Well at least a large base of people are fairly happy from getting a boost in credits from the standard client. Since some of the others in the thread were unable to explain it, I'll take a stab at it. Almost the entire base of std clients are seeing a rise due to a few things, averaging from higher powered machines, eveninig the OS disparity, and the clincher.. finally getting credit for MMX instruction sets. I know Rosetta isn't compiled to use SSE, but I believe it does make some use of MMX instructions.

Since the std client's benchmark isn't programmed to use MMX instructions to measure the cpu, it is not reflected in the original scores, or in the original claims. Thus, it has been underreporting the cpu all this time. With the work based system, you DO see the effect, and are thus granted credit for the extra "work" that these instructions have been doing all this time. This should give the entire stock client base a boost in points, so probably 90 - 95+% of people get a boost, while the 5-10% using the opt clients take a decline. Either way, the power was always there, but the std benchmark never accounted for it. That would be my theory anyway, as to why you are not seeing a massive credit drop. The opt clients are being erased by the boost given to the standards.

[DPC]Division_Brabant~OldButNotSoWise wrote:
So rosetta lost say 20% of their CPUpower, just to keep a few people satisfied.
Even a scientist must see that he was betting on the wrong horse.

The credit system is fair now. There are 500+ new hosts joining every day. That is almost 200,000 hosts per year. Think about where Rosetta will be in one year, or two or five years, considering TFLOPS. Fixing the credit system was betting on the right horse.

---

list of my results

There are 500+ new hosts joining every day. That is almost 200,000 hosts per year. Think about where Rosetta will be in one year, or two or five years, considering TFLOPS. Fixing the credit system was betting on the right horse.

I don't think there are 500 NEW hosts every day. More likely the same PCs with new host ids, from moving around, stopping and starting, re-attaching, changing user accounts because of changing teams, or whatever. I seriously doubt that 500 a day is any kind of sustainable number.


I'd rather see numbers on ACTIVE hosts per day.

edit: Looks like a short-duration bump caused by the Czech National team moving from Predictor while it's down. They were averaging about 25,000 a day on Predictor, so I don't see them picking up the 500,000 credits/day that XS was doing.

---

Proudly Banned from Predictator@Home and now Cosmology@home as well. Added SETI to the list today. Temporary ban only - so need to work harder :)



"You can't fix stupid" (Ron White)

Angus wrote:
I'd rather see numbers on ACTIVE hosts per day.

Yes, okay true. The number of active hosts is currenntly increasing nevertheless.

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

---

list of my results

Well at least a large base of people are fairly happy from getting a boost in credits from the standard client. Since some of the others in the thread were unable to explain it, I'll take a stab at it. Almost the entire base of std clients are seeing a rise due to a few things, averaging from higher powered machines,

Why shoud you benifit if i choose to pick my equipment and finely tune it and it does nothing else but power crunch? The science benifits as i tune my rigs for the project nothing else.(this is a hobby for me and i spend a lot of time with my equipment)
heater

Edit: i run opteron 170's watercooled on linux and 1 p4 windows box on 5.5.0.

---

[DPC]Division_Brabant~OldButNotSoWise wrote:
So rosetta lost say 20% of their CPUpbower, just to keep a few people satisfied.
Even a scientist must see that he was betting on the wrong horse.

The credit system is fair now. There are 500+ new hosts joining every day. That is almost 200,000 hosts per year. Think about where Rosetta will be in one year, or two or five years, considering TFLOPS. Fixing the credit system was betting on the right horse.

yes 500 new "needy children" every day, according to Doc Baker, I saw that video too.

But how many of those "new" hosts are modern machines? How many are tourists? How many are dedicated crunchers. Are they mostly P-2's that are giving a fraction of their daily output, or are they Conroes, putting out 24/7 for rosetta. Most of the losses in big "horses" were 24/7 machines. it gong to take a while to see teh real drop in work output, and you won't really be able to tell from the credits, especially if the majority of clients are getting a boost. It just won't show up unless you look at the work. Of course, those that have left won't have any record of their work done.

