FAD: May-09 mondobyte's doing the TOP-DAWG dance today!

[Freewolf]

Originally posted by: GeoffS
I get the impression that there's more to it than just the CPU speed... the hardware also comes into play based on what I'm seeing with my machines... for example (and I don't have any of the cool hardware you guys do :roll: ) I have 3 XP1700+ machines... they are averaging 135, 140, and 145 respectively. My 2000+ machines are running 159, 161, 163 and 168... my dual 1900 does 154 on each, whereas my single 1900s are doing 149 and 152... go figure...

What cool hardware would that be? I don't have any cool hardware, most of it gets pretty darn warm 😛

 

[BadThad]

Originally posted by: GeoffS
I get the impression that there's more to it than just the CPU speed... the hardware also comes into play based on what I'm seeing with my machines... for example (and I don't have any of the cool hardware you guys do :roll: ) I have 3 XP1700+ machines... they are averaging 135, 140, and 145 respectively. My 2000+ machines are running 159, 161, 163 and 168... my dual 1900 does 154 on each, whereas my single 1900s are doing 149 and 152... go figure...

Yes, it's quite odd. I have two "over-performers" in my herd:

1) An XP2000+ running at 1750MHz, CPU rating = 167. Old Via KT266A chipset w/ PC2100 ram!
2) Dell P4 2.53GHz (stock), Win 2000, 256MB RAM, CPU rating 158. I have two other 2.53GHz systems, both overclocked in the 2.6GHz range and they only give me a rating of 135-140.

This makes NO SENSE! 😕

 

[JTWill]

You cant go by the difference of AMD and Intel, different design. AMD is better for floating point calculations which is what FAD is.

 

[mondobyte]

Originally posted by: BadThad
Didn't mondo mention (or hint) of some sort of hack to Linux that can inflate your points? I don't understand how that would work, according Geoff's (and mine) calculations CPU Rating is nothing more than Points per Hour, which is calculated from gigaflops. Gigaflops is primarily based on the CPU's FPU power, secondarily it's based on the subsystem surrounding the CPU since data has to be moved around to keep the CPU "fed".

So, from my understanding, most of "CPU Rating", in theory, should be derived purely from the CPU clock speed, FPU efficiency and L1/L2 cache. The rest should not make that significant difference unless something on the system is totally FUBAR.

Oh man.....😕

Ok teammates ... I will not teach you how to "fudge" your CPU Rating with Linux/Windows. I will however say that it is well documented that most Linux distros will give a significantly rating higher than Windows 2K/XP/2003. There are certain Linux distros that are profoundly higher even when compared with other Linux distros. I will also say that there are certain hardware tweaks that really are a "g spot" for the FaD rating calculation in the irrespective of the OS. From my tweaking, CPU speed is important but I get higher numbers at less than the max stable overclock I for a CPU. It is consistently the case with various motherboards and various processors.

I would be curious as to the distro that you were running that gave the different CPU ratings. Which one and which version?

mondo

 

[mondobyte]

Originally posted by: BadThad

Originally posted by: GeoffS
I get the impression that there's more to it than just the CPU speed... the hardware also comes into play based on what I'm seeing with my machines... for example (and I don't have any of the cool hardware you guys do :roll: ) I have 3 XP1700+ machines... they are averaging 135, 140, and 145 respectively. My 2000+ machines are running 159, 161, 163 and 168... my dual 1900 does 154 on each, whereas my single 1900s are doing 149 and 152... go figure...

Yes, it's quite odd. I have two "over-performers" in my herd:

1) An XP2000+ running at 1750MHz, CPU rating = 167. Old Via KT266A chipset w/ PC2100 ram!
2) Dell P4 2.53GHz (stock), Win 2000, 256MB RAM, CPU rating 158. I have two other 2.53GHz systems, both overclocked in the 2.6GHz range and they only give me a rating of 135-140.

This makes NO SENSE! 😕

It makes perfect sense to me ...

I have some "overperforming" windows machines myself. There's some very interesting tweaks that can be made to the hardware that really sends the FaD CPU Rating calculation into orbit so to speak ... and AMD or Intel is a factor but the tweaks are more of a factor!

 

[Freewolf]

What start the whole thing is I happened to see a thread where the statement was made that linux was faster and Fad gave higher cpu rating in linux than in windows. Gentoo in this case. As an old seti cruncher I was surprised to see this since in seti windows is the faster of the two.
I thought this info could be of use to anyone on the team who knew how to setup and use linux so I posted it here with no thoughts of "fudging" points indeed if I wanted to do that I happened upon a thread a while back that would be much easier to do than trying to learn linux would be.

 

[mondobyte]

Perhaps I was misunderstood ...

Linux and tweaking are not "fudging" or cheating. I do know how to get CPU ratings that cannot be supported based on current hardware. I won't go there in this discussion.

I am more than willing to help others with Linux though and if you are going to run Linux, you may as well run one of the most advantageous distros.

I am more than willing to share how to tweak the hardware for the "sweet spot".

Unfortunately, most of my clients must run Windows so I have no options.

 

[JTWill]

I still cant install fad by the instructions.
bash:mkdir /usr/local/think: No such file or directory.................I am very angry thats what the mkdir is supposed to do is make a directory
SuSE 8.2 pro. You know the best distro I'd like to know what it is.

 

[GeoffS]

Are the other dirs there too? /usr and /usr/local? That's the only reason that comes to mind that the mkdir would fail...

 

[JTWill]

Its because this idiot didnt put a space in......... mkdir(space)/ 😀

 

[BadThad]

Originally posted by: mondobyte

Originally posted by: BadThad

Originally posted by: GeoffS
I get the impression that there's more to it than just the CPU speed... the hardware also comes into play based on what I'm seeing with my machines... for example (and I don't have any of the cool hardware you guys do :roll: ) I have 3 XP1700+ machines... they are averaging 135, 140, and 145 respectively. My 2000+ machines are running 159, 161, 163 and 168... my dual 1900 does 154 on each, whereas my single 1900s are doing 149 and 152... go figure...

Yes, it's quite odd. I have two "over-performers" in my herd:

1) An XP2000+ running at 1750MHz, CPU rating = 167. Old Via KT266A chipset w/ PC2100 ram!
2) Dell P4 2.53GHz (stock), Win 2000, 256MB RAM, CPU rating 158. I have two other 2.53GHz systems, both overclocked in the 2.6GHz range and they only give me a rating of 135-140.

This makes NO SENSE! 😕

It makes perfect sense to me ...

I have some "overperforming" windows machines myself. There's some very interesting tweaks that can be made to the hardware that really sends the FaD CPU Rating calculation into orbit so to speak ... and AMD or Intel is a factor but the tweaks are more of a factor!

Well? What's the secret? I'm an avid hardware tweeker, just look at my Rigs link. Every PC I own is overclocked and tweeked to the max. If it boosts the rating, then it's doing more work since rating is just points per hour....that's not cheating.

 

[Freewolf]

bty if there is anything you don't want to say here on anandtech we can always move to the RA forums.

 

[Wolfsraider]

sounds good 😛