This article provides a list of motherboard chipsets made by Intel, divided into three main categories: those that use the PCI bus for interconnection (the 4xx series), those that connect using specialized "hub links" (the 8xx series), and those that connect using PCI Express (the 9xx series). The chipsets are listed in chronological order.
Pre-chipset situation
An earlier chipset support for Intel 8085 microprocessor can be found at MCS-85 family section.
To integrate the functions needed on a mainboard into a smaller number of ICs, Intel licensed the ZyMOS POACH chipset for its Intel 80286 and Intel 80386SX processors (the 82230/82231 High Integration AT-Compatible Chip Set). The 82230 covers this combination of chips: 82C284 clock, 82288 bus controller, and dual 8259A interrupt controllers among with other components. The 82231 covers this combination of chips: 8254 interrupt timer, 74LS612 memory mapper and dual 8237A DMA controller among with other components. Both set were available USD $60 for 10 MHz version and USD $90 for 12 MHz version in quantities of 100.[2] This chipset can be used with an 82335 High-integration Interface Device to provide support for the Intel 386SX.[3][4]
82310 MCA family chipset - announced in April 1988.[10] This chipset also supports the 80386SX based machines as well.[11] Which it does includes:[4][12]
82306 Local Channel Support Chip
82307 DMA Controller/Central Arbiter
82308 Micro Channel Bus Controller
82309 Address Bus Controller
82706 VGA Graphics Controller
82350 EISA - announced in September 1988.[13][14] This chipset supports the i486 microprocessor. It was expected to be available in the later half of 1989.[15]
82311 MCA - announced in November 1988.[16][17] Includes: 82303 and 82304 Local I/O Channel Support Chips, 82307 DMA Controller/Central Arbiter, 82308 Micro Channel Bus Controller, 82309 Address Bus Controller, 82706 VGA Graphics Controller, 82077 Floppy Disk Controller.[4][14]
82320 MCA - announced in April 1989.[18] This chipset supports the i486 microprocessor. It was expected to be available in the later half of 1989.[19]
82340SX PC AT - announced in January 1990, it is the Topcat chipset licensed from VLSI.[20]
82340DX PC AT - announced in January 1990, it is the Topcat chipset licensed from VLSI.[20]
82360SL - announced in October 1990.[22] It was a chipset for the mobile 80386SL and 80486SL processors. It integrated DMA controller, an interrupt controller PIC, serial and parallel ports, I/O Control, NMI, Real Time Clock, Timers and power-management logic for the processor. This chipset contains 226,000 transistors using the one-micron CHMOS IV technology. It was available for USD $45 in quantities of 1,000.[23]
82365SL - PC Card Interface Controller. This support PCMCIA 2.0 standard using the Exchangeable Card Architecture which supports both I/O and memory ExCA-compliant cards. It uses the Intel386SL power-management features. This was available for USD $35 in samples of quantities of 1,000-unit.[24]
82380 - High Performance 32-Bit DMA Controller with Integrated System Support Peripherals. This chipset has 20-level programmable interrupt controller a superset of Intel's 82C59PIC. It also has four (x4) 16-bit programmable internal timers which its superset Intel's 82C54PIT. It also has built-in DRAM refresh controller as well. It is available for USD $149 and USD $299 for 16 MHz and 20 MHz respectfully in quantities of 100.[25] The Intel M82380 met under MIL-STD-883 Rev. C standard. This military device was tested which includes temperature cycling between -55 and 125 °C, hermeticity and extended burn-in. This military version can have transfer rate of 32 Mbytes per seconds at 16 MHz. This military version were available in 132-lead CPGA and 164-lead CQPK. This military version were available for USD $520 100-unit of quantities for the PGA version.[26]
82384 - Clock Generator. The available version for USD $15 in quantities of 100.[27]
82385 - High Performance 32-Bit Cache Controller.[28] This chipset was introduced in February 1987. It was available for 20 MHz version.[29] There is 33 MHz version available for the 386DX processor.[30] Paired with 33 MHz 386 CPU and 64-Kbyte memory subsystem, it performed up to 7.8 MIPS.[31] There is 82385SX version for the 386SX microprocessor.[32]