Why shoud you benifit if i choose to pick my equipment and finely tune it and it does nothing else but power crunch?

Because you are a philanthropist by nature?

---

list of my results

XtremSystem is unfortunatly (though fortunatly to Rosetta) a small amount of the work contributed (about 10% in old credits, so max 5% in new credits).

You take one team, please take the whole complete top30

We will see, I'll take a look to the outputstats of the top10 (33% in old credits, 15 in new ?) and all teams are going down to almost zero, I don't looked at the teams beneath them, but I think you see this trend in almost the complete top30.

So rosetta lost say 20% of their CPUpower, just to keep a few people satisfied.
Even a scientist must see that he was betting on the wrong horse.

I have nothing against the reasons for why they want to leave (since I agree with most of them :-)) but I can get over it and hope bakerlabs will learn from it.


I don't like the fact they told team members to leave there team, go to other teams trying to rally support and leave Rosetta etc..
Oh and I only replied to that post.


(33% in old credits, 15 in new ?)
Not sure really, Crunch3rs was usualy 300% over the norm (well that what I generally got across a range of CPU's) so I would assume they move to an average somewhere between 1/3 and 1/2 of their total.

I'm for sure that not all teammmembers used a optimized clients.
I'm using the optimized client for my game PC, all others (family PC's) are on the standard client. (sorry not are, I must learn speaking in the past)

I just don't give my E6600 power to a project that not appreciated it.

---

Angus wrote:
I'd rather see numbers on ACTIVE hosts per day.

Yes, okay true. The number of active hosts is currenntly increasing nevertheless.

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

That isn't a graph of how many active hosts today vs yesterday - it's the number of hosts that have earned credit in the last 30 days. It's takes a long time to reflect big movements in and out with that type of statistic.

---

Proudly Banned from Predictator@Home and now Cosmology@home as well. Added SETI to the list today. Temporary ban only - so need to work harder :)



"You can't fix stupid" (Ron White)

yes 500 new "needy children" every day, according to Doc Baker, I saw that video too.

No, most of the new users are not needy children. They are the average joes that install the program on their one or two machines at home and then about forgets it and don't think that the credit game is a matter of life and death.

---

list of my results

don't think that the credit game is a matter of life and death.

You know my wife say`s the same thing about football..... I say at least you can win at football, you never win the life game, alway`s the same result.

---

Not all Czech`s bounce but I`d like to try with Barbar ;-)

Make no mistake This IS the TEDDIES TEAM.

As it appears most top teams suffer, because many of their member use the optimized clients - that was to be expected. Many of the smaller teams though gain as most "standard" crunchers. I for myself am pleased to see the mighty power of a conroe box in a correct way now and don't have to compare it to some anonymous and falsified benchmarks which made the leaderbord for hosts completely meaningless. Now the true power of a highly tuned system can be much better shown and compared on the basis of real numbers. For example Hymay has a nice Conroe 6600, which made almost 1400 credits/day (I mean new work-based credits) - my Athlon 64 @ 2.4 GHz makes merely 400. With the old credit system I could make any credit per day, so there was no big difference between P2 and Conroes.

In fact I don't understand why the big guys don't like this change, since their numbers can't be any longer achieved by cheaters who just edit their benchmarks. Now you can get big numbers only if you have big hardware.

I assume the disappointment of some has to do with the fact that now less points are awarded than before for those who use the optimized client, but think of it like a currency changeover, the new currency is much more valuable than the old, so a RAC of 1000 is like a RAC of 3000 previously - I for myself feel doing more work than before, althoug I got 900 credits/day with 5.5.0 with the old credit system. Now I get 400 and feel those credits are much more valuable than before.

Just an offer to develop a positive emotion towards the new credit system.

I think a lot of people are missing the point and again going on and on about the points issue(Which has yet to be proved, so please wait).

