[Shaktai]

FYI: From the DNETC page....

RC5-72 is now actively running! Clients with version v2.9001 or higher support both RC5-72 and OGR. Older clients that are versioned v2.8 support only OGR (RC5-64 is no longer active).

The download page is: http://www.distributed.net/download/clients.html Mac clients (OS-X only for RC5-72) are available. OS-9 and 68k clients will only run OGR at this time.

It should be noted that the STATS server is not yet working for this project, though it is expected soon. Units crunched now will apply when the stats server comes online per the FAQ.

IMPORTANT UPDATE: The Mac version does not yet appear to be Altivec enhanced. It runs only with Core #0, regardless of G3 or G4.

As a result, a Power Mac 800 only benchmarks 1,626,924 keys/sec (core #0 only). An Athlon 1900 under win XP benchmarks 1,693,256 with core 0, but will utilize core 5 to do 4,850,777 keys/sec.

No G4 advantage yet. No Altivec yet.

[Scotttheking]

AND THE RESULTS ARE IN.

This project will take up to 256 times as long as rc5-64.
I solved it, I solved it!!!

Really, it's that simple. The project is officially a waste of time.

I can't stop you from running it, but that's my opinion. There really is no point to it. RC5 isn't even used anymore.

[The_Equivocator]

I agree with Scott. Don't waste your computer's time.

[jarling]

Here Here!

[unregistered]

RC5 ROCKS!

[Scotttheking]

This is why I find RC5-72 to be worthless.

These challenges were issued over 5 years ago when RSA was trying to prove that their encryption was very hard to break. Well, in the last 5 years, the encryption has been replaced. It is no longer their latest and greatest.

RC5-72 will take up to 256X longer then RC5-64. This means that, even accounting for increases in speed and more computers running, we are probably talking about 10-20 years, or longer. All that to try and crack an encryption that is no longer the big encryption to use.

RC5 is all RC5. The only thing they are doing is increasing the key length. It serves no new purpose.

The project really is worthless. Trying to crack an old encryption scheme with longer and longer keys really is wasting cycles. Proved once, it's a great test. Now it takes a simple spreadsheet, plus the history of RC5-64, to determine how long any key length could take to crack.


Now, I am not saying that Team MacNN will not participate. Some of you want to, and I will not tell you you can't. For those that do, we will have a team, stats, and install pages (which should all be the same as what we have now, I haven't looked yet). What I am saying is that, if you have the choice, I'd like you to choose a different project, I am not forcing you to.
My opinion is that, of all the DC projects, RC5 is the only that has no purpose but stats.
Will I run it? Yes, on a few machines that I can install nothing else on that will be turned on anyway. I will switch them from RC5 as soon as I am able to run something else. It will be a "if I can do nothing else and they will sit idle otherwise, I will run it" project.

Some comments that I have heard:


It is optimized for altivec, nothing else is. I want to run something that uses the most of my processor.

It is still only using one part, the vector component. Your integer and floating point units are still idle. It just happens that RC5 is sped up by the altivec unit. Other projects may or may not receive a significant speed boost by coding for altivec, or they may have other reasons for not, like code integrity. It is much easier to use the same code on all platforms, as it keeps bugs from appearing in one version.
New RC5 cores are very easy to test, as you know what output you want. It is easy to feed a few hundred blocks in with known output, and make sure their output is correct.

I like the client.

What do you mean?
If it is in reference to the good design of the client, with it's multitude of features, I agree. It is the best written client. However, unless it offers a feature you have to have, I do not see why this is a reason to run the client.
If it is in reference to the low memory footprint, that has nothing to do with the client itself, it has to do with the crunching core. RC5 does not have a dataset like any of the scientific projects, and as a result, it has a very small footprint. It is not doing any data analysis, so it needs no data, and has minimal output.


The above is my point, my stance. Whether or not you agree is your choice. Again, I encourage you to run something else, if not, that is your choice, and I will not stop you from running it.

As a side note, rc5.macnn.com will still be the proxy address, and it will be working whenever dnet releases a linux version of the proxy, which has not happened yet. I don't know if the stats generator that I was using works with the new proxy, but assuming it outputs the same way, it should.

--Scott

[Jsnuff1]

Originally posted by Scotttheking:

RC5-72 will take up to 256X longer then RC5-64.

Not necessarily so, if they exhaust most of the keyspace again, yea it will take 256 times longer, but remember the key could be found on the first day how long rc5 lasts really is all based on chance.

[Shaktai]

Well I was just sharing the info because others have expressed interest and I stumbled onto the information, but have little interest in the project myself, outside of curiosity and preliminary testing.

I think that we have too many other good projects that can provide more real world benefit to work on. Besides, what I am really waiting for is SETI 2 (astropulse/southern hemisphere) and BOINC. I like the potential I see there and the fact that it goes beyond just the search for ET into other significant astronomical research as well.

