Intel’s 12th Generation Core processors, code-named “Alder Lake,” debut today, and a big part of the rollout party, as always, is a new high-end chipset. But this is not another mere ordinary “Z”-series chipset launch.
Intel’s new Z690 marks the biggest change in Intel’s motherboard chipsets in nearly a decade. With the introduction of Alder Lake, Intel is transitioning from DDR4 to the newly emerging DDR5 RAM, with a host of new capabilities and tweaking possibilities. The Z690 chipset is facilitating this move with the introduction of a new version of the classic Extreme Memory Profiles—XMP 3.0—and a brand-new memory-speed technology that Intel has named Dynamic Memory Boost. The chipset has several other new features baked in, including fresh overclocking tools to handle the Alder Lake CPUs’ unusual multi-core design.
Before diving into this article, it would be helpful to first read through our coverage of the new Intel 12th Generation Alder Lake CPUs. To summarize the information in that article, Alder Lake processors utilize a multi-core design with large CPU cores that are optimized for performance and smaller cores that are designed to be more energy-efficient. This is similar to the “big.LITTLE” paradigm on ARM-based SoCs, but with Intel hardware and some software elements to help manage the workload between the two types of cores. Alder Lake also ships with native support for DDR5 memory (though some motherboards will ship with DDR4 support), as well as support for the nascent PCI Express 5.0. Let’s start with that.
PCI Express 5.0 Is Here…for the Future
In total, in the new chipset the CPU is allocated 20 PCI Express (PCIe) 5.0 lanes, 16 of which are dedicated for graphics cards and cannot be used for anything else. By default, all 16 of these will be tied to the primary PCIe x16 slot. Z690 motherboards that have a second PCIe x16 slot can be configured to have this split in half and shared between the two PCIe x16 slots to facilitate a multiple-GPU configuration.
The remaining four PCIe 5.0 lanes are intended for use with M.2 NVMe SSDs. Currently, of course, there aren’t any PCIe 5.0 NVMe SSDs on the market, though prototypes have been shown. But if there were (and we assume they will eventually come to the mainstream consumer market), this new platform would support them with a maximum theoretical bandwidth of up to 128Gbps.
That said, until PCIe 5.0 NVMe drives show up, this connection can still be used for any M.2 PCIe 3.0 or PCIe 4.0 NVMe SSDs. It’s essentially a form of future-proofing for buyers of high-end motherboard platforms with the new Intel chips.
Chipset PCIe 4.0 and SATA
The Z690 chipset itself now has support for up to 12 PCIe 4.0 lanes, which the outgoing Z590 chipset lacked. Like the extra four PCIe 5.0 lanes on the CPU, these will mostly be useful for M.2 NVMe SSDs and other high-speed interfaces that the motherboard maker may integrate, such as Thunderbolt controllers. For less data-hungry devices (such as Wi-Fi chips or sound cards or chipsets), the Z690 has up to 16 PCIe 3.0 lanes available that should supply plenty of bandwidth to these devices.
Like with previous generations of Intel chipsets, some of the PCIe lanes share a data connection with other devices. For example, the Platform Controller Hub (PCH) can support eight Serial ATA (SATA) 3.0 ports, but using some of these will remove either a PCIe 3.0 or PCIe 4.0 lane from use elsewhere.
Given the total number of lanes available, though, and the declining use of SATA 3.0 devices as M.2 and NVMe devices take over the storage space (SATA is mostly just used for spinning-platter hard drives, at this point), this doesn’t appear problematic. It’s actually more surprising to see that the Z690 can natively support eight SATA 3.0 devices, as essentially all Intel chipsets in the last decade have supported no more than six SATA devices.
DMI 4.0: Double the Bandwidth
A bit of background first: DMI, or Direct Memory Interface, is the connection used to glue together virtually everything except RAM and the graphics card to the CPU. The DMI lanes function similar to PCIe lanes, with DMI 4.0 having bandwidth parity with PCIe 4.0. In turn, the Z690 chipset functions essentially as a PCIe device connected to the CPU just like a graphics card…except instead of sending graphics data, the chipset enables the CPU to connect to dozens of other devices. That includes all USB, audio, internet, and storage devices through the chipset and the DMI 4.0 connection. Any data passing from any of these devices to the system’s RAM also has to pass through this same connection, which is why having adequate bandwidth for the DMI connection is critically important.
