Hardware tipster HXL (@9550pro) has now released mold screenshots of AMD Phoenix 2, showing the processor design of the new generation Zen 4 core + Zen 4c core, which obviously uses a 2+4 core 12-thread design.

From the picture we can see that this chip uses a large L3 cache (located in the middle on the left, green structure), two high-performance Zen 4 cores (below the L3 cache), four smaller Zen 4c cores (three one above the L3, one to the right of the Zen 4 core), and a large built-in GPU (on the right).

In addition, there are multiple DDR5/LPDDR5 PHY interfaces on the top of the die, and there are physical interfaces such as PCIe and USB on the bottom and left and right sides of the APU.

These cores are significantly smaller than similar products, suggesting AMD is positioning the Phoenix 2 as a low-cost APU aimed at budget laptops.

While we don’t yet know exactly what AMD cut out of Phoenix to make it so small and cheap, AMD is likely to reveal these new models in the near future, and may also include them in the Ryzen 8000 series launching in 2024 product lineup, we will be able to get a clearer picture of this product then.

AMD officially released the 7040U series of processors in May this year, including four models including the R3 7440U, covering 4-core to 8-core models. Among them, the R5 7640U and R5 7540U use the 6C12T design.

AMD later confirmed to xda-developers that the company does have a six-core APU and that it will be used in the 7440U. Although AMD did not confirm the underlying design, other details are perfectly consistent with the revelations. Unlike the previously announced eight-core R3 7440U, it is almost certainly the Phoenix2 processor.

In addition, up master @金 Pig Upgrade Pack also previously released real shots of the appearance of Phoenix2. According to various leakers, this Phoenix 2 chip will feature two Zen 4 cores and four Zen 4c cores, for a total of 6 cores and 12 threads.

AMD said that it is not just a Phoenix product that switches from 8 cores to 6 cores, but also has adjustments in CPU, GPU and even AI.

Initial rumors suggested that Phoenix 2 would be present in the Ryzen 3 7440U and Ryzen 5 7540U, and AMD said the Ryzen 5 7540U is not a similar design, at least “just the 7440U for now.”

In addition, it was previously reported that “Phoenix 2” will use 2 full-blooded Zen4 cores and 4 low-power Zen4C cores for processing background tasks. The RDNA 3 GPU has been reduced from six WGP (12 CU) to two (4 CU), and the TDP has dropped from “35W+” to a maximum of 28W.

It was reported in March this year that information about the Phoenix “hybrid structure” processor appeared in AMD’s official documents at the time. This processor corresponds to the Phoenix 2 mobile APU, uses TSMC N4 process, has a power consumption of 15-28W, and is used for Products such as thin and light laptops or game consoles.

According to AMD’s official PPT, various indicators of AMD Zen 4c and Zen 4 are basically the same. The difference is that the Zen 4 core has 4MB L3 cache per core, while Zen 4c has 2MB.

AMD data shows that Zen 4c has the exact same IPC (performance at a given frequency) as Zen 4 because its front end, execution stages, load/store components and internal cache hierarchy are exactly the same.

The post Zen 4 + Zen 4c core, AMD Phoenix 2 CPU mold exposed with large L3 cache appeared first on TechGoing.

The AMD Instinct MI300 APU adopts AMD’s Zen 4 and CDNA 3 architecture and will be used in the El Capitan supercomputing in the United States. The supercomputing is expected to be put into use in 2024, and the FP 64 peak computing performance will exceed 2 ExaFLOPS.

AMD’s Instinct MI300 is a multi-chip APU featuring Zen 4 general-purpose x86 cores and a CDNA 3-based compute GPU that share a unified pool of packaged memory, including Infinity Cache cache and HBM shared memory design. AMD will use TSMC’s N5 (5nm-class) manufacturing process to produce CPU/GPU chipsets for the MI300, but the company did not disclose the number of CPU cores and GPU stream processors for the Instinct MI300.

Papermaster also revealed that AMD is expected to launch an EPYC Bergamo processor competing with Arm servers such as the AWS Graviton series in the first half of 2023, using the Zen 4C architecture.

AMD EPYC Bergamo processors will feature up to 128 cores and will be optimized for HBM-powered Xeon chips, as well as server products from Apple, Amazon and Google, using the SP5 socket, optimized for higher-throughput workloads.

Additionally, the AMD Genoa-X CPU, expected to be in production late Q3/early Q1 2023 and to launch around mid-2023, will feature a similar architecture to the Milan-X chip with 3D V-Cache design method. The Genoa-X CPU will be equipped with more than 1GB of L3 cache and up to 96 cores based on the Zen 4 architecture.

The post AMD Instinct MI300 APU successfully running with Zen 4 and CDNA 3 architecture appeared first on TechGoing.

