Here's the results from one of ted3's machines - the new credit system with credits/second. Prior to Aug 23, it hovered around 0.003 credit/sec, and since then, it's been hovering around 0.00375. David Kim has already mentioned there being a problem with the WU that spiked on the graph (it has a name that reminds me of John "Stamos").

Looking at this shows that the new credit system is working.

---

I don't see the logic why someone should claim that a 386 should get the same credit as a Core 2 Dual.

I don't see the logic why someone should use a top of the hill machine crunching rosetta, when he doesn't get the credits for that.

So what you're saying, I give my CPU's 24/7 (hence the higher electrical bill) to rosetta, and they only used en little of the potential of my PC, because they don't want to spend some money to rewrite their clients, which only can deal with museapieces right now ?????

If i need another good reason!

---

I don't see the logic why someone should claim that a 386 should get the same credit as a Core 2 Dual.

I don't see the logic why someone should use a top of the hill machine crunching rosetta, when he doesn't get the credits for that.

So what you're saying, I give my CPU's 24/7 (hence the higher electrical bill) to rosetta, and they only used en little of the potential of my PC, because they don't want to spend some money to rewrite their clients, which only can deal with museapieces right now ?????

If i need another good reason!

It's not that easy. It's not done with a some money and some programming effort. A real app is way more complex than a benchmark and Rosetta is way more complex than those apps over at Einstein and Seti, which search for nothing in noise. Many people looked over the code of Rosetta and there is no way to optimize it for specific CPUs that would not hinder further development - at least not yet found. With your top of the hill machine you get more credits as with a slow machine - maybe not your full potential but there is no app which utilizes the full potential. If you want fully optimized clients, which perform simple and easy to optimize tasks go to Seti.

If you want fully optimized clients, which perform simple and easy to optimize tasks go to Seti.

That's not a medical project.
I'm already switched my home PC's to D2OL, my teammates needs some help there.
And when I've decided which project earn my support/money, then I'll visit my relatives to change their clients too.

---

I don't see the logic why someone should use a top of the hill machine crunching rosetta, when he doesn't get the credits for that.

So what you're saying, I give my CPU's 24/7 (hence the higher electrical bill) to rosetta, and they only used en little of the potential of my PC, because they don't want to spend some money to rewrite their clients, which only can deal with museapieces right now ?????

If i need another good reason!

Crunch3r created an optimized application for Seti that crunched through WUs 3 times as fast as the default Seti application. He then created a Boinc client that gave 3 times the benchmarks of the default Boinc client. The "optimized" Boinc client didn't optimize anything, it doesn't have any magical properties, it doesn't turn on anything - it just increased the benchmarks so a WU got the same amount of credits as it did before. When tied together with the Optimized Seti Application, it gave you 3 times the credit for 3 times the work.
You imply that because you have a cpu with enhanced abilities (64 bit, SSE, SSE2, SSE3, 3DNow, etc) that you should get 3 times the normal credit and that Rosetta should produce an optimized client that operates at 3 times the standard client's rate. The fallacy with that idea is the fact that optimizing a program doesn't produce identical speedups. Take a look at the thread where we discussed 64 bit mode; some programs double their performance while some actually get slower moving from 32 bit mode to 64 bit mode. Turn on 64 bit mode, SSE, SSE2, SSE3, and 3DNow optimizations and compile the infamous "Hello World" app in C. Does it work 3 times, 2 times, or even 50% faster with the switches for 64 bit mode, SSE, SSE2, and SSE3, or 3DNow optimizations? NO. (Perhaps with a DirectX call that allows writing directly to the video ram will speed it up, although it would have to be done thousands? of times to produce a measureable time difference.)
If you've got a top of the line computer, then you'll get more credits than a normal computer since you'll be producing more work than the normal computer. If you produce 6 times the work as another system, you get 6 times the credit the other system gets. If you produce 2 times the work of another system, you get double the credit that system does.

Ethan posted a link to how to get access to the (older?) Rosetta code and a few people interested in trying their hand at optimizing it for the above cpu enhancements (and hopefully Altivec, as well) have volunteered and we may see a few optimizations added to the Rosetta code.

Come to think of it.. we had this discussion at most of the projects I've been a part of. At FaD it was stated that it had already been optimized 40x from its original form.. and SSE, 64 bit mode, etc wouldn't benefit the application enough to justify it. At D2oL, during the 6 months I was there, the programmer was basically non existant and there was a long list of 2? year old promised features for the client that still had not been added - let alone thinking about hand optimizing the code.

---

I don't see the logic why someone should use a top of the hill machine crunching rosetta, when he doesn't get the credits for that.

You DO get credit for that if your tweaks make your machine faster in ways that Rosetta can utilize. So, overclocking is one example of such a tweak.

So what you're saying, I give my CPU's 24/7 (hence the higher electrical bill) to rosetta, and they only used en little of the potential of my PC, because they don't want to spend some money to rewrite their clients, which only can deal with museapieces right now ?????

