Sign in with
Sign up | Sign in
Your question

Prime95 - Different Torture Test Options?

Last response: in Overclocking
Share
Anonymous
a b K Overclocking
March 31, 2004 6:29:17 PM

Archived from groups: alt.comp.hardware.overclocking (More info?)

Hi,

I see Prime95 now has different test options (Blend, etc. .) Would anybody
care to say a few words as to which test is good for what?

Also how do you get a P4 with Hyperthreading to run two *instances* of
prime95? Do I need to install the program twice or what?
--
Wayne ][
<Sign on door reads: Please Do No Disturb! Pentium 4 assembly in progress!>
Anonymous
a b K Overclocking
March 31, 2004 6:29:18 PM

Archived from groups: alt.comp.hardware.overclocking (More info?)

"Wayne Youngman" <waynes.spamtrap@tiscali.co.uk> writes:
>I see Prime95 now has different test options (Blend, etc. .) Would anybody
>care to say a few words as to which test is good for what?

On a similar subject, beating up on processors trying to factor numbers,
if anyone would take a few minutes to run the following test on some
AMD parts, maybe 2600, 3000 and one of the 64 parts of about the same
speed, and P4 at about the same speed, I'd greatly appreciate it.

The code to do this is a Java applet that is behind this web page:
http://www.alpertron.com.ar/ECM.HTM

After the screen loads type
(2^2281+1)/(3*22811*95803)
into the "number or expression to factor..." box
and type
400
into the box next to "New Curve"
and then click the "New Curve" button.

That should start it running the Java applet on your box, trying
to find the next factor for this. It will take perhaps 20-30 minutes
of furious crunching to complete Step 1 up to 100% and Step 2 up to
100%. The elapsed time to get to 100% on Step 2 is the benchmark.
(Then it will start over with Step 1, so you have to watch it)

With an AMD 2000+ I'm getting just under 40 minutes and I'd greatly
appreciate some numbers to see how much I would gain with a 2600
and with a -64 processor. Or if I were to switch to a P4 at about
the same speeds or better.

Thank you
(email address is valid, if that matters)
Anonymous
a b K Overclocking
April 2, 2004 4:07:43 AM

Archived from groups: alt.comp.hardware.overclocking (More info?)

"JB" <highlinex@yahoo.com> writes:
>> > "Don Taylor" <dont@agora.rdrop.com> wrote in message
>> > news:tLSdnRCiX7uglfbd4p2dnA@scnresearch.com...
>> > > On a similar subject, beating up on processors trying to factor numbers,
>> > > if anyone would take a few minutes to run the following test on some
>> > > AMD parts, maybe 2600, 3000 and one of the 64 parts of about the same
>> > > speed, and P4 at about the same speed, I'd greatly appreciate it.
>> > >
>> > > The code to do this is a Java applet that is behind this web page:
>> > > http://www.alpertron.com.ar/ECM.HTM
>> > >
>> > > After the screen loads type
>> > > (2^2281+1)/(3*22811*95803)
>> > > into the "number or expression to factor..." box
>> > > and type
>> > > 400
>> > > into the box next to "New Curve"
>> > > and then click the "New Curve" button.
>> > >
>> > > That should start it running the Java applet on your box, trying
>> > > to find the next factor for this. It will take perhaps 20-30 minutes
>> > > of furious crunching to complete Step 1 up to 100% and Step 2 up to
>> > > 100%. The elapsed time to get to 100% on Step 2 is the benchmark.
>> > > (Then it will start over with Step 1, so you have to watch it)
>> > >
>> > > With an AMD 2000+ I'm getting just under 40 minutes and I'd greatly
>> > > appreciate some numbers to see how much I would gain with a 2600
>> > > and with a -64 processor. Or if I were to switch to a P4 at about
>> > > the same speeds or better.
>>
>> I got 52 minutes with a P4-1.8A @ 2.771 GHz and 512 MB.<

>This test includes interaction with the site so the connection speed
>contaminates the results. I am on dial-up and maybe that explains why my
>time is so high.

It seems to me that once it starts crunching there is almost no
interaction. You can even drop your connect to the net and if
your web browser doesn't complain about the dropped connection
it will just keep crunching away.

Now, DISCLAIMER, I'm NOT kicking dirt on anybody or anything.
As a comparison I tried running the identical test on a Dell
2400 with a 2.4 Ghz real P4, not a Celeron. The bus speed to
memory is faster than the AMD I tried it on and got the 40
minute figure. The clock is 2.4/2.0, there is almost zero
video I/O with this test, there is zero hard drive activity,
from what I can tell, it doesn'use much memory to run.