Zen 4 EPYC server processor vs. Zen4 Ryzen processor

The fourth-generation AMD EPYC processors include 14 models ranging from 16-core to 96-core, all using Zen4 architecture, and also support DDR5 memory and PCIe Gen 5 storage, support CXL1.1+ memory expansion, in the safe The performance has been greatly enhanced, such as the expansion of AMD Infinity Guard, and the number of encryption keys has been increased by a factor of 2.

AMD says that with 4th Gen AMD EPYC processors, businesses can:

• Performance and efficiency support. With up to 96 cores in a single processor, the new AMD EPYC processors help customers further meet their computing needs by deploying fewer, more powerful servers. This enables them to achieve greater flexibility in the data center to support corporate sustainability goals and drive real-world dividends.

・Data security protection. Through the “Security by Design” system, AMD can further provide stronger security features in x86 CPUs. The 4th Gen AMD EPYC processor family is enhanced with AMD Infinity Guard, an advanced feature set that provides both physical and virtual layer protection. The 4th Gen AMD EPYC processor has 2x the number of encryption keys compared to its predecessor, helping customers keep data safe whether it’s stored locally, in the cloud, or residing in memory.

• Adopt the latest industry features and architectures. These processors will further provide customers with a comprehensive feature set, simply picking the number of cores and frequencies that suit their workload needs. 4th Gen AMD EPYC processors introduce DDR5 memory and 5th Gen PCIe support critical to AI and ML applications. Additionally, 4th Gen AMD EPYC processors support CXL1.1+ for memory scalability to help customers achieve greater in-memory workload capacity needs.

4th Generation AMD EPYC Processor Product List

The post AMD Zen 4 EPYC server processor picture tour: with 96-core appeared first on TechGoing.

AMD has announced that it will unveil the fourth-generation EPYC (Xiaolong) server CPU code-named Genoa at 10:00 a.m. Pacific Time on November 10.

In the third quarter, TSMC’s 5-nanometer revenue accounted for 28%, ranking the largest source of revenue. Some people in the industry have assessed that TSMC’s 5nm capacity utilization rate in the fourth quarter is still at full load. Although the capacity utilization rate in the first and second quarters of next year is slightly lowered, it still remains above 90%. As for the second half of next year, it will return to full load.

TSMC pointed out in a recent legal meeting that the destocking of consumer chips has caused TSMC’s 7-nanometer and 6-nanometer utilization rates to begin to decline significantly, but TSMC is still optimistic about the continued increase in demand for the 5-nanometer process in the fourth quarter, which will balance the weakening end market demand. And the impact of customers’ continued inventory adjustments, TSMC’s fourth-quarter results were flat with the previous quarter.

AMD’s RDNA 3 architecture graphics card released last week is the world’s first chiplet design GPU, GCD and MCD combine 5nm and 6nm process, providing up to 54% performance per watt improvement over the previous generation RDNA 2 architecture.

In addition, in order to reduce the negative impact of weak demand for consumer electronics products on operations, AMD’s product strategy in the next 2-3 years will focus on the development of artificial intelligence (AI) and high-performance computing (HPC), and expand in 5G infrastructure, automotive Electronics, aerospace defense, industrial and medical, data centers and other market layouts, and TSMC will be the most important partner.

AMD will officially announce the launch of the Zen 4 architecture Genoa processor this week, which will be shipped to major OEMs by the end of the year. Next year, it will launch Bergamo for native cloud computing, Genoa-X for technology and database computing, as well as edge terminals and telecom. The infrastructure Siena and other processors use TSMC’s 5nm or 4nm process.

In addition, AMD also plans to launch the data center acceleration processor MI300 for AI computing, equipped with Zen 4 and RDNA 3 cores, and is expected to use TSMC’s 5nm mass production next year.

The post TSMC will produce 5nm Zen 4 Genoa processors for AMD, new products launch this week appeared first on TechGoing.

The highest-end Ryzen 9 PRO 5945 reaches 12 cores, but the strange thing is that the thread specification is left blank, and there is a high probability that multi-threading is not supported, but it is the first time to hide in this way.

It has a 6MB L2 cache, a 64MB L3 cache, a base frequency of 3.0GHz, and a maximum acceleration of 4.7GHz, which is 700/100MHz lower than the Ryzen 9 5900X.

The Ryzen 7 PRO 5845 has 8 cores and 16 threads, the second-level cache is 4MB, the third-level cache is 32MB, and the frequency is 3.4-4.6GHz, which is 400/100MHz lower than the Ryzen 7 5800X.

The Ryzen 5 PRO 5645 has 6 cores and 12 threads, the second level cache is 3MB, the third level cache is 32MB, and the frequency is 3.7-4.6GHz, which is exactly the same as the Ryzen 5 5600X, because the latter is also 65W.

