Intel's 12th-Gen Alder Lake chips will bring the company's hybrid architecture, which combines a mix of larger high-performance cores paired with smaller high-efficiency cores, to desktop x86 PCs for the first time. Intel has shared many of the architectural details at its Architecture Day 2021, but recently whipped the covers off the full details, including specifications, pricing and performance benchmarks.
Intel's initial Alder Lake lineup starts with three chips and their graphics-less variants, with the flagship $589 Core i9-12900K, which Intel bills as the 'world's fastest gaming processor,' leading the charge with 16 cores and 24 threads. Intel claims this chip provides an average 13% generational jump in gaming performance, beats AMD's Ryzen 9 5950X flagship by up to 30% in gaming, and offers up to twice the performance of its predecessor in content creation workloads. Intel even claims up to an 84% generational gain in fps in some game streaming scenarios. We have the full breakdown of those claims in the performance section below. You can pre-order the Alder Lake processors now, but we suggest holding off until you can see independent testing when our review lands on November 4, 2021.
Alder Lake represents a massive strategic shift as Intel looks to regain the uncontested performance lead against AMD's Ryzen 5000 series processors. AMD's Zen 3 architecture has taken the lead in our Best CPUs for gaming and CPU Benchmarks hierarchy, partly on the strength of Ryzen's higher core counts. That's not to mention the pressure from Apple's M1 processors that feature a similar hybrid design to Alder Lake and come with explosive single-threaded performance improvements of their own.
Intel's Alder Lake brings disruptive new architectures and supports features like PCIe 5.0 and DDR5 that leapfrog AMD and Apple in connectivity technology and also outstrip Ryzen's core counts in mobile designs.
It all starts with a new way of thinking for x86 chips by pairing high-performance and high-efficiency cores within a single chip. That well-traveled design philosophy powers billions of Arm chips, often referred to as Big.Little (Intel calls its implementation Big-Bigger), but it's a first for x86 desktop PCs. The Golden Cove architecture powers Alder Lake's 'big' high-performance cores, while the 'little' Atom efficiency cores come with the Gracemont architecture. Intel etches the cores on its 'Intel 7' Enhanced SuperFin process, marking the company's first truly new node for the desktop since 14nm debuted six long years ago.
Intel is going all-in: the company will reunify its desktop and mobile lines with Alder Lake. Intel says it will tune Alder Lake for high-performance, a must for desktop PCs and high-end notebooks.
Potent adversaries challenge Intel on both sides. Apple's Arm-powered M1 processors have set a high bar for hybrid designs, outperforming all other processors in their class with the promise of more powerful designs to come. Meanwhile, AMD's Ryzen 5000 chips have taken the lead over Intel's aging Skylake derivatives. Intel's Rocket Lake chips overtook AMD in single-threaded performance, but they still trail in multi-core workloads due to Rocket Lake's maximum of eight cores, while AMD has 16-core models for the mainstream desktop.
Intel certainly needs a come-from-behind design to thoroughly unseat its competitors, swinging the tables back in its favor like the Conroe chips did back in 2006 when the Core architecture debuted with a ~40% performance advantage that cemented Intel's dominance for a decade. Intel's Raja Koduri has already likened the transition to Alder Lake with the debut of Core, suggesting that Alder Lake could indeed be a Conroe-esque moment.
Intel's 12th-Gen Alder Lake At a Glance
- Alder Lake comes to market on November 4, 2021. Preorder is now.
- Six models: $589 Core i9-12900K, $409 i7-12700K, $289 i5-12600K. All three are also available as graphics-less KF models.
- LGA1700 socket requires new motherboards
- The Alder Lake SoC will span from desktop PCs to ultramobile devices with TDP ratings from 9W to 125W, all built on the Intel 7 process. The desktop PC comes with up to eight Performance (P) cores and eight Efficient (E) cores for a total of 16 cores and 24 threads and up to 30 MB of L3 cache for a single chip.
- Alder Lake supports either DDR4 or DDR5 (LP4x/LP5, too). Desktop PC supports x16 PCIe Gen 5 and x4 PCIe Gen 4, while mobile supports x12 PCIe Gen 4 and x16 PCIe Gen 3, Thunderbolt 4, and Wi-Fi 6E.