But the Dell took almost 60 minutes. Repeat, I'm not criticizing
ANYBODY. But I'm really confused why it wouldn't take more
like 33 minutes. There were no other jobs running on either
machine (but that doesn't seem to matter a great deal), both
were running XP, Home on the Dell, Pro on the AMD. I can't
figure out why it seems somewhere like half the speed I would
expect. It is even slower than the 52 minutes quoted for the
P4 above, but if I compare the clock rates those don't look
wildly out of line with each other.

I had a pro check the BIOS info on the Dell and he agrees
that it is a real P4, not a celeron. The only thing I've been
able to find that is different thus far is that the Dell is
running Java 2 from Sun and the AMD is running Microsoft Java.
I haven't taken the leap trying to swap Java's yet.

I never had any intention of turning this into bragging about
"mine is faster than yours" or criticizing either brand.
But I'm wondering if there is something wrong with the Dell.
I did the usual virus and spyware checks, etc. Nothing.
Is it possible that this particular Java app just runs slower
on P4's than on AMD's? This certainly isn't the sort of
streaming video crunching or massive floating point calcs
where the P4 might shine. Or maybe there is some other
explanation that I haven't figured out yet. Any ideas would
be greatly appreciated.

And I'm wondering where I should go from my AMD 2000. All
my time is spent sitting inside Mathematica doing symbolic
crunching, other than this little factoring problem came
up and the new version of elliptic curve factoring code
in Mathematica isn't ready yet. Few can afford multiple
CPU Mathematica licenses, I can't. So I am thinking of
the AMD 64's, Wolfram has included specific optimizations
to take advantage of 64 bit CPU's now. If it were feasible
it would be cheaper to just have 4 cheap AMD 2600's for the
price of a fast 64 bit system today.

Thanks for any suggestions, looking for > 2x current speeds.
Related resources
April 2, 2004 5:59:41 AM

Archived from groups: alt.comp.hardware.overclocking (More info?)

"Don Taylor" <dont@agora.rdrop.com> wrote in message
news:n-WdnTIUscKynvDdRVn-tw@scnresearch.com...
> "JB" <highlinex@yahoo.com> writes:
> >> > "Don Taylor" <dont@agora.rdrop.com> wrote in message
> >> > news:tLSdnRCiX7uglfbd4p2dnA@scnresearch.com...
> >> > > On a similar subject, beating up on processors trying to factor
numbers,
> >> > > if anyone would take a few minutes to run the following test on
some
> >> > > AMD parts, maybe 2600, 3000 and one of the 64 parts of about the
same
> >> > > speed, and P4 at about the same speed, I'd greatly appreciate it.
> >> > >
> >> > > The code to do this is a Java applet that is behind this web page:
> >> > > http://www.alpertron.com.ar/ECM.HTM
> >> > >
> >> > > After the screen loads type
> >> > > (2^2281+1)/(3*22811*95803)
> >> > > into the "number or expression to factor..." box
> >> > > and type
> >> > > 400
> >> > > into the box next to "New Curve"
> >> > > and then click the "New Curve" button.
> >> > >
> >> > > That should start it running the Java applet on your box, trying
> >> > > to find the next factor for this. It will take perhaps 20-30
minutes
> >> > > of furious crunching to complete Step 1 up to 100% and Step 2 up to
> >> > > 100%. The elapsed time to get to 100% on Step 2 is the benchmark.
> >> > > (Then it will start over with Step 1, so you have to watch it)
> >> > >
> >> > > With an AMD 2000+ I'm getting just under 40 minutes and I'd greatly
> >> > > appreciate some numbers to see how much I would gain with a 2600
> >> > > and with a -64 processor. Or if I were to switch to a P4 at about
> >> > > the same speeds or better.
> >>
> >> I got 52 minutes with a P4-1.8A @ 2.771 GHz and 512 MB.<
>
> >This test includes interaction with the site so the connection speed
> >contaminates the results. I am on dial-up and maybe that explains why my
> >time is so high.
>
> It seems to me that once it starts crunching there is almost no
> interaction. You can even drop your connect to the net and if
> your web browser doesn't complain about the dropped connection
> it will just keep crunching away.
>
> Now, DISCLAIMER, I'm NOT kicking dirt on anybody or anything.
> As a comparison I tried running the identical test on a Dell
> 2400 with a 2.4 Ghz real P4, not a Celeron. The bus speed to
> memory is faster than the AMD I tried it on and got the 40
> minute figure. The clock is 2.4/2.0, there is almost zero
> video I/O with this test, there is zero hard drive activity,
> from what I can tell, it doesn'use much memory to run.
>
> But the Dell took almost 60 minutes. Repeat, I'm not criticizing
> ANYBODY. But I'm really confused why it wouldn't take more
> like 33 minutes. There were no other jobs running on either
> machine (but that doesn't seem to matter a great deal), both
> were running XP, Home on the Dell, Pro on the AMD. I can't
> figure out why it seems somewhere like half the speed I would
> expect. It is even slower than the 52 minutes quoted for the
> P4 above, but if I compare the clock rates those don't look
> wildly out of line with each other.
>
> I had a pro check the BIOS info on the Dell and he agrees
> that it is a real P4, not a celeron. The only thing I've been
> able to find that is different thus far is that the Dell is
> running Java 2 from Sun and the AMD is running Microsoft Java.
> I haven't taken the leap trying to swap Java's yet.
>
> I never had any intention of turning this into bragging about
> "mine is faster than yours" or criticizing either brand.
> But I'm wondering if there is something wrong with the Dell.
> I did the usual virus and spyware checks, etc. Nothing.
> Is it possible that this particular Java app just runs slower
> on P4's than on AMD's? This certainly isn't the sort of
> streaming video crunching or massive floating point calcs
> where the P4 might shine. Or maybe there is some other
> explanation that I haven't figured out yet. Any ideas would
> be greatly appreciated.
>
> And I'm wondering where I should go from my AMD 2000. All
> my time is spent sitting inside Mathematica doing symbolic
> crunching, other than this little factoring problem came
> up and the new version of elliptic curve factoring code
> in Mathematica isn't ready yet. Few can afford multiple
> CPU Mathematica licenses, I can't. So I am thinking of
> the AMD 64's, Wolfram has included specific optimizations
> to take advantage of 64 bit CPU's now. If it were feasible
> it would be cheaper to just have 4 cheap AMD 2600's for the
> price of a fast 64 bit system today.
>
> Thanks for any suggestions, looking for > 2x current speeds.<


