Reddit user chain-77 discovered that a $95 Ryzen 5 4600G APU can do respectable AI work by telling Linux to see it as a 16GB GPU. Although the processor doesn't compare to dedicated cards in traditional graphics rendering, AI relies heavily on memory, where an APU's ability to allocate shared memory freely becomes an advantage.

>Interpretation one says that high-end GPUs are irrelevant to 99 percent of gamers and that AMD focussing on affordable graphics cards is actually a good thing. Give us 75 percent of the performance of a high-end Nvidia GPU for half the money, AMD, and everything looks pretty sweet. >

You’ve probably seen several YouTube videos where people claim they have heard from, “someone in the know” that AMD isn’t making high end versions of their next video cards.

https://youtu.be/va8nyj6y7d4

https://youtu.be/YZtEcPadqrs

This is so damned much bullshit, chicken little is laughing at the sky falling…

It’s HALF a story.

The FACT is AMD has just released a paper regarding a more efficient method of computing object locations in a scene when ray tracing.

Currently the method used is called the Monte Carlo method. It is computationally intensive and requires… HARDWARE.

AMD has come up with a method (called Neural Intersection Function) that is 150% more efficient. This means that EXISTING hardware will… do what?

AMD can build cards that use less hardware, use less power, and what do you suppose that means?

LESS COST FOR MORE PERFORMANCE! So when you see the nonsense that “AMD is exiting building GPU’s…” No, they are building 150% more efficient GPU’s, at lower cost while the cost of graphics cards is going back up.

An opinion on why AMD graphics cards get overlooked.

I’ve upgraded to a more powerful system with a Ryzen 9 7950x3D processor on a x670e Tomahawk WiFi motherboard, complemented by a 64GB kit of 4800MHz DDR5 RAM (the profile on it supports 6GHz), two 2TB M.2 NVMe SSDs, and an AMD 6750xt graphics card, powered by a Corsair 1kW supply. My choice of motherboard brand avoids potential security and heat/overvoltage issues with Gigabyte and Asus respectively. With additional slots for expansion, potential improvements might include considering higher-end graphics for better gaming or heavy-duty tasks, or utilizing the empty M.2 (pci-e 5) slot for additional storage if needed. What would you do differently?

The upcoming AMD components are projected to incorporate a blend of Zen 5 and Zen 5c cores, with the 'Strix Halo' APU, designed for the AM5 platform, expected to offer up to 16 cores. The 'Strix Halo' is positioned as a standalone desktop processor, intended to raise the APU segment's standards and challenge existing discrete graphics cards by 2024. Meanwhile, the 'Strix Point' Ryzen 8000 APUs are likely to succeed the 'Phoenix' Ryzen 7000 series, with rumored variants including a 12-core, 24-thread version. AMD also anticipates launching a 'Strix Halo' series for laptops, potentially featuring up to sixteen Zen 5 cores, reinforcing speculations about its interest in hybrid architectures. AMD is also reportedly developing Phoenix 2 APUs, combining Zen 4 cores with smaller, power-efficient Zen 4c cores, expected to operate at lower clock speeds. Notably, these new chips are projected to be based on the Zen 5 architecture, incorporating an integrated 'RDNA 3.5' GPU and a refined Ryzen AI engine, with production slated for an advanced node, likely TSMC 3nm, in 2024. Rumored TDP values for Ryzen 8000 'Strix Point' APUs range from 15 to 54 watts across different series, hinting at the possible efficiency of the Zen 5 and Zen 5c architecture. Despite uncertainties in precise specifications and release dates, these speculations provide insights into AMD's future directions and technological innovations.

An explanation of the difference between EPYC Zen 4 and Zen 4c

For over a decade, the expectation that computers would keep becoming faster under Moore's Law is being questioned as we near the physical limits of shrinking transistors. The consistent miniaturization of transistors has been pivotal in enhancing computer power, but as these transistors approach atomic sizes, their reliability as conductors of electrical current diminishes. However, this doesn't spell the end of technological advancement. While we continue the process of transistor miniaturization, the future of processor technology lies in innovation beyond transistors, with 3D fabrication presenting as a promising next step.

AMD's Ryzen 8000-series CPUs, based on the new Zen 5 microarchitecture, have reached a crucial point on their way to market, with major industry benchmarking utility AIDA64 now recognizing the architecture. The update came from FireWire's latest release notes for the AIDA64 tool, a popular system health and performance evaluation software among professionals and PC enthusiasts. FireWire would have needed detailed information from AMD to implement this preliminary support for Zen 5.

However, this does not necessarily give an exact timeline for when the Zen 5 architecture will hit the market in the form of AMD Ryzen 8000-series desktop processors. AIDA64's preliminary support for Intel's Meteor Lake processors was introduced back in 2021, and these chips are not expected to launch until later this year. According to AMD's roadmap, the release of the Ryzen 8000-series is set for 2024, indicating that the company is on track with its launch plan.

