New C1 stepping adds more speed to P4's

[Belial]

Just reading through some stuff about the new 2.8 p4 and it seems like they get a bigger boost then just from having more mhz and it's from the new stepping. Good news is that it's not just on the 2.8's; it's also going to be used on the slower p4's.

[jamesavery22]

Unfortunately the lowest multiplier for a P4 with C1 stepping is 20 =( no more 3ghz with 1400mhz rdram the 2.26 was perfect for that =( 3ghz w/ pc1200 isnt bad though

[shurleydude]

will they add this to the 2.6A 400mhz processors?

[jamesavery22]

Originally posted by shurleydude
will they add this to the 2.6A 400mhz processors?

The only 2.6 p4s are C1s

[shurleydude]

so what are you saying? a 2.6A automatically has C1 stepping in it? Does that mean a 2.6A would be faster than a 2.53B without C1?

[jamesavery22]

Originally posted by shurleydude
so what are you saying? a 2.6A automatically has C1 stepping in it? Does that mean a 2.6A would be faster than a 2.53B without C1?

Uh, I think you are a little confused.

The Stepping is the revision of the actual manufacturing process.

All 2.6ghz P4s are made on the C1 Stepping. New steppings can mean numerous things. In this case it is just better tolerances. Back in the PIII days there were like 8 different steppings for the 550. One was on a .25u tech with 512k of half speed L2 and there was one on a .18u tech with 256k of full speed L2 so the performance was different between steppings. But there are no spec differences between C1 and B0 P4s other than power consuption(edit: this is probably because there are no P4 C1s at the same stock speed as a B0) and default core V. A 2.6 will outperform a 2.53 at stock speeds in terms of processing power but the 2.53 has higher memory bandwidth so it could outperforum the 2.6 in memory intense apps. I dont really understand your question, hope this answers it...

[Belial]

The new stepping just makes it a wee bit faster then the older ones; they added some more instruction cache or something like that(can't remember exactly what it is now). But even if I can't remember what it is now the C1 does add something that improves performance. I saw on another site that many of the older processors would be being made with this new stepping also; like the 2.53's etc...

The new stepping is good for maybe a 2% or so improvement from what I've read so far. Not a lot but every little bit helps right?

[marv]

Originally posted by Belial
The new stepping just makes it a wee bit faster then the older ones; they added some more instruction cache or something like that(can't remember exactly what it is now). But even if I can't remember what it is now the C1 does add something that improves performance. I saw on another site that many of the older processors would be being made with this new stepping also; like the 2.53's etc...

The new stepping is good for maybe a 2% or so improvement from what I've read so far. Not a lot but every little bit helps right?

This should help on which have the C1 stepping:

http://www.sharkyforums.com/showthre...hreadid=152373

[jamesavery22]

Originally posted by Belial
The new stepping just makes it a wee bit faster then the older ones; they added some more instruction cache or something like that(can't remember exactly what it is now). But even if I can't remember what it is now the C1 does add something that improves performance. I saw on another site that many of the older processors would be being made with this new stepping also; like the 2.53's etc...

The new stepping is good for maybe a 2% or so improvement from what I've read so far. Not a lot but every little bit helps right?

....The C1 does not have an more cache then the B0. Its kinda hard to say that the C1s are faster than the B0s because none of the C1s have the same clock speed as the B0s... So you are just saying a 2.66 is faster than a 2.53. I dunno about more instructions because that would be on the architecture side. But Im pretty sure this new stepping does not add more instructions. And I do not believe any 2.53s will be C1s since none are listed as of right now. It seems like the differences between the C1s and the B0s are just voltage and higher yields.

[vertices]

Somebody had an unlocked 2.8 and brought the multiplier down to the same as 2.53. Basically they tested with a "new" 2.53 and an "old" 2.53 and the new one had a 2% advantage.

So it was an even test of new vs old.

[Belial]

These are the new c1 stepping processors and #'s they have listed at Hardocp:

C1 stepping

1.8A SL6LA
2A SL6GQ
2.20 SL6GR
2.26 SL6DU
2.40 SL6GS
2.40B SL6DV
2.53 SL6DW
2.66 SL6DX
2.80 SL6DY(?)


The C1 stepping is faster because they doubled the # of entries for the instruction TLB. This is under the L1 cache info according to the Hardocp article.

[Belial]

those where the listing I found at that site for the lower speed C1 steppings...for those that care.

[Moridin]

Originally posted by Belial
they added some more instruction cache or something like that(can't remember exactly what it is now).

Yes it is a little faster, even with the same speed and multiplier. Intel has only said it’s due to minor architectural changes, but I’ve seen at least one reliable person claim there are some improvements in cache coherency. Apparently it’s mostly to improve SMT performance in the XEON versions, but it does help the desktop P4 keep data in it’s L1 cache a little better, which accounts for the speed increase.

[Belial]

It's probably worth hunting down the C1 stepping if your in the market for a new P4 processor...More speed for the same price is a good deal.

[KJY9]

Originally posted by Belial
These are the new c1 stepping processors and #'s they have listed at Hardocp:

C1 stepping

1.8A SL6LA
2A SL6GQ
2.20 SL6GR
2.26 SL6DU
2.40 SL6GS
2.40B SL6DV
2.53 SL6DW
2.66 SL6DX
2.80 SL6DY(?)


The C1 stepping is faster because they doubled the # of entries for the instruction TLB. This is under the L1 cache info according to the Hardocp article.

Hmmmm funny...Intel's spec sheet doesn't have the 2.53 listed as being one with the C1 stepping?