I think your'e right. Looking at the meter on winxp reveals little going
over the phone line. I also ran benchmarks on the CPU and found the xp2000
much faster in both integer and fpu tests. Mine beats the xp in the memory
benchmark but the cpu performance seems to be the key for this particular
application.

Perhaps the same would hold in Mathcad. My copy of Mathcad is v2 (DOS) so
it's not a candidate for comparison.
Anonymous
a b K Overclocking
April 2, 2004 6:34:31 AM

Archived from groups: alt.comp.hardware.overclocking (More info?)

Don Taylor wrote:
> "Wayne Youngman" <waynes.spamtrap@tiscali.co.uk> writes:
>> I see Prime95 now has different test options (Blend, etc. .) Would
>> anybody care to say a few words as to which test is good for what?
>
> On a similar subject, beating up on processors trying to factor
> numbers, if anyone would take a few minutes to run the following test
> on some AMD parts, maybe 2600, 3000 and one of the 64 parts of about
> the same speed, and P4 at about the same speed, I'd greatly
> appreciate it.
>
> The code to do this is a Java applet that is behind this web page:
> http://www.alpertron.com.ar/ECM.HTM
>
> After the screen loads type
> (2^2281+1)/(3*22811*95803)
> into the "number or expression to factor..." box
> and type
> 400
> into the box next to "New Curve"
> and then click the "New Curve" button.
>
> That should start it running the Java applet on your box, trying
> to find the next factor for this. It will take perhaps 20-30 minutes
> of furious crunching to complete Step 1 up to 100% and Step 2 up to
> 100%. The elapsed time to get to 100% on Step 2 is the benchmark.
> (Then it will start over with Step 1, so you have to watch it)
>
> With an AMD 2000+ I'm getting just under 40 minutes and I'd greatly
> appreciate some numbers to see how much I would gain with a 2600
> and with a -64 processor. Or if I were to switch to a P4 at about
> the same speeds or better.


31 min 51 sec 32-bit Barton @ 2200MHz
April 2, 2004 6:34:32 AM

Archived from groups: alt.comp.hardware.overclocking (More info?)