Your point is well taken, that an application that can more FULLY utilize your machine would seem to allow you to contribute more to science. But stop for a moment to consider that most other applications use a quarum system, so net/net they're only using 1/3rd of your systems abilities. Or, put another way, 2 out of 3 WUs your expend electricity and time into crunching, are done simply to validate the work of other participants, and to even out their credit claims.

Now, even without fully utilizing all of the hardware functions you mention, I think Rosetta still comes out on top so far as using the donated resources wisely. Since the Rosetta application IS changing and improving constantly, it is more important to consider maintainability of any code changes made. Code full of if defs for various compilation and platform-specific optimizations is definately harder to maintain. Perhaps that is another reason more specific optimizations have not been devised. In the end, it is the compiler that decides how the program is implemented on the CPU. So... if there are specific compiler suggestions, I'm sure they are open to hearing them.

---

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/

BennyRop, THINK (FaD) ended up being about 100x faster than the original client was (I can give you a link to a presentation Keith did in the October before he closed FaD).

It was originally helped along by Intel. It was pretty much single floating point operations, and SSE (don't even think MMX was turned on) did little to help it along. I think Intels tuning did the 40% and scientific improvements did the rest.

---

Team mauisun.org

FWIW, my team (Team MacNN!!) shows it's members a rolling 24hr stat. When the new credit method kicked in, I switched to 1hr (minimum) run time, hoping to hit the credit lotto. My rolling 24 hr average was ~5k/24hrs. Then we had the verification server glitch. We first thought that the PTB had instituted a delay in granting credits to remove the lotto driver. So at that time, I increased my runtime to 6 hours. In the couple of days since then, my average has increased to ~7k/24hrs.

Of course, with this new credit method, this could mean something, or absolutely nothing. No way to know.

I just increased my run time to 8 hours. We'll see what happens.

Update: After switching to 8 hour run time, and waiting 8 hours + 24 hours, my rolling 24 hour credits had dropped to 6.9k, down from ~7.4k. I'm going back down to 6 hour run time to see if it goes back up.

---

Reno, NV
Team: SETI.USA

In no way I wanted to imply that the opt clients delivered falsified results. I stated so on many occasions and will do so in future. You explained very nicely how those impressive benchmarks can be achieved by utilizing all the power modern processors offer. However as you said, Rosetta can't use all those features yet so in fact the benchmark offers more of a potential speed which could be gained from such a computer if the app could utilize all the features. Whether credit should be based on the real work done or the potential a host offers is another question and I'm undecided about that. However I was and am against a system in which every host just gets what he claims. That led to laughable claims of poor hosts and as I remember the most absurd were dealt with but every day new absurd appeared and the more subtle cheats were never caught. It was just a hole in the credit system, which required constant action from the project staff and prevented them from the important tasks.

Actually, Crunch3r's 5.5.0 client did return benchmarks that were higher than the physical abilities of a processor even if perfect efficiency was achieved.

This is my 2.3 GHz Athlon XP. It's a Barton core running at 11.5 * 200 MHz. Now look at my integer benchmarks, as given by Crunch3r's 5.5.0 client - "Measured integer speed 11476.56 million ops/sec". Thing is, an Athlon XP can only issue 3 integer instructions per clock cycle, meaning that my processor is only capable of 3 * 2,300,000,000 integer instructions per second. That means that with absolute perfect efficiency, which you don't get ever, my processor can issue 6,900 million ops/sec. Where did the extra 4,576,56 million ops/sec come from?!

This is my 2.14 GHz Duron. It runs Trux's 5.2.13 or something like that. Note it's integer benchmark - "Measured integer speed 6364.27 million ops/sec". But that is physically possible because 3 * 2,140,000,000 = 6,420 million ops/sec.

I am sick and tired of people saying that Crunch3r's 5.5.0 client rewards them based on what they can contribute. No, you can't contribute something that is physically beyond the capabilities of your processor.

---

BennyRop, THINK (FaD) ended up being about 100x faster than the original client was (I can give you a link to a presentation Keith did in the October before he closed FaD).

It was originally helped along by Intel. It was pretty much single floating point operations, and SSE (don't even think MMX was turned on) did little to help it along. I think Intels tuning did the 40% and scientific improvements did the rest.

The SSE optimization attempts by Intel were done on think version 1.03 and earlier when it was being used at United Devices. The 100x (up to to 160x depending on the query) improvement in speed was due to programming changes after the Find-a-Drug project started on 18 April 2002. FaD started with think version 1.03c and fadv1.31b was released in November 2005. http://www.treweren.com/

The points couldn't be manipulated at Find-a-Drug because the benchmark was part of the encrypted think.exe.

