AMD Athlon 64 X2 4800, O/C to 2.58GHz (7.5% O/C)
2GB Ram (clocked at CL2.5)
Windows XP
32.69 Credits/Hr (Most recent 10 results)
hostid=288946

AMD Athlon 64 X2 4400, Stock Clock (2.2GHz)
1GB RAM (forget what type, think it's CL3)
Linux (SUSE 10.1 SMP)
28.92 Credits/Hr (most recent 12 results)
hostid=288952

--miw

That gives you a result of 6.35 (2.58GHz) and 6.57 (2.2GHz) credits per core per GHz.

--
Mats

---

Bird-Dog,

you are running WUs for 1 hour which gives you a lot of overhead for initialization especially on WUs with bigger proteins. Longer runtimes will surely increase your output if you have your PC on for several hours.

However, if you only have it on shorter then initialization will take place each time you start up the machine and will cost time. Longer runtimes per WU will only help you little in that case. But they will save you online time/volume.

--Christoph
[/quote]
My computers on 24/7 and crunch other projects. Does the initialization take place every time its preempted then starts again.
thx

Does the initialization take place every time its preempted then starts again.

Only if you don't keep in memory while preempted. That's a configuration option in the General Preferences. In general, you want it to say YES keep in memory (virtual memory, swapped out actually). This retains as much crunched data as possible. If you remove from memory, the WU not only has to go through initialization, but it also has to reset to a prior checkpoint to continue on.

---

Add this signature to your EMail:
Running Microsoft's "System Idle Process" will never help cure cancer, AIDS nor Alzheimer's. But running Rosetta@home just might!
https://boinc.bakerlab.org/rosetta/

AMD Athlon 64 X2 4800, O/C to 2.58GHz (7.5% O/C)
2GB Ram (clocked at CL2.5)
Windows XP
32.69 Credits/Hr (Most recent 10 results)
hostid=288946

AMD Athlon 64 X2 4400, Stock Clock (2.2GHz)
1GB RAM (forget what type, think it's CL3)
Linux (SUSE 10.1 SMP)
28.92 Credits/Hr (most recent 12 results)
hostid=288952

--miw

That gives you a result of 6.35 (2.58GHz) and 6.57 (2.2GHz) credits per core per GHz.

--
Mats

Yup. I noticed that as well. one of these days when I get back home to the farm, I'll have a closer look at this. Since the 4400 and 4800 have the same L2 cache (1M/core) I sort of expected to get credit/core/GHz about the same for each. But notice that one box is under Linux and the other is under XP. The XP box does some other tasks such as virus scans and backups on a daily basis which may influence things. Memory timings may also have an impact here.

I'll also have a look at this from a SETI and Einstein point of view....

--miw

---

--miw

AMD Athalon XP 2400+ (2.0 GHZ)
768 MB DDR 2100
Windows XP Home
My WU time is set to 10 hours
Without typing all of the info, 90.34 hours and credits and stuff,
I am getting 8.185 credits/hour

You seem to be doing better than me with the same system although you have more RAM and longer wu pref. Do either af these make any differance.

It might have to do with my pagefile optimization. If I left my computer running Rosetta over the weekend, my memory would get so loaded that the computer slowed way down. Here is a link to a page I used to tweak my virual memory funcion for better system performance over the long term. But you need two hard drives to use this optimization.
Pagefile Optimization

And you may want to increase your crunching time (cpu time) to at least the default 3 hours. It's a fast processor, give it some to time really do it's thing.

---

AMD Athlon 64 X2 4800, O/C to 2.58GHz (7.5% O/C)
2GB Ram (clocked at CL2.5)
Windows XP
32.69 Credits/Hr (Most recent 10 results)
hostid=288946

AMD Athlon 64 X2 4400, Stock Clock (2.2GHz)
1GB RAM (forget what type, think it's CL3)
Linux (SUSE 10.1 SMP)
28.92 Credits/Hr (most recent 12 results)
hostid=288952

--miw

That gives you a result of 6.35 (2.58GHz) and 6.57 (2.2GHz) credits per core per GHz.

--
Mats

That is probably the ballpark for all Athlon 64 CPUs. I have a Venice single-core @2.4 GHZ and are getting between 15 and 16 credits/hour, so about 6.25 and 6.66 per GHz per hour.