"S.Heenan" <sheenan@wahs.ac> wrote in message
news:Xi4bc.8410$Pk3.5381@pd7tw1no...
> Don Taylor wrote:
> > "Wayne Youngman" <waynes.spamtrap@tiscali.co.uk> writes:
> >> I see Prime95 now has different test options (Blend, etc. .) Would
> >> anybody care to say a few words as to which test is good for what?
> >
> > On a similar subject, beating up on processors trying to factor
> > numbers, if anyone would take a few minutes to run the following test
> > on some AMD parts, maybe 2600, 3000 and one of the 64 parts of about
> > the same speed, and P4 at about the same speed, I'd greatly
> > appreciate it.
> >
> > The code to do this is a Java applet that is behind this web page:
> > http://www.alpertron.com.ar/ECM.HTM
> >
> > After the screen loads type
> > (2^2281+1)/(3*22811*95803)
> > into the "number or expression to factor..." box
> > and type
> > 400
> > into the box next to "New Curve"
> > and then click the "New Curve" button.
> >
> > That should start it running the Java applet on your box, trying
> > to find the next factor for this. It will take perhaps 20-30 minutes
> > of furious crunching to complete Step 1 up to 100% and Step 2 up to
> > 100%. The elapsed time to get to 100% on Step 2 is the benchmark.
> > (Then it will start over with Step 1, so you have to watch it)
> >
> > With an AMD 2000+ I'm getting just under 40 minutes and I'd greatly
> > appreciate some numbers to see how much I would gain with a 2600
> > and with a -64 processor. Or if I were to switch to a P4 at about
> > the same speeds or better.
>
>
> 31 min 51 sec 32-bit Barton @ 2200MHz
>
>
Anonymous
a b K Overclocking
April 2, 2004 3:55:18 PM

Archived from groups: alt.comp.hardware.overclocking (More info?)

"JB" <highlinex@yahoo.com> writes:
>"Don Taylor" <dont@agora.rdrop.com> wrote in message
>news:n-WdnTIUscKynvDdRVn-tw@scnresearch.com...
>> "JB" <highlinex@yahoo.com> writes:
>> >> > "Don Taylor" <dont@agora.rdrop.com> wrote in message
>> >> > news:tLSdnRCiX7uglfbd4p2dnA@scnresearch.com...
<snip>
>> >> > > The code to do this is a Java applet that is behind this web page:
>> >> > > http://www.alpertron.com.ar/ECM.HTM
>> >> > >
>> >> > > After the screen loads type
>> >> > > (2^2281+1)/(3*22811*95803)
>> >> > > into the "number or expression to factor..." box
>> >> > > and type
>> >> > > 400
>> >> > > into the box next to "New Curve"
>> >> > > and then click the "New Curve" button.
<snip>
>> >> > > With an AMD 2000+ I'm getting just under 40 minutes and I'd greatly
<snip>
>> >> I got 52 minutes with a P4-1.8A @ 2.771 GHz and 512 MB.<
>>
>> >This test includes interaction with the site so the connection speed
>> >contaminates the results. I am on dial-up and maybe that explains why my
>> >time is so high.
>>
>> It seems to me that once it starts crunching there is almost no
>> interaction. You can even drop your connect to the net and if
>> your web browser doesn't complain about the dropped connection
>> it will just keep crunching away.
>>
>> Now, DISCLAIMER, I'm NOT kicking dirt on anybody or anything.
<snip>
>> I never had any intention of turning this into bragging about
>> "mine is faster than yours" or criticizing either brand.
<snip>
>> Is it possible that this particular Java app just runs slower
>> on P4's than on AMD's? This certainly isn't the sort of
>> streaming video crunching or massive floating point calcs
>> where the P4 might shine. Or maybe there is some other
>> explanation that I haven't figured out yet. Any ideas would
>> be greatly appreciated.

>I think your'e right. Looking at the meter on winxp reveals little going
>over the phone line. I also ran benchmarks on the CPU and found the xp2000
>much faster in both integer and fpu tests. Mine beats the xp in the memory
>benchmark but the cpu performance seems to be the key for this particular
>application.

Thank you for a polite and reasoned response. I'm glad this hasn't
turned into an argument like so many thing seem to do. I really
tried to make it clear that I wasn't being critical of anyone in
this and maybe that worked.

>Perhaps the same would hold in Mathcad. My copy of Mathcad is v2 (DOS) so
>it's not a candidate for comparison.

It seems that for any of these it really takes a fairly intense set
of calculations to take any more than the blink of an eye.

There in a carefully hand tuned assembly code->.exe file for elliptic
curve factoring done by a guy named Zimmerman that is specifically
crafted for the P4. Perhaps that would be more representative for
the P4. But when I tried to compare a similar supposedly tuned
..exe for the AMD against the Java applet on the AMD I was very
surprised to see that the times were about the same. The only thing
I can imagine is that most of the time in the Java code is spent
inside the long integer routines and that they must have been fairly
well written. But we don't really know whether the two algorithms
might be somewhat different internally.

Thanks again
!