There were a few issues not just points and going on about how marvelous the new system appears to be may come back and bite you.

The big guy`s as you so put it could not understand the fuss, could not understand why the little guys were moaning when 5.5 was out there and caused a lot less fuss and work for the project Devs. Cannot understand when you talk Boinc cross parity cos there isn`t any. Cannot understand why the Devs failed to communicate the new system or implement it properly. Definately cannot understand why it seemed that a few with such small RAC`s got such a big say. They gave their all 24/7 to this one project.

Now here`s a graph for you to do.

Count the callers for a new system, then count the "big guys" most of whom were and are silent.

I think it`s time to quieten down until true values are known.

---

Not all Czech`s bounce but I`d like to try with Barbar ;-)

Make no mistake This IS the TEDDIES TEAM.

I think a lot of people are missing the point and again going on and on about the points issue(Which has yet to be proved, so please wait).

There were a few issues not just points and going on about how marvelous the new system appears to be may come back and bite you.

The big guy`s as you so put it could not understand the fuss, could not understand why the little guys were moaning when 5.5 was out there and caused a lot less fuss and work for the project Devs. Cannot understand when you talk Boinc cross parity cos there isn`t any. Cannot understand why the Devs failed to communicate the new system or implement it properly. Definately cannot understand why it seemed that a few with such small RAC`s got such a big say. They gave their all 24/7 to this one project.

Now here`s a graph for you to do.

Count the callers for a new system, then count the "big guys" most of whom were and are silent.

I think it`s time to quieten down until true values are known.

Your post belongs more to the communication thread, nevertheless I answer here: I disagree that a so-called vocal minority of little guys got such a big say. In fact the contrary is true: The project team explicitly stated that using 5.5.0 was and is not considerd cheating and stopped any discussion about it (although only a minority was/is using it). They stopped the discussion about backdating as well. I don't really recall anyone suggesting keeping the old credit system, since there is a broad consensus that it was flawed and easy to cheat. The only thing what some big guys didn't get was a mysterious message they kept demanding from David Baker about - I really don't know. He stated that he is grateful for all the contribution and that he won't alter previous credits. He stated very clearly that calling users of the optimized clients cheater is not tolerated on this board.

We'll see in a few weeks more clearly what the project gained and lost due the switch to the new credit system. At the moment it seems there is a (short-term) loss but not as great as some would have expected (or wish to would have happened).

As it appears most top teams suffer, because many of their member use the optimized clients - that was to be expected. Many of the smaller teams though gain as most "standard" crunchers. I for myself am pleased to see the mighty power of a conroe box in a correct way now and don't have to compare it to some anonymous and falsified benchmarks which made the leaderbord for hosts completely meaningless. Now the true power of a highly tuned system can be much better shown and compared on the basis of real numbers. For example Hymay has a nice Conroe 6600, which made almost 1400 credits/day (I mean new work-based credits) - my Athlon 64 @ 2.4 GHz makes merely 400. With the old credit system I could make any credit per day, so there was no big difference between P2 and Conroes.

In fact I don't understand why the big guys don't like this change, since their numbers can't be any longer achieved by cheaters who just edit their benchmarks. Now you can get big numbers only if you have big hardware.

I assume the disappointment of some has to do with the fact that now less points are awarded than before for those who use the optimized client, but think of it like a currency changeover, the new currency is much more valuable than the old, so a RAC of 1000 is like a RAC of 3000 previously - I for myself feel doing more work than before, althoug I got 900 credits/day with 5.5.0 with the old credit system. Now I get 400 and feel those credits are much more valuable than before.

Just an offer to develop a positive emotion towards the new credit system.

The only "falsified" benchmarks were dealt with, and deleted with help from an XS team member if I recall correctly. You seem to also be implying that the opt clients also falsified their benchmarks, I would point out that those machines were perfectly capable of delivering those benches. The std client simply wasn't programmed to use MMX, SSE 1 or SSE2 to compute. Programs are quite literal beasts, If you do not specifically TELL them to use those extensions.. they will not.

