Penryn
Intel's 45 nm shrink of the Core microarchitecture, from 2007
Penryn is Intel's 2007/2008 "Tick" step in its Tick-Tock cycle: the Core microarchitecture shrunk to 45 nanometers as CPUID model 23. Sold across Celeron, Pentium, Core 2 and Xeon brands, it added SSE4.1 instructions and larger caches, and was succeeded by Nehalem.
- 45 nm die shrink of the Core microarchitecture, CPUID model 23
- First Intel mainstream line with SSE4.1 instructions
- L2 cache raised to 6 MB per chip, 24-way associativity
- Transistor counts from 228M to 820M across the family
- Six-core model 29 die measures 503 mm2
What it is
Penryn is a family of x86 CPUs from Intel, introduced in November 2007. It is the 45 nanometer die shrink of the Core microarchitecture, identified as CPUID model 23. The design appeared under several code names depending on socket and core count: Penryn for Socket P parts, Wolfdale for LGA 775 dual-core, and Yorkfield for multi-chip module quad-core parts. Chips were sold as Celeron, Pentium, Pentium Dual-Core, Core 2 and Xeon. Clock speeds ran from 1.06 GHz to 3.33 GHz, with one to four cores in most lines and up to six in Xeon. It brought SSE4.1 instructions to Intel's mainstream processors.
How it came about
Penryn arrived as the "Tick" half of Intel's Tick-Tock programme, where one step shrinks an existing design to a finer process and the next introduces a new microarchitecture. The shrink to 45 nm came in 2007/2008, following the 65 nm Core 2 generation and preceding Nehalem. Intel marketed full-cache and reduced-cache steppings side by side under model 23, giving them identical CPUID values. A later stepping, E0/R0, added the XSAVE and XRSTOR instructions and replaced earlier ones. Model 29, stepping A1, appeared by February 2008 in the high-end Xeon 7400 series, code-named Dunnington.
Inside the chip
Penryn keeps the Core microarchitecture but improves several units. A new divider cuts latency, and a new single-cycle shuffle engine speeds data movement and helps enable the added SSE4.1 instructions. The instruction set covers x86-16, IA-32 and x86-64, with MMX through SSE4.1 plus VT-x and VT-d virtualisation support. Each core has 64 KB of L1 cache. L2 grows to a maximum of 6 MB per chip, up from 4 MB, and its associativity rises from 16-way to 24-way. Reduced parts carry 3 MB of L2. Xeon models add a shared L3 cache of 8 MB to 16 MB. Front-side bus speeds span 533 to 1600 MT/s.
How it was made
Penryn is built on Intel's 45 nanometer process, with transistor counts from 228 million to 820 million depending on the part. Quad-core processors and Wolfdale-DP are multi-chip modules that combine two dies in one package. Packages include Socket M and Socket P (both μPGA 478), Socket T (LGA 775), and FCBGA in μBGA 479 and μBGA 965 forms. Model 29 with six cores and an L3 cache reaches a die size of 503 mm2.
Where it went
Socket P Penryn parts went into mobile systems on the Intel Mobile 965 Express (Santa Rosa refresh) platform, while stepping E0/R0 supports the later Intel Mobile 4 Express (Montevina) platform. Wolfdale and Yorkfield served desktops on LGA 775. Xeon versions used LGA 771: Wolfdale-DP and Harpertown are multi-chip modules with two or four active Wolfdale cores. The six-core model 29 appeared in the Xeon 7400 series, known as Dunnington, aimed at very high-end servers.
Versions and family
Code names track core count, cache and socket: Penryn (Socket P), Wolfdale (LGA 775 dual-core) and Yorkfield (LGA 775 quad-core MCM). Cut-down parts with 3 MB of L2 are called Penryn-3M, Wolfdale-3M and Yorkfield-6M. Penryn-L is the single-core version, made from the Penryn-3M model with one core left active rather than a separate design like Merom-L. On the Xeon side, Wolfdale-DP and Harpertown cover two- and four-core LGA 771 modules. Penryn follows Merom and is followed by Nehalem.
Why it matters
Penryn was Intel's first high-volume 45 nm processor family and carried the Core design through the 2007/2008 period across laptops, desktops and servers. It introduced SSE4.1 to the mainstream x86 line and raised cache sizes while lowering divider latency. The same core design was reused across many brands and sockets, with features switched on or off for each market. It also set up the transition to Nehalem, the next step in the Tick-Tock cycle.
Problems and reception
Stepping C0/M0 is limited to the Santa Rosa refresh platform in mobile parts, and stepping C1/M1 was produced as a bug fix for quad-core processors only. Stepping E0/R0 supersedes all earlier steppings and adds XSAVE/XRSTOR. Penryn is listed as unsupported.
Questions people ask about Penryn
What exactly is Penryn?
It is Intel's 45 nanometer version of the Core microarchitecture, released from November 2007. The design was sold under many names, including Core 2, Pentium, Celeron and Xeon, with different core counts, caches and sockets for each market.
What did it change compared with earlier Core 2 chips?
Intel rebuilt the divider for lower latency, added a single-cycle shuffle engine, and introduced SSE4.1 instructions. Maximum L2 cache per chip rose from 4 MB to 6 MB, and cache associativity went from 16-way to 24-way.
What do the code names mean?
Penryn refers to Socket P mobile parts, Wolfdale to dual-core desktop chips on LGA 775, and Yorkfield to quad-core multi-chip modules on the same socket. In Xeon, Wolfdale-DP and Harpertown are LGA 771 modules with two or four active Wolfdale cores.
How was it manufactured?
Intel built it on a 45 nm process, with between 228 million and 820 million transistors. Quad-core parts combine two dies in one package. The largest variant, model 29 with six cores and L3 cache, has a die area of 503 mm2.
What came before and after it?
Penryn followed Merom, the 65 nm Core design, and was replaced by Nehalem. In Intel's Tick-Tock scheme it was the "Tick", the process shrink step, while Nehalem brought the new microarchitecture.
Good reading and sources
The facts on this page come from the Wikipedia article "Penryn (microarchitecture)", written by its contributors and used under the CC BY-SA 4.0 licence, and from its Wikidata record (CC0). EveryChip rewrote the text in its own words, with AI assistance, and compares every page with the article to make sure no passage is copied. It may be reused under the same CC BY-SA 4.0 licence with a link back.
It may contain mistakes, and figures change over time. Nothing here is investment or buying advice. If something is wrong, tell us. Last refreshed 2026-11-14.