Thanks for clarifying that, FluffyChicken and Vester. :)

---

Is there already a discussion started over the competition between Teams. With the new system it will take way to long to close a gap of 4 milion point. The daily difference is to small with the new sytem. I think for the competition it will be very wise to reset the counters and proceed with Rosetta 2 because a lot of user which a here for the good cause but also for the competition will leave

---

BurnHard (sounds like a [H]ard|OCP name), there were many teams and members such as Morphy375 and Hoogie who did not use optimized clients and it would be unfair to them to reset the stats. I don't often mention my credits because some of them were done with optimization. The Dutch Power Cows and teams like TSC Russia can easily catch up. Some of the complaints from the hard core optimizers indicate that some of them have left already.

It would be possible for the stats sites to produce stats tables for before and after optimization.

Following the lead of BennyRop. I created similar charts for my five puters. I've included both claimed and granted credit as show in a per hour basis. It seems like lately granted credits have trended higher than the first few days of wus. This doesn't appear tied to any one particular type of wu. I tried the charts with the wu names in them, but that made them too tall to display here. Variations in the "claimed credit" might be explained by the fact the benchmarks had been rerun. Yes, that one big one were the SRAMANs and were mistakenly submitted with the 2cr/model (Ralph) payment system.

This is not a comparison between new and old credit system, but rather it shows the credits earned by various flavors of (my) HW & SW. None of the boards are overclocked.



I combined the latest 5 results, and took the average just to smooth things out. Some things look very odd to me:

1) wonder woman has been consistantly (slightly) faster than batman crunching SETI. Yet it is earning almost HALF the credit of slower batman.

2) superman is ranked too low compared to the Pentium machines. The G5 should be faster than the Pentium on a per-thread basis, right?

3) What the heck is rogue doing below *all* the P4s??? Something is not right here. I thought athlons were supposed to stomp all over P4s.

4) The Core chips are sweet! I can't wait to get my Core 2 Duo on line!

Bottom line, something is not adding up right with the new credit system in my mind.

Edit: 5) There is no way a Pentium D 940 is as fast as a Core Duo 2.16. Something is not right here.

---

Reno, NV
Team: SETI.USA

mmciastro, how do you get your data into spreadsheets? Do you type by hand, or import it somehow? TIA.

---

Reno, NV
Team: SETI.USA

I'm confused. Someone please explain this graph:

http://www.boincstats.com/charts/chart_uk_rosetta_project_new_credits.gif



Personally, my chart shows a drop by about 50%, and my team's chart shows the same. I have looked at many of the top 25 teams out there, who have contributed 45% of the total credits, up to 7% each, and they are *way* down with the new credit system.

So how is it that the total credits have not dropped off? Increase in membership? Can it really be that much?

Well, there it is afterall. A drop of about 13% of credits so far.

---

Reno, NV
Team: SETI.USA

BurnHard (sounds like a [H]ard|OCP name), there were many teams and members such as Morphy375 and Hoogie who did not use optimized clients and it would be unfair to them to reset the stats. I don't often mention my credits because some of them were done with optimization. The Dutch Power Cows and teams like TSC Russia can easily catch up. Some of the complaints from the hard core optimizers indicate that some of them have left already.

It would be possible for the stats sites to produce stats tables for before and after optimization.

Its fair to start over for the ones that didn't use the optimized client. These folks now have the chance to be top ranked but is we just go on it will take a very long time for them to close the gap between. I believe it will be very simple just reset the STATS database and start over. There is notting else that has to be changed but it will be a fair solution

---

Zombie:

mmciastro, how do you get your data into spreadsheets? Do you type by hand, or import it somehow? TIA.

Here's a way - although after posting this, someone will point out ways of making it more streamlined. :)

Get to the results pages of the computer you wish to chart. Select all the lines you want to deal with. I highlight the bottom right hand corner's granted credit, and then select the upper left hand corner's workunit ID#. Choose copy, and then paste into notepad.
If you have more than 20 results, go to the next page, and repeat until you've got them all copied into Notepad.

Select all the data in Notepad, and then paste it into Excel. Delete the lines with waiting results, ghost WUs, and errored WUs.

Select cell K1 and use the formula (=i1/h1*3600) for old credits/hour. Select cell L1 and use the foruma (=j1/h1*3600) for new credits/hour. Copy and paste the formulas into K2 through K20 (if you have 20 lines) and L2 through L20 (if you have 20 results). Then graph K1 through L20. Or just L1 through L20.

Have fun, and experiment..

---

mmciastro, how do you get your data into spreadsheets? Do you type by hand, or import it somehow? TIA.

I copy and paste the "results Page" for each individual computer into its' own worksheet. Then sort the entire data and go back and delete any duplicates (this is especially needed in projects with a quorum as a wu from a week ago may have been pending, and if you don't copy and paste them all you'll miss them). I have the columns set up so copy and past works without extra manipulation (I.E I labelled the columns the same as is seen on the results page), then I added the extra columns I felt were needed. Some have "functions" which automatically calculate "granted and claimed/hour". Then since I copied the entire results page, I click on the "result ID" link and copy over the number of decoys, the WU name, application version, and boinc version.



tony

the formula for claimed credit/hour is: =SUM(K29/(J29/60/60))
Granted credit is: =SUM(L29/(J29/60/60))