- Intel's new hyper-threaded Performance (P) core, which comes with the Golden Cove microarchitecture designed for low-latency single-threaded performance, comes with an average of 19% more IPC than the Cypress Cove architecture in Rocket Lake. It also supports AVX-512 and AMX (a new AI-focused matrix-multiply ISA) for data center variants (both are disabled on consumer chips).
- Intel's new single-threaded Efficiency (E) core comes with the Gracemont microarchitecture to improve multi-threaded performance and provide exceptional area efficiency (small footprint) and performance-per-watt. Four small cores fit in the same area as a Skylake core and deliver 80% more performance in threaded work (at the same power). A single E core also delivers 40% more performance than a single-threaded Skylake core (at the same power) in single-threaded work (caveats apply to both).
- Intel's Thread Director is a hardware-based technology that assures threads are assigned to either the P or E cores in an optimized manner. This is the sleeper tech that enables the hybrid architecture.
- Alder Lake does not support AVX-512 under any condition (fused off in P cores, not supported in E cores) to ensure an even ISA application.
- Four variants: -S for desktop PCs, -P for mobile, -M for low-power devices, -L Atom replacement, -N educational (probably Chromebooks)
- Intel will hold the inaugural Intel Innovation event October 27-28. The event is largely thought to be the official unveiling of the Alder Lake processor stack.
Intel Alder Lake Release Date
Alder Lake will ship on November 4, 2021 and reviews will arrive the same day. You can preorder the chips now. Intel is only bringing its priciest parts to the retail market at first, but it is also shipping 28 more SKUs to OEMs for delivery in systems that will debut in early 2022. Intel hasn't shared the details of those models.
Intel Alder Lake-S Desktop PC Specifications and Pricing
| U.S. Price | Cores | Threads | P-Core Base/Boost | E-Core Base/Boost | TDP / PBP / MTP | DDR4-3200 | L3 Cache | |
| Ryzen 9 5950X | $799 | 16P | 32 threads | 3.4 / 4.9 GHz | - | 105W | DDR4-3200 | 64MB (2x32) |
| Core i9-12900K / KF | $589 (K) - $564 (KF) | 8P + 8E | 16 Cores / 24 threads | 3.2 / 5.2 GHz | 2.4 / 3.9 GHz | 125W / 241W | DDR4-3200 / DDR5-4800 | 30MB |
| Ryzen 9 5900X | $549 | 12P | 24 threads | 3.7 / 4.8 GHz | - | 105W | DDR4-3200 | 32MB (1x32) |
| Core i9-11900K | $549 | 8P | 16 threads | 3.5 / 5.3 GHz | - | 125W | DDR4-3200 | 16MB |
| Core i7-12700K / KF | $409 (K) - $384 (KF) | 8P + 4E | 12 Cores / 20 threads | 3.6 / 4.9 GHz | 2.7 / 3.8 GHz | 125W / 190W | DDR4-3200 / DDR5-4800 | 25MB |
| Core i7-11700K | $409 | 8P | 16 threads | 3.6 / 5.0 GHz | - | 125W | DDR4-3200 | 16MB |
| Ryzen 7 5800X | $449 | 8P | 16 threads | 3.8 / 4.7 GHz | - | 105W | DDR4-3200 | 32MB |
| Core i5-12600K / KF | $289 (K) - $264 (KF) | 6P + 4E | 20 Cores / 16 threads | 3.7 / 4.9 GHz | 2.8 / 3.6 GHz | 125W / 150W | DDR4-3200 / DDR5-4800 | 16MB |
| Core i5-11600K | $272 | 6P | 12 threads | 3.9 / 4.9 GHz | - | 95W | DDR4-3200 | 12MB |
| Ryzen 5 5600X | $299 | 6P | 12 threads | 3.7 / 4.6 GHz | - | 65W | DDR4-3200 | 32MB |
The Alder Lake chips use the Intel 7 process, which used to be referred to as '10nm Enhanced SuperFin' before Intel recently renamed its process nodes during its latest process and packaging roadmap update. The Golden Cove cores support Hyper-Threading, allowing two threads to run on a single core, while the smaller Gracemont cores are single-threaded. That means some models could come with seemingly-odd distributions of cores and threads. All Alder Lake chips support DDR4-3200 or DDR5-4800.
