New, more stable WinSMP client in open beta

[uncle_fungus]

Read the announcement here: http://foldingforum.org/viewtopic.ph...p=14926#p14926

Make sure you read the instructions carefully.

[Slade]

OOh, thanks for the heads-up.

I've installed it, which was easy enough, and now it's running.
I don't know whether the FAH process was causing excess drive thrashing at all, but it's stopped with the install of the new client.

I'll keep an eye on it over the next few days.

[uncle_fungus]

One thing to note, this new client does NOT work with x64 Windows.

[Cdr Zero]

Cheers for the notice, running now.

[mitsimonsta]

Excellent.. been waiting for this.

Unfortunately I find out about it at work!! No worries, will swap it onto one of the quads at lunchtime.

[chestRcopRpot]

Mitsi when you test it can you report if running 2 instances will interfear with each other? nothing worse than closing 1 client only to have the other one hose the wu at 95%

[mitsimonsta]

I will replace BOTH instances on one box, and compare to my other quad still on V5.91

Frame times, PPD % increase/decrease, and general stability reports coming.

[mitsimonsta]

I have just installed two instances of WinSMP V5.92 on my 'Enterprise' box. Seemed fairly painless... couple of possible traps for young players. Full two-instance install instructions coming later tonight... not too hard really.

EDIT: TO INSTALL TWO INSTANCES: Install first instance as normal. DO NOT FIRE UP FIRST INSTANCE YET.
Install Second instance, when running install.bat, you need to cancel the DeinoMP install. Do not create a new credential store. Add your username and password. Then fire up the first instance. Once it is crunching, fire up the second instance.

Only just got things running, so no frame times yet. Folding Forum users are reporting 60secs/frame slower on P2653, but I would save hours and hours of downtime from MPICH stuffing up. I expect that the 36XX units may run faster however.

[mitsimonsta]

Losing 50 secs/frame on p2653 proteins, dual instances, on a Q6600 OC'd to 3.6Ghz, Memory 800Mhz @ 4-4-4-4.

This is a loss of 170ppd per instance.

[Azrael]

Hm, maybe i wont upgrade it then.

[mitsimonsta]

I think I would gain 300-400ppd with the client actually shutting down properly after a unit, and firing up again on a new unit in any case. I often have a machine sitting for 8 hours (while I am asleep) sitting on a FILE_IO_ERROR.

Losing 250ppd on the 3.2Ghz clocked C2D's.....not so good. Anyway, will get them swapped over and then check how my average is heading in a few days time. If I start averaging 14Kppd instead of 13Kppd, then it's worth it.

Just the E6420 and D945 to swap over now.

[chainbolt]

This is an *early* beta release--please expect some rough edges.

No, thank you.

I am getting increasingly fed up with the constant trouble and their incapability to fix the problems. Just lost again of couple of SMP WU because of client communication errors, a non responding server, or whatever.

[mitsimonsta]

Originally Posted by chainbolt

No, thank you.

I give it the thumbs up Chainy. Slight loss of ppd for alot more stability.

Upgrade instructions in other thread for dual instances.

[Azrael]

Yeah, i lost two SMP WUs today. Im debating upgrading on one of the machines. Since i use MPIEXEC for other stuff on my work desktop i think ill leave 5.91 on there, but put 5.92 on the other machine.

[mitsimonsta]

Well Az, keep your main box running as is, migrate another one to the new version. I think you will be pleasantly surprised.

It has been rock solid since installing earlier, and that was with a WU at 95% on one box. The end of the unit was quick shutting down, so it was into the next unit much faster. I think it tends to load the cores, decompress the WU and start crunching faster too.