In the meantime I will keep a little power on SETI now that I have passed the 1100 mark, and the rest will probably go to Folding@Home. Not enough team interest to make a serious run at D2OL right now, but I look forward to the CASP 5 results for dFold, which should be annouced very soon.

[reader50]

Originally posted by Shaktai:
...
In the meantime I will keep a little power on SETI now that I have passed the 1100 mark...

I am saved!

[Scotttheking]

Originally posted by Jsnuff1:


Not necessarily so, if they exhaust most of the keyspace again, yea it will take 256 times longer, but remember the key could be found on the first day how long rc5 lasts really is all based on chance.

note the "up to"

Originally posted by Shaktai:
Well I was just sharing the info because others have expressed interest and I stumbled onto the information, but have little interest in the project myself, outside of curiosity and preliminary testing.

I think that we have too many other good projects that can provide more real world benefit to work on.
SNIP

Not a problem. Information is good.

That is what I feel, also.

[l008com]

A note, the 68K Client has no OGR core, which means currently can't do any projects at all. Classic PPC and 68K clients should be out soon though. And hopefully an optimised PPC G4 core will be too.

[Scotttheking]

rc5.macnn.com is up and running.
It accepts WU for RC5-72 and OGR.

I will turn on stats for it as soon as it is being used by a few people.

If you are running a proxy at home, which I figure a few people will since there is now an OSX version, set it to buffer a low amount, and use the max days setting and set it to 1 unless you have a reason to set it higher (modem, flakey connection, etc), and change the keyserver options to have rc5.macnn.com as the keyserver.

[Shaktai]

Originally posted by reader50:
I am saved!

A little means my two slowest boxes, still probably good for 3+ units per day average heh heh!!!. Figure I ought to catch you before years end still.

[proux]

Well just to voice another opinion I was all on RC5 because it was the only one that was Altivec optimized. Moved back to SETI since it was solved and will stay there. Cracking a 72 bit key has no use, especially when you can have 256 bits keys now.
Now if only the new optimized SETI was here...

[[APi]TheMan]

So what are the majority of Team MacNN users running now? A few days ago I tried switching to Folding@home... and then I think it screwed my laptop up... Upon rebooting I had kernal_tas taking 99% CPU usage.

I thought it might be the core.exe or something, I can't remember, but I read about some .exe taking tons of CPU usage if you didn't kill the process correctly...

Oh well, so I'm not running and distributed computing efforts as of now.

[chris v]

I'm on the fence. I downloaded the client, but I'm only getting 2-3mkeys on my dual gig. Don't know what's up with that, but i seem to recall getting something like 10mkeys per processor before, with RC5 -64. They do have the best stats, but "advancing the idea of distributed computing" is not seeming like such a worthwhile goal anymore. The utility of DC has been amply demonstrated, and people are beginning to put it to practical use.

I'll probably keep up SETI for now-- it's not like there's not going to be PLENTY of time to get in on RC5-72. Hope they altivec-enhance the client at some point.

CV

[Shaktai]

Originally posted by [APi]TheMan:
So what are the majority of Team MacNN users running now? A few days ago I tried switching to Folding@home... and then I think it screwed my laptop up... Upon rebooting I had kernal_tas taking 99% CPU usage.

I thought it might be the core.exe or something, I can't remember, but I read about some .exe taking tons of CPU usage if you didn't kill the process correctly...

Oh well, so I'm not running and distributed computing efforts as of now.

'

Hmm! never heard of F@H messing up like that, and I have never had any problems with getting it to shut down properly, either with the GUI or CLI. In fact, it is about as friendly of a low priority program as you can get. The biggest problem is increasing its priority. I've got my computers split between Folding@Home (should have a big dump tomorrow if their stats update properly) and SETI. However Distributed Folding (dfold) is a decent and friendly project as well. CASP 5 results should be announced in a few days, and that will give an idea of how effective Distributed Folding can be. There is also a MacNN D2OL team which is Mac Friendly, but not a lot of interest in it it seems.

Folding@Home has had some real world successes already, and shows a lot of promise. Their status are sluggish updating though, and they seem to have an ongoing battle with their data servers, but are slowly improving. Of course SETI is the old standby and SETI 2 plus BOINC may just start a whole new Distributed Computing revolution, plus lots more interesting possibilties.

When I am uncertain I just jump into something and maybe move around between projects until I find one I can be comfortable with for the longer haul. Just feels like I am wasting my investment if my computers aren't doing something.

[RealMac]

I too am having difficulty justifying participation in the RC5-72 effort. Even in a couple years, I doubt CPUs will be able to handle this type of challenge at a speed several hundred times faster than presently. Work has been done on dedicated devices for these type of challenges, but not much has come out of that.

Devices with FPGAs (field-programmable gate arrays) were attempted, but deemed too expensive for the kind of performance they were offering to the average joe. Once Altivec enabled PowerPC products were available, we began to see a quantum leap with RC5 crunching speeds. To continue seeking alternatives for speeding up the process shifted from using dedicated silicon to having more processors.

