FAH - x2-3800 v C2D 6300 = AMD Wins?

FatRakoon

FatRakoon

FatRakoon

Soldato
Joined
18 Oct 2002
Posts
10,428
Location
Behind you... Naked!
Got a funny one here.

Im playign about with a few of my PCs setups, and I have run a few tests on them with FAH. ( Bloody takes ages, but why not )

Anyway, im having a look at my stats with FAHMON

Now, I have seen somethign very interesting..

Ok, what I have been messing about with the last few days is simply getting them to run the same projects and simply comparing them

Funny one this, but I have found, that my conroe is the slowest CPU for folding according to FAHMON anyway.

What I have is this...These are a couple of the current projects FWIW

Project : 1809

Opteron 144
= Avg. Time / Frame : 10mn 14s - 215.30 ppd
Conroe ( Core 1 )
= Avg. Time / Frame : 17mn 21s - 126.99 ppd

Project : 2125

X2-3800 ( Core1 )
= Avg. Time / Frame : 44mn 11s - 127.76 ppd
Conroe ( Core 2 )
= Avg. Time / Frame : 45mn 07s - 125.12 ppd

Can you see what I mean?

The opteron 144 is doing the EXACT same project as CORE 1 on the Conroe, and one of the cores from the 3800 are running the very same project as CORE 2 on *** conroe.

On both Projects, the AMD is proving the quicker, the x2-3800 by a very small fraction, but the Opteron is almost double the speed of the conroe!!!

These are 2 projects I know, but even the last time, I ran the same project on these 3 systems, with both cores on the conroe and the x2-3800 so, 5 different cores running the same project and in that project, which IIRC was number 1809 ( Amber ) the opteron had also completed that project a good few hours before either the 3800 or the C2D had.

Does this then prove that for folding, an Opteron is a better choice than a conroe? Also that the x2-3800 is a tiny tiny bit quicker than the conroe???

Surely not?

That cannot be right can it?

Anyone care to verify these results? Perhaps if oyu have both an AMD and a Conroe to run the very same project on both and tell me which CPU completed their project first?
 
i think if you check out: www.fahinfo.org
you can look at each project on each type of CPU

and i think you'll find the general trend, mhz to mhz, is:

tinker = athlon 64 wins
amber = athlon 64 wins
gromacs = conroe wins
double gromacs = conroe wins

i could be wrong but im pretty sure im right, but thats only mhz to mhz, ie a 2.5ghz athlon64 vs a 2.5ghz core duo, and obviously you can only realistically clock an athlon so far, to say 3ghz, but a core duo will do 3.5ghz with relative ease, so they will win on the gromacs core WU's anyway, and if you OC far enough they will win on the ambers through sheer mhz power

essentially, if your OCing both, a conroe will win at everything, but it will be closer with amber cores
 
Not all c2d's do 3ghz even fewer do 3.5.

X2 is better at the old cores iirc because it has better cache and register management mainly due to the core isolation (independent cache for each core) and a few other technicalities.
But when it comes to more modern cores and cores that use a cpu 'better' without a doubt the c2d pulls ahead.
 
Thanks for that reply.

Project 2125 is Gromacs, and the conroe is 1 minute slower over 45 minutes.

Project 1809 in an Amber and the Opteron is a full 7 minutes quicker than the Conroe.

As you say however, its Mhz v Mhz and different projects are better for their CPUs then...

That would indeed bring things closer perhaps.

The only double gromacs Im runnign right now is the P4 2.6 which is only switched on for a few minutes here and there, and dont have any tinkers

The Conroe is currently at stock, so I will slow the Opteron down then, and see if I cant grab tiny/quick projects for all 4 of the different types and re-give it another go perhaps?

----

Sorry, I didnt get your Reply Lay-z-boy till I posted this.

I suppose it could be just a chance then, that perhaps I managed to get projects that happen to be much better on AMD than on Intel?

If thats the cae then, then surely if one can be selective n what projects you run, you can in theory, use only very efficient projects on a poor spec AMD and actually get much better high scores than someone with a top-end intel then?

...In theory at least!
 
Last edited:
The project you're running are as old as the hills... well not quite but they have been around a good while and long before C2D was released.

May I suggest keeping an eye for some newer WUs? Dunno if there is a release date you can find for each one, but as said above older WUs are likely faster on older CPUs because it is what they optimised for.

SiriusB
 
Yes, Im not selective on what comes down... If I catch it early, I will stop it and simply copy the data over to another PC rather than anything else... I just wait for FAH to give me a new project on the cores otherwise.

How can I select specific projects?

I should run a couple of older ones and a couple of newer ones then.

Old and new on each core perhaps?
 
You can't really chose specific projects, unless you get yourself 8 clients and see what project each one gets and keep the ones you want.

Old and New on each CPU would be the best test.

SiriusB
 
lazyboy i mean most athlon64 chips top at a lower speed to what most conroes will top out at
so averagely they will clock higher and therefore process WU's faster on average

for every user who can only clock his c2d to 3ghz, there will be an athlon user who can only clock his A64 to 2.5ghz
 
Last edited:
example comparisons from www.fahinfo.org

project 1809: (an amber project)

Athlon 64 X2 - Manchester - 84.48 PPD [per 1ghz clock speed]
Core 2 Duo -- Conroe ------ 75.13 PPD [per 1ghz clock speed]


project 2404: (a gromacs project)

Athlon 64 X2 - Manchester - 113.69 PPD [per 1ghz clock speed]
Core 2 Duo -- Conroe ------ 171.53 PPD [per 1ghz clock speed]
 
Last edited:
Project : 2125
Core : Gromacs
Frames : 100
Credit : 392


-- e6600-3 --

Min. Time / Frame : 25mn 11s - 224.15 ppd
Avg. Time / Frame : 25mn 21s - 222.67 ppd


-- e6600-2 --

Min. Time / Frame : 25mn 41s - 219.78 ppd
Avg. Time / Frame : 25mn 44s - 219.36 ppd

Dunc
 
Project : 2125
Core : Gromacs
Frames : 100
Credit : 392

-- Folding 1 --

Min. Time / Frame : 27mn 56s - 202.08 ppd
Avg. Time / Frame : 35mn 18s - 159.91 ppd


-- Brenda 1 --

Min. Time / Frame : 27mn 18s - 206.77 ppd
Avg. Time / Frame : 28mn 11s - 200.29 ppd


both core 2 duos at about 3.1GHz (both now 3.2)
 
You must log in or register to reply here.
Back
Top Bottom