New A2 Core work units

S

SmokeRngs

[H]ard|DCer of the Month - April 2008
Joined
Aug 9, 2001
Messages
18,291
It seems some new A2 core work units for Linux have been released. Currently running 2 of the 4433s worth 893 points each and I do not like them at all. They are running on 2 cores of a [email protected] and only doing between 1700 and 1900 PPD. From the look of things, they aren't optimized very well considering they are only using about 70-80% of the processing power available to them. I find this very sad considering the previous work units used almost 100% of the processing power and did a much better job overall. :(

 
Maybe they needed a bit of real world to optimize?

Ohhh I can hope can't I :(

 
You guys with the Linux SMP are like the beta beta testers. Give it a couple days at least. They might give them a points bump or find something like they did with those 430 point GPU units.

 
Wheresatom said:
You guys with the Linux SMP are like the beta beta testers. Give it a couple days at least. They might give them a points bump or find something like they did with those 430 point GPU units.

Click to expand...

The points for the work unit aren't really the problem. It's the horrible use of resources that is a problem. As an average, it's only using 75% of the two CPU cores. Get that back to 100% or close to it and the PPD would jump by the same percentage which would put it at or better than the 2605 and 2653 work units. As those are the best A1 core work units to get, I don't see that as a bad thing.

The funny thing is that the Linux SMP client is at least 10 times more stable than the WinSMP client.

 
SmokeRngs said:
The funny thing is that the Linux SMP client is at least 10 times more stable than the WinSMP client.
Click to expand...
It's very stable but I have frequent hangs with the A2 WUs after completion, already mentioned in a couple of recent threads. There doesn't appear to be any solution in sight from what I can tell either. So, if the new WUs turn to be very poor PPD-wise I will just revert to folding only A1 units (if I can). I'm shocked that Stanford hasn't cone up with a solution to such a significant problem. It's easy for them to release new WUs by the boat load, but fixing what would seem should be a high priority issue is less of a concern. :confused: :rolleyes:
 
APOLLO said:
It's very stable but I have frequent hangs with the A2 WUs after completion, already mentioned in a couple of recent threads. There doesn't appear to be any solution in sight from what I can tell either. So, if the new WUs turn to be very poor PPD-wise I will just revert to folding only A1 units (if I can). I'm shocked that Stanford hasn't cone up with a solution to such a significant problem. It's easy for them to release new WUs by the boat load, but fixing what would seem should be a high priority issue is less of a concern. :confused: :rolleyes:
Click to expand...

I'm wondering if the new A2 units aren't going to hang and that's why they have released them. I guess I will find out tomorrow.

And I'm off to bed as it's well past my bedtime.

 
I also got that 4433 unit and it frankly stink in terms of PPD. We'll see if they will adjust the points since it's obvious they benchmarked it wrong.

 
About the only good thing I've seen about the 4433s is that the upload is a small fraction of the other work units. However, it still seems to like to hang after the upload is finished. I don't have any more of them right now to keep my eye on so we'll see what happens to them in the future.

I've actually gotten a bunch of 2668 and 2669 work units instead. This should help in the PPD department as long as I catch all of the ones that hang after uploading.

 
The only solution I've found for the hang is a restart :(

 
Xilikon said:
No, just find a struck FahCore process, kill it with kill -9 pid and it will resume fine.

Click to expand...

This is what I do except I have the GUI running so I just load up Gnome System Monitor or KDE Sysguard to kill the stuck A2 core process(es). Normally it's just one stuck process but I have seen up to three of them.

 
Here are my stats for the 4433 WU:
E2160 @ 3.0 GHz: 1,872 PPD
X2 4000+ @ 2.1 GHz: 990 PPD
 
Eek :eek: I must be living right (nahh...:rolleyes:) I just configured a E6600 to "dual boot" between WinXP 64 and Ubuntu 64 v8.04 (only a 80 GB SATA). After getting the WinSMP to work (in other words forget the .NET thing and upgrade to something called NetFx64), I did the password thing and downloaded a 2665 WU). Seeing I prefer the stableness and easiness of LinSMP after letting the WinSMP do a couple of steps I Ctrl + C'ed it and loaded up the Ubuntu 64 v8.04 CD. After configuring it like I wanted it (ie Gkrellm, lm_sensors, Tweak Ubuntu, etc). I did the 32 bit libs thing (so I can run 32 bit programs under a 64 bit OS in Linux 64, kinda' like the NetFx64 in Windose 64 bit), the libwxgtk2.8-dev, wx-common and the g++ thing for FAHmon v2.3. 2b (love that FAHmon :p) I then proceeded to DL a 2669 A2 WU and away I went. I don't even want to see any of those WU's mentioned above (1700 to 1900 PpD for a OC'ed quad) :eek: :( :mad: :eek:

Folding and WCGing for a CURE

 
capreppy said:
The only solution I've found for the hang is a restart :(

Click to expand...

I don't run them as a service. When (not if) the client hangs a simple ctl+c followed by hitting the up arrow key (to bring up the run command), then hit enter. Works every time although baby sitting clients is less than exciting.:rolleyes:

 
BillR said:
Works every time although baby sitting clients is less than exciting.:rolleyes:

Click to expand...

I've babysitted two clients within the last hour and I'm waiting on another one to finish within the next 45 minutes so I don't go to bed and have a bunch of clients waiting for no reason at all. I also have another one which should be done in about an hour and a half. I might stay up for that one or I might not. These hangs are not fun.

 
Strange enough, those hangs is very very very rare for me. I get 1-2 per week at most and I often catch it when I get up in the morning.

IcedEmotion, the hang at end is the same for A1 and A2 with any WU. It's good to know what to do in those cases since a restart is laborious.

 
Xilikon said:
Strange enough, those hangs is very very very rare for me. I get 1-2 per week at most and I often catch it when I get up in the morning.

IcedEmotion, the hang at end is the same for A1 and A2 with any WU. It's good to know what to do in those cases since a restart is laborious.

Click to expand...

I just use the kill -9 pid you mentioned in another thread and it seems to kick things back into gear. I was commenting on the point value though, not the hangs; I remember when 2665 first started coming out everyone said it was benchmarked wrong but Pande said it was supposed to to be that way... I hope it isn't the same thing for these new A2 WUs.

 
IcedEmotion said:
I just use the kill -9 pid you mentioned in another thread and it seems to kick things back into gear. I was commenting on the point value though, not the hangs; I remember when 2665 first started coming out everyone said it was benchmarked wrong but Pande said it was supposed to to be that way... I hope it isn't the same thing for these new A2 WUs.

Click to expand...

Actually, the 2665 was I think originally worth around 1500 or so points and the PPD on them were absolutely horrible. Stanford went back and "rebenchmarked" the unit to 1920 points. The PPD is still horrible but it's not as bad as it used to be. As it is, the 2665 isn't a Windows only work unit. I've gotten plenty of them on my Linux systems as well.

 
Yeah one of my linux smp vms is doing a 2665 right now for a whopping 1400ppd. Thats about 1200 ppd less than the a2 stuff.
 
Ya, I've had the odd 2665 on linux as well but it's a rare occurence. With Windows it hits 90% of the time for me which is why I normally think of it as a Windows WU.

What's the deal with project 510x? I have two crunching now and they're only getting around 1300 ppd down from 2000-2600 on other WUs.

 
You must log in or register to reply here.
Back
Top