Msg from VJ

Here's a couple regarding the v7 client on linux.
It looks like the client is hard coded to run at least a screen res of 1024x768. My linux boxes only run 800x600, it makes configuring the client difficult at best, can we please have the client auto detect the screen res and resize itself accordingly.
The config file seems to store its self in several places and keeps all the old ones. My experience it is difficult to know which file the client is actually running from.

 

Here's one on v7 on windows, I think this has already been raised over at FF but am not sure

The GPU whitelist seems to take forever to be updated and should not need a download of a whole new client. Is it possible to split the list off into a seperate file that can be downloaded as needed.

 

Proper GPU support in Linux would be another one

 

From what i can gather, the only way to run any GPU's in Linux is under wine, even then it doesn't always work. I'm not sure but it seems to me to be a core issue as I think drivers exist for Linux.

Native support would rock, especially if you can get no hit on SMP processing

 

I believe he is asking for an actual Linux GPU client, native support in the V7 client, so we don't have to use WINE anymore which is a PITA to setup and does not support ATI GPUs.

 

Here's one on cores.

A3 runs at the same speed on both Linux and Windows. A5 doesn't anymore, Linux is way faster - is that going to be fixed or is it a windows problem?

 

Are they going to reconsider their decision not to do a 64-bit Windows client?

 

I'll second GPU3 on Linux. I tried running in wine with my pair of gtx460s, and the performance was horrid when it worked at all. Despite the increased efficiency of the Linux A5 core, I actually lost PPD going from Windows to Linux because I had to shut off the GPUs. Since I have to run in Linux more often lately for non-folding stuff, I'd really appreciate a supported way to run those.

 

^^this

 

s/client/cores/

^^^ this, actually

I've got a few items, will post weekend time.

 

What DooKey said...I have machines that can run big bigadv, but since I need Windows, it's not a possibility at this time.

 

To me I've seen a lack of bigadv units the last 2 weeks. I keep getting 6099's on all my big adv rigs in windows, rarely got any 6900's, but a month ago I was getting constant 26xx's. I run v7 client on all my boxes and have seen dependency issues crop up on 2 where the fahclient will not load.

 

I had the same problem on v7 last week. I loaded up v6 and immediately got 6903 on both rigs.

 

I hope thats not an indication that they are thinking of a points nerf again. Perhaps they could make some of the larger SMP projects "bigadv" so that marginal systems that would struggle to complete them in time don't get them and they get done by the bigger iron.

VJ has to bear in mind that points are our only "visible payback" on what can be expensive hardware. To nerf the points again when there are machines costing thousands being built for this purpose would not bode well for the project.

 

Is AMD GPU folding ever going to be competitive...ever???

last I checked even a single client just cripples SMP, cuts it off at the knees, is this just the nature of the architecture and there is no fix??

also +1 on recent lack of -bigadv in v7

 

I thought I saw someone mention on the folding forums(Kasson?), that they were doing some server moves during the next 2 weeks....so bigadv might be scarce from time to time.

 

Instead of nerfing bigadv again, it seems they're increasing SMP points. This makes sense if they want an increase in SMP WU's folded. My old C2D E8200 went from ~2500 PPD to ~5500 PPD when folding these new WU's: 7000 through 7028.

 

What projects are simulating what diseases? Can I dedicate my contribution to Alzheimer's (in my personal situation) or any specific cause? This was one of the majors when I started.

 

I've got one for the V7 client. Please fix ticket #395 already. PPD reporting has been unreliable and, in many cases, doesn't work at all. This feature has never worked as intended.

 

64 bit windows

Better ATI performance

Did the v7 log ever get cleaned up?

For the love of god, no more points nerfs

 

No on the logs and ATI performance sucks ass still. Just fired up some core 16 WUs on two unlocked 6950's... trying to figure out this roller coaster GPU usage I am seeing.

 

Then they need to prioritize SMP appropriately. IF they have enough doing BA then bump up points for SMP. Simple as that. Please pass that on and don't filter it or give us opinions. Your position is expected to pass on our concerns and their answers. Sorry to be so blunt, but that's the simple facts.

 

Ok, when you said "I don't think" it set me off. Plus I can't stand PG after the nerf for no good reason.

 

Kendrak, don't take this wrong, but don't give them an excuse with your question.

It would have been better to ask them straight out.......Why are you not doing a 64bit Windows client since so many people are running 64bit windows now compared to 64bit Linux?

 

I disagree completely, Kendrak's tactful language and professional writing is not only the best approach, but the only one that will see our opinions heard. If you were to storm in there accusations flying and e-peen pissing all over the place like so many of the keyboard cowboys that frequent tech forums, you would only earn their scorn and disregard for your message.

Thanks for taking the initiative on this Kendrak, my feedback for Stanford is the following:

-Please don't ever nerf points again. If you must devalue a certain work unit class, do it through inflation of the point 'currency' by increasing the other work unit returns.

-When will the team release the A5 core for OS X clients? There is a growing percentage of Apple hardware out there but most of those users will not run a VM just to fold SMP.

-What can be done to improve the Windows client's efficiency deficit compared to Linux? Regardless of if we get the extra bigadv units on windows, the ppd delta currently applies to SMP and normal bigadv as well.

 

I am pretty sure the 32bit Windows client only refers to the client itself which is the v7 GUI interface. I do believe the 32bit Windows client chooses the 64bit core which is what does the folding, if the WU has a 64bit Windows core available. I believe that is what the developer stated in one of his post on the FF when he was answering a question as to why there was no 64bit Windows Client.

Edit I found the post by the devoliper over at the FF and below is a quote from it.

http://foldingforum.org/viewtopic.php?f=67&t=19648&start=60#p195905

 

I hate to muddy the water, but I oppose point inflation, it devalues historic point bases built up with great effort and expense on older hardware.

I really respect the black and blue folders that are in the top thousand...that took some doing.

 

Notes from an old curmudgeon...

I was here before the advent of bigadv. bigadv was, and still is, an experiment and never considered to be a permanent fixture on the folding@home scene. There was never, ever, a guarantee that points would be consistent, stable, or any other noun you want to impose.

Recognizing how crack addicts must feel, I am strongly beginning to believe there are many too wrapped up in points and maintaining their status quo on the various top e-peen lists. Within reason, given their financial investment in this hobby, I can understand their concerns when there is a lack of bigadv work units. However, these people need to understand that bigadv is really a niche aspect to the research that F@H does.

The shortage that some are now experiencing with bigadv units is a result of Pande Group prioritizing where they need the work the most and right now that is on SMP work units. Vijay has stated this himself so please do not lay blame on Kendrak or anyone else. If there is no work available for bigadv, there is no work available. Plain and simple. Yes, you invested significant financial dollars to enable bigadv calculations on your rig, but no one ever stated bigadv would be a plentiful, and permanent, fixture.

Going forward, and rather than more points nerfing, I'd almost prefer seeing the thread restriction on bigadv increase from 8 to 16 on -bigadv. 8 threaded CPUs are extremely plentiful right now, for fair prices, and can crunch through some SMP units very quickly.

On Stanford's end, points for SMP units need to increase a bit. We really don't want/need those with silver spoons stuck in all orifices to run away with all of the points. I believe BA points are fine where they are but SMP could use some loving and buffing.

Anyway, these are a few my ramblings after listening to several over the past few days. Thanks to Kendrak for picking up the DAB Chalice when I unexpectedly left earlier in the year.