In the early hours of September 3rd, Beijing time, Intel officially released the 11th generation of Low Power Core processors codenamed Tiger Lake, which are the biggest leap forward for Intel in recent years, from the process process to the underlying architecture, from specifications to application performance. Taking advantage of this opportunity, Intel not only redesigned core, nuclear LOGO, but also created a new Intel logo image, but also finally for Athena notebook prepared a unique name EVO, the future Athena notebook will have this label.
Tiger Lake’s enhanced 10nm process, the first addition of a new SuperFin transistor technology, claims to deliver performance improvements comparable to full node conversion.
The CPU section is a new Willow Cove architecture with a maximum of 4 core 8 threads and a maximum frequency of 4.8GHz, a qualitative leap from the Sunny Cove 3.9GHz in Ice Lake 10 Generation Core, thanks to architectural upgrades and 10nm SuperFin process technology.
The GPU nuclear display is the new Xe LP, the first of its kind in the Xe architecture, a low-power version with up to 96 EU execution units, up to half, up to 2x performance improvements, and support for deep learning acceleration instruction DP4A to accelerate AI applications.
Memory supports up to 32GB LPDDR4/4X-4266, 64GB DDR4-3200, and will support LPDDR5-5400 in the future, with Intel full memory encryption.
For the first time in connectivity, PCIe 4.0 is natively supported, both standaled graphics or SSDs are available, USB4, Thunderbolt 4, fully integrated with 4x DP/USB/PCIe, with bidirectional bandwidth of up to 40Gbps per port.
Multimedia and display, with two MEDIA encoders, up to 4K60fps 10-bit, 8K30fps 10bit, and four display lines for up to 8K60Hz or four units 4K60Hz screen with DisplayPort 1.4 HBR3, BT.2020, and IPU6 image processing unit for up to 4K90fpf resolution video, 42 megapixel photos ( Initial support for 4K30fps, 27 megapixels).
Ai, with the integration of the new Gaussic network accelerator GNA 2.0, is equivalent to the role of the NPU and can efficiently perform neurodescientization calculations with low power consumption and CPU occupancy.
Among them, the Xe LP core is completely new, with efficient thread control, 8 width appropriate units, first-level data caching, 3.8MB three-stage cache (increase of 25%), end-to-end compression, VRS variable coloring rate, up to 24 pixels per clock cycle, 48 textures.
FP16, FP32 floating-point performance improved by 84%, but also for the first time added INT8 integer processing power, performance 8.29TOPS.
It is worth noting that in addition to hardware improvements, Intel also attaches great importance to drive and application software construction, such as a complete rewrite of the DX11 driver, lower overload, while supporting game sharpening, configuration optimization, instant game tuning, naturally can better play the potential of hardware performance, the official repair can run most games smoothly at 1080p resolution.
In terms of chipsets, integrated Gig-Wi-Fi 6 (AX201), FIVR (fully integrated voltage regulator) and DVFS (dynamic voltage and frequency regulation), programmable four-core audio DSP (low-power voice wake-up/USB and Bluetooth audio), 4th generation CSME (improved decryption and side Channel protection), touch master (low power/touch and stylus synchronization), IO supports 4 USB 3.0, 10 USB 2.0, 12 PCIe 3.0, 2 SATA 6Gbps, 8 GWIOs.
Intel 11-generation Core low-power version debuted a total of 11 models, can be divided into two series, where the number number 5 end of the thermal design power consumption of 12-28W, 0 end is 7-15W, are no longer limited to a static power level, but the entire SoC in one range, according to the actual scene needs and real-time dynamic adjustment, notebook design is more flexible.
Intel recommends that the 7-15W can be used for 12-13 inch fanless silent ultra-thin book, 12-28W is suitable for 14-15 inch active cooling light book, light game book or design book.
In contrast, Ice Lake 10 Core numbers 8, 5, and 0 are fixed at 28W, 15W, and 9W, respectively.
First, summarize some of the rules of model naming and specifications:
1, i7 series 4 core 8 threads, 12MB three-level cache, i5 series 4 core 8 threads, 8MB three-level cache, i3 series 2 core 4 threads, 6MB three-level cache, but next year there will be new changes.