All the opt client did was add the use of those extensions into the calculation. Nothing was falsified.. which is why the staff stated their use was NOT cheating. Modern cpu's ARE capable of delivering those benchmarks and that power when you program to USE the hardware that is in EVERY modern cpu, from the P2 onward. Rosetta itself is not optimized for SSE or above yet, however in a cross project environment that it irrelevant.. others may be. Anyway, enough on that, this is not a opti/not opti thread, just felt compelled to distinguish what has been FALSIFIED.

Secondly, my "nice" 6400 putting out a lovely 1400 points. Yes, that would be nice if it were @ stock clocks of 2.13 GHZ. Sadly, it was running at 3.4 Ghz when it produced those results. Not quite so impressive over your A64 @ 2.4 now.. when you consider its dual cores, and 1ghz faster. The current system is a nice beginning, but the current scoring is nowhere close to accurate. It was tuned to mirror the std client, with all its underreporting (notice the boost you now get by getting credits for MMX work done?) and is therefore just as inaccurate. All the upper end machines are being averaged down, yes they get more points, but at a reduced rate. And since Rosetta isn't programmed to use sse or above, no work gets done thanks to those innovations, and no credit is granted to the power they might provide. Whether it be potential or actual.

---

Tralala you said in previous post

In fact I don't understand why the big guys don't like this change, since their numbers can't be any longer achieved by cheaters who just edit their benchmarks. Now you can get big numbers only if you have big hardware.

Which you have now stated that D.Baker said was not allowed on this forum.

I don't really recall anyone suggesting keeping the old credit system

I do recall quite a few post`s asking why doesn`t everyone get 5.5 and be done.

---

Not all Czech`s bounce but I`d like to try with Barbar ;-)

Make no mistake This IS the TEDDIES TEAM.

At the moment it seems there is a (short-term) loss but not as great as some would have expected (or wish to would have happened).

I think that's true, but what can I say, I need to visit my family to install another client, that takes time (but this is something positive with the new creditsystem, you must visit your relatives :P )

And if you are a big cruncher with a computerpark/school whatever ,changing the clients on that isn't also easy.

And when all the optimized clients are down, the avarage credits, that they're pushing up now, wil be lowered.
At this moment the not optimized client become a little more points, so they get lesser and lesser when more optimized clients are down, maybe lesser then they had before.
That's when they started the discusion that it's unfair that a Core 2 Dual gets more points for a WU then a 386.

---

As it appears most top teams suffer, because many of their member use the optimized clients - that was to be expected. Many of the smaller teams though gain as most "standard" crunchers. I for myself am pleased to see the mighty power of a conroe box in a correct way now and don't have to compare it to some anonymous and falsified benchmarks which made the leaderbord for hosts completely meaningless. Now the true power of a highly tuned system can be much better shown and compared on the basis of real numbers. For example Hymay has a nice Conroe 6600, which made almost 1400 credits/day (I mean new work-based credits) - my Athlon 64 @ 2.4 GHz makes merely 400. With the old credit system I could make any credit per day, so there was no big difference between P2 and Conroes.

In fact I don't understand why the big guys don't like this change, since their numbers can't be any longer achieved by cheaters who just edit their benchmarks. Now you can get big numbers only if you have big hardware.

I assume the disappointment of some has to do with the fact that now less points are awarded than before for those who use the optimized client, but think of it like a currency changeover, the new currency is much more valuable than the old, so a RAC of 1000 is like a RAC of 3000 previously - I for myself feel doing more work than before, althoug I got 900 credits/day with 5.5.0 with the old credit system. Now I get 400 and feel those credits are much more valuable than before.

Just an offer to develop a positive emotion towards the new credit system.