If more desktop PC manufacturers went from single processors to duals, ala Apple, the time it takes to crack rc5-72 would be considerably cut. The chances of this happening are slim to none though.

Pentium 4 systems running the client are at a disadvantage, as it takes several more clock cycles to perform the same amount of work an XP or G4 system can do.

Not saying it won't happen soon or eventually, but the people who handle the process of writing the cores on the mac platform seem somewhat indifferent towards distributed.net. The traffic on one of the lists I'm subscribed to seems to have shrunken to nil. A year ago it was hustling...

One of my proposals is that we have a support the mac distributed.net programmers fund. I don't think they know how much we appreciate the hard work they do optimizing cores for our platform of choice. They aren't paid for their work. I doubt many of us would want to put hours upon hours into something when we weren't even sure the userbase as a whole would be interested.

I personally agree with the consensus here. RC5-72 sounds like an ambitious project, but why? Far more attractive alternatives are available in the present, performance needs work right now and the amount of time it would take to complete in a worst case scenario isn't particularly attractive.

Just my two cents.

[unregistered]

In case anyone is interested RC5-72 beta stats are up, we are currently # 19
http://stats.distributed.net/team/tm...team=865184860

[Shaktai]

Originally posted by Unregistered:
In case anyone is interested RC5-72 beta stats are up, we are currently # 19
http://stats.distributed.net/team/tm...team=865184860

Wow! I am #4 on the team and wasn't even trying. What really happened is that when i tested the clients, I thought I had quit them before resuming SETI. However, the client's kept running in the background. I didn't discover it for many hours until I realized that my SETI progress was moving at a snails pace. Had to force kill them but apparently had already done 30 billion keys.

Might be something to note. If you tested the client, and then quit, you may need to use the process monitor or TOP to make sure it "really" quit. It failed to quit on both of my Macs.

[Arkham_c]

Originally posted by Shaktai:


Wow! I am #4 on the team and wasn't even trying. What really happened is that when i tested the clients, I thought I had quit them before resuming SETI. However, the client's kept running in the background. I didn't discover it for many hours until I realized that my SETI progress was moving at a snails pace. Had to force kill them but apparently had already done 30 billion keys.

Might be something to note. If you tested the client, and then quit, you may need to use the process monitor or TOP to make sure it "really" quit. It failed to quit on both of my Macs.

Yeah, all I have running is one full-time PC (1.6GHz) and one part time (750MHz). I'm not running it on my Macs at all (been playing too many games on them to slow 'em down).

With common encryption keys being 5120 bytes in length, this challenge is really a joke. My cracking keys on it is just joining the lottery to try to win the $2k.

[l008com]

Originally posted by RealMac:
One of my proposals is that we have a support the mac distributed.net programmers fund. I don't think they know how much we appreciate the hard work they do optimizing cores for our platform of choice. They aren't paid for their work. I doubt many of us would want to put hours upon hours into something when we weren't even sure the userbase as a whole would be interested.

Agreed. I know I check every day, because as soon as the new G4 core comes out, there will be a HUGE shift in the RC5 stats from where they are settling in right now. Who are these people and how can we tell them that their work IS needed?

[Scotttheking]

Hmm, I know a way to shift the stats.
I have a farm that I can't run anything else on until the Win2k upgrade is approved, which won't happen for a while or else won't happen.
The farm has been idle for a few months.
I guess I could drop the dnet client on them for now

[l008com]

The core in the current OS X client is REALLY bad. My 900MHz Celeron Running XP at work is just as fast, some days it even outperforms a tad, as my Dual 533 G4 PowerMac!!! If someone writes a real nice core for the G4, I should start getting about 4X faster keyrates. As should all other Mac users. Windows clients are already off the default cores and on to some optimised, or at least partially optimised cores.

[l008com]

And to make things worse, I was just talking to dnet employees on thier IRC thing and they said it will probably be weeks until there are optimized PPC cores, as far as this guy knew, no one was working on any.

[Scotttheking]

For anyone who wants to use the proxy, which it at rc5.macnn.com and runs on port 2064 (the default port), I turned on some stats.
They are still weird, which could be either that the dnet proxy hasn't gone final and is outputting something weird, or that these stats are beta, but here they are.
If you want the email listing to show your username, not email, you have to send me the email addy and matching username, and I set it up.

http://teamstats.macnn.com/~distcomp/rc5stats/

[RealMac]

I advise you not to get hostile with him (remember he's doing this on his own spare time), but kindly asking him might be worth the effort.

d a n _ o e t t i n g@qwest.net

Scott edit: Added the space between each letter to foil the spambot that combs these boards.

[l008com]

Wow he's probably goona be pissed you posted his email there, but I emailed him anyway, I was real nice and everything though.