Intel's $589 16-core Core i9-12900K comes with eight P-cores that support hyper-threading, and eight single-threaded E-cores for a total of 24 threads. That's a 33% increase in thread count over the previous-gen Core i9-11900K. The P-cores have a 3.2 GHz base, and peak frequencies reach 5.2 GHz with Turbo Boost Max 3.0 (this feature is only on P-cores). This chip comes with 125W PBP (base) and 241W MTP (peak) power ratings.
The 12900K has a 100 MHz reduction in peak clock frequency compared to the 11900K, but that isn't too meaningful given the entirely new hybrid architecture — these chips will realize performance gains from using different core types for different tasks. Speaking of which, the E-cores have a 2.4 GHz base and stretch up to 3.9 GHz via the standard Turbo Boost 2.0 algorithms. The chip comes armed with 30MB of L3 cache and 14MB of L2.
At $589, the Core i9-12900K comes at a $40 premium over its prior-gen counterpart, squeezing in between the $799 16-core Ryzen 9 5950X and $549 Ryzen 9 5900X. That could be attractive if Intel's performance claims pan out (more later), but it leaves a sizeable $185 gap between the Core i9 and i7 families that Intel inadequately plugs with the graphics-less $564 Core i9-12900KF. It's logical to expect a filler product between Core i7 and i9 in the future (possibly like the Core i9-10850K).
The $409 Core i7-12700K comes with the same $409 tray pricing as the previous-gen Core i7-11700K and has eight P-cores and four E-cores, for a total of 20 threads. The P-cores run at a 3.6 / 5.0 GHz base/boost, while the E-cores weigh in at 2.7 / 3.8 GHz, all fed by 25MB of L3 cache and 12MB of L2. The graphics-less $384 Core i7-12700KF comes with a $25 price reduction.
The 12700K's $409 price point means that Intel has kept the Core i7 flagship at its same price point, where it lands between the $449 Ryzen 7 5800X and $399 Ryzen 5 5600X. The 12700K/F's increased performance could make it a more attractive part than its lackluster previous-gen counterpart, the hard-to-recommend Core i7-11700K.
The Core i5-12600K's $289 price point remains the same as the prior-gen Core i5-11600K, meaning it lands right smack dab in gamer country, going toe-to-toe with the $299 six-core Ryzen 5 5600X and representing the lowest point of entry to the Alder Lake family (at least for now). This chip comes with six threaded P-cores that operate at 3.7 / 4.9 GHz and four E-cores that run at 2.8 / 3.6 GHz, for a total of 16 threads. That's paired with 20MB of L3 and 9.5MB of L2 cache.
AMD's competing Ryzen 5 5600X currently leads our Best CPU for gaming list, but it faces a stiff challenge from Intel's 12600K.
All the Alder Lake chips support both DDR4 and DDR5 memory, but there are several caveats to the listed DDR5 support. As a default, DDR5 runs in Gear 2 mode, resulting in higher latency, and standard motherboards only support DDR5-4800 if the motherboard has only two physical slots. Therefore, at stock settings, the chip will only support DDR5-4400 on any motherboard with four slots, even if only two slots are populated.
Intel has discarded its 'TDP' (Thermal Design Point) nomenclature, and now assigns a Processor Base Power (PBP) metric in its place. The company also added a secondary Maximum Turbo Power (MTP) metric to its spec sheets to quantify the highest power level during boost activity (typically called PL2).
Alder Lake's new memory controllers support four different memory types: DDR5-4800 and LP5-5200, along with DDR4-3200 and LP4x-4266. This single design's broad memory support enables different types of memory configurations for different use-cases. It appears that Intel will split its memory support into DDR4 for lower-end Z690 motherboards, B- and H-series models, and mobile systems, while DDR5 will only slot in for the highest-end Z-series motherboards. This makes sense given the expected high pricing for DDR5 memory in the early days of adoption, though it's notable that Intel hasn't confirmed its approach yet.