AMD Athalon XP 2400+ (2.0 GHZ)
768 MB DDR 2100
Windows XP Home
My WU time is set to 10 hours
Without typing all of the info, 90.34 hours and credits and stuff,
I am getting 8.185 credits/hour

You seem to be doing better than me with the same system although you have more RAM and longer wu pref. Do either af these make any differance.

It might have to do with my pagefile optimization. If I left my computer running Rosetta over the weekend, my memory would get so loaded that the computer slowed way down. Here is a link to a page I used to tweak my virual memory funcion for better system performance over the long term. But you need two hard drives to use this optimization.
Pagefile Optimization

And you may want to increase your crunching time (cpu time) to at least the default 3 hours. It's a fast processor, give it some to time really do it's thing.

A single thread of Rosetta should not use up 768MB of memory, so any page-file optimization will only be used "once", as the first time Rosetta loads up it may need more memory than what is currently available, but as long as there's no other projects eating memory as well (I run Einstein, Seti and Rosetta on a Windows machine with 512MB of RAM with "stay in memory", and yes it does use the swap-file, but Rosetta is generally in the 50-150MB range, so shouldn't cause a problem on a machine with more than 256MB of RAM - unless the machine is active doing other work too, of course).

--
Mats

---

AMD Athalon XP 2400+ (2.0 GHZ)
768 MB DDR 2100
Windows XP Home
My WU time is set to 10 hours
Without typing all of the info, 90.34 hours and credits and stuff,
I am getting 8.185 credits/hour

You seem to be doing better than me with the same system although you have more RAM and longer wu pref. Do either af these make any differance.

It might have to do with my pagefile optimization. If I left my computer running Rosetta over the weekend, my memory would get so loaded that the computer slowed way down. Here is a link to a page I used to tweak my virual memory funcion for better system performance over the long term. But you need two hard drives to use this optimization.
Pagefile Optimization

And you may want to increase your crunching time (cpu time) to at least the default 3 hours. It's a fast processor, give it some to time really do it's thing.

A single thread of Rosetta should not use up 768MB of memory, so any page-file optimization will only be used "once", as the first time Rosetta loads up it may need more memory than what is currently available, but as long as there's no other projects eating memory as well (I run Einstein, Seti and Rosetta on a Windows machine with 512MB of RAM with "stay in memory", and yes it does use the swap-file, but Rosetta is generally in the 50-150MB range, so shouldn't cause a problem on a machine with more than 256MB of RAM - unless the machine is active doing other work too, of course).

--
Mats

Except that you have a little process called CCAPP.EXE in your Windows task manager that will grow very large if you run Rosetta straight for a couple of days, or even a week. Then you will get a nice little message saying that your virtual memory is too low and you need to end some programs to free up memory, and you won't be able to open other programs. I found that before I changed my virtual memory settings, I had to reboot my computer every two days just to clear the memory.

---

Except that you have a little process called CCAPP.EXE in your Windows task manager that will grow very large if you run Rosetta straight for a couple of days, or even a week. Then you will get a nice little message saying that your virtual memory is too low and you need to end some programs to free up memory, and you won't be able to open other programs. I found that before I changed my virtual memory settings, I had to reboot my computer every two days just to clear the memory.

Yes, I have one of those. After 14 days and a bit [we had a power-cut in my office 14 days ago], the size of it is 208KB, which isnt'a cause for concern. It's part of Symantec Antivirus, and I guess if it grows to large size quickly, then it's a problem in Symantec Antivirus, and may be a good reason to use a different version of SAV, or use a differet AV solution altogether [I have no choice in the matter if I wish to connect my machine to the corporate network - but then I don't seem to have the problem you're seeing either... I'm using version 9.0.2.1, if tha'ts of any use to you].

--
Mats

---

Except that you have a little process called CCAPP.EXE in your Windows task manager that will grow very large if you run Rosetta straight for a couple of days, or even a week. Then you will get a nice little message saying that your virtual memory is too low and you need to end some programs to free up memory, and you won't be able to open other programs. I found that before I changed my virtual memory settings, I had to reboot my computer every two days just to clear the memory.

