⏱️ Lectura: 10 min
Xiaomi’s new flagship processor, the Xring O3, scored 3,945 points on Geekbench’s single-core test and 15,221 on multi-core, according to screenshots shared on August 24, 2026 by the Ice Universe account on X. These figures put it on par with Apple’s cores in single-thread tasks and well ahead in multi-thread workloads.
📑 En este artículo
The figure caught the attention of researcher Daniel Lemire, who pointed out that the design reveals something more important than the final number: the mobile chip industry is betting on massively parallel cores, with more execution ports and more cache than many laptop processors.
TL;DR
- Xiaomi’s Xring O3 scored 3,945 points on Geekbench single-core and 15,221 on multi-core, according to screenshots from August 24, 2026.
- The chip packs 44 MB of total cache, more than most laptop processors.
- The large C1-Ultra cores support SME2 (Scalable Matrix Extension 2) for AI and SVE2 for SIMD.
- The design has 21 execution ports, six of which support 128-bit SIMD.
- Built on TSMC’s 3nm process, with a prime core that exceeds 4 GHz.
- Google’s Tensor G6 also uses C1-Ultra cores but only reaches about 2,700 points single-core.
- AMD Zen 5 still leads in raw SIMD width (4×512 bits) compared to ARM’s maximum of 6×128 bits.
- Daniel Lemire notes that the industry trend is toward massively parallel cores with more cache.
Introduction
Xiaomi has spent years competing with Apple and Samsung in the high-end phone market. Now the Chinese company is opening another front: in-house silicon design under the Xring brand, its bid to stop relying exclusively on Qualcomm and MediaTek for the chips in its flagship models.
The Xring O3 is the latest generation in that family. Unlike previous announcements focused solely on power efficiency, this time the conversation centers on a leap in raw performance that, if confirmed, would bring Xiaomi closer to Apple’s level in single-thread computing.
What Happened
On August 24, 2026, the Ice Universe account posted Geekbench screenshots on X attributed to the Xring O3, showing 3,945 points on single-core and 15,221 on multi-core. Daniel Lemire picked up the thread to analyze the architecture behind those numbers.
The technical community’s response was mixed. User Jon Clegg called the leak “part legit, part BS,” noting that the architecture (TSMC’s 3nm process, a prime core above 4 GHz, plenty of cache, and a very wide execution engine) is indeed consistent with a genuine performance leap. Another user, under the alias l1bertypr1me, pointed out a contradiction: Google’s Tensor G6 uses the same C1-Ultra core but only reaches about 2,700 points single-core, well below the Xring O3.
Context and History
Xiaomi is no stranger to designing its own chips. The company introduced its first in-house processor, the Surge S1, in 2017, but didn’t continue that line for nearly a decade. The return to in-house design under the Xring brand reflects a regional trend: Chinese manufacturers are looking to reduce their dependence on external silicon suppliers amid trade restrictions with the United States.
That context explains why every Xring announcement draws so much scrutiny. It’s not just about phone performance: it’s a signal of how far China’s semiconductor industry can go without relying on Western architecture licenses, even though the core design is still based on ARM.
Xiaomi is already one of the world’s largest phone makers by sales volume. Having its own high-end processor would let it negotiate better terms with Qualcomm for its other models, without depending entirely on a single external supplier for its flagship line.
Xring O3: Technical Details and Performance
The Xring O3 is built on TSMC’s 3-nanometer process. Its largest core, called C1-Ultra, is where nearly all the novelties are concentrated: it supports SME2 (Scalable Matrix Extension 2) to accelerate matrix and AI workloads, and SVE2 (Scalable Vector Extension 2) for SIMD-style data parallelism.
The most striking feature is the width of the execution engine: 21 execution ports, six of which support 128-bit SIMD operations. That’s wider than most Intel or AMD desktop processors. By comparison, AMD Zen 5 still holds the edge in raw SIMD throughput thanks to its 4×512-bit units, while 6×128 bits is the most any ARM chip offers today, according to Lemire himself.
The declared total cache is 44 MB, a figure higher than most current laptop processors. That combination of more ports, more SIMD, and more cache is, according to Lemire, “where all the transistors are going” in 2026 mobile core design.
💭 Key takeaway: the trend isn’t just higher clock speed, it’s more parallel execution units per cycle, which allows many independent additions or multiplications to happen at the same time.
Comparison Table: Execution Width
| Chip | Process | Total Cache | Execution Ports | Max SIMD |
|---|---|---|---|---|
| Xiaomi Xring O3 (C1-Ultra) | TSMC 3nm | 44 MB | 21 | 6×128 bits |
| Google Tensor G6 (C1-Ultra) | N/A | N/A | N/A | 6×128 bits (same family) |
| AMD Zen 5 | TSMC 4nm | Varies by SKU | N/A | 4×512 bits |
The table shows the asymmetry: Zen 5 moves fewer units, but each one processes a much wider block of data, while the Xring O3’s ARM design bets on multiplying the number of smaller units.