Alder Lake CPUs connect to the Z690 chipset over an updated interface that Intel has dubbed “DMI 4.0.” The peak theoretical bandwidth of the DMI 4.0 x8 connection is rated at 15.75GBps, which effectively gives it the same bandwidth as a PCI-E 4.0 x8 interface. That’s a two-fold increase over the DMI 3.0 x8 connection utilized by 11th Generation Core (“Rocket Lake”) CPUs and the Z590 chipset.
Interestingly, information we received prelaunch denoted the possibility of an DMI 4.0 x4 connection that would be roughly equal to the old DMI 3.0 x8 connection in terms of bandwidth. Intel told us that OEMs would be free to configure either an x4 or x8 connection as they saw fit, but we doubt any OEMs will do this with Z690 chipset motherboards.
Why? The reduction in bandwidth could potentially hamper the performance of devices connected to the PCH. This will all depend on how many devices are in use, however, and boards with relatively few connections (such as Mini-ITX motherboards) may see little to no impact in a DMI 4.0 x4 connection. Regardless, there’s no real logical reason why an OEM would choose not to utilize the full connection on a Z690 chipset, which leaves us to wonder if this could instead be put to use on lower-end chipsets in the future that don’t support a DMI x8 connection. This will likely be clarified when the inevitable lower-cost “B” and “H” 600-series chipsets roll out.
RAM Support: DDR5, XMP 3.0, and Dynamic Memory Boost Technology
Alder Lake and the Z690 chipset are the first Intel CPU and PCH to support DDR5, and this is an area that Intel has clearly put a great deal of time and effort into. DDR5 was designed to permit data to be transmitted four times each clock cycle, versus only twice each clock cycle on all previous iterations of DDR RAM. This has the potential to generate significantly more bandwidth between the RAM and CPU, but like with all new types of RAM, it may take some time for this benefit to be realized.
Alder Lake officially supports DDR5 RAM with an effective clock speed of up to 4,800MHz. The platform, however, retains support for DDR4, as well, with a maximum effective clock speed of up to 3,200MHz. (To be clear: Motherboards will come with support for one of the two memory standards, DDR4 or DDR5, not both.)
At face value, this makes DDR5 look like the clear, decisive choice for most buyers, but this fails to take into consideration the existence of higher-clocked DDR4 RAM. We’ve seen DDR4 operating at speeds over 5,000MHz in the past, and while that’s still quite uncommon for DDR4, DDR4 RAM clocked above 4,000MHz is readily available from several major memory makers. DDR5 also has significantly higher timings than DDR4, which could cause it to perform worse than DDR4 until the technology matures and speeds increase.
Arguably, Intel’s new Dynamic Memory Boost technology is even more interesting than DDR5 is itself. This is essentially the RAM equivalent of Intel’s Turbo Boost Max technology—and if you’ve been working in the computer industry for a long time, that might well sound bananas to you.
Why? Historically, RAM issues have been one of the leading causes of system instability and BSODs. Using RAM from different vendors, RAM of mixed rated speeds, or even RAM from the same vendor bought in different sets has been known to cause system-stability issues in the past.
So what did Intel do? The company built a “turbo mode” feature into its new platform to raise and lower this relatively crash-prone component’s timings and clock speed, depending on operating conditions. Intel said this was extensively tested to ensure it was highly stable, and at this time we have no reason to doubt this claim. (We’ll see!)
Intel said this feature was designed to boost performance, and while we don’t see how this would improve performance over the RAM running at full speed at all times, we do see a few situations in which it could be beneficial. But before going over that, let’s talk about how exactly this technology works.
According to Intel, Dynamic Memory Boost is auto-trained at boot using the XMP profiles built into either DDR4 or DDR5 RAM. DDR5 RAM utilizes XMP 3.0 and can store up to five memory profiles at a time, including two rewritable profiles. One of these will be the JEDEC DDR5 standard that is set at 4,800MHz; the other two will be higher-clocked profiles. At boot, the system will run a simple check to ensure it’s able to set the correct settings for each memory profile. If, for some reason, the RAM is not able to use a memory profile, it will simply default to the JEDEC standard.
