RIBOs Woor!

[AlabamaCajun]

I've got 3 of them RIBOs in the house this morning. Two on the 170 and one on my P4. Looks like the AMD rigs have something to feast on. The QMDs are still great but crunching a bunch of those 600 pt Gromys is well worth my AMD investment. :burn: :burn: :burn:

[godofgorks]

gotta love them 600.... been getting squat on on my little duron.. out of commision atm... it's ppd doubles on those 600s

no QMD crunching on that P4 of yours?

[AlabamaCajun]

gotta love them 600.... been getting squat on on my little duron.. out of commision atm... it's ppd doubles on those 600s

no QMD crunching on that P4 of yours?
1141 Ribo and they are slower on a P4 by a few hours. The 1140s are a little faster. My 170 can put one on in about 24-26 hours depending on how much I interupt it by doing system tweeks. These AMD processors have much bigger caches that run less efficient code better. I asked on F@H and Vijay said stanford and the team doesn't have the time and resources needed to refactor the code which my design uses more cache. I've seen the difference between the two chips. QMDs are optimized for Intel which helps them out. If they were to refactor the code, you would see a reduction by anywhere from 25-50%. What they are running is getting the job done, if only I could help. I tried rebuilding the Tinker Source but what I have is in fortran so I'll have to build it on Linux which I need to really get back into. While fortran is good for scientific work, it's probably not for handling huge volumes of data. This s where the object oriented languages if designed and implemented well can take advantage of efficiency. These things are not on the minds is scientists, they are only interested in the results. Efficiency would just get them there faster.

[pscout]

sounds like my next farm critters will be a toss up between new p4's and opterons!

These p4's are just too easy with a steady qmd flow ... we'll have to see what the future holds between some hopefully, code optimizations, and WU workloads :shrug:

[dwschoon]

Ive gotten 3 600's in a row on my tualatin 1.75ghz, whereas before it has gotten nothing but small gromacs. It has been great.

[AlabamaCajun]

Ive gotten 3 600's in a row on my tualatin 1.75ghz, whereas before it has gotten nothing but small gromacs. It has been great.
A Tua at 1.75, lead me to your thread. I have one stable at 1.35!

[dwschoon]

A Tua at 1.75, lead me to your thread. I have one stable at 1.35!

It is a celeron 1300 running 13x135. Im using a powerleap adapter on an abit be6-II v2.0. Cooling is an Vantec FCE-62540D with a delta screamer on it. I have the fastest tualatin in the cpu database, since i got it to windows at 1803mhz. I also have a celeron 1000 that will do 1500 with good memory.

[dwschoon]

Validation
http://valid.x86-secret.com/show_oc?id=61102

[harlam357]

My 170 can put one on in about 24-26 hours

Are you sure? Is that with one instance of F@H or two? My 170 takes ~39hrs to complete a RIBO per core & I'm OC'd to 2.7GHz on each core. I think 24-26hrs is a little favorable my friend. ;)

[AlabamaCajun]

It is a celeron 1300 running 13x135. Im using a powerleap adapter on an abit be6-II v2.0. Cooling is an Vantec FCE-62540D with a delta screamer on it. I have the fastest tualatin in the cpu database, since i got it to windows at 1803mhz. I also have a celeron 1000 that will do 1500 with good memory.
Makes sense, the multi on this one is 9 with the FSB at 133 stock, non Celeron. I had it at 150 and booted to 166. Because it goes to the wife, it's now at 144 are around 1300. Tried to get her a 775 socket but she was not interested yet and I had this one. I put in half a gig and a 120G drive, silent boost copper and a nice quiet 90mm fan. Coolest rig in the house, thats celcius not bling factor.

[AlabamaCajun]

Are you sure? Is that with one instance of F@H or two? My 170 takes ~39hrs to complete a RIBO per core & I'm OC'd to 2.7GHz on each core. I think 24-26hrs is a little favorable my friend. ;)
I want to say it was an 1140 but the 1141s take over 36 hours. I'll have to check the log. This is a Linux rig but I don't see that making any big difference.

[AlabamaCajun]

Figured out what happened that 170 picked up am 1808 PMD core that blew through while core 2 was running the 1140. They do take 40+ hours and that would explaine the bouncing around of scores. The 1141s do seem to take a little longer. Either way the 400+ points per day would be close the same PPD/Day as a P4 on QMDs I think. Glad you caught that but now I disappointed that the Opterons are not smoking the P4s.
Good Catch Harlam! :p

[dmatko]

my newark uninterrupted takes about 34 hours to go through one of those ribos. i had a bit of a set back a few days ago with one. power failed, system shutdown. i tried to restart, found out windows corrupted due to unsafe shutdown, lost a file. did a complete repair, went back in but unfortunately it was a loss. funny thing is it was at 240/250. DEFINATELY lookin into pikin up a backup UPS for safe shutdowns.

[AlabamaCajun]

my newark uninterrupted takes about 34 hours to go through one of those ribos. i had a bit of a set back a few days ago with one. power failed, system shutdown. i tried to restart, found out windows corrupted due to unsafe shutdown, lost a file. did a complete repair, went back in but unfortunately it was a loss. funny thing is it was at 240/250. DEFINATELY lookin into pikin up a backup UPS for safe shutdowns.
Yes, UPSs are a must have. I've got get one for my folding tower.

[ChasR]

An A64 clawhammer (1MB Cache) produces RIBOs at about 131ppd/GHz. An x2 or dual core opteron should be able to nearly double this as the RIBOs aren't particularly bandwidth intensive. Cache size has a small effect on production of these WUS. An A64 Winnie (512KB cache) produces 125ppd/GHz on them.

[harlam357]

My 170 seems to do ~135ppd/Ghz on these... so ChasR, your estimate is definitely much more inline with what I'm actually seeing.

[dmatko]

off topic...im considering picking up the APC br1500 to use on 3 systems. its rated up to 865 watts, hope its enough for 3 computers.

[pscout]

off topic...im considering picking up the APC br1500 to use on 3 systems. its rated up to 865 watts, hope its enough for 3 computers.
depends what kind of rigs they are ... based on my sig rig with its WC etc, you are good for at least 2. 3 will likely work, esp if they are power light.

I have a couple of p4 folders running on an old 600w apc - 1 is a northwood, 1 is a prescott - unfortunately no monitoring software on it so i don't know what they are drawing ... but they are minimal systems , 1 sata , 1 low power pci video, no floppy or cdrom. But they are both overclocked with some vcore and vdimm boosts i think. So on an 865 w ups i am sure i could add a third p4 folding rig.

Of course if you have amd, you can probably go with even more as long as you don't have power hungry periipherals/video on them.

I also have an apc 1000 (600W) for my sig rig .. with its monitoring you can watch the power draw as you add rigs. I assume yours will have software as well, so just monitor consumption as u add each rig ... make sure they are under load to get each rigs max reading.

[harlam357]

Iffy... again, I hope you don't mind if I call you Iffy. :D I have one of these running my 170 rig, 20" & 15" CRTs, router, DSL modem all on full battery backup. If memory serves me correctly, all that eats ~350-400 watts of power. Leaving 15-20 minutes of uptime when power is lost. I think if you limit the number of monitors on the backup to one and run no extra extra stuff, you could probably put three rigs on one of these. I will probably do the same when I move my rigs to their new home... also have a br800 that will be moved to my 170 rig. The nice thing about the br1500 is the capability to add another powercell. Even though that won't up the watt output, it will definitely add to you runtime when the power goes out. IMO the br1500 is the right choice.