2, memory support, 12-28W version also supports DDR4-3200, LPDDR4X-4266 (i7)/3377 (i5/i3), 7-15W version only supports LPDDR4X-4266.
3, nuclear display level, i7 series are G7 96 units, i5 series are G7 80 units, i3 series are G4 48 units, no longer G1 level.
4, nuclear frequency, 12-28W version of 1.25-1.3GHz, 7-15W version of the unified 1.1GHz.
The flagship model is also expected to see the i7-1185G7 most often, the reference frequency is 3.0GHz, full-core Turf 4.3GHz, single-core Turf reaches 4.8GHz, and the core display frequency is 1.35GHz, after the news indicates that it can be overclocked to 1.6GHz, but it is not clear exactly how to exceed.
The i7-1165G7 CPU frequency is reduced to 2.8/4.1/4.7GHz and the core display frequency is reduced to 1.3GHz.
The i5-1135G7 CPU frequency continues to drop significantly to 2.4/3.8/4.2GHz.
The i3-1115G4 CPU frequency reference is the highest 3.0GHz, while the all-core and single-core Turf are 4.1GHz and the core display frequency is 1.25GHz.
The flagship model of the 7-15W version is the i7-1160G7 with a CPU frequency of 1.2/3.6/4.4GHz. The i5-1130G7 frequency is reduced to 1.1/3.4/4.0GHz, and the i3-1110G4 reference frequency is the highest 1.8GHz in the series, with full-core and single-core Turf at 3.9GHz.
In addition, next year will add two special i3, actually 4 core 8 threads, one is i3-1125G4 2.0/3.3/3.7GHz, the other is i3-1110G4 1.1/3.0/3.5GHz.
In other words, their CPUs are the highest specifications, while GPUs are the lowest specifications, suitable for devices and scenarios that do not require graphical performance but need to calculate performance.
Intel also revealed that the 11th-generation Core family’s entry-level Pentium, Saiyan models, as well as the Bolly platform, will follow later, as well as DG1 indesibilly graphics.
Acer, Asus, Dell, Dynabook, HP, Lenovo, LG, MSI, Thunder Snake, Samsung and others are expected to launch more than 150 notebooks with 11 Core processors, more than 20 of which are EVO Athena certified.
Next look at some of Intel’s official performance comparison data, not only compared to their own last generation of products, but also compared to AMD Ryron’s, flagship i7-1185G7 against the flagship R7 4800U, this practice is really quite rare, of course, the data is for reference only.
Based on the benchmark, computing performance is 28% higher, graphics performance is 67% higher, and AI performance is four times higher.
Of course, Intel now emphasizes real-world performance, working, authoring, gaming, connecting all-round crushing.
AI is also an aspect of Intel’s full commitment and promotion, with the addition of DP4A, VNNI dedicated instruction sets, and a separate GNA 2.0 module, which is still basically zero next door.
On the CPU side, the dynamic power range has just been talked about, with officials claiming that performance at 28W can be 33-37% higher than at 15W.
This generation of 28W limit state compared to the last generation of 25W limit state, performance can also generally lead 10-30% of the look, while crushing the limit performance mode of the Sharp Dragon.
AI performance is even more out of the same language, the 10th generation is more than 40% higher than next door, the 11th generation can lead by about 77% with CPU calculations, based on the new hardware units and instruction sets are several times more than the lead, up to more than 11 times.
Performance losses are smaller when powered, at about 5%, up to 38% next door.
GPU, Iris Xe core display at 1080p resolution, 3A can generally exceed 30FPS, lightweight games stable over 60FPS, e-games are generally above 100FPS.
Compared to the 10 generation performance improvement is less than 50%, can be doubled at most.
Compared next door can also lead up to 80%, although exaggerated, but these two generations of Sharp Dragon in the GPU really progress is very few, the architecture does not upgrade, the computing unit is still less, or rely on high frequency support.
Even if Iris Xe can already get rid of stand-alone graphics cards like the MX350, it’s no wonder NVIDIA has rushed out the new architecture of the MX450 and supported PCIe 4.0.
VRS variable coloring rate, Iris Xe also supports this time, can increase performance, AMD will have to wait until the RDNA2 architecture.