Yes, I have one of those. After 14 days and a bit [we had a power-cut in my office 14 days ago], the size of it is 208KB, which isnt'a cause for concern. It's part of Symantec Antivirus, and I guess if it grows to large size quickly, then it's a problem in Symantec Antivirus, and may be a good reason to use a different version of SAV, or use a differet AV solution altogether [I have no choice in the matter if I wish to connect my machine to the corporate network - but then I don't seem to have the problem you're seeing either... I'm using version 9.0.2.1, if tha'ts of any use to you].

--
Mats

Whats this CCAPP.EXE thing all about. In the task manager its 68300kb and using no cpu. Is this ok.
thx

CCAPP.EXE is a process of Symantec Norton AntiVirus. If it grows while Rosetta is running something is wrong with Norton AntiVirus. If you google around a bit you'll find that Norton products are some of the worst software written (no this is not an exaggeration). If you can, switch to a better AntiVirus solution (I use F-Prot and am happy).

with my last 10 results it shows
8,551 credit/hour
****************************************
P4 2.66GHz, 1GB Ram,WinXPHome

---

I've taken all the reported data, consolidated it into one spreadsheet, and here it is. Let me know if I made a mistake so I can correct it. This has been sorted by credit/hour/host

tony

My system (https://boinc.bakerlab.org/rosetta/results.php?hostid=268485) is generating roughly 65 credits per hour. The 18.1 that you quote is per core, and the complete system has two processors with two cores each.

--
Mats

---

My system (https://boinc.bakerlab.org/rosetta/results.php?hostid=268485) is generating roughly 65 credits per hour. The 18.1 that you quote is per core, and the complete system has two processors with two cores each.

--
Mats

updated, thanks, 72.4/hour

>>> Intel Celeron 2.40GHz, 512 GB RAM?, Windows XP
15 results, Cobblestones/Hour range 6.12 to 8.90, Ave = 7.56
Decoys made range 16 to 43, Ave = 27

>>> Intel P4 2.53 GHz @ 2.75 GHz, 2 GB RAM, Windows XP Pro
19 results, Cobblestones/Hour range 6.92 to 9.80, Ave = 8.11
Decoys made range 10 to 49, Ave = 31

>>> AMD Dual Opteron 848 (2 cpus) 2.2 GHz, 2 GB RAM, Linux Fedora Core 3
16 results, Cobblestones/Hour range 9.99 to 18.25, Ave = 13.96 (27.92)
Decoys made range 17 to 83, Ave = 53

>>> AMD Dual Core Opteron 275 (2 cpus/4 cores), 2.2 GHz, 4 GB RAM, Fedora Core 3
24 results, Cobblestones/Hour range 10.66 to 18.15, Ave = 14.34 (57.36)
Decoys made range 19 to 94, Ave = 54

>>> AMD 4800+ X2 2.4 GHz @ 2.52 GHz, 4 GB RAM, Windows XP Pro
18 results, Cobblestones/Hour range 12.94 to 15.42, Ave = 14.36 (28.72)
Decoys made range 22 to 86, Ave = 51

To show what we used to get, I still have a WU showing on my account page that was done back on the 21/2/06 when I had WU time at 8 hours (currently 6). Claimed and granted Cobblestones were 113.90 with 385 decoys for 14.26 C/h. This was done on the 4800+ machine not optimised and Rosetta 4.82 WU's.
My 4800+ machine is a bit down on a couple of others that have reported in this post, I have had a few Windows issues and possibly heating issues with the AMD 4800+ so a week ago I reformatted and have since added new fans so we will see what happens.

---

>>> Intel Celeron 2.40GHz, 512 GB RAM?, Windows XP
15 results, Cobblestones/Hour range 6.12 to 8.90, Ave = 7.56
Decoys made range 16 to 43, Ave = 27

>>> Intel P4 2.53 GHz @ 2.75 GHz, 2 GB RAM, Windows XP Pro
19 results, Cobblestones/Hour range 6.92 to 9.80, Ave = 8.11
Decoys made range 10 to 49, Ave = 31

>>> AMD Dual Opteron 848 (2 cpus) 2.2 GHz, 2 GB RAM, Linux Fedora Core 3
16 results, Cobblestones/Hour range 9.99 to 18.25, Ave = 13.96 (27.92)
Decoys made range 17 to 83, Ave = 53