All three are AM4 package interfaces, support PCIe 4.0, and dual-channel DDR4-3200 memory, and cannot be overclocked.

The post AMD’s three Ryzen PRO 5000s are online: 12 cores do not support multi-threaded models appeared first on TechGoing.

(From: Phoronix)

It is clear that with CPU vulnerability mitigation enabled by default – the Zen 4 chip still maintains a pretty good overall advantage compared to disabling mitigation.

Phoronix wrote.

"On Zen 4 platforms, you can still use mitigations=off to disable SSB and Spectre (V1/V2) mitigations while the kernel is booting, leaving the system in a vulnerable state.

Although many of the CPU security vulnerability mitigations we've seen before, have taken some performance hits. But on the Zen 4 Ryzen R9-7950X, you don't actually have to disable it."

Across the 190 tests conducted in this evaluation, the default state of enabling CPU vulnerability mitigations won an average of nearly 72%.

Zen 4 CPU Vulnerability Mitigations – GeoMean Test Comparison

Neowin points out that while Zen 4 is largely an improvement on Zen 3, its test scores using the retpoline patch on the previous-generation R9-5950X platform are still slightly behind Intel CPUs.

The post Tests show AMD Zen 4 carefully deployed CPU vulnerability mitigations appeared first on TechGoing.

The 7000 series processors, including the Ryzen 9 7950X and Ryzen 9 7900X, Ryzen 7 700X and Ryzen 5 7600X, are based on TSMC’s 5nm process and feature the new Zen 4 architecture.

The Zen 4 architecture exceeds the original preset goal of increasing the clock instruction count (IPC) by 10%. The new architecture increases the IPC by 13%, and the Zen 4 architecture improves single-threaded performance by 29% and multi-threaded performance by over 35% compared to the previous generation Zen 3 architecture.

And thanks to the partnership with TSMC, the Zen 4 architecture also features higher clock speeds and better thermal design, with more than 25% performance improvement per watt over the previous generation.

It is worth noting that the Zen 4 architecture will also add support for the AVX-512 instruction set. Intel just announced in March this year that the instruction set would be disabled in subsequent production of the 12th generation Core processors. Previously Intel had found that running such instruction sets would make the CPU power consumption increase significantly.

Su Zifeng said that the 7000 series Rex scores on the Geek bench exceeded Intel’s flagship processor i9-12900K.

At the same time, Su also claimed that the 7000 Series Rex flagship model, Ryzen 9 7950X, is by far the most powerful gaming CPU in the world.

Ryzen 9 7950X has 16 cores and 32 threads. With a base frequency of 4.5GHz and a peak frequency of 5.7GHz, it is 800MHz higher than its predecessor and 200MHz higher than Intel’s 12th generation Core flagship processor, the i9-12900KS.

In terms of actual performance, AMD claims that in gaming scenarios, the AMD Ryzen 9 7950X will be able to provide a 35% performance boost compared to the Core i9-12900K in games such as Tomb Raider. And in scenarios running v-ray, the Ryzen 9 7950X is 62 percent faster than the Intel period CPU i9-12900K and delivers 47 percent more performance per watt.

The Ryzen 9 7900K, on the other hand, is equipped with 12 cores, 24 threads, and a maximum frequency of 5.6GHz, positioning it on par with the AMD Ryzen 9 5900X, and at the same price as the latter, so it can be said to be an increase in volume without an increase in price.

The entry model Ryzen 5 7600X is equipped with 6 cores and 12 threads. With a base frequency of 4.7GHz and a maximum frequency of 5.3GHz, AMD claims that this entry-level chip still has a 5% performance increase compared to Intel’s 12th generation Core flagship products.

The flagship AMD Ryzen 9 7950X is priced at $699, the AMD Ryzen 9 7900X at $549, the Ryzen 7 7700X at $399 and the AMD Ryzen 5 7600X at $299.

AMD said the Raider 7000 series processors will be available on September 27.

AMD also debuted a new GPU with the RDNA 3 architecture at the event, which, according to Su Zi Feng, takes a small-chip approach to stacking and uses TSMC’s 5nm process to increase performance per watt by 50% compared to existing Radeon GPUs.

A demo video of a game using Ryzen 9 7950X with RNDA 3 architecture GPU was shown live, but no further details of the GPU’s performance in the game were announced.

In addition to the product announcements, AMD also showed off its blueprint for the coming period.

At the launch AMD showed a more detailed roadmap of the Zen architecture. According to the roadmap, AMD will release the Zen 5 architecture processor with 4nm/3nm process in 2024.

Su Zifeng also said that in the next few quarters, AMD will also release Zen4C, RDNA3 and XDNA three new architectures.