Alder Lake also supports up to PCIe 5.0 with 64 GB/s of throughput across a x16 lane connection. The desktop PC chips support a x16 PCIe Gen 5 connection with an additional x4 PCIe Gen 4 connection (it is unclear if this x4 connection is used for the chipset or exposed to the user), while lower-power models support a x12 PCIe Gen 4 config paired with a x16 PCIe Gen 3 connection.
The first chips based on the design come in three different packages, each for a different segment: The desktop PC chip that will drop into new motherboards with an LGA 1700 CPU socket (yes, 115x coolers with converters are compatible), a high-performance BGA Type3 package for mobile applications (this is likely a 12-28W UP3 package, though Intel hasn't confirmed), and a high-density BGA Type4 HDI package for Ultra Mobile applications (likely a 7-15W UP4 equivalent for ultra-thins).
Intel Alder Lake Gaming Benchmarks and Performance
As with all vendor-provided results, take these with a grain of salt. Also, be aware that some of these benchmarks come from Intel-sponsored game titles. We have included Intel's test notes here. Intel tested gaming with DDR5-4400 memory in Windows 11 but says that we can expect a comparable level of performance with DDR4 memory and/or Windows 10. Intel claims that Alder Lake retains the performance lead over Ryzen regardless of the type of memory or OS used.
For testing, the company used an Nvidia RTX 3090 GPU at the 1080p resolution — standard fare for CPU gaming benchmarks. Intel tested with Windows Defender and the performance-sapping Virtualization Based Security (VBS) feature active (you can see our testing of the impact of VBS on gaming here).
Intel admitted that it tested without the Windows 11 patch that fixes an L3 cache and boost clock issue with AMD's chips. Intel told us during briefings that it will retest with the patch once it is available (it has since become available) and update these benchmarks if there are any material changes. (We've tested the Windows 11 and AMD patches and haven't seen any drastic improvements.)
Intel's results point to the 12900K having a sizeable lead over the 5950X in seven titles, ranging from an 8% to 30% lead, along with a 3% loss in Shadow of the Tomb Raider and a tie in Crysis Remastered. Most of these titles are lightly threaded, so they don't benefit too much from Ryzen 9 5950X's core counts. Intel says its losses/ties are typically in GPU-bound titles.
Intel pointed out that some titles, like Hitman 3, are already optimized to support running physics and background tasks on Alder's E-cores, boosting performance. Intel expects optimizations for the hybrid architecture to filter out to more game titles in the future, as latency-sensitive tasks like rendering perform best on P-cores while hybrid-aware engines can shuffle background tasks like physics and audio to the E-cores. Optimizations for hybrid architectures can vary, though: For instance, Mount and Blade is optimized to run on all cores for the best performance.
Intel avoided Ryzen 5000 in the content creation benchmarks. Here we can see that the company put the Core i9-12900K head-to-head with the previous-gen 11900K and tested 31 game titles, with the 12900K coming out on top by a geometric mean of 13%. Intel also displayed a multi-tasking workload that consisted of simultaneously gaming while streaming and recording with OBS, with the latter two functions running on the E-cores while the P-cores ran the game code. Intel claimed that the Core i9-12900K provided 84% higher fps than the 11900K.
Intel's content creation benchmarks avoided any direct comparisons with AMD's competing processors and consisted entirely of applications that benefit from optimizations for hybrid architectures. That means these tests aren't indicative of what you can expect with most applications, but we'll suss that out in our coming review.
There will be teething pains, though. As we reported, Denuvo DRM falsely identified Intel's E-cores as a separate system, and thus 91 Denuvo-enabled game titles wouldn't work with Alder Lake chips. Intel has worked with Denuvo, and the software maker issued a flurry of game patches to fix the issue. However, 32 titles are still not patched. Intel says 18 of those game titles land in the 'Top 100' list of Denuvo titles. Of those, sixteen have already been patched, with the remaining two slated for a fix by the end of the month. Intel says that all games should eventually work with Alder Lake.