>>> AMD Dual Core Opteron 275 (2 cpus/4 cores), 2.2 GHz, 4 GB RAM, Fedora Core 3
24 results, Cobblestones/Hour range 10.66 to 18.15, Ave = 14.34 (57.36)
Decoys made range 19 to 94, Ave = 54

>>> AMD 4800+ X2 2.4 GHz @ 2.52 GHz, 4 GB RAM, Windows XP Pro
18 results, Cobblestones/Hour range 12.94 to 15.42, Ave = 14.36 (28.72)
Decoys made range 22 to 86, Ave = 51

To show what we used to get, I still have a WU showing on my account page that was done back on the 21/2/06 when I had WU time at 8 hours (currently 6). Claimed and granted Cobblestones were 113.90 with 385 decoys for 14.26 C/h. This was done on the 4800+ machine not optimised and Rosetta 4.82 WU's.
My 4800+ machine is a bit down on a couple of others that have reported in this post, I have had a few Windows issues and possibly heating issues with the AMD 4800+ so a week ago I reformatted and have since added new fans so we will see what happens.

Just checked and the Pentium 4 machine only has 1 GB RAM. This same machine takes 34+ hours to do an Einstein WU but has no trouble with Seti and Rosetta. No longer doing Einstein with this machine.

---

FYI. My new Mac Pro (2 dual core Intel Woodcrest 2.66 with 2 gig mem) gets about 22.84 credit per hour per core (4 total) doing rosetta under Mac OS X 10.4 with the stock boinc client. (https://boinc.bakerlab.org/rosetta/workunit.php?wuid=32456733). This seems a little low and I believe it is because the rosetta app itself is poorly optimized for new Mac Pro with OS X. I dropped the stock client because it was also poorly optimized (was badly under reporting, I know does not affect score but poorly optimized code bothers me and wont run on my systems, other than Mac OS X itself :). I will at some point use the boot camp option and see what kind of scores it gives under windoze native. My guess is at least a %20 improvement unfortunately. Kinda sucks the worse OS gets the most developer love.

---

FYI. My new Mac Pro (2 dual core Intel Woodcrest 2.66 with 2 gig mem) gets about 22.84 credit per hour per core (4 total) doing rosetta under Mac OS X 10.4 (https://boinc.bakerlab.org/rosetta/workunit.php?wuid=32456733). This seems a little low and I believe it is because the rosetta app itself is poorly optimized for new Mac Pro with OS X.

I just fired up my new homebrew linux box with Core 2 Duo E6600 (2.4ghz) last night. You can see it here. The core 2 duo is pretty much the same thing as the new Xeon, with the major difference is that the new Xeon works in a dual chip setup.

I'm getting about 20.2 credits per hour per core. That works out to about ~970 credits/day. So your slightly higher c/hr number seem about right to me with your slightly faster clock speed.

Oh yeah, you will get a slightly better memory performance by using 4 DIMMs instead of 2. Not sure what impact that would have on running Rosetta.

---

Reno, NV
Team: SETI.USA

FYI. My new Mac Pro (2 dual core Intel Woodcrest 2.66 with 2 gig mem) gets about 22.84 credit per hour per core (4 total) doing rosetta under Mac OS X 10.4 (https://boinc.bakerlab.org/rosetta/workunit.php?wuid=32456733). This seems a little low and I believe it is because the rosetta app itself is poorly optimized for new Mac Pro with OS X.

I just fired up my new homebrew linux box with Core 2 Duo E6600 (2.4ghz) last night. You can see it here. The core 2 duo is pretty much the same thing as the new Xeon, with the major difference is that the new Xeon works in a dual chip setup.

I'm getting about 20.2 credits per hour per core. That works out to about ~970 credits/day. So your slightly higher c/hr number seem about right to me with your slightly faster clock speed.

Oh yeah, you will get a slightly better memory performance by using 4 DIMMs instead of 2. Not sure what impact that would have on running Rosetta.

Thanks for the info. Hmm seems Mac app speed is comparable to linux app at least. Not sure how much the 4 DIMMS buys me for sure but probably any advantage is erased by the Macs poor handling of threads and stupid old legacy microkernel design. Oh well at least I can drop into the bsd command line and ignore the dumb mac gui fluff. Still have to say it is the best workstation going for the price.

---