New distributed.net Pre-Release Clients

[Moose]

The Following Clients have been placed/updated for pre-release:

• Windows 95/98/NT [x86/Zipped] v2.8011.464 2000-10-25
• AmigaOS [m68k] v2.8011.464 2000-10-25
• AmigaOS [PPC/PowerUp] v2.8011.464 2000-10-25
• AmigaOS [PPC/WarpOS] v2.8011.464 2000-10-25

The Pre-Release Page can be found at:
http://www.distributed.net/download/prerelease.html

Please remember to report bugs at http://www.distributed.net/bugs/

Updates of interest:
imp: many: OGR cores improved - should now be slightly faster
new: all: command line options may be passed via a 'dnetc_opt=' (or 'DNETC_OPT=') environment variable.
chg: rc5: random-prefix is no longer written to the .ini
chg: all: If a dialup connection had been previously initiated by the client AND the link drops while doing network i/o, the client will dial again.
new: the crunch-o-meter displays a twirling baton (by default) if not running with a GUI window and output is not redirected.
imp: networking layer rewritten from scratch. It is also a lot more chatty, and prints very specific messages on error.
new: client can be configured with multiple keyserver names.

See the changes.txt file for a full listing.

Enjoy!
moose

 

[RC]

Please tell me that there will soon be a new Linux client that has acceptable performance under OGR? Please? lol!!

 

[Moose]

I could tell you that... but I don't know if its true or not.

But seriously it really depends on how much feedback Cyrus gets on win32 client.

You can feel free to mail bug reports for the win32 client directly to Cyrus Patel cyp@fb14.uni-mainz.de

moose

 

[RC]

Are you aware that your latest Linux client is 30% less efficient compared to the latest WIN32 client under OGR?

 

[Engineer]

Thanks Moose!

Running as I type:

[Oct 25 03:48:37 UTC] Benchmark for RC5 core #6 (RG/HB re-pair II)
0.00:00:17.67 [3,561,438.37 keys/sec]



and...

[Oct 25 03:53:19 UTC] Benchmark for OGR core #0 (GARSP 5.13)
0.00:00:16.01 [6,703,418.59 nodes/sec]



1GHz TBird...

Hey Viz (below),
I'm not sure if that is an improvement or not...I'm just reporting that I'm trying it out ....Sorry about your troubles

After running for a full night....3.48 Average Mkeys/sec...Nothing else but client running. That's seems just a tad slower than what I normally get on the 463b client....

 

[Viztech]

Whoa Nellie!

Network problems here on the LAN!

I will advise when I figure it out, but it won't find my Pproxy or break through the firewall on port 2064 like the 463b clients will.

viz

Engineer-
So is that an improvement over the 463b client?

 

[TOOCOOL]

interesting benchmarks for k62-550
running winblows se

463 client rc5 930,472.31 keys/sec
new client rc5 945,768.70 keys/sec

463 client ogr 3,320,603.91 nodes/sec
new client ogr 2,285,996.10 nodes/sec :Q

ran benchmarks 3 times each and used middle #

 

[RC]

Using win32 463b client and Athlon 1000@1100 OGR 8.8Mnodes/sec

If a 1.0GHZ Tbird now benches at 6.7Mnodes/sec I think I'll wait for an update.

In all fairness it's the first release (beta) so it's possible that
performance for OGR will improve greatly in the next updated beta release.

Rather odd that improved OGR performance was one of the improvements listed.

 

[Moose]

Please don't go by Benchmarks alone. The benchmarks might not be currently accurate. Benchmarks are an odd science, I could tell you all about them and how they are different from running the client normally, but I won't bore you with those details. All I ask is you run the beta for a full run or 2 and then see the noderate. Report any slow cores to cyrus and report any inaccuracies in the benchmark to him as well.

Using a beta client I would hope you test it fully instead of just seeing you are getting a better rate out of it.

thanks
moose

 

[Viztech]

I can't make it connect by IP address at all, either on the LAN or my dual node server (one node on the LAN, the other on the 'net) I don't have DNS on my LAN, so it is worthless. I even took a shot at using NETBios names, still no luck.

Similar to RC and TOOCOOL, I had a huge performance DECREASE on my K6 233, from about 1.5 mnodes to .97

My Celeron also saw a slight decrease, from 3.7 to 3.55 mnodes

Waiting for the next release, but I'll drop Cyp an email.

viz

edit-
I am going to run a WU or 2 on the K6 too see if that rate is correct. I agree that the benchmarks are not quite the same, but this AMD difference is something.

 

[RC]

Moose - I agree. I will be glad to try the new win32 client and report my findings.

If I understand your remark, the sooner Cyrus receives feedback on the new WIN32 client the sooner we might see a new Linux client.

I will report to Cyrus our findings on the current Linux client and OGR. I have sent email on the matter previously, but I do not recall whether I sent the info to Cyrus.

We appreciate all the work you fellows put into this project.

Without you folks there would be no RC5 or OGR.

Regards,

RC

 

[sciencewhiz]

Moose,

Is there a new changes.txt that lists the changes in the pre-release clients? The one linked to from the pre-release page is the standard one that only covers the released clients.

 

[stappel]

Is there an experation date on these clients? The AIX prerelease has a internal experation date of 5 nov. Lucky I saw it before installing it on more machines.

The client runs oke on my P3-550. is about 0,8% slower on multiple units.
It look like it getting the wrong size packets. I request size 30, but it crunches 28 packets. I shall test some more.

 

[nukefarmer]

the ogr benchmark is around 15% faster with the new client (2.60 Mnodes/s <-> 2.99 Mnodes/s) on Cel@466 here.
The 463 client benchmarked speed is quite close to what i get in real life, so i suppose the 464 benchmark won't be far off also.
I'll give it full run later when current stub is completed (don't want to waste a day cracking )

 

[Moose]

Cyrus Updated the beta bin. I've placed it on the pre-release download page. v2.8011.464b. For those that asked where a new changes.txt file is, if you download the client you will find it in the download. I won't be placing a new one on the web page untill the client is released

moose

 

[Engineer]

Moose (and Viz),

I tried the client (464b) at work this morning and I can't get to a keyserver either. We are using MS Proxy Server 2.0 and I have the MS Proxy Client wrapper installed which usually gives almost unrestricted access to the net. The client will resolve the keyserver name to IP address then simply time out...Not sure why? 4.63b works just fine.

 

[Moose]

Please make sure you email cyrus about this

thanks
moose

 

[Viztech]

OK-

The second beta has resolved some of the issues, but I am having the same trouble with it as Engineer-being able to get out from behind the firewall where we have been able to with the 463b client.

Cyp mentioned in his return mail that there will be two OGR cores - one for PPro and above, and one for P5 and below.

It looks as though we are in a true beta test of the network functionality at this point, and this second beta version is moving in the right direction.

BTW people, you can download the betas into a seperate directory, shut down your normal client, and execute the beta for testing. This will have no adverse affect on your existing buffers and ini.

viz

edit- Yes Moose, contacting Cyp direct on these at Cyrus Patel cyp@fb14.uni-mainz.de

 

[Engineer]

Update:

While the client wont connect to the network normally (with the MS proxy client), it will still connect through the main firewall here using the Gateway address, login and password and http port 80 pproxies. Hope this helps.