Titanic questions

[snme2pm1]

Originally Posted by Madpoo

How about this one then... instead of a solder bridge, just a "whatever" length of wire, where the length is determined by what resistance you want? After all, we're talking about some small resistance anyway, so a coil of 20 feet of whatever gauge copper, perhaps?

It is lots of years since I have looked up wire resistance tables also, but the 20 feet figure you speculated sounded alarming.
A coil (as you suggested) created from such a length might to some extent be described as an inductor and its reactive component might frustrate the purpose of current sensing.
Another aspect is that the resistivity varies with temperature, so you don't want it to warm significantly.
A quick google suggests 18AWG, being diamater 0.0403" or 1.024mm, has resistance of about 20.9 milli ohms per metre.
So, if my arithmetic is in tact, based on that, for example, a 5 mOhm section would be about 240 mm.
Sorry for the SI units as used in this part of the world.

Now a further cautious comment, and without wanting to highlight and draw further attention to what I regarded as an unwarranted interpretation of my comments back a little way.
I do not regard any participant at this forum in a disrespectful manner.
If anyone has felt any offence from anything I have written, please believe that there was no hurt intended at all.
My participation is intended to be helpful and further discuss things, and learn more from others also.

[LaurV]

Originally Posted by TObject

Shortly after these fabulous GPUs were introduced it was shown, on mersenneforum.org, that to yield reliable LL results, Titans have to be down clocked (particularly in the memory department). Is this now outdated information?

No, the cards still run better (more efficient "output results vs power" ratio) when downclocked, and it is more stable. But some people don't care about consumption and want to push it to the limits. And the "new tech" in cooling helps us.

Otherwise, why all the talk about overclocking: to churn out incorrect residues faster?

My opinion (and it may be general opinion too) is that it only makes sense if you have two, and run then in parallel for the same (LL+DC) test, to compare/self-check residues at every checkpoint. Otherwise you will lose time with tests you will have to repeat in case of mismatch (or waste other people's time to DC/TC/QC etc). Now PrimeNet accepts two results from cudaLucas, assuming they have different shifts (which may not be so good, because such results are sent manually, and some idiots can try to cheat the system, for credit, but we like the way it is now, it helps the project more, and we and are optimistic that the PrimeNet police will catch/discourage such guys, and that is another subject anyhow ).

The talk about overclocking in this thread is not really a talk, we only play "bâza". (sorry for the quality of the link, and how do you call this game in English?)