Intel Alder Lake-P and Alder Lake-M Mobile Processor Specifications
| Big + Small Cores | Cores / Threads | GPU |
| 6 + 8 | 14 / 20 | GT2 Gen12 96EU |
| 6 + 4 | 10 / 14 | GT2 Gen12 96EU |
| 4 + 8 | 12 / 16 | GT2 Gen12 96EU |
| 2 + 8 | 10 / 12 | GT2 Gen12 96EU |
| 2 + 4 | 6 / 8 | GT2 Gen12 96EU |
| 2 + 0 | 2 / 4 | GT2 Gen12 96EU |
*Intel has not officially confirmed these configurations. Not all models may come to market. Listings assume all models have Hyper-Threading enabled on the large cores.
The Alder Lake-P processors are listed as laptop chips, so we'll probably see those debut in a wide range of notebooks that run the gamut from thin-and-light form factors up to high-end gaming notebooks. Intel recently released a game developer guide that mentioned a 6 + 8 and 2 + 8 configuration, but a recent benchmark posting revealed the existence of Intel's 14-core model for laptops. There is precious little information available for the -M variants, but they're thought to be destined for lower-power devices and serve as a replacement for the recently-retired Lakefield chips.
As you'll notice above, all of these processors purportedly come armed with Intel's Gen 12 Xe architecture in a GT2 configuration, imparting 96 EUs across the range of chips. That's triple the execution units over the desktop chips and could indicate a focus on reducing the need for discrete graphics chips. What we've seen from the current 96 EU Xe solutions (i.e. Tiger Lake) suggests that performance might be on the level of a GT 1030, however, so count on gaming laptops including dedicated GPUs.
Finally, an Alder Lake-L version has been added to the Linux kernel, classifying the chips as '"Small Core" Processors (Atom),' but we haven't seen other mentions of this configuration elsewhere. Alder Lake-N has also been listed by Intel, and it will target the educational segment.
Intel Alder Lake 600-Series Motherboards, LGA1700 Socket, DDR5 and PCIe 5.0
Intel's incessant motherboard upgrades, which require new sockets or restrict support within existing sockets, have earned the company plenty of criticism from the enthusiast community — especially given AMD's long line of AM4-compatible processors. That trend will continue with a new requirement for LGA1700 sockets and the 600-series chipset for Alder Lake (we've already seen plenty of listings and pictures of Z690 motherboards). Still, if rumors hold true, Intel will stick to the new socket for at least the next generation of processors (7nm Meteor Lake) and possibly for an additional generation beyond that, rivaling AMD's AM4 longevity.
Intel also introduced its 14nm Z690 chipset as part of today's release, and you can read about the chipset and some of the first 60+ motherboards in our Z690 motherboard roundup here. There's a wide selection of DDR5 motherboards spread among the various motherboard makers' high- and lower-end Z690 families. However, DDR4 models appear to be confined to the lower-end Z690 boards (unlike previous generations, no motherboard supports both DDR4 and DDR5). We expect pricing for DDR5 to be substantially higher than DDR4, currently projected to be a 50 to 60% markup, for some time.
Alder Lake supports up to 16 lanes of PCIe 5.0 (technically for storage and graphics only, no networking devices) and an additional four lanes of PCIe 4.0 from the chip for M.2 storage. Intel has also added 12 lanes of PCIe 4.0 that hang off the chipset, a nice step up from the Z590 chipset's PCIe 3.0 support. That provides a total of 28 PCIe platform lanes for Alder Lake systems. Intel also doubled the throughput of the DMI connection between the chip and chipset from an x8 DMI 3.0 pipe, which clocks in at 7.88 GB/s, to an x8 DMI 4.0 connection that delivers 15.66 GB/s. Intel also added support for the Volume Management Device feature that enables PCIe SSD management and the ability to create bootable RAID configurations.
Because the LGA1700 socket is bigger than the current sockets used in LGA1151/LGA1200 motherboards, existing coolers may be incompatible, but cooler conversion kits, which most cooler makers will provide for free for existing customers, can accommodate the larger socket. (Coolers that support both LGA11xx and LGA2066 already exist, so an in-between option isn't too difficult.)
The larger socket is needed to accommodate 500 more pins than the LGA1200 socket. Those pins are needed to support newer interfaces, like PCIe 5.0 and DDR5, among other purposes, like power delivery. Intel has also listed Alder Lake-S BGA support documentation, indicating that soldered-down models will also come to market.
