A GHz-Day

[LiquidNitrogen]

OK, I am expecting I will be flogged as a result of asking this, but I searched around and couldn't find an answer.

What the heck is a "GHz-Day" unit of measure?

[fivemack]

It's half a day of calculation on (I think) a 2GHz Northwood Pentium-4, or eight hours on a 3GHz Northwood Pentium-4.

The prime95 code stays sufficiently resident in cache that it scales well with CPU speed, so multiplying by the gigahertzage and the number of cores is not an irredeemably dreadful way to compare throughput of different machines. Though when the CPU architecture changes a lot and George improves the FFT code, you can get another constant factor coming in.

[R.D. Silverman]

Originally Posted by fivemack

It's half a day of calculation on (I think) a 2GHz Northwood Pentium-4, or eight hours on a 3GHz Northwood Pentium-4.

The prime95 code stays sufficiently resident in cache that it scales well with CPU speed, so multiplying by the gigahertzage and the number of cores is not an irredeemably dreadful way to compare throughput of different machines. Though when the CPU architecture changes a lot and George improves the FFT code, you can get another constant factor coming in.

Everyone should read my article:

Exposing the Mythical MIPS Year
IEEE Computer Vol 32 1999

[Uncwilly]

http://www.mersennewiki.org/index.php/Computing_power

Has the GHz-day and the venerable P90-year

Originally Posted by R.D. Silverman

Exposing the Mythical MIPS Year
IEEE Computer Vol 32 1999

Handy link http://academic.csuohio.edu/yuc/perf...tical_MIPS.pdf

[davieddy]

Originally Posted by LiquidNitrogen

OK, I am expecting I will be flogged as a result of asking this, but I searched around and couldn't find an answer.

What the heck is a "GHz-Day" unit of measure?

P90 years forever!

Doesn't anyone remember cgs units?

David

[ATH]

Originally Posted by Uncwilly

http://www.mersennewiki.org/index.php/Computing_power

Has the GHz-day and the venerable P90-year

The use of TFLOPS and GFLOPS as the plural of TFLOP/GFLOP is very confusing. This is a measure of work done, while normally TFLOPs/GFLOPs is a measure of computation speed (TFLOP/s and GFLOP/s).

[science_man_88]

Originally Posted by LiquidNitrogen

OK, I am expecting I will be flogged as a result of asking this, but I searched around and couldn't find an answer.

What the heck is a "GHz-Day" unit of measure?

every hear of a kilowatt-hour ?

if so you know what that means you should be able to find out what a GHz-Day is.

kilowatt-hour = running at 1 kilowatt for 1 hour. a GHz-day last I checked would be a processor running at 1 GHz for 1 full day. my outdated and supposedly obsolete processor from what I've founds can max out at 2.8 GHz holding it there for 1 Day = 2.8 GHz* 1 day = 2.8 GHz-Days. but not all that can go to prime 95 as the OS has to run.

[science_man_88]

Originally Posted by science_man_88

every hear of a kilowatt-hour ?

if so you know what that means you should be able to find out what a GHz-Day is.

kilowatt-hour = running at 1 kilowatt for 1 hour. a GHz-day last I checked would be a processor running at 1 GHz for 1 full day. my outdated and supposedly obsolete processor from what I've founds can max out at 2.8 GHz holding it there for 1 Day = 2.8 GHz* 1 day = 2.8 GHz-Days. but not all that can go to prime 95 as the OS has to run.

http://answers.yahoo.com/question/in...1100447AAeZ0E9 is similar.

[LiquidNitrogen]

Originally Posted by science_man_88

a GHz-day last I checked would be a processor running at 1 GHz for 1 full day.

I understood the concept, I was wondering based on what architecture. The link showed this:

GHz-days

The work accomplished by one core of a hypothetical 1GHz Core 2 Duo CPU in one day.



BTW, thanks for the link Uncwilly.


The chess community uses a 1.0 GHz Pentium III for a great deal of their hardware comparisons. See the link below for more info:


http://www.jens-hartmann.at/Fritzmarks/


Notice our machine at #2 by a mere 0.64 GHz on the PIII scale achieved the result with 6 cores x 2 with hyperthreading on, while the #1 machine had 12 real cores (Dual Westmere).

Folks,
I've been thrashing some spares servers for a while now and have seen some wild flucuations in my GHz/Days total. Has anyone experienced this before and is there any point caring about it?

regards,
andrew

[Uncwilly]

Originally Posted by Unregistered

I've been thrashing some spares servers for a while now and have seen some wild flucuations in my GHz/Days total. Has anyone experienced this before and is there any point caring about it?

If you are doing primality testing (the L-L test) and you started your machines nearly the same time, they all would report their results at about the same time. This would cause your GHz-days to jump up quite quickly and then seem to idle until the next round is reported.