The post AMD Zen 4 architecture to kill the crazy! appeared first on TechGoing.

She replied, “Indeed, if you look back over the last 18 months, it does vocalize a lot of issues, both in terms of capacity constraints and logistics. From AMD’s perspective, we have dramatically increased our overall tea capacity in terms of wafers, substrates and back-end. So with the launch of Zen 4, we don’t anticipate any supply shortages.”

“On a logistical level, it takes longer to ship to each region. Much so that we held the event at the end of August, but will go on sale on September 27. Truthfully, one of the reasons for doing that is to make sure we have the product in every region so that people can actually consider a full purchase.”

A number of high-end motherboards will be available on September 27, as well as four different Ryzen 7000 series processors. Prices and parameters for the processors are shown below.

The post Dr. Su: There will be no shortage of AMD Zen 4 chip supply appeared first on TechGoing.

LbrExtV2 is expected to be integrated into the “perf” subsystem of the current branch before the release of Linux Kernel 6.1. The merge window for Linux 6.1 will open in October 2022.

The Last Branch Record (LBR) function determines and analyzes the control flow of software by recording branch information to registers in real-time. Branch information will allow the system to decide which paths “hot code” should point to, such as compiler optimizations for a particular profile or other optimizations seen in the application.

AMD’s patchwork will allow the LbrExtV2 functionality to be incorporated into the latest “perf” subsystem in Linux and allow the observation of performance events. The new LbrExtV2 includes “LBR-Freeze-on-PMI for better correlation of PMC overflow events, new speculative information, and new hardware-based filtering support if data for specific branch types is required.” Larabel reports. The filters included in the latest branch will include “kernel only, userspace only, conditional branches, near-close calls, near-indirect calls, near-returns, near-indirect jumps, near-relative jumps, and far branches”.

The post AMD Zen 4 Last Branch Record Extension v2 Coming Soon to Linux 6.1 appeared first on TechGoing.

Compared to Zen 3, the Zen 4 architecture is expected to deliver an 8-10% increase in IPC performance. And with a higher TDP and clock rate, the Raider 7000 series desktop CPUs are expected to deliver even greater surprises.

"AMD highlights that Zen 4 delivers >15% single-threaded, >35% multi-threaded, and >25% performance per watt improvements over Zen 3.

The next generation CPUs feature optimally reorganized cache, doubled L2 (from 512KB to 1MB) + shared L3 cache similar to the previous generation.

There is also support for DDR5 via AMD EXPO memory overclocking expansion profiles, and PCIe 5.0 graphics / M.2 SSDs."

(1) To cut to the chase, the AMD R9-7950X continues the 16C / 32T design with a base frequency of 4.5 GHz / acceleration of up to 5.7 GHz.

"Such impressive numbers are even 200 MHz higher than the Intel Core 12 i9-12900KS with a single core 5.5 GHz RWD.

Total cache is 80MB, divided into 16MB L2 (1MB per core) + 64MB L3 (32MB per CDD), with a nominal 170W TDP / 230W PPT power consumption."

While it’s unclear what AMD’s pricing strategy will be, the R9-7950X will clearly be a strong successor to the R9-5900X and leave the Core i9-12700K far behind.

(2) The second R9-7900X chip is a 12C / 24T design with a base frequency of 4.7 GHz / acceleration up to 5.6 GHz. The total cache is 76MB, divided into 12MB L2 + 64MB L3.

Its positioning is similar to the R9-5900X, but thermal design power consumption is also relaxed to 170W, and it is expected to take Core i7-12700K off the table.

(3) The third R7-7700X chip is an 8C/16T design, officially positioned as the “gamer’s choice” with a base frequency of 4.5 GHz / acceleration up to 5.4 GHz.

The total 40MB cache is divided into 8MB L2 + 32MB L3, but with a lower TDP of 105W (142W PPT).

(4) The fourth R5-7600X chip is a 6C / 12T design that is expected to be a hit with mainstream gamers with a base frequency of 4.7 GHz / single core acceleration up to 5.3 GHz.

The total 38MB cache is divided into 6MB L2 + 32MB L3, and the TDP is greatly relaxed from 65W in the R5-5600X to 105W (142W PPT), which is probably the necessary price to pay for hitting higher frequencies.

As for the legendary successor to the R7-5800X, AMD has not revealed anything so far. One speculation is that the company intends to use a SKU equipped with 3D V-Cache cache to take its place (similar to the R7-5800X3D).

As for the truth, please wait for the official announcement later this year. If all goes well, both the Zen 4 Raider 7000 desktop CPU and the accompanying 600-series AM5 motherboard are expected to arrive in the fourth quarter of 2022.

The post AMD Zen 4 sharp dragon 7000 series debut four CPU SKU specifications burst summary appeared first on TechGoing.