Version 1.0.6 - Few questions

[Minty]

Hi
I've just installed version 1.0.6 and found that it's much faster than 1.0.4 on my P4 2.26GHz (windows XP). It will now run at a bit above 9 Megamd5, whereas before it would run at 6.6-6.7 - this is a MAJOR improvement! (I feel like a bit of a lemon running so slow for a couple of weeks - didn't read anywhere about speed boosts - just thought they were bug updates) I have a few questions though - when I copy the new .exe into my old folder to preserve the old start points, I get only 7.2 whereas if it's in the installed folder 9 no probs - why is this - especially as the control files are only text files? Also if running the test switch in the old folder, it is a bit lower than new install (10.66 max down to 10 using SSE2ALU2). Are there going to be any more clients coming soon with similar speed boosts?

Also when running the client I hear the hard drive chuntering maybe once every 5 seconds, but with longer delays sometimes - it's not transferring much data (not even enough to bring the light on mostly) but definitely down to md5crk, is this normal and why access the HD so often - couldn't it run in memory just writing the DPs to disk?

Finally, from the website I get the impression that the Linux x86 client is faster than the Windows client per MHz - if so why is this?
Just to round things off my PC takes 0.16 Amps (@ 245V) more on 100% CPU than 0% - 40W so in a working day this could be as much as half a unit of leccy. Not much but I guess if ur meter watching it makes a difference!

Please help

[alpha]

Originally posted by Minty
when I copy the new .exe into my old folder to preserve the old start points, I get only 7.2 whereas if it's in the installed folder 9 no probs - why is this - especially as the control files are only text files?

The first time you start the client (from a fresh installation), it benchmarks all the cores and selects the fastest one. If you overwrite the client with a newer version, it does not repeat these benchmarks. It sounds the client is selecting a different core on a fresh installation on your computer, hence the difference in speed.

Or maybe you were running another CPU-intensive task the first time you started the client, and it affected the results of the initial benchmark, therefore selecting the wrong core?

[Minty]

Thanks for the reply alpha - everything was re/benchmarked using the test switch and the right core was selected. I took a look at it again today and found that it runs properly (bout 1-1.5 Megamd5 faster) if u just copy over the md5_user,send,save etc. to the new dir when upgrading instead of switching .exe and recalibrating - dunno why it's a bit bizarre really. Am over the moon to have 9 - it's blasting along and a whole lot better than 6.7!
However, I'm still a li'l green eyed bout those Mac G5's pulling 36 - might just leave em to it whilst I go retro and have a go at the DES cracking proj. (sorry for the plug!)

[brif8]

Where are people getting 36's for G5, then there must be something wrong with mine BIG time I only get 18's

[Mystwalker]

I guess that are Dual G5s, ain't it?

[Minty]

Yes - dual G5's are the order of the day! Is your G5 a dual processor machine? (it sounds like 2GHz from the numbers) If so then just install md5crk again in a different directory and run it from there - voila - you should now have two instances running at about the 18megamd5/sec each! If you don't have dual processors then there's not a lot to be done - but you're still double me and flying so no worries!
I'm no expert on Macs, but this should work fine.

(I think somewhere someone mentioned that they ran 2 instances like this and after a while speed dropped - memory leak or temperature increase or something - could be to do with long periods of time running or some other software prob etc. so just keep an eye on speed at least at the beginning and maybe restart if a problem)

Good luck!