Intel Q3'05 Roadmap: Conroe Appears, Speculation Ensues

Besides all the talk about the future architecture, there's not much to say about the desktop parts. Here's the current roadmap.

Intel Desktop Performance Roadmap
Processor	Core Name	Clock Speed	Socket	Launch Date
Pentium 673	Cedar Mill	3.8 2MB	LGA 775	2H'06
Pentium 672	Prescott 2M + VT	3.8 2MB	LGA 775	Q4'05
Pentium 663	Cedar Mill	3.6 2MB	LGA 775	Q1'06
Pentium 662	Prescott 2M + VT	3.6 2MB	LGA 775	Q4'05
Pentium 653	Cedar Mill	3.4 2MB	LGA 775	Q1'06
Pentium 643	Cedar Mill	3.2 2MB	LGA 775	Q1'06
Pentium >= 633	Cedar Mill	3.0 2MB	LGA 775	Q2'06
Pentium 631	Cedar Mill (no VT)	3.0 2MB	LGA 775	Q1'06
Pentium 571	Prescott	3.8 1MB	LGA 775	Now
Pentium 561	Prescott	3.6 1MB	LGA 775	Now
Pentium 551	Prescott	3.4 1MB	LGA 775	Now
Pentium 541	Prescott	3.2 1MB	LGA 775	Now
Pentium 531	Prescott	3.0 1MB	LGA 775	Now
Pentium 521	Prescott	2.8 1MB	LGA 775	Now

The single core Pentiums remain unchanged from last month, with the exception of the 673 showing up at the top. Processor models ending with a 3 will use the new Cedar Mill core, the single core version of Presler. They will be based on 65nm process technology and will include all the same extra technologies we mentioned earlier. They will also have HyperThreading enabled, where the dual core Presler chips do not. There is also a potential lower end 633 model scheduled to be introduced in Q2'06, though it may or may not be released, likely depending on demand and yields.

One final update for the mainstream desktop market is that the EM64T enabled 5x1 Pentium chips are finally available. We've been talking about them for a few months, and retail availability was expected before now. They were probably held back to let inventory of the earlier versions clear out. You can see the new chips in our Pricing Engine.

Intel Desktop Value Roadmap
Processor	Core Name	Clock Speed	Socket	Launch Date
Celeron D ???	Cedar Mill 512K + EM64T	???	LGA 775	2H'06
Celeron D 355	Prescott 256K + EM64T	3.33 256K	LGA 775	Q4'05
Celeron D 351	Prescott 256K + EM64T	3.2 256K	LGA 775	Now
Celeron D 346	Prescott 256K + EM64T	3.06 256K	LGA 775	Now/Soon
Celeron D 341	Prescott 256K + EM64T	2.93 256K	LGA 775	Now/Soon
Celeron D 336	Prescott 256K + EM64T	2.8 256K	LGA 775	Now/Soon
Celeron D 331	Prescott 256K + EM64T	2.66 256K	LGA 775	Now/Soon
Celeron D 326	Prescott 256K + EM64T	2.53 256K	LGA 775	Now/Soon

The Celeron picture is similar to the single core Pentium market. The EM64T enabled Celeron D chips are all starting to ship, after a month or two of waiting. Once again, you can check the current prices and availability on our Pricing Engine - at present the 351 is available, but we aren't picking up any of the slower "+1" parts. Current generation Celerons do not have VT, HT, or EIST support, but they do include XD (as have all Celeron D chips since the "J" variants started shipping almost a year ago).

Once Intel transitions to 65nm, a new version of the Celeron based off the Cedar Mill core will arrive. Clock speeds are not yet set, but we do know that it will continue to use a 533 MHz FSB, and it will increase the amount of L2 cache to 512K. That will make the chip relatively interesting, as the old Northwood core also included 512K of cache. Of course, the pipeline of Northwood was only 20 stages rather than 31, so clock for clock Northwood may still be faster. With a 65nm process, however, we expect the chips to be able to hit relatively high clock speeds, and architectural tweaks may help them to surpass Northwood performance. You're still looking at relatively equivalent performance for around $100 a CPU, which compares favorably to the higher end Northwood cores of the past.

If you're in the market for a value system and you feel the need to purchase a 64-bit processor, we'd recommend that Intel buyers get a motherboard that uses the 945P or 945G chipsets. If you don't feel the support for dual core processors is important, we'd still recommend 915G/P as being the next best alternative. We would stay away from the 915GV/GL/PL chipsets as they have limitations that make them unattractive; not to mention the fact that those chipsets are almost EOL by now too. 915GV is like 915G, but no external graphics port is provided. 915GL has the same problem, but it only supports DDR memory instead of DDR or DDR2. Finally, 915PL supports an external X16 PCIe slot but eliminates DDR2 support and allows the use of 1 DIMM per memory channel, limiting you to a maximum of 2GB RAM. The bottom line is that the P or the G versions are what most mainstream users will desire. There's also the 955X chipset on the high end, which allows the use of up to 8GB of RAM. Those who really need the increased address space of a 64-bit OS will likely want the option to use more than 4GB of RAM as well.

Final Thoughts

Most of the major releases of 2005 have now occurred, and other than a speed bump or two, there's not much else awaiting Intel owners (at least in the desktop market) this year. 2006 has quite a bit more in store. We haven't covered the mobile section, but many expect to see the dual core Pentium M Yonah available for the desktop as well as mobile markets. Beyond that, we're all waiting for Intel's answer to the performance conundrums of the past year, where they officially lost the performance crown to AMD and haven't been able to regain it. The change to 65nm processes will likely have some advantages, but the real counterattack is going to come in the form of the next Intel architecture. We've given some speculation as to what may be present, but most of the real details are still closely guarded secrets.

For nearly 25 years Intel was the leader in PC processor technology and performance, and we have a suspicion that they will be pulling out all the stops to regain that lead in 2006. The fact that they still lead in profits and market share gives them a lot of resources they can apply to that goal, and we eagerly await additional details of the new architectures. AMD will naturally have their own new processors and architectures, but we know even less about K9 than we do about Conroe. Waiting is the hardest part, unfortunately.