82395DX - High Performance Smart Cache. This chipset contains internal 16-Kbye of SRAM and 1,000 cache tags. This controller supports up to 128-Kbytes of cache memory subsystem which it features four-way set associativity; a 16-byte line size; a four, double-word write buffer; and concurrent line-buffer caching. This also support write-buffer memory update protocol and maintains cache coherency during bus snooping. Paired with 33 MHz 386 CPU and the controller can perform up to 8.3 MIPS. This was available in 196-pin PQFP for USD $90 and $109 for 25- and 33-MHZ version in quantites of 1000 respectively.[33] There is Intel 82395SX version which it contains 8-Kbyte of cache memory for the 80386SX microprocessor family which it performs as much as 7% better than the 82385SX version. It was available for USD $44 in quantities of 1000 units housed by 132-pin PQFP. The Intel 82396SX version contains 16-Kbyte of cache memory which were available in second quarter of 1991.[34]
82495DX - Cache Controller. This support zero-wait-state with two-way set associative cache with several configurable parameters. This support MESI protocol and bus snooping. It is available for USD $198.[36]
82490DX - 32-Kbyte Dual Port Intelligent Cache SRAM. Providing second level write-back cache with dual-ported buffers and registers. It is available for USD $41.[37]
Pentium chipsets
While not an actual Intel chipset bug, the Mercury and Neptune chipsets could be found paired with RZ1000 and CMD640 IDE controllers with data corruption bugs. L2 caches are direct-mapped with SRAM tag RAM, write-back for 430FX, HX, VX, and TX.
[*] Remapping of PCIE/APIC memory ranges not supported,[67][68] some physical memory might not be accessible (e.g. limited to 3.5 GB or similar).
[1] Some later revisions of motherboards based on 945P,945G and 945PL chipset usually supports some Core 2 processors (with later BIOSes). Core 2 Quad is not supported. Only Core 2 Duo, Pentium Dual-Core, and Core2 based Celerons.
Summary:
915P (Grantsdale)
Supports Pentium 4 on an 800 MT/s bus. Uses DDR memory up to 400 MHz, or DDR2 at 533 MHz. Replaces AGP and CSA with PCI Express, and also supports "Matrix RAID", a RAID mode designed to allow the usage of RAID levels 0 and 1 simultaneously with two hard drives. (Normally RAID1+0 would have required four hard drives)
Sub-versions:
915PL - Cut-down version of 915P with no support for DDR2 and only supporting 2 GB of memory.
915G (Grantsdale-G)
915P with an integrated GMA 900. This core contains Pixel Shader version 2.0 only, it does not contain Vertex Shaders nor does it feature Transform & Lighting (T&L) capabilities and therefore is not Direct X 8.1 or 9.0 compliant.
Sub-versions:
915GL - Same feature reductions as 915PL, but supports 4 GB of memory. No support for external graphics cards.
915GV - Same as 915G, but has no way of adding an external graphics card.
910GL - No support for external graphics cards or 800 MT/s bus.
925X (Alderwood)
Higher end version of 915. Supports another PAT-like mode and ECC memory, and exclusively uses DDR-II RAM.
Sub-versions:
925XE - Supports a 1066 MT/s bus.
945P (Lakeport)
Update on 915P, with support for Serial ATA II, RAID mode 5, an improved memory controller with support for DDR-II at 667 MHz and additional PCI Express lanes. Support for DDR-I is dropped. Formal dual-core support was added to this chipset.
Sub-versions:
945PL - No support for 1066 MT/s bus, only supports 2 GB of memory.
945G (Lakeport-G)
A version of the 945P that has a GMA 950 integrated, supports a 1066 MT/s bus.
Sub-versions:
945GC - Same feature reductions as 945PL but with an integrated GMA 950.
945GZ - Same as 945GC but only supports DDR2 memory at 400/533 MT/s. No support for external graphics cards (some boards, like Asus P5GZ-MX, support through ICH7 on PCIe ×16 @4 lanes mode).
955X (Lakeport)
Update for 925X, with additional features of "Lakeport" (e.g., PAT features and ECC memory), and uses DDR2.
