

 

I would say that even after the reduction they are still are a good upgrade for folding, you'd get the new supersized bigadvs that require 12 threads, at 4.2ghz they do around 90K in Linux, 60K for the normal bigadvs.
Though Stanford have proved that they can make drastic changes seemingly at the drop of a hat with little or no regard for their most loyal, enthusiastic folders.

 

What have you bought? and more importantly what procs are going in it !?

 

Indeed, but I've still not had a P6904

 

Are you using -bigbeta? they do seem a bit thin on the ground recently, one of my hex machines has done 4 normal ones in a row now, tho' my other is dropping a 6904 in a few hrs, a day late though a combination of a bad OC and children turning the computer off

 

I bought a supermicro board and 4 low power 8 core opterons (6128HEs). The intention being to play around in preparation for potentially dropping in Interlagos chips, or discounted 12 core opterons, when they're released. The chips cost me just over £500 altogether and, if everything goes to plan, should come in a bit below SR2 performance.

From his recent post over at ff, it looks like Dave_Goodchild is underway with multiple, and no doubt, more powerful versions of the 4p approach.

I got some last week but haven't picked one up since Friday. My last 6903s where sent today so just regular bigadv for me right now. Having a plentiful supply of these would have lessoned the blow somewhat and I could again comment on planning but .....

 

Where did you buy the board from? I've been considering something similar myself, did you notice the 4 x Opteron 6176 SE on ebay recently? I considered putting in a silly cash offer as I knew who the guy was selling them and he was in the next county to me, but I chickened out, seemed too daunting, I had no idea where to buy the board from, no idea what case to use, just generally no idea about that kind of machine.

 

Yes, using -bigbeta, and getting P6903 WUs with no trouble. Just the luck of the draw I guess.

 

I've had them both and according to HFM the p6903s give a slightly better ppd.

 

No it's not that, I have a few things that I am wanting and can only afford one of them. The never using a GPU thing is I am an ATI/AMD graphics person and don't really want to be dropping any cash on nVidia cards. Plus I like tinkering with Linux although when it comes to the folding I'm still a novice so that sorts of fits in with the SMP, bigadv route.

I would not have been overclocking it so would have left it at 3.33 GHz as my room is boiling hot in the summer. I don't know what points that would get. Maybe in the winter but........

 

On the points change issue - it's happened before, it will happen again. The advice from Stanford in the past has always been not to base any builds on current WU types, as they change. It's something to bare in mind .......

On the upgrade discussion - I'm in the process of upgrading a number of my rigs, partly because I still Fold on quite a few GPU's and the leccy bill has been getting a bit out of hand (that's an understatement ), and partly because I still have Q6600's and CPU's have moved on a bit since then .... hex cores seem like a nice step change improvement ...

 

Yes, I saw them too and guessed that it may be ocuk's Baba disposing of his chips. After splashing out a bit too much on the SR2s I thought I'd be cautious and experiment with less expensive processors. I'm expecting prices of the 12 cores to drop soon and if I bought fast chips I'd just end up building another rig when Interlagos is released

I bought the board from here. It's not cheap but the least expensive I could find in stock when the blood rush hit. 'In stock' can be a bit of loose term with this supplier as it's on a slow boat from somewhere in the Far East. Memory is unbuffered ECC from the same place for the same reason. I think I could have bought non-ECC with lower timings but didn't want to take the risk.

For a case, I'm afraid I can't help there as it's going to run naked using some plywood, plastic spacers and screws. The heatsink are relatively cheap 2u supermicro jobbies until I can find my preferred heatsink, the 92mm Noctua, in stock somewhere. There seems to be plenty of the 120mm around but not the 92mm.

It's all a bit new to me and am keeping my fingers crossed that it all comes together OK. I'm afraid that I have the technical nouce of a cucumber so best not to make any decisions until this configuration is up and running. Hope this has helped.

Guess so. Just picked up a 6904 this morning.

This isn't aimed at you Doc but rather Stanfords standing advice.