FireWire's release notes also show enhanced support for Intel's upcoming processor generations: Meteor Lake, Arrow Lake, and Lunar Lake. Lunar Lake, with its new microarchitecture, is expected to be released late in 2024 or early in 2025. The upcoming 14th-gen Raptor Lake refresh and Meteor Lake chips, both under the 14th-gen label and expected to launch later this year, will stick with Intel's current hybrid performance-efficiency core design.

Arrow Lake, expected in 2024 and set to be built on TSMC's 3nm process node, has reportedly been dealing with delays. This is not unusual for TSMC, which manufactures the processor wafers for numerous tech giants including AMD, Nvidia, and Apple. With the 3nm node predicted to be highly popular, production schedules are likely to be tight.

AMD's Zen 5 will also be utilizing TSMC's 3nm node, which could be a double-edged sword for the company. Intel secured a significant portion of 3nm production capacity in 2021, ostensibly for an Arc GPU and several server chips. If Intel substitutes Arrow Lake chips into this capacity, it is likely to have a larger stock of its chips at the launch of Arrow Lake compared to AMD's stock of 3nm Zen 5 chips. However, it appears that AMD placed orders for Zen 5 chips before Intel, which means Zen 5 chips may roll off the production line before Intel's Arrow Lake chips. Whether AMD's first-to-market advantage will result in a more successful launch, as seen with the previous Zen 4, remains to be seen.

Now for the big question. You’ve got an AMD CPU/APU and GPU. You want to overclock… Now what? What videos do you watch? What websites do you go to? What tools do you use?

Thankfully this isn’t a current product. However, I will point out…

Let’s just let Linus have the floor.

During a recent "Meet the Experts" webcast, AMD shared some key design aspects of its forthcoming Ryzen 8000 series processors for desktops, specifically targeting workstations and value servers. These AM5 socketed chips are set to blend the next-generation Zen 5 CPU architecture with an upgraded graphics architecture, labeled Navi 3.5. A roadmap for AM5, spanning from 2022, was shared during the webcast, highlighting the transition from the Ryzen 5000 series that combines Zen 3 and Vega graphics, to the current Ryzen 7000 series integrating Zen 4 CPU cores and Navi 3.0 graphics, and finally the AMD Ryzen 8000 series debuting in 2024, which will pack Zen 5 CPU cores and Navi 3.5 graphics.

The Ryzen 8000 series chips provide those willing to hold out for another generation of upgrades or new PC builds with a concrete target to anticipate. The presentation slide also confirmed that the AM5 will continue into 2026, and at this pace, we could expect two more Ryzen desktop generations before the socket is replaced by a new one, possibly named AM6.

Previous roadmaps for next-gen AMD Ryzen desktops hinted at the Ryzen 8000 series being codenamed 'Granite Ridge' (or 'Strix Point' for laptops), featuring up to 16 Zen 5 CPU cores across two CCDs. Also part of the SoC will be the Navi 3.5 GPU, speculated to be a die shrink of Navi 3 for enhanced clocks/efficiency but with largely the same features. An intriguing point from the slide suggests the current Ryzen 7000 series processors for AM5 use 'Navi 3.0' graphics, although chips released in 2023 utilize RNDA 2 (Navi 2.x) integrated GPUs. Whether a refresh is in the cards to align with the roadmap remains to be seen.

RDNA 4, following the RDNA 3 GPU architecture that launched in 2023, represents the first preview of Radeon RX 8000-series gaming GPUs, codenamed 'Navi 4x'. While there aren't any specific details on the RX 8000 graphics cards, it's expected that the flagship RX 8900 XT will likely utilize a dual-node architecture using 3nm and 5nm processes. Rumors suggest that the Navi 4x family will exclusively employ the RDNA 4 architecture, and that the upcoming Navi 3x (RX 7000 series) may use a combination of RDNA 3 for high-end GPUs and a refreshed RDNA 2 for more budget-friendly options.

AMD's new roadmap confirms the release of next-gen Ryzen 8000 CPUs featuring Zen 5 technology and an updated "Navi 3.5" graphics architecture next year. Speculations indicate that Navi 3.5 might enable integrated graphics with performance comparable to an Nvidia RTX 3070. Although AMD has previously mentioned Zen 5, this is the first official mention of Ryzen 8000 and the new Navi 3.5 graphics architecture.

The official roadmap, however, doesn't delve into any details regarding Navi 3.5. But according to a recent Twitter post from serial leaker Kepler_L2, Navi 3.5 is essentially the RDNA 3 architecture with a few elements from RDNA 4. In particular, it's speculated to have revised shader ALUs with support for new FP32 instructions and enhancements to the geometry engine, but it won't incorporate new RDNA 4 scheduler and improved RT cores. The actual impact on performance from these updates remains uncertain.