All Core 2 chipsets support the Pentium Dual-Core and Celeron processors based on the Core architecture. Support for all NetBurst based processors was officially dropped starting with the Bearlake chipset family.[70] However, some motherboards still support the older processors.[71]
[*] Remapping of PCIE/APIC memory ranges not supported,[67] some physical memory might not be accessible (e.g. limited to 3.5 GB or similar). Operational configuration is 4 ranks - 2× 2 GB dual rank modules or 4× 1 GB single rank modules - depends on number of motherboard DDR2 slots.
Summary:
946PL (Lakeport)
Update on 945PL, supports 4 GB of memory.
946GZ (Lakeport-G)
A version of 946PL with GMA 3000 graphics core.
P965 (Broadwater)
Update on 945P, no native PATA support, improved memory controller with support for DDR2 memory up to 800 MHz and official Core 2 Duo support.
G965 (BroadwaterG)
A version of P965 that has a GMA X3000 integrated graphics core.
Q965 (Broadwater)
Expected G965 intended for Intel's vPro office computing brand, with GMA 3000 graphics instead of GMA X3000 graphics. Supports an ADD2 card to add a second display.
Sub-versions:
Q963 - Q965 without an external graphics interface or support for ADD2.
975X (Glenwood)
Update of 955, with support for ATI Crossfire Dual Graphics systems and 65 nm processors, including Core 2 Duo.
P35 (Bearlake)
The P35 chipset provides updated support for the new Core 2 Duo E6550, E6750, E6800, and E6850. Processors with a number ending in "50" have a 1333 MT/s FSB. Support for all NetBurst based processors is dropped with this chipset.[70]
G33 (BearlakeG)
A version of P35 with a GMA 3100 integrated graphics core and uses an ICH9 South Bridge.
Sub-versions:
G35 - G33 with a GMA x3500 integrated graphics core and uses an ICH8 South Bridge, no DDR3 support.
Q35 (BearlakeG)
Expected G33 intended for Intel's vPro office computing brand, no DDR3 Support.
Sub-versions:
Q33 - Q35 without vPro support.
P31 (BearlakeG)
A version of P35 with an ICH7 South Bridge, supports only 4 GB of DDR2 memory and does not support DDR3 memory.
Operational configuration is 4 ranks - 2× 2 GB dual rank modules or 4 × 1 GB single rank modules - depends on number of motherboard DDR2 slots. 4GBs modules are not supported.
G31 (BearlakeG)
A version of P31 with a GMA 3100 integrated graphics core. It supports a 1333 MT/s FSB with Core 2 Duo processors, but Core 2 Quad processors are only supported up to 1066 MT/s.[80]
G41 (EaglelakeG)
Update of G31 with a GMA X4500 integrated graphics core and DDR3 800/1066 support.
P45 (Eaglelake)
Update of P35, with PCIe 2.0 support, Hardware Virtualization, Extreme Memory Profile (XMP) and support for ATI Crossfire (x8+x8).
Sub-versions:
P43 - P45 without Crossfire support.
G45 (EaglelakeG)
A version of P45 that has a GMA X4500HD integrated graphics core and lacks Crossfire support.
Sub-versions:
G43 - Same feature reductions as P43, but with a GMA X4500 integrated graphics core.
Q45 (EaglelakeQ)
Expected G43 intended for Intel's vPro office computing brand. Also supports Hardware Virtualization Technology and Intel Trusted Platform Module 1.2 feature.
Sub-versions:
Q43 - Q45 without vPro support. Also lacks Intel Trusted Platform Module 1.2 support.
B43 - Q43 with an ICH10D South Bridge.
[1] The 975X chipset supports only ×16 PCI Express (electrically) in the top slot when the slot below it is unpopulated. Otherwise it and the lower slot (both attached to the Memory Controller Hub) operate at ×8 electrically.
[2] Only later revisions of the 975X chipset boards support Core 2 processors. See MSI 975X Platinum (MS-7246) rev 1.0 (first release), and MSI 975X Platinum Powerup revision (MS-7246) rev 2.1 (released autumn 2006) as example. source: https://web.archive.org/web/20210515170458/http://ixbtlabs.com/articles2/mainboard/msi-975x-platinum-powerup-edition-i975x.html
Officially 975X supports a maximum of 1066 MT/s FSB. Unofficially, third-party motherboards (Asus, Gigabyte) support certain 1333FSB 45 nm Core2 processors, usually with later BIOS updates.