PCIe 5.0 and DDR5 support give Intel a connectivity advantage over competing chips, but there are a lot of considerations involved with these big technology transitions. As we saw with the move from PCIe 3.0 to 4.0, a step up to a faster PCIe interface requires thicker motherboards (more layers) to accommodate wider lane spacing, more robust materials, and retimers due to stricter trace length requirements. All of these factors conspire to increase cost.
We can expect those same PCIe 4.0 requirements to become more arduous for motherboards with a PCIe 5.0 interface, particularly because they will require retimers for even shorter lane lengths and even thicker motherboards. That means we could see yet another jump in motherboard pricing over what the industry already absorbed with the move to PCIe 4.0. Additionally, PCIe 5.0 also consumes more power, which will present challenges in mobile form factors.
Intel has announced that Alder Lake will support DDR5 memory, but that will cause pricing pressure. Notably, every transition to a newer memory interface has resulted in higher up-front DIMM pricing, which is concerning in the price-sensitive desktop PC market. DDR4 for example first came to the HEDT segment on Intel's X99 platform in 2014, and pricing at the time was more than double the cost of DDR3. Skylake brought DDR4 to the mainstream segment in 2015, but it still carried a 25-50% price premium. Current signs point to a 50% to 60% premium for DDR5 memory.
DDR5 is in the opening stages; some vendors, like Adata, TeamGroup, and Micron, have already begun shipping modules. The inaugural modules are expected to run in the DDR5-4800 to DDR5-6400 range. The JEDEC spec tops out at DDR5-8400, but as with DDR4, it will take some time before we see those peak speeds.
We have, however, seen signs that only the higher-end Alder Lake desktop PC platforms, like Z-series motherboard, will support DDR5, while lower-end boards will use DDR4 for a friendlier price of entry.
Intel has long listed the TDP of a processor as its guaranteed rating at base frequencies, also known as PL1. However, the chip can also opportunistically (meaning this isn't guaranteed) boost to higher frequencies and thus consume far more power, but only if it is safely within certain power, temperature, and current limits. This is called the PL2 power state, and Intel hasn't included this metric on its standard spec sheets.
Now Intel has redefined its power nomenclature to have a 'Processor Boost Power' (PBP) value representing the guaranteed base performance level (PL1). This replaces TDP. Intel will also now list a 'Maximum Turbo Power' (MTP) specification that quantifies the power consumption during Turbo Boost, also known as PL2. That means you'll no longer see a TDP rating on the spec sheet.
Intel's processors have a 'Tau Duration' setting that dictates how long the processor can stay in the boosted MTP state (PL2) before it drops back down to the PBP state (PL1 – base power). Intel specified this duration as 58 seconds for Rocket Lake chips, but this is only a guideline. Motherboard vendors are free to alter this value to any length of time if their motherboard can handle the power delivery required to sustain the boost. As shown in the graphic above, most motherboard vendors change the Tau setting to infinite to stay within boost for an infinite amount of time. Given that Intel's Tau settings are only recommendations, the chip remains inside of the warranty regardless of boost duration. As an infinite Taue is a common practice on nearly every enthusiast motherboard, Intel will now set the Tau to a default of 'infinite' for all of its K-series (overclockable) models, but retain the same 58-second duration for its locked chips. This means that the Core i9-12900K's MTP (PL2) is now the same as its PBP (PL1). In other words, the chip will always operate at 241W.
Additionally, the new desktop PC motherboards for Alder Lake chips will herald the arrival of mainstream ATX12VO motherboards that leverage a new lower-power PSU specification. Both systems with support for standard power supplies and the ATX12VO spec are planned, but Intel is on a full-court press to push the adoption of the new standard. However, despite Intel's fondness for the standard, we have yet to see a significant number of 600-series boards that support ATX12VO. Almost all boards work with standard power supplies.
Intel 12th-Gen Alder Lake Xe LP Integrated Graphics
The media engine, in this case the same Gen12 Xe LP architecture found in Tiger Lake but ported to the Intel 7 process, comes in two variants: one with 32 EUs (GT1) for desktop PCs, and another GT2 variant with 96 EUs for the mobile variants. The desktop PC models come with 33% more EUs than the current desktop chips with Gen9.5 UHD 630 Graphics, but that's a far cry from the 96 EUs found in 11th Gen Tiger Lake. But this is on the desktop, where most users that care about graphics performance will simply use a dedicated GPU.