The only "falsified" benchmarks were dealt with, and deleted with help from an XS team member if I recall correctly. You seem to also be implying that the opt clients also falsified their benchmarks, I would point out that those machines were perfectly capable of delivering those benches. The std client simply wasn't programmed to use MMX, SSE 1 or SSE2 to compute. Programs are quite literal beasts, If you do not specifically TELL them to use those extensions.. they will not.

All the opt client did was add the use of those extensions into the calculation. Nothing was falsified.. which is why the staff stated their use was NOT cheating. Modern cpu's ARE capable of delivering those benchmarks and that power when you program to USE the hardware that is in EVERY modern cpu, from the P2 onward. Rosetta itself is not optimized for SSE or above yet, however in a cross project environment that it irrelevant.. others may be. Anyway, enough on that, this is not a opti/not opti thread, just felt compelled to distinguish what has been FALSIFIED.

Secondly, my "nice" 6400 putting out a lovely 1400 points. Yes, that would be nice if it were @ stock clocks of 2.13 GHZ. Sadly, it was running at 3.4 Ghz when it produced those results. Not quite so impressive over your A64 @ 2.4 now.. when you consider its dual cores, and 1ghz faster. The current system is a nice beginning, but the current scoring is nowhere close to accurate. It was tuned to mirror the std client, with all its underreporting (notice the boost you now get by getting credits for MMX work done?) and is therefore just as inaccurate. All the upper end machines are being averaged down, yes they get more points, but at a reduced rate. And since Rosetta isn't programmed to use sse or above, no work gets done thanks to those innovations, and no credit is granted to the power they might provide. Whether it be potential or actual.

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.

The new system puts a stop to that and credits the actual amount done. One could imagine a credit system which would credit based on the potential of the specific host however I don't know how one could make it accurate and cheat-proof. Therefore I think granting credit on the actual work done is the better way. I disagree that the new credit system isn't accurate. Indeed there were old WU which deviated a lot from the mean but new WU seem all in line. My results vary only by 10% to 20%, which is in the range of the "natural" variation due to different run times of models. This natural deviation averages out over time and can't be manipulated.

Remains the question about how many points to award. Sure the project team could have adapted the new system towards the claimed credit of the opt client but for me this makes no difference. Important is the intra-project parity which means tow credits from one host are comparable to 2 credits from another host. Whether I get 10 credits/day or 1000 is not so much a difference as long as everyone is awarded credit on the same basis.

Tralala you said in previous post

In fact I don't understand why the big guys don't like this change, since their numbers can't be any longer achieved by cheaters who just edit their benchmarks. Now you can get big numbers only if you have big hardware.

Which you have now stated that D.Baker said was not allowed on this forum.

It is not allowed to call users of the opt client cheaters but people who manually falsify their benches were and are cheaters. In the past I could easily edit my CPU type to "Future CPU whatsoever" and put in some absurd benchmarks and match your real Conroe.

I don't really recall anyone suggesting keeping the old credit system

I do recall quite a few post`s asking why doesn`t everyone get 5.5 and be done.[/quote][/quote]

Which would have created intra-project parity but not prevented manual cheating. Furthermore it is not practical to ask everyone to install the opt client. A system which does not verify the claims is just flawed.

At the moment it seems there is a (short-term) loss but not as great as some would have expected (or wish to would have happened).

I think that's true, but what can I say, I need to visit my family to install another client, that takes time (but this is something positive with the new creditsystem, you must visit your relatives :P )

And if you are a big cruncher with a computerpark/school whatever ,changing the clients on that isn't also easy.

And when all the optimized clients are down, the avarage credits, that they're pushing up now, wil be lowered.
At this moment the not optimized client become a little more points, so they get lesser and lesser when more optimized clients are down, maybe lesser then they had before.
That's when they started the discusion that it's unfair that a Core 2 Dual gets more points for a WU then a 386.

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

Credits may go down as more opt clients leave but as long as the real work does not decrease it is only a problem of inter-project parity and can be easily solved (by using a divider like CPDN, Einstein and SETI).