Wednesday, September 2nd 2026

Possible AMD RDNA 5 GPU IP Ships Inside a Samsung SoC on 2 nm Node

Today marks the day we see AMD's next-generation RDNA 5 GPU IP inside working silicon, but not in the form you might expect. Thanks to SemiAnalysis's die annotation of the Exynos 2700 SoC from Samsung, we see AMD's new GPU IP shine in a mobile form with the Xclipse 970 GPU. As earlier leaks reported, AMD and Samsung co-developed the Xclipse 970 GPU using AMD RDNA technology and adapted it for mobile. However, the GPU IP has now been upgraded to the latest RDNA 5, making this SoC technically the first RDNA 5-equipped chip that will be shipping. Built on Samsung Foundry's SF2 node, Samsung has designed this SoC with a unique 10-CPU core configuration and a strong GPU to compete with Qualcomm Snapdragon Elite SoCs.

On the CPU front, Samsung uses Arm's off-the-shelf CPU core IP, which includes C2-Ultra and C2-Pro. This setup features a single C2-Ultra big core running at 4.24 GHz, one C2-Ultra running at 3.36 GHz, four C2-Pro cores running at 3.74 GHz, and four C2-Pro cores running at 2.88 GHz, making a total of 10 CPU cores. On the GPU side, the Xclipse 970 packs eight GPU WGPs, each containing two Compute Units. This results in 16 CUs on the GPU front if everything remains the same with RDNA 5 and no changes to the WGP/CU ratio have been made by AMD since the last-generation RDNA 4.
For Samsung, this is a significant launch as the company tests the limits of its SF2 2 nm node with a highly heterogeneous design. It's also a big launch for AMD since the GPU is possibly RDNA 5, about which we haven't had much information for a while. Externally, Samsung connects the Exynos 2700 with a 24-bit LPDDR6 connection, ensuring satisfactory bandwidth for both CPU and GPU core clusters. We are awaiting the first benchmarks of this mobile SoC, which could give us a preview of what the desktop RDNA 5 will be capable of.
Source: SemiAnalysis
Add your own comment

10 Comments on Possible AMD RDNA 5 GPU IP Ships Inside a Samsung SoC on 2 nm Node

#1
john_
This setup features a single C2-Ultra big core running at 4.24 GHz, one C2-Ultra running at 3.36 GHz, four C2-Pro cores running at 3.74 GHz, and four C2-Pro cores running at 2.88 GHz, making a total of 10 CPU cores.
Soon this setup will be describing desktop processors.

Anyway, it will be interesting to see performance, clocks and power consumption of that GPU.
AleksandarKBuilt on Samsung Foundry's SF2 node
I wonder if AMD could choose this node for the RDNA5 GPUs. I mean if they already have one - whatever GPU - made in this node, isn't this also a nice test vehicle for RDNA5 manufacturing?
Posted on Reply
#2
jesdals
Well this may point towards next Gen GPU from AMD being UDNA
Posted on Reply
#3
JohH
Oh, that's silly. AMD likely won't ship any RDNA5 products for about 12 months. Despite, apparently, the design being finished.
Posted on Reply
#4
ARF
john_I wonder if AMD could choose this node for the RDNA5 GPUs. I mean if they already have one - whatever GPU - made in this node, isn't this also a nice test vehicle for RDNA5 manufacturing?
JohHOh, that's silly. AMD likely won't ship any RDNA5 products for about 12 months. Despite, apparently, the design being finished.
Probably they will not, because they have nothing faster than the RX 9070 XT | RX 7900 XTX 24 GB 2022, so releasing very overclocked and overvolted GPUs from the factory at super steep prices will net them zero sales.
Posted on Reply
#5
igormp
AleksandarKSamsung connects the Exynos 2700 with a 24-bit LPDDR6 connection
Minor nit: from the screenshot it looks like 4x24-bit (96-bit total), rather than just a mere 24-bit bus.
Posted on Reply
#6
Denver
96Bit*

A single DDR6 channel will be 96-bit. It's going to be quite the upgrade.
Posted on Reply
#7
dogwitch
rumor with lots of salt..
even title is lets get around being sued legal speak.
Posted on Reply
#8
AusWolf
We are awaiting the first benchmarks of this mobile SoC, which could give us a preview of what the desktop RDNA 5 will be capable of.
Not necessarily. Let's not forget that scaling isn't linear between different power targets and applications. It could give us a clue about features, but nothing much more.
Posted on Reply
#9
kondamin
Memory controller way behind the npu and gpu away from the cpu cores.
That’s interesting
Posted on Reply
#10
T1beriu
Wccftech speculated that 2700 might use RDNA 5, but didn't provide any proof. They just used one small sentence. TPU takes that a uses RDNA 5 12 times in this article. Gotta get them clicks at any cost.
Posted on Reply
Nov 12th, 2026 04:39 CST change timezone

New Forum Posts

Popular Reviews

TPU on YouTube

Controversial News Posts