Intel says the Xe LP engine supports 1080p gameplay and features a 12-bit end-to-end video pipeline. The desktop PC models don't have Thunderbolt 4 connectivity or an image processing unit (IPU), with those features being used only for mobile variants.
The UHD 770 engine clocks in at (up to) 1550, 1500, and 1450 MHz for the 12900K, 12700K, and 12600K, respectively.
We've also seen Alder Lake-P benchmarks (the mobile chips) with the GT2 configuration, with 96 EUs (768 shaders). The early Xe LP iGPU silicon on the -P model runs at 1.15GHz, but as with all engineering samples, that could change with shipping models.
Alder Lake's integrated GPUs support up to five display outputs (eDP, dual HDMI, and Dual DP++), and support the same encoding/decoding features as both Rocket Lake and Tiger Lake, including AV1 8-bit and 10-bit decode, 12-bit VP9, and 12-bit HEVC.
Intel Alder Lake CPU Architecture
Intel pioneered the x86 hybrid architecture with its Lakefield chips, with those inaugural models coming with one Sunny Cove core paired with four Atom Tremont cores.
Compared to Lakefield, both the high- and low-performance Alder Lake-S cores take a step forward to newer microarchitectures. Alder Lake-S actually jumps forward two 'Cove' generations compared to the 'big' Sunny Cove cores found in Lakefield. The big Golden Cove cores come with increased single-threaded performance, AI performance, Network and 5G performance, and improved security features compared to the Willow Cove cores that debuted with Tiger Lake.
Alder Lake's smaller Gracemont cores jump forward a single Atom generation and offer the benefit of being more power and area efficient (perf/mm^2) than the larger Golden Cove cores. Gracemont also comes with increased vector performance, a nod to an obvious addition of some level of AVX support (likely AVX2). Intel also lists improved single-threaded performance for the Gracemont cores.
You can read our deep-dive coverage of the Golden Cove Performance Core architecture here. In summary, the Golden Cove microarchitecture is designed for low-latency single-threaded performance and comes with an average of 19% more IPC than the Cypress Cove architecture in Rocket Lake. It also supports AVX-512 and AMX (a new AI-focused matrix-multiply ISA) for data center variants (both are disabled on consumer chips).
You can also read our deep dive coverage of the Gracemont Efficiency Core architecture here. In summary, the single-threaded Efficiency (E) core, which comes with the Gracemont microarchitecture, is designed to improve multi-threaded performance and provide exceptional area efficiency (small footprint) and performance-per-watt. Four of these small cores fit in the same area as a Skylake core and deliver 80% more performance in threaded work (at the same power). A single E core also delivers 40% more performance than a single-threaded Skylake core (at the same power) in single-threaded work (caveats apply to both).
Lakefield served as a proving ground not only for Intel's 3D Foveros packaging tech but also for the software and operating system ecosystem. At its Architecture Day 2020, Intel outlined the performance gains above for the Lakefield chips to highlight the promise of hybrid designs. Still, the results come with an important caveat: These types of performance improvements are only available through both hardware and operating system optimizations. Let's look at Intel's solution to that problem.
Intel Thread Director
Intel unveiled the answer to the software challenge at its Architecture Day 2021 — the new Thread Director. Due to Alder's use of both faster and slower cores that are optimized for different voltage/frequency profiles, unlocking the maximum performance and efficiency requires the operating system and applications to have an awareness of the chip topology to ensure workloads (threads) land in the correct core based on the type of application.
The current thread scheduling systems are based entirely on static rules (priority, foreground, background) and tend to be inefficient and create software programming overhead. That's where Intel's Thread Director technology comes in. This hardware-based technology provides enhanced telemetry data to Windows 11 to assure that threads are scheduled to either the P or E cores in an optimized and intelligent manner, but in a way that's transparent to software.
This technology works by feeding the Windows 11 operating system with low-level telemetry data collected from within the processor itself, thus informing the scheduler about the state of the core, be it power, thermal or otherwise. (As we covered here, Intel has integrated a new power microcontroller in each Gracemont core, a first, that collects similar data on the order of microseconds instead of milliseconds, so it might be part of the new telemetry system.)
