Future CPUs for T series

[choiyc]

Supposedly, all the missing CPU's are due within Q1 2005...
Does anybody know how much faster 533 mhz FSB is compared to 400 mhz? I do a lot of numerical modeling, not too heavy, but it runs for a few days on my T23.....

Thanks

Pentium M
770 2.13GHz 533 2MB
765 2.10GHz 400 2MB
760 2GHz 533 2MB
755 2GHz 400 2MB
750 1.86GHz 533 2MB
745 1.80GHz 400 2MB
740 1.73GHz 533 2MB
735 1.70GHz 400 2MB
730 1.60GHz 533 2MB
725 1.60GHz 400 2MB
715 1.50GHz 400 2MB

LV
758 1.50GHz 400 2MB
738 1.40GHz 400 2MB

ULV
753 1.20GHz 400 2MB
733 1.10GHz 400 2MB
723 1GHz 400 2MB
713 1.10GHz 400 1MB

[awolfe63]

Well it's 25% faster....

I think you really wanted to ask how much faster your system would be - but it doesn;t really matter since the current chipset can't handle a 533 FSB and the new chipset is a completely different design.

[choiyc]

Thanks for the reply.

Yes, my question was how faster will 533 FSB will be compared to 400 mhz, in terms of number crunching operations, not graphics etc.

Also, anybody happened to know if the form factor of current T's will change when 533mhz FSB CPUs come out? Or would it be like T41 and T42 change?

Thanks again,

YC

[Kenn]

The Pentium IV has traditionally benefited tremendously from FSB increases.

Of course, that begs the question of whether the Penitum-M is more similar to the long-pipeline P4 or, for example, a lower-clock, lower-latency AMD Athlon (I guess that's kind of a rhetorical question), where an FSB increase applied to the Barton core certainly didn't make as much of a difference as on the Intel chip.

IMO, a 25% FSB speed bump, where the CPU is outpacing it by a multiplier of 5x, will make for a pretty noticeable improvement, certainly more than what you'd get going from 1.8Ghz to 2.0Ghz.

[sktn77a]

Most number crunching consists of integer and floating point operations which are performed internally to the processor. However, depending on the software you use, intermediate results could be temporarily stored in RAM. For integer and floating point operations, the 533MHz FSB will have no effect on computation speeds. However, the software may perform operations which may be sped up by the faster memory bus speed. I doubt the overall effect will be significant, though.

Then again, if you're doing a lot of nonlinear mixed effects modelling on large databases, you might notice the difference.

[Kenn]

Most number crunching consists of integer and floating point operations which are performed internally to the processor. However, depending on the software you use, intermediate results could be temporarily stored in RAM. For integer and floating point operations, the 533MHz FSB will have no effect on computation speeds. However, the software may perform operations which may be sped up by the faster memory bus speed. I doubt the overall effect will be significant, though.

Then again, if you're doing a lot of nonlinear mixed effects modelling on large databases, you might notice the difference.

OTOH, it's likely that anything that takes a few days per run is going to be shuttling a lot of data back and forth across the memory subsystem.

I guess in the end we'll just have to wait and see the benchmarks!