Scoring question for test client

[Meadmaker]

I've just uploaded my first 50 generations using the new test client, and I see a slight problem with the scores awarded.

The test client runs 100 structures/generation, so the scoring should be 100*(sq rt of generation #), but it seems to be only crediting 50*(sq rt of generation #)

Can someone please clarify this for me?

TIA.

[Mikus]

Originally posted by Meadmaker
The test client runs 100 structures/generation, so the scoring should be 100*(sq rt of generation #), but it seems to be only crediting 50*(sq rt of generation #)

For me, 100 structures (one generation) with the test client seem to take about the same amount of CPU execution time as 50 structures (one generation) took with the normal client. So I don't mind that the 'points' per generation for the test client are the same as the 'points' per generation for the normal client.

mikus

[Brian the Fist]

Originally posted by Meadmaker
I've just uploaded my first 50 generations using the new test client, and I see a slight problem with the scores awarded.

The test client runs 100 structures/generation, so the scoring should be 100*(sq rt of generation #), but it seems to be only crediting 50*(sq rt of generation #)

Can someone please clarify this for me?

TIA.

I will clarify - it is a TEST client. Once it is official, the points will adjust accordingly. Right now points are awarded by the server which does not care how many structures are in a generation (i.e. it cannot tell).

[Meadmaker]

No problem, Howard - thank you for letting me know so promptly.

The new client does seem to be about 15-25% faster on my rigs, but so far I'm not seeing any lower scoring - hopefully they will come as it gets nearer to generation 250.

[Anteraan]

Thanks for the update there, Howard. However, I have another question is this same vein. Is there any rough timeframe for when the test client might go official? I've run my first 250 through, and am well on my way to the next 250. I can happily report no problems whatsoever, and my scoring has increased, despite the 50/100 issue. For the record I'm running the Windows version (text client) on an AMD 2000+ system with XP and a gig of TWINX.

I realize that my situation is not the same as many others', but I wanted to turn in some positive feedback. This new client has made this wickedly twisted protein much more bearable. I've gone from to in just a few days. Yes, I do it for the science (I'd run the program anyway), but I also like to see progress on some level, and it was coming slowly with the previous (technically current) algorithm. The new one just moves much more quickly, especially in terms of the tight spot processing. It works more quickly, and seems to be needed far less often. So far, I'm nothing but pleased with this client. I wish I'd have gone to it the first day it came out. Well done!