If the system is able to apply the settings for at least two XMP profiles, it will store the information for these two profiles and alternate between them while the system is on based on workload. By default, the system will boot with the slower of the two memory profiles enabled, but any time 30% or more of the RAM’s bandwidth is being used for a period of 10ms or longer, the RAM will switch to the faster settings for more performance.
Both the memory timings and the operating frequency change when this happens, but the operating voltage remains fixed at the power level required for the faster of the two RAM settings. Intel said that the voltage was not adjusted due to a latency penalty that would be incurred by doing so. This likely means that if Intel wanted to change the voltage, as well, it would slow down the process of switching between the two memory profiles and cause issues. This works in a similar fashion for both DDR4 and DDR5 RAM.
Now that we’ve covered how this works, the possible benefits that we see with this system should be more clear. First, when overclocking the RAM, this should provide a bit of a safety net and make recovering from a bad overclock easier. A bad RAM overclock can sometimes require a BIOS reset to recover from. But as this system will always boot with the slower memory profile in use, getting into the BIOS to fix the overclock should be as simple as just booting into the BIOS like usual.
Now, this is all well and good on desktops. But this also appears to have potential benefits for eventual laptop implementations, whenever Intel eventually takes Alder Lake mobile. Lower-clocked RAM will use less power and produce less heat. Saving power and generating less heat is a plus for any system, but in a laptop this could have an added benefit when it comes to extending battery life, which is always a hard-fought battle.
Overclocking Alder Lake, and Other Changes
Overclocking is dead? Long live overclocking, if you ask Intel.
Intel discussed overclocking Alder Lake CPUs more than it has for quite some generations during its prerelease briefings. As always, the top-of-the-line Z-series chipset for Intel’s new platform has options for overclocking the CPU, the RAM, and various other components such as the cache. Also, Hyper-Threading and entire cores can be disabled on an individual, per-core basis. All CPU cores can be set with their own individual multiplier, as well, to help eke out extra performance when not all cores are in use. AVX, which can be problematic while overclocking, has its own clock settings and can be disabled entirely to avoid causing instabilities. More on overclocking to come, for sure, as we start to dig into the first enthusiast Z690 boards in our hands.
In addition to the more notable improvements mentioned above, Intel also made a few more subtle updates to other aspects of the PCH. Compared to the Z590 chipset, Z690 will ship with implementation support for up to four USB 3.2 Gen 2×2 ports (up from three). It will also have an upgraded Wi-Fi 6E controller integrated into the Z690 chipset.
There are also a host of other features carried over from Z590 that remain unchanged. A Gigabit MAC is still built into the chipset with a faster 2.5Gbit MAC included as an optional add-on. Intel’s Smart Sound Technology, Optane Memory support, and Rapid Storage Technology (RST) support all remain unchanged from the last generation.
…and a New Socket: LGA 1700
New generation, new socket! The Alder Lake platform requires a new motherboard, of course, and the pin count on these CPUs is way up.
The new socket is named LGA 1700, reflecting the number of pins (up from 1,200 in the LGA 1200 socket of Comet Lake and Rocket Lake). The socket itself differs a quite a bit from the last-generation LGA1200 socket, but cooler support is a mixed bag.
Technically, there is a new cooler-mounting outline for LGA 1700 that will require a specifically LGA 1700-compatible bracket, but some Z690 motherboards include mounting holes for both LGA 1700 and LGA 1200 coolers. If you’re planning on using an older cooler with a new build, you’ll want to make sure that your new board has LGA 1200 mounting holes…or the cooler’s maker offers a conversion bracket that you can get your mitts on at the same time you order your motherboard and CPU.
A New Dawn Over Alder Lake…and Future ‘Lakes’
Consider just the introduction of future-looking PCIe 5.0 support and now-shipping DDR5 memory, and on those points alone, the new Z690 chipset is unquestionably the most envelope-pushing platform that Intel has released in the last decade. “Haswell,” way back in the 4th Generation Core days, may have seen the introduction of DDR4, and 11th Gen Rocket Lake was the first CPU family from Intel with support for PCIe 4.0. But Alder Lake and the Z690 chipset look to outdo them both.
After years of baby-step, incremental feature updates, the improvements in the Z690 platform feel much more real and meaningful. It will be a treat to dig into deeper as we move toward reviewing the first Alder Lake motherboards.
