64-bit Pentium 4 "F", 2MB Prescotts and 925XE Updates

[Sheik]

:eek: A few days ago we reported on dual core LGA775 processors but we held back on some other information until we could get better sources for confirmation. Before we get to the real juice this week, here are a couple tidbits to soak up and put things in perspective.

First, it looks like we are going to see our first 1066FSB processor for third week in October. The 2MB L3, 3.4GHz Pentium 4 Extreme Edition will be quickly followed by a 3.73GHz 2MB Prescott based CPU with XD stack protection before the end of the year. Intel's enterprise division labels 2MB L2 Prescott processors as "Iriwindale" cores, and it is probably safe to say that the 3.73GHz P4EE is little more than an overclocked Iriwindale anyway. The 925XE chipset, which will launch as the only chipset capable of 1066FSB, should at least paper launch by the second week of October.

Of course the larger news here is the under the table announcement of EM64T Prescott "F" CPUs. These Pentium 4s are not labeled as 5xx or 7xx series CPUs, but rather relying on the old MHz naming again. As you can see below, the 3.60F Pentium 4 already has been penciled in Intel's Q3'04 timeslot. Unusually, back in April we noticed Intel was going to ship workstation Pentium 4s with EM64T capabilities (along with 5xx product nomenclature), but we constantly overlooked it as it was but a footnote in an 80 page document. Perhaps more interesting is Intel's decision to bypass its Model Number name scheme for workstation Pentium 4s in favor of the MHz Model again.

All Pentium 4 "F" processors have the E-0 stepping which makes them capable of the execute disable (XD) bit. The new "F" series Pentium 4s are already shipping in discreet channels.

Don't be surprised if 64-bit enabled Cranford (Nocona with 8MB L3 cache) shows up by Q1 of 2005 as well. The Copper River successor, "Mukilteo", has also been scheduled for Q3'05 and has been labeled as a Pentium 4 workstation CPU; it appears NetBurst is still here to stay for a while.

Thats a lot of cash err cache. :D

Source
http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=2152

[-=TriX=-]

all the codenames get confusing :\

1066 fsb, nice, bet we won't see many people oc'ing that 1:1

[Salkcin]

-=TriX=--->

That's why we got (high latency) DDR2. They have the potential of reachng DDR800 wich normal DDR never would... DDR2-533 has already shown up - their bandwidth in dual channel are equal with 1066FSB... it wont be 1:1 (since DDR2 4-doubles the bandwidth of the host frequency) - the normal ratio will be 2:1 now.

[Lithan]

Those 3.73 prescotts are going to be cherry. I'm surprised they have a yield good enough to put them out. I wonder if they will be improving the stock cooler.

[lutjens]

Those 3.73 prescotts are going to be cherry. I'm surprised they have a yield good enough to put them out. I wonder if they will be improving the stock cooler.

They don't...that's why even the 3.6 hasn't showed up yet. They can't produce enough 3.6's to satisfy market demand, so rather than put them on allocation, they paper launch it and hoard all the 3.6's produced to build stock. Once it reaches an acceptable level, they'll begin shipping them.

I can't see the 3.73 as being any different. Intel's hoping that yields will have improved as they get closer to launch, but if they're so bad to not be able to ship the 3.6, they have some work to do to be able to launch the 3.73 GHz part.

What will likely happen is that the 3.6 will ship in quantity very shortly before the paper launch of the 3.73 GHz part. And the cycle will repeat itself...personally I expect to see 3.73 GHz chips with retail availability sometime in the new year at the earliest...I'm guessing beginning of February (my prediction).

[Lithan]

Originally posted by lutjens
They don't...that's why even the 3.6 hasn't showed up yet. They can't produce enough 3.6's to satisfy market demand, so rather than put them on allocation, they paper launch it and hoard all the 3.6's produced to build stock. Once it reaches an acceptable level, they'll begin shipping them.

I can't see the 3.73 as being any different. Intel's hoping that yields will have improved as they get closer to launch, but if they're so bad to not be able to ship the 3.6, they have some work to do to be able to launch the 3.73 GHz part.

What will likely happen is that the 3.6 will ship in quantity very shortly before the paper launch of the 3.73 GHz part. And the cycle will repeat itself...personally I expect to see 3.73 GHz chips with retail availability sometime in the new year at the earliest...I'm guessing beginning of February (my prediction).

Bad news for people buying 3.0e's and hoping for a gem to LN2 and superpi if intel is skimming the top chips off for their 3.6's and 3.73's.

[-=TriX=-]

Originally posted by Salkcin
-=TriX=--->

That's why we got (high latency) DDR2. They have the potential of reachng DDR800 wich normal DDR never would... DDR2-533 has already shown up - their bandwidth in dual channel are equal with 1066FSB... it wont be 1:1 (since DDR2 4-doubles the bandwidth of the host frequency) - the normal ratio will be 2:1 now.

DDR2 4 folds the host frequency bandwidth? thatnks for the info didn't realize that. Wonder how much of a nimprovement more bandwidth and the new cache will give...

[HawainPanda]

man, i can't wait for these chips...they have everything going for them and with the 90nm they're gonna reach speeds that the athlon fx wouldn't have until amd switches to 90nm process

[Perc]

Originally posted by Lithan
Bad news for people buying 3.0e's and hoping for a gem to LN2 and superpi if intel is skimming the top chips off for their 3.6's and 3.73's.

you got that right man!!! man im afraid to buy anything right now that has intels name on it! i think im going to take a break from intel and go have some fun with the amd's for a while.... damm i wanted one of these new intels too... oh well its always best to wait for the yield to improve then to be the poor sap stuck with an earlyer one that ocs like shizit....

peace perc,