Monday, November 27th 2023
AMD Phoenix AM5 APUs to Get Ryzen 8000 Series Branding, Company Readies 5000GT Series for AM4
AMD is giving final touches to its first APUs for the Socket AM5 desktop platform. A report by Sakhtafzar Magazine suggests that the company could give processor models in the series Ryzen 8000G numbering, instead of the previously thought 7000G series. The company is preparing as many as 14 processor models spanning the 4 nm "Phoenix" and "Phoenix 2" monolithic dies. Both chips combine "Zen 4" CPU cores with an iGPU based on the RDNA 3 graphics architecture. While the current Ryzen 7000 series "Raphael" desktop processors feature integrated graphics, AMD doesn't consider them APUs, as their iGPU are just about enough for non-gaming desktop use cases. APUs are designed for entry-level gaming.
The "Phoenix" silicon has up to 8 "Zen 4" CPU cores, and an iGPU with up to 12 RDNA3 compute units. This chip is powering the Ryzen 5 8600G, Ryzen 7 8700G, their PRO variants, and their respective "GE" (energy efficient) sub-variants. The "Phoenix 2" silicon barely qualifies as an APU, as its iGPU only has 4 RDNA3 compute units (compared to the 2 RDNA2 CUs on the "Raphael" iGPU. It also has a maximum CPU core count of 6, from which two are "Zen 4" cores that can sustain higher boost frequency bins, and four are "Zen 4c" cores which run at lower clock speeds (albeit with an identical IPC and ISA). AMD is using "Phoenix 2" on the desktop platform to carve out several sub-$150 class processor models across the Ryzen 5 and Ryzen 3 brands; a package with a monolithic "Phoenix 2" die probably has a lower bill of materials (BOM) than a "Raphael" multi-chip module.
The source also claims to have gaming performance comparisons of the Ryzen 8000G "Phoenix" desktop APU's iGPU, compared to the Ryzen 7 5700G "Cezanne" Socket AM4 desktop APU, where it posts performance gains between 40% to 200% higher. This is because "Cezanne" packs a much older iGPU based on the Vega graphics architecture, while "Phoenix" uses the 3 generations ahead RDNA3.
Sticking with the AM4 platform, and AMD is planning to release several new processor models for the older platform, including the Ryzen 7 5700X3D and 5500X3D that feature the 3D Vertical Cache technology that benefits gaming performance; and a handful new APUs, namely the 5700GT and 5600GT. At this point, it's not known what the "T" brand extension signifies in AMD nomenclature. Intel uses "T" to denote energy-efficient SKUs, but AMD uses "E" for that job.
AMD is expected to announce the new Socket AM5 and AM4 processors on January 31, the article says.
Sources:
Sakhtafzar Magazine, Wccftech, VideoCardz
The "Phoenix" silicon has up to 8 "Zen 4" CPU cores, and an iGPU with up to 12 RDNA3 compute units. This chip is powering the Ryzen 5 8600G, Ryzen 7 8700G, their PRO variants, and their respective "GE" (energy efficient) sub-variants. The "Phoenix 2" silicon barely qualifies as an APU, as its iGPU only has 4 RDNA3 compute units (compared to the 2 RDNA2 CUs on the "Raphael" iGPU. It also has a maximum CPU core count of 6, from which two are "Zen 4" cores that can sustain higher boost frequency bins, and four are "Zen 4c" cores which run at lower clock speeds (albeit with an identical IPC and ISA). AMD is using "Phoenix 2" on the desktop platform to carve out several sub-$150 class processor models across the Ryzen 5 and Ryzen 3 brands; a package with a monolithic "Phoenix 2" die probably has a lower bill of materials (BOM) than a "Raphael" multi-chip module.
Sticking with the AM4 platform, and AMD is planning to release several new processor models for the older platform, including the Ryzen 7 5700X3D and 5500X3D that feature the 3D Vertical Cache technology that benefits gaming performance; and a handful new APUs, namely the 5700GT and 5600GT. At this point, it's not known what the "T" brand extension signifies in AMD nomenclature. Intel uses "T" to denote energy-efficient SKUs, but AMD uses "E" for that job.
AMD is expected to announce the new Socket AM5 and AM4 processors on January 31, the article says.
99 Comments on AMD Phoenix AM5 APUs to Get Ryzen 8000 Series Branding, Company Readies 5000GT Series for AM4
"10th gen" can mean Amber Lake Y, Comet Lake, or Ice Lake. lol
A lot of 13th gen desktop chips are derived from Alder Lake, however all that differentiates Alder Lake and Raptor Lake is cache and count of e-cores, which are also features Intel uses to segment their products, so does it make any difference at all if a lower-end 13th CPU is actually an Alder Lake die?
If you want another example, 11th gen included Rocket Lake i7 and i5 desktop chips (Cypress Cove 14nm), Comet Lake i3 desktop chips (Skylake 14nm), and Tiger Lake Mobile chips (Willow Cove 10sf), which makes for the most confusing Intel generation I can remember.
But even Intel's 11th gen was easier than what AMD is doing now on mobile. That generation every mobile chip was Tiger Lake, and every i3 on desktop was Comet Lake, and every i5 and i7 on desktop was Rocket Lake.
AMD 7xxx series mobile chips, on the other hand, include 4 different generations in the 15-28W range and two in the 35-45W range.
(Source: www.anandtech.com/show/18718/amd-2023-ryzen-mobile-7000-cpus-unveiled-zen-4-phoenix-takes-point)
Numbers mean something if you know the numbers...
Searching generically "10th Gen" will give you at least three architectures... What are we talking about here?
Get that shoe out of your mouth.All Raptop Lake Cores from 13600 and below are Alder Lake dies. Learn something, I beg you...
videocardz.com/newz/intel-confirms-core-i5-13600-cpu-and-below-are-based-on-alder-lake-dieIt's one year old discussion. The world has moved on.
I saw a couple asus tuf with 7735hs and I bought a acer 17 nitro with 7840hs but those are about all I was interested in seeing the Intel lappy's were way more than those two and battery life was crap on intel as well
The acer is said to get nearly 8hrs using eco mode.
So discarding ? nope it was just right at 1k.us though.
Gottcha read up a little.
I think any of these solutions would work and be cheaper than HBM. I suspect the reason they're not used isn't about manufacturing cost. Maybe OEMs want the freedom to pair different CPUs and dGPUs together. Maybe Intel and AMD believe the R&D costs are best spent elsewhere. Maybe Nvidia would file an anti-trust suite if they tried it.
Plus, this is about bandwidth, I don't see how AI could come up with any additional.
At least what I didn't take into account is that Iran is freer to trade with China than the USA which China has a decently sized tech & software base.