makes me curious whether HT is really an advantage for E@H ...
BTW, what's the clock of your P4?
==========
@ Steffen, I don't know if the AMD CPU take better advantage of the Einstein WU's than the Intel HT CPU's or not, but my feeling is the Intel HT CPU's are definitely able to keep up with the AMD's and in most cases are still faster than the AMD's when running 2 at a time in HT Mode.
I run 7 HT PC's here at Einstein and from looking at a lot of other peoples Computer Times for the WU's I don't see where I'm at a disadvantage at all.
..I run 7 HT PC's here at Einstein and from looking
====================================================
The point is, PoorBoy, when you would put beside all your 7 really fast INTEL HT based crunchers and put 7 (relatively cheap) AMD CPUs, average, not top end,
they will do the same work, or perhaps even more.
With reduced power consumation and less "GHz"
This is valid for Einstein and LHC, for seti it's just the other way.
When a PC is used not only used for crunching only, (burning cdroms, type writing, playing videos/sounds,...) the HT based CPUs can play their more advantages.
Only my modest opinion.
friendly
ric
In a couple of mounth we will see the dual core desktop CPUs!!!
=1 physical CPU will have 2 physical cores, each will have 2 logical (HT) CPUs
gives 4 active slots;)
=1 physical CPU will have 2 physical cores, each will have 2 logical (HT) CPUs
gives 4 active slots;)
=========
We will have to wait and see what that sort of CPU can do with the Projects. I CPU crunching 4 WU's might be a case of to much work with to little CPU power to work with, I don't know ...
>
> > my P4 has two WUs at ten Hours, with HT
> > and my Athlon64 3.200+ takes about 5:30 for every WU
> > and my XP 2.500+ takes 6:30 for every WU
>
> makes me curious whether HT is really an advantage for E@H ...
> BTW, what's the clock of your P4?
>
> Steffen
>
>
the P4 3.2 is running on an I875 Mobo with Dualchannel and 2 512 MB Twinmos Syster at 2.0 3 3 8
> On the other side, it smells a bid that the Einstein Applications are "better
> optimized" for AMD based hosts??
Not better optimized, the AMD processors are just a better match to the task. The P4 FPU (Floating Point Unit) is extremely poor compared to the AMD Athlon [XP/64], a quick glance at the CPU Benchmark in BOINC will show you that my 2.2Ghz Athlon XP shows 2039Mflops (about 0.93ops/clock) while my 2.66Ghz P4 shows 1383Mflops (0.52 ops/clock) giving the AMD an almost 80% faster FP rate per clock cycle. The intel chips do very well when they can utilize the SSE/2/3 units (which seems to have been why they decided not to invest more transistors in the FPU). Take a look at other programs which require heavy floating point operations (SciMark, POVRay, many CADD and FEA programs) and you'll find the same result.
makes me curious whether HT
)
makes me curious whether HT is really an advantage for E@H ...
BTW, what's the clock of your P4?
==========
@ Steffen, I don't know if the AMD CPU take better advantage of the Einstein WU's than the Intel HT CPU's or not, but my feeling is the Intel HT CPU's are definitely able to keep up with the AMD's and in most cases are still faster than the AMD's when running 2 at a time in HT Mode.
I run 7 HT PC's here at Einstein and from looking at a lot of other peoples Computer Times for the WU's I don't see where I'm at a disadvantage at all.
..I run 7 HT PC's here at
)
..I run 7 HT PC's here at Einstein and from looking
====================================================
The point is, PoorBoy, when you would put beside all your 7 really fast INTEL HT based crunchers and put 7 (relatively cheap) AMD CPUs, average, not top end,
they will do the same work, or perhaps even more.
With reduced power consumation and less "GHz"
This is valid for Einstein and LHC, for seti it's just the other way.
When a PC is used not only used for crunching only, (burning cdroms, type writing, playing videos/sounds,...) the HT based CPUs can play their more advantages.
Only my modest opinion.
friendly
ric
In a couple of mounth we will see the dual core desktop CPUs!!!
=1 physical CPU will have 2 physical cores, each will have 2 logical (HT) CPUs
gives 4 active slots;)
=1 physical CPU will have 2
)
=1 physical CPU will have 2 physical cores, each will have 2 logical (HT) CPUs
gives 4 active slots;)
=========
We will have to wait and see what that sort of CPU can do with the Projects. I CPU crunching 4 WU's might be a case of to much work with to little CPU power to work with, I don't know ...
> > > my P4 has two WUs at
)
>
> > my P4 has two WUs at ten Hours, with HT
> > and my Athlon64 3.200+ takes about 5:30 for every WU
> > and my XP 2.500+ takes 6:30 for every WU
>
> makes me curious whether HT is really an advantage for E@H ...
> BTW, what's the clock of your P4?
>
> Steffen
>
>
the P4 3.2 is running on an I875 Mobo with Dualchannel and 2 512 MB Twinmos Syster at 2.0 3 3 8
Greetings from Germany NRW
Ulli
[img]http://boinc.mundayweb.com/one/stats.php?userID=380 [/img]
> On the other side, it
)
> On the other side, it smells a bid that the Einstein Applications are "better
> optimized" for AMD based hosts??
Not better optimized, the AMD processors are just a better match to the task. The P4 FPU (Floating Point Unit) is extremely poor compared to the AMD Athlon [XP/64], a quick glance at the CPU Benchmark in BOINC will show you that my 2.2Ghz Athlon XP shows 2039Mflops (about 0.93ops/clock) while my 2.66Ghz P4 shows 1383Mflops (0.52 ops/clock) giving the AMD an almost 80% faster FP rate per clock cycle. The intel chips do very well when they can utilize the SSE/2/3 units (which seems to have been why they decided not to invest more transistors in the FPU). Take a look at other programs which require heavy floating point operations (SciMark, POVRay, many CADD and FEA programs) and you'll find the same result.