How to Test It
Until a commercial phone with the Xring O3 is available outside China, the way to verify figures like these on any Android chip is to run Geekbench 6 directly and read the CPU flags exposed by the kernel. With a device connected via ADB, you can inspect the hardware without installing anything extra:
adb shell cat /proc/cpuinfo | grep -E "Features|CPU part"That command lists the ARM extensions the kernel detects, such as sve2 or sme2 if the firmware exposes them, along with the core identifier. It’s the first step to confirm whether a phone actually has the architecture the manufacturer claims.
To run the actual benchmark, Geekbench 6 has a free version for Android:
adb install com.primatelabs.geekbench6.apk
adb shell am start -n com.primatelabs.geekbench6/.HomeActivityThe result uploads automatically to Geekbench Browser, where it leaves a public link with a subtest breakdown. That public link is what allows figures to be compared independently, rather than trusting a lone screenshot.
⚠️ Heads up: a Geekbench screenshot without a public link to the result on geekbench.com isn’t verifiable. The gap between the Tensor G6 (about 2,700 points) and the Xring O3 (3,945 points) using the same base core is reason enough to ask for the original link before accepting the figure as accurate.
Impact and Analysis
If the Xring O3 figures are confirmed by independent tests, Xiaomi would have a selling point that until now was almost exclusive territory for Apple: single-thread performance comparable to an iPhone’s. That matters because much of a phone’s daily experience, like opening apps or scrolling through lists, depends more on single-core performance than on having many cores working at once.
The other side of the analysis is the technical doubt raised by the community itself. It’s not impossible for the Tensor G6, built on the same C1-Ultra core, to perform noticeably worse in single-core than the Xring O3: cache configuration, silicon binning, and power firmware can all significantly change the final result. Even so, it calls for reviewing the data carefully before treating it as final.
SME2 support in the C1-Ultra core is also relevant for on-device AI workloads, such as summarizing text, generating automatic captions, or running small vision models without relying on a server. If sustained performance matches the peak reported on Geekbench, the Xring O3 could compete with Apple’s neural engine in those kinds of local tasks, not just in generic computing.
For the Latin American market, where Xiaomi is already one of the best-selling manufacturers, the eventual arrival of phones with the Xring O3 will depend on whether the company decides to take its in-house chip outside China, something that so far has been the exception rather than the rule.
What’s Next
Daniel Lemire noted that Apple could announce its next processor soon, which would narrow the Xring O3’s momentary single-core lead. He also pointed out that getting a phone with the new Xiaomi chip could be difficult outside the markets where the company launches it first.
What to watch next is Geekbench results with a verifiable public link, third-party benchmarks like AnTuTu or GFXBench for sustained load, and power consumption data under multi-thread load, which is where these kinds of wide designs tend to lose efficiency compared to simpler ones.
Core Architecture Diagram
flowchart TD
A["C1-Ultra Core"] --> B["21 execution ports"]
B --> C["6 SIMD 128-bit ports"]
B --> D["Integer and load/store ports"]
A --> E["SME2 support (matrices/AI)"]
A --> F["SVE2 support (scalable SIMD)"]
A --> G[("Total cache 44 MB")]The diagram sums up why the design is described as “wide”: it’s not one fast path, it’s many parallel paths sharing the same cache.
Try it yourself: run adb shell cat /proc/cpuinfo on your own phone and compare it against the figures in this article before accepting any Geekbench screenshot you see circulating on social media.
📖 Summary on Telegram: See summary
Frequently Asked Questions
What is the Xring O3?
It’s the latest flagship processor designed in-house by Xiaomi, built on TSMC’s 3nm process, intended for its high-end phones.
Is the 3,945-point Geekbench single-core score officially confirmed?
No. The figure comes from a screenshot shared on X by the Ice Universe account, without a verifiable public link to the result on Geekbench Browser at the time of this publication.
What are SME2 and SVE2?
SME2 (Scalable Matrix Extension 2) is an ARM extension for accelerating matrix operations, useful for AI workloads. SVE2 (Scalable Vector Extension 2) is its equivalent for SIMD-style parallel data processing.
Why does the Tensor G6 perform worse if it uses the same C1-Ultra core?
The technical community doesn’t have a definitive answer. Factors like cache configuration, power management firmware, and silicon binning can explain part of the difference, but the reported gap caught the attention of several developers.
When will phones with the Xring O3 be available outside China?
There’s no confirmed date. Xiaomi hasn’t announced international availability for models with this chip at the time of this publication.
References
- Daniel Lemire’s thread on X: original analysis of the Xring O3’s architecture and the Geekbench figures shared by Ice Universe.
- Geekbench: official site for the benchmark used to measure the chip’s single-core and multi-core performance.
- ARM: documentation on the base architecture, including the SVE2 and SME2 extensions mentioned in the chip.
- TSMC on Wikipedia: context on the 3nm manufacturing process used in the Xring O3.
📱 Enjoy this content? Follow @programacion on Telegram for daily tech content in Spanish: quick summaries, fresh content every day.
Imagen destacada: Foto de Jorge Salvador en Unsplash
0 Comments