2^p-1
Internet PrimeNet Server
Parallel Technology for the Great Internet Mersenne Prime Search

Internet PrimeNet Server Statistics Charts
Parallel Technology for the Great Internet Mersenne Prime Search

Between 23 November 1997 and 04 February 2006, PrimeNet has handled 11,579,649,914 P90 machine-hours (1,320,973 P90 machine-years) of Mersenne number tests. PrimeNet's handled throughput rate is currently best characterized by a fitted, exponential trend line, however 2005 showed a temporary 6-month period of decline before resuming its climb.

Notice that after most deep dips in the curve there's a high spike right after it. The deep dips are service outages and the spikes are recovery activity, reflecting PrimeNet's ability to automatically recover from outages.


Earlier Charts, 1997/11/23 to:
1998/03/13, 1998/03/31, 1998/07/08, 1998/10/04, 1998/11/30,
1999/02/18, 1999/06/28, 1999/08/25, 1999/11/24,
2000/02/10, 2001/03/10,
2003/11/15, 2004/04/26

26 May 2004 - Discovery of M#40 in November 2003 doubles global GIMPS participation, rocketing PrimeNet's aggregate performance to an amazing sustained throughput of 14 teraflops!

PrimeNet has handled 4,223,723,933 P90 machine-hours (481,830 P90 machine-years) of Mersenne number tests since starting 23 November 1997.

15 November 2003 - Continued strong global GIMPS participation, new CPU models and matching code updates have brought PrimeNet's aggregate performance to a sustained throughput of nearly 9 teraflops!

10 March 2001 - PrimeNet's statistics show a current sustained throughput of over 1.4 teraflops!

10 February 2000 - PrimeNet has handled 276,743,000 P90 machine-hours (30328 P90 machine-years) of Mersenne number tests, and shows a sustained throughput of 1.15 teraflops!

24 November 1999 - PrimeNet completes its second year serving GIMPS on the Internet! Its statistics show a current sustained throughput of 0.95 teraflop!

25 August 1999 - PrimeNet's statistics show a current sustained throughput of over 0.8 teraflop , proceeding into September at well over 0.85 teraflop with spikes over 0.9 teraflop (measured in calibrated P5 90Mhz, 32.98 MFLOP units: 25658999 FPO / 0.778s using 256k FFT).

9 June 1999
- PrimeNet's statistics show a current sustained throughput of over 0.7 teraflop (measured in calibrated P5 90Mhz, 32.98 MFLOP units: 25658999 FPO / 0.778s using 256k FFT).

30 April 1999
- PrimeNet's statistics show a current sustained throughput of over 0.6 teraflop (measured in calibrated P5 90Mhz, 32.98 MFLOP units: 25658999 FPO / 0.778s using 256k FFT).

19 February 1999
- PrimeNet's statistics show a current sustained throughput of 0.5 teraflop, or over 40 P90 CPU years per day (measured in calibrated P5 90Mhz, 32.98 MFLOP units: 25658999 FPO / 0.778s using 256k FFT).

31 January 1999
- At 37 P90 CPU years/day, 7500 accounts are actively testing Mersenne numbers. There are over 13,000 machines applied to GIMPS, with 1.3 machines per account, and an 'average machine' of 130Mhz Pentium/II running 18 hours/day.


4 October 1998 - At 24.2 P90 CPU years/day (210240 P90 CPU hours/day), 4900 accounts are actively testing Mersenne numbers. 10 accounts are producing over 1000 CPU hours/day, which is about 16 computers.  4 of those accounts are over 2000 CPU hours/day, and one is over 3000 CPU hours/day.    Ignoring the fact that many machines run only during business hours, this is an average of roughly one P133 computer per account, or 43 CPU hours/day.

27 June 1998 -
At 17.0 P90 CPU years/day, GIMPS PrimeNet is has had steady new user growth, 760 active new accounts. This period marked a major transition milestone, where a client software change is needed to test larger Mersenne exponents.

1 May 1998
- At 12.8 P90 CPU years/day, PrimeNet is handling about 99% of the GIMPS total workload.   The slowing in the CPU throughput rate is in part due to the larger Mersenne exponents recently assigned - the machine that would have returned a result in a few weeks now require about a month. If this is accurate, the CPU rate should start to rise a bit more quickly again, later in May, and continue increasing through mid-June.

1 April 1998 - This data is assembled from PrimeNet's transaction logs.  Prior to 23 November 1997 the logs did not have entries for CPU time, even though the server tracked it in its tables. Time resolution is to the nearest second in the raw data, but here we see detail only to the nearest day. The jumpiness of the red throughput rate line is in part due to binning the CPU time into particular days.

UTC	PST	EST	CPU yrs

0:00	16:00	19:00	342.8
1:00	17:00	20:00	357.6
2:00	18:00	21:00	346.3
3:00	19:00	22:00	338.6
4:00	20:00	23:00	326.1
5:00	21:00	0:00	303.9
6:00	22:00	1:00	297.9
7:00	23:00	2:00	305.1
8:00	0:00	3:00	277.7
9:00	1:00	4:00	266.3
10:00	2:00	5:00	266.8
11:00	3:00	6:00	281.0
12:00	4:00	7:00	304.7
13:00	5:00	8:00	326.7
14:00	6:00	9:00	344.0
15:00	7:00	10:00	343.4
16:00	8:00	11:00	362.5
17:00	9:00	12:00	370.2
18:00	10:00	13:00	370.7
19:00	11:00	14:00	360.8
20:00	12:00	15:00	362.0
21:00	13:00	16:00	358.9
22:00	14:00	17:00	368.7
23:00	15:00	18:00	334.0
cpu_hist1.gif (29547 bytes)

[ World Test Status | Account Reports | Top Producers | News & Info | Statistics | FAQ ]
[
Manual Testing | GIMPS Mersenne Prime Search ]


Web support, contact primenet@mersenne.org.
© 1998-2006
Entropia.com