As for Celeron and Celeron D support, some boards and revisions support it, some not. (see upper example, MSI Powerup Edition has reintroduced back Celeron support, probably due to later released Core2-based Celerons, which were often more powerful than higher clocked Netburst Pentiums 4.
[3] The 975X chipset technical specification shows only DDR2-533/667 memory support. Actual implementations of 975X do support DDR2 800.
[4] VT-d is inherently supported on these chipsets, but may not be enabled by individual OEMs. Always read the motherboard manual and check for BIOS updates. X38/X48 VT-d support is limited to certain Intel, Supermicro, DFI (LanParty) and Tyan boards. VT-d is broken or non existent on some boards until the BIOS is updated. Note that VT-d is a chipset Memory Controller Hub technology, not a processor feature, but this is complicated by later processor generations (Core i3/i5/i7) moving the MCH from the motherboard to the processor package, making only certain I series CPUs support VT-d.
The Nehalem microarchitecture moves the memory controller into the processor. For high-end Nehalem processors, the X58 IOH acts as a bridge from the QPI to PCI Express peripherals and DMI to the ICH10 southbridge. For mainstream and lower-end Nehalem processors, the integrated memory controller (IMC) is an entire northbridge (some even having GPUs), and the PCH (Platform Controller Hub) acts as a southbridge.
Chipsets supporting LGA 1156 CPUs (Lynnfield and Clarkdale).
Not listed below is the 3450 chipset (see Xeon chipsets) which is compatible with Nehalem mainstream and high-end processors but does not claim core iX-compatibility. With either a Core i5 or i3 processor, the 3400-series chipsets enable the ECC functionality of unbuffered ECC memory.[82] Otherwise these chipsets do not enable unbuffered ECC functionality.
Chipsets supporting LGA 1155 CPUs (Sandy Bridge and Ivy Bridge). The PCIe 2.0 lanes from the PCH ran at 5 GT/s in this series, unlike in the previous LGA 1156 chips.[83]
The Cougar Point Intel 6 series chipsets with stepping B2 were recalled due to a hardware bug that causes their 3 Gbit/s Serial ATA to degrade over time until they become unusable. Stepping B3 of the Intel 6 series chipsets will have the fix for this. The Z68 chipset which supports CPU overclocking and use of the integrated graphics does not have this hardware bug, however all other ones with B2 did.[84] The Z68 also added support for transparently caching hard disk data on to solid-state drives (up to 64 GB), a technology called Smart Response Technology.[85]
1 For Sandy Bridge mainstream desktop and business platforms. Sandy Bridge CPUs provide 16 PCIe 2.0 lanes for direct GPU connectivity.
2 For Ivy Bridge mainstream desktop platform. Ivy Bridge CPUs provide 16 PCIe 3.0 lanes for direct GPU connectivity and additional 4 PCIe 2.0 lanes.[86]
Chipsets that support LGA 1150 CPUs are listed below. Haswell and Haswell Refresh CPUs are supported by all listed chipsets; however, a BIOS update is usually required for 8-Series Lynx Point motherboards to support Haswell Refresh CPUs.[87]Broadwell CPUs are supported only by 9-Series chipsets, which are usually referred to as Wildcat Point.[88]
The C1 stepping of the Lynx Point chipset contains a bug – a system could lose connectivity with USB devices plugged into USB 3.0 ports provided by the chipset if the system enters the S3 sleep mode.[89]
2 X58 TDP includes the X58 IOH TDP in addition to the ICH10/ICH10R TDP.
3 For Sandy Bridge enthusiast desktop platform. Sandy Bridge CPUs will provide up to 40 PCIe 3.0 lanes for direct GPU connectivity and additional 4 PCIe 2.0 lanes. NOTE : This reference number 4 is on X79, which is a Sandy bridge -E, not Sandy Bridge, and PCIe 3.0 only is enabled when an Ivy Bridge-E CPU or Xeon E-5 series is used.
