PSI, mpg, GHz P4-days and P90 years

[Batalov]

I now that pounds per square inch are related to atmospheric pressure, something like 15psi, right?
Miles per gallon is another ludicrous monster... apparently can be related to liters per 100km, but I care much more about what is a GHz P4-days in P90 years (which are as most of you know, forever)?

I did f'ing google!

[Batalov]

I'll answer it myself, then.

1 P90year = 6.085 GHz P4-days
0.16434 P90years = 1 GHz P4-day

Established empirically (by observing credits for similar jobs on V4 and V5 servers)

[Prime95]

// In Primenet v4 we used a 90 MHz Pentium CPU as the benchmark machine
// for calculating CPU credit. The official unit of measure became the
// P-90 CPU year. In 2007, not many people own a plain Pentium CPU, so we
// adopted a new benchmark machine - a single core of a 2.4 GHz Core 2 Duo.
// Our official unit of measure became the C2GHD (Core 2 GHz Day). That is,
// the amount of work produced by the single core of a hypothetical
// 1 GHz Core 2 Duo machine. A 2.4 GHz should be able to produce 4.8 C2GHD
// per day.
//
// To compare P-90 CPU years to C2GHDs, we need to factor in both the
// the raw speed improvements of modern chips and the architectural
// improvements of modern chips. Examining prime95 version 24.14 benchmarks
// for 640K to 2048K FFTs from a P100, PII-400, P4-2000, and a C2D-2400
// and compensating for speed differences, we get the following architectural
// multipliers:
//
// One core of a C2D = 1.68 P4.
// A P4 = 3.44 PIIs
// A PII = 1.12 Pentium
//
// Thus, a P-90 CPU year = 365 days * 1 C2GHD *
// (90MHz / 1000MHz) / 1.68 / 3.44 / 1.12
// = 5.075 C2GHDs

[cheesehead]

Originally Posted by Prime95

// Thus, a P-90 CPU year = 365 days * 1 C2GHD *
// (90MHz / 1000MHz) / 1.68 / 3.44 / 1.12
// = 5.075 C2GHDs

... and how much is that in grams-turkey per fortnight?

If GIMPS had adopted my suggested PC-independent standard a few years ago, you wouldn't be at the mercy of techology's forcing you to do all that repeatedly. We'd already have a work-unit that can stand the test of time. For which we could be thankful each and every succeeding year.

- - - -

Hey, how come a forum search for turkey or microwave doesn't find a certain thread from a few years ago, like it used to?

[davieddy]

Having recently been reminded that a US gallon
differs from a UK one, I am still pondering whether this
makes petrol more or less expensive here.

BTW what was your PC-independent standard?

[petrw1]

Originally Posted by Prime95

In 2007, not many people own a plain Pentium CPU, so we adopted a new benchmark machine - a single core of a 2.4 GHz Core 2 Duo. Our official unit of measure became the C2GHD (Core 2 GHz Day). That is, the amount of work produced by the single core of a hypothetical 1 GHz Core 2 Duo machine.

When I look at the V5 statistics page I get the impression the new unit of measure is going to be GFLOPS, since that is what I see on the graphs,while seeing no mention of C2GHD.

As long as we continue to base it on the CPU of the day are we not setting ourselves up for another needed adjustment in another few years? Would not something more generic like GFLOPS be more time persistent?

Just my 2 cents (1.9 Canadian Pennies) worth.

[ATH]

From v4 server stats:

Test Type CPU yr/day GFLOP/s
------------ ---------- ----------
Lucas-Lehmer 2527.820 30429.135

You can calculate how many GFlop 1 P90 cpu year is:
30429.135 GFLOP/s * 86400 s/day / 2527.820 CPUyr/day = 1,040,057 GFLOP/CPUyr(P90)

This calculator on Team Prime Rib can calculate how many P90 yrs per LL test, but I'm curious what the formula is. Its based on FFT size:
http://www.teamprimerib.com/rr1/bin/calc_p90.php

[S485122]

The formula is known see thread Number of P90 years per exponent

Jacob

[Batalov]

From that thread

Originally Posted by leifbk

...
    if exp > 35100000: exp_p90 = exp *  6.743
    if exp > 40250000: exp_p90 = exp *  9.02
    if exp > 50000000: exp_p90 = exp * 10.945
    if exp > 59400000: exp_p90 = exp * 13.09
....

I wonder if thе 39,500,000 limit (where 2560K FFT kicks in) is used now instead of 40,250,000 in the formula on the server. If it isn't, it may be unfair to the people doing the numbers between these bounds - they receive 25% less credit than they did work (they almost inevitably use 2560K FFT, not 2048K). All other limits are shifted, similarly.

[Batalov]

Originally Posted by petrw1

...Just my 2 cents (1.9 Canadian Pennies) worth.

Here's an example that one cannot rely on anything historically "axiomatic" (USD vs CAD) - these days, not only USD is lower than a euro, -- it is ridiculously lower. Now it is lower than CAD and australlian dollar, too... and the yen is next... Well, maybe first the rouble, then the yen.

Just my 0.02 yen.

[S485122]

Originally Posted by Batalov

I wonder if thе 39,500,000 limit (where 2560K FFT kicks in) is used now instead of 40,250,000 in the formula on the server. If it isn't, it may be unfair to the people doing the numbers between these bounds - they receive 25% less credit than they did work (they almost inevitably use 2560K FFT, not 2048K). All other limits are shifted, similarly.

As far as I know the formula has not been updated to reflect the new exponent limits for chosing the FFT size for SSE2 processors. But who said the system should be fair ? Look at what you earn for trial factoring when not successfull. And even worse : when doing P-1 factoring and not finding a factor one earns 0,001 P90 CPU year ! (about 700 times less credit per time unit !)

Some of the parameters will change with the V5 server.

Jacob