GHz-Days

[nomad]

On my account summary the sum of the GHz-days of the last 365 days is 267.
But when i look at the CPU-section the sum is above 300 GHz.
(I´ve waited a few hours for updating but nothing happened.)

[cheesehead]

Originally Posted by nomad

On my account summary the sum of the GHz-days of the last 365 days is 267.
But when i look at the CPU-section the sum is above 300 GHz.

On the http://mersenne.org/cpus/ page, titled "Computer Details", the totals under "GHz-Days" are not restricted to the most recent 365 days.

E.g., my own 365 day total on my account summary page is 833 GHz-days, but adding up the totals on my "Computer Details" table gives 3275 GHz-days.

[borgyg]

Over the past several months it seemed that my results page displayed more GHz-days results than what was being added to my total under Top Producers. Now, over the past week, I've added 991.3 GHz-days when I added up my results (Feb 23-Feb 28) yet my Top Producer total increased from 28910 to 29398, an increase of only 488, less than half.

In fact, I just moved my results over the past 365 days into a spreadsheed and added up my GHz-days: 29398. I've been a member since around 1999 and seem to be missing years of results being credited to my account. What can be done, if anything, to get my results over the years credited properly?

[cheesehead]

Originally Posted by borgyg

Over the past several months it seemed that my results page displayed more GHz-days results than what was being added to my total under Top Producers.

On your own results page are totals for both "Stats for the Last 365 Days" and "Lifetime Stats", right?

Now, over the past week, I've added 991.3 GHz-days when I added up my results (Feb 23-Feb 28) yet my Top Producer total increased from 28910 to 29398, an increase of only 488, less than half.

In fact, I just moved my results over the past 365 days into a spreadsheed and added up my GHz-days: 29398. I've been a member since around 1999 and seem to be missing years of results being credited to my account. What can be done, if anything, to get my results over the years credited properly?

On the Top Producers page, under the "Customize..." button is a subtitle:

Top producers for the last year.

So, on that page only the totals for the last 365 days are shown, not lifetime stats. The match with your results over the past 365 days is exactly what one would expect.

That also explains why the addition of 991.3 GHz-days raised your (Last 365 Days) total by only 488 GHz-days -- Apparently, 366 days or so ago you had reported 503 GHz-days. So 991 GHz-days were added to your past-365-days total, but 503 GHz-days became a more-than-365-days-ago contribution, so were subtracted.

Yours isn't the first misunderstanding we've had.

I'll recommend that the subtitle on the Top Producers page be changed to:

Top Producers During the Last 365 Days.

so that it matches the time period description on the individual account result page ... and perhaps put it in bold lettering.

[borgyg]

That explains it...and puts my "lifetime" GHz-Days to 63841. I guess I'm glad that I'm not the only one to miss it. Thank you for the clarification!

I was wondering if there was an easy-to-understand explanation of how a user gets 23GHz days from ONE trial factor (or more accurately, 92 from 4 TFs)

I've seen some wildly high numbers in P1 as well.

(pardon hijak, but first GHz days thread available)
k

[Prime95]

Originally Posted by Unregistered

I was wondering if there was an easy-to-understand explanation of how a user gets 23GHz days from ONE trial factor (or more accurately, 92 from 4 TFs)

If you have more details please post them. It will make it easier for us to investigate and explain.

[Mini-Geek]

Originally Posted by Unregistered

I was wondering if there was an easy-to-understand explanation of how a user gets 23GHz days from ONE trial factor (or more accurately, 92 from 4 TFs)

I've seen some wildly high numbers in P1 as well.

(pardon hijak, but first GHz days thread available)
k

It depends on what sort of TF or P-1 assignments they were. Both can certainly take that much time and give that much credit. The simple explanation, then, is that you could get 23 GHz days from one TF because it took that much time/work, because the bit level was high enough to make it so. This would happen in the normal course of pre-LL checks if the number was large enough, e.g. for numbers over 100 million digits.
Note that since factors of Mersenne numbers must be of the form 2kp+1 where p is the exponent and k is any integer over 0, the larger the exponent is, the fewer possible factors there are to check in a given bit range. For example, with p=4M, factoring from 2^69 to 2^70 takes about 30 GHz-days, and with p=332M, (which is ~100M digits) 2^75 to 2^76 is about 23 GHz-days. Even though the factors are larger in the latter case, there's fewer to check than in the first case because they're spaced out over multiples of 332,000,000 instead of 4,000,000.
Without further details of the boundaries of the tests, I can't give a more exact answer.
These pages might interest you:
http://mersenne-aries.sili.net/credit.php
http://mersenne-aries.sili.net/factorbits.php

[firejuggler]

to give you an example, with a gtx460, running an exponent of the billion digits project from 81 to 82 bits take me 48 hours of constant running (and i'm cpu bound).
it would be equal to 147.424382 GHz-day . for just one exponent (see mini-geek link)

[Uncwilly]

That is about what one would get taking an exponent in the 332,000,000 range from 75 to 76. It is 5.7 from 73 to 74.

[firejuggler]

I realize that my above post is not really understandable.
In 48 hours, thanks to mfaktc, I TF'ed M3 321 928 241 from 81 to 82 bits. This would grant me about 147 GHz-day. All that in less than 48 hours (and i'm CPU bound).