Although I'm aware of the Stanford advice, we have to base our upgrade decisions on something. For instance, you may be purchasing hex cores whilst I'm going to be decommissioning them. My logic is that there may be a shortage of 6903/6904s and the next move may be to increase the core count required for these. Who knows? We can't both be right and some transparency from PG would really help us out.

I'd hoped that as the project matured that it would become more professional and not just use their standing advice as licence to keep projects in 'beta' for over two years and then conduct overnight changes. If a fire station can forecast how many fires there will be to get the number of vehicles right, surely PG can discuss, with caveats, what changes they've got in the pipeline and what they'll be focusing on next. Any successful enterprise has to be able to effectively forecast and plan. It's just that we tend to forgive the Project's shortcomings because they're a group of scientists working for a good cause.

 

Well said, that man!

 

Not folding ATM as it gets too hot and needed to give leccy bill a bit of a rest. Is there not "some" over reaction here? Folding is a lot of fun, it lets us (all in the name of a good cause) learn about extreme overclocking - if an OC is folding stable - its stable! Push hardware to the limit, discuss configurations, work out best practice etc etc.

It seems to me that all Stanford seem to be saying is - we have a project that is active on numerous fronts, we wanted to encourage people into one of the areas - this has been done more successfully that they wanted, and they are now carrying out the digital equivalent of a cull hoping resources get re deployed.

Some of the huge contributors on here have made significant hardware investments because they love folding and they want to max out the points for out team - which is brilliant, but these changes will affect everyone so its still going to be a level playing field.

Maybe its a lesson in not putting all your folding eggs in one basket? A balanced approach is maybe the best approach?

Anyhow, only my 2 penneth. Looking forward to folding again in the winter

LP

EDIT - Just seen your daily points phoenics, insane! - well done.

 

Most folding rigs are not overclocked as folding is probably one of the biggest stress-tests possible (apart from Prime95) and any flaw in an O/C will run the risk of the mathematics being compromised by a single miscalculation. If you want to benchmark an overclocked system, run benchmarks like the CPC Benchmark Suite, 3D-Mark, SuperPi etc.

IMO the reaction is justified, people have spent a lot of money building specialist folding rigs that are designed to give the best PPD possible. Stanford have moved the goalposts with little notice and immediate effect, giving people no chance ro react to the changes. Now the previously mentioned systems are no longer what the creator intended, which is especially annoying if that was the sole purpose of the build.

Thankfully I'm not one of those people (98% of my points are from spare work kit that is being stress-tested or burned in or is simply unused) but I can imagine what it would be like to spend a vast amount of cash for a system that is no longer what it was designed to be.

 

[Not a rant, more a quiet explanation] The problem (a problem?) is that scientific research is not a straightforward journey from A to B, which can be planned in advance, nor is it like a fire-station where you have good, relevant figures from the recent past. It is more like a maze in that you don't know which route will get you to your destination, or which are dead ends. Some routes may be easier going than others, but have unexpected pitfalls lurking. Plus when you reach point B you may very well find that despite all the available evidence when you started, being there (at point B) is not helpful for the overall cause. Or there may be several different point Bs. An earlier post referred to the PG group 'making it up as you go along' but that is the best you can do. A number of the projects are designed to gain knowledge of the most efficient ways of doing the calculations, and still end up with computed results that match the experimentally derived, real-world results. Not all the proteins in PGs work-units are directly related to the diseases mentioned ~ the project is aiming at methods and understanding of the underlying reasons for misfolding.

And you have to remember that the PG are not working in isolation, other scientists are working in the area of understanding protein folding. So if an important result is published in the scientific literature, priorities may indeed change overnight, and then may change back later. Or go off in some totally unexpected direction.

Also, please bear in mind that even the PG group failed to imagine that donors would start to build machines purely for folding, let alone the multi-core and multi-GPU mchines that now exist. You can not predict the appearance of someone like Atlas_folder or James_ (our recent weekend stress-tester). [/Not a rant, more a quiet explanation]

And I speak as one who has a dual-Xeon machine, and 980Xs, and GTX 560 Ti s, even though I don't run them all 24/7