Additionally, Thread Director can also detect the instruction mix (scalar/vector) used in any given thread at a nanosecond granularity, and then communicate with the Windows 11 scheduler to steer the thread to the correct execution core, be that a high-performance P-Core or an efficient E-Core. Typically, vector/AI workloads will be prioritized to performance cores while scalar instructions and background tasks are moved to efficiency cores. However, the system is dynamic, so thread placement decisions can vary based on the dynamic mix of conditions and workloads present on the processor at any given time.
Additionally, threads can go through various phases and instruction mixes over their lifetime, so the scheduler constantly re-adjusts based on the real-time telemetry data. This is helpful when the number of threads designated for 'performance' outnumber the available cores, for instance. In that case, less demanding 'performance' threads, such as a program in a spin loop, can be moved off to the efficiency cores while more deserving workloads are assigned to the performance core.
Previously, the operating system didn't have access to this type of telemetry data to inform scheduling decisions, instead using simple data like whether the process was a foreground or background task. This enhanced system allows the operating system and processor to work in tandem to assure correct scheduling in real-time, thus avoiding costly software re-coding. This is a promising sign that existing code will run well on the Alder Lake processors.
Programmers can access more granular control, too, by specifying that certain threads are used in a certain manner through an expansion of the PowerThrottling API that allows developers to assign a QoS attribute to their threads. Additionally, a new EcoQos classification lets software tag threads that respond best on the efficiency cores to assure they are prioritized to execute on the E-Cores. Microsoft says that the Edge browser and 'various' Windows 11 components now take advantage of the EcoQos classification system, and we can expect support to broaden quickly.
This looks to be a promising and less-intrusive (at least from a coding standpoint) method of ensuring that the correct threads land on the correct cores, thus delivering optimal performance. That said, we'll have to see it in action before we can pass judgement on its efficacy – much of its potency will boil down to the latency involved with the process of communicating telemetry data and moving the thread, and intel isn't sharing those details yet. Additionally, it's possible that an excess of communication between the Thread Director and the Windows 11 scheduler could create a challenging workload of its own, so finding the right amount of granularity will be key to assuring both timely thread placement and a minimum of system overhead.
The system is already far in development, and Microsoft says that further enhancements to the engine are already underway and in planning for Windows 11, with more details to be shared at a later date.
Alder Lake chips will also work fine with a bog-standard Windows 10 operating system – existing thread-scheduling techniques continue to work with the processors, just not as well. While the chips work, you'll miss out on the enhanced capabilities of Thread Director (that's Windows 11 only), which will have a varying impact on performance and power consumption based on instruction type and application usage models. In other words, your mileage will vary.
The hybrid architecture could still result in some teething pains, as Intel itself recently divulged that some older games with DRM might not work with the new chips unless developers add specific software optimizations.
Finally, it has long been known that the Gracemont cores do not support the AVX-512 instruction set, and speculation has been rife about how the code would work on Alder Lake processors, if at all. Intel's answer is simple: AVX-512 will not work on either type of core present in Alder Lake. The high-performance cores do feature the Golden Cove architecture that supports AVX-512 natively, but Intel has fused that feature off (yes, the 512-bit FMA is still present and consumes die area) for the consumer chips. In contrast, server chips with Golden Cove have two 512-bit FMAs and fully support AVX-512. Meanwhile, the Gracemont cores are simply not AVX-512 capable, and disabling support allows the Alder Lake chip to have uniform ISA support.
All that's left is third-party testing, but it appears that the AMD vs Intel battle has been reignited. Come back on November 4, 2021. We'll have all the details then.
You may be interested in:
>> Is a Chromebook worth replacing a Windows laptop?
>> Find out in detail the outstanding features of Google Pixel 4a
>> Top 7 best earbuds you should not miss
Related Posts:
>> Recognizing 12 Basic Body Shapes To Choose Better Clothes
>>Ranking the 10 most used smart technology devices
>> Top 5+ Best E-readers: Compact & Convenient Pen
0 Comments:
Post a Comment