4 For Haswell enthusiast desktop platform. Haswell CPUs will provide up to 40 PCIe 3.0 lanes for direct GPU connectivity and additional 4 PCIe 2.0 lanes.
LGA 2066
Chipsets supporting LGA 2066 socket for Skylake-X processors and Kaby Lake-X processors.
The C621 Chipset also supports FCLGA3647 socket for Skylake-SP as well as Cascade Lake-W and Cascade Lake-SP processors.
The 100 Series chipsets (codenamed Sunrise Point), for Skylake processors using the LGA 1151 socket,[95] were released in the third quarter of 2015.[96]
The 200 Series chipsets (codenamed Union Point) were introduced along with Kaby Lake processors, which also use the LGA 1151 socket;[97] these were released in the first quarter of 2017.[98]
While Coffee Lake shares the same socket as Skylake and Kaby Lake, this revision of LGA 1151 is electrically incompatible with 100 and 200 series CPUs.
The 300 Series chipsets were introduced along with Coffee Lake processors, which use the LGA 1151 socket; the enthusiast model was released in the last quarter of 2017,[99] the rest of the line was released in 2018.[100]
C232 and C242 chipsets do not support CPU integrated GPUs, as they lack FDI support. Officially they support only Xeon processors, but some motherboards also support consumer processors (6/7th generation Core for C230 series, 8/9th generation Core for C240 series and its Pentium/Celeron derivatives).
LGA 1200 is a CPU socket designed for Comet Lake and Rocket Lake desktop CPUs. Like its predecessors, LGA 1200 has the same amount of pins its name would suggest: 1200. Under the hood, LGA 1200 is a modified version of LGA 1151, its predecessor. It features 49 additional protruding pins that are used to improve power delivery and provide support for eventual updates with I/O features.
^When applied to computer memory (RAM or cache) the quantities KB, MB and GB are defined as: 1 KB = 1024 B, 1 MB = 1024 KB, 1 GB = 1024 MB, consistent with the JEDEC memory standard.
^The Pentium Pro, Pentium II/III, and the Celerons based on them are essentially the same design with minor internal revisions and varying cache designs. Because of this, the same chipset can be used for Socket 8, Socket 370, Slot 1, or Slot 2 designs with any CPU in the P6 family. In practice however, newer chipset designs are usually made only for the newer processor packages, and older ones may not be updated to accommodate for recent package designs. In addition, certain chipsets may be implemented in motherboards with different processor packages, much like how the 440FX could be used either with a Pentium Pro (Socket 8) or Pentium II (Slot 1). A new feature for the latest Intel chipsets is hardware virtualization support (Intel VT-d).[46] The chipset support for this technology is not very clear for the moment.[47]
^The Intel 82943GML mobile chipset unofficially supports Core Duo, Core 2 Duo, and Pentium Dual Core processors as well as 667 MHz FSB, which is a popular upgrade for many older notebook computers such as certain models of Acer Aspire 3680.
^Remapping of PCIE/APIC memory ranges not supported,[67] some physical memory might not be accessible (e.g. limited to 3.5 GB or similar).
^Ormsby, John, Editor, "New Product Focus: Components: Intel's 82X3X Chip-set Handles Logic Functions That Once Required The Services Of Sources Of Chips", Intel Corporation, Microcomputer Solutions, January/February 1988, page 13
^Lewnes, Ann, "Welcome 80386SX", Microcomputer Solutions, September/October 1988, page 2
^ abBrownstein, Mark (November 13, 1989). "Uncorking the future". InfoWorld. p. 101. Archived from the original on June 27, 2014. Retrieved October 15, 2016.
^Intel Corporation, "NewsBits: Intel Introduces EISA and MCA Support for i486 Microprocessor", Microcomputer Solutions, July/August 1989, page 1
^Intel Corporation, "New Product Focus Component: A 32-Bit Microprocessor With A Little Help From Some Friends", Special 32-Bit Issue Solutions, November/December 1985, page 13
^Intel Corporation, "New Product Focus Components: The 32-Bit Computing Engine Full Speed Ahead", Solutions, May/June 1987, page 10
