What the listing actually shows
Geekbench's public record for the run is more specific than the headlines. It lists a Lenovo 22ETPF6DTA with an "AMD Eng Sample: 100-000001713-32_Y" processor, tested with Geekbench 7.1.0 for Windows AVX2. The system details include:
- A 10-core, 20-thread topology, with one cluster of 4 cores and one of 6.
- A 2.00 GHz base frequency and a maximum frequency of 1,996 MHz.
- The codename "Medusa Point", 1 MB of L2 per core, and 16 MB of L3.
A second Geekbench record backs up the Lenovo link. It is a Geekbench AI 1.7.0 run uploaded on November 5, 2026. It names the same processor identifier on a Lenovo 22ESFVT014 with Windows 11 Pro, a Balanced power plan and 16 GB of RAM. That run used the CPU backend through OpenVINO, so it says nothing about NPU or GPU AI performance. It does suggest that Lenovo-associated systems with this chip are running more than one benchmark.
How the comparison works
VideoCardz reports that the single-core score is 1.6% higher than the Ryzen 9 9950X3D, which averages 3,051 points in Geekbench 7. The sample also beats the Ryzen 9 9950X at 3,069 and the Ryzen 9 9900X at 3,023. It still trails the Ryzen 7 9850X3D (3,131) and the Ryzen 9 9950X3D2 (3,164).
In multi-core, the 18,313 score is 5.3% above the Ryzen AI MAX 390 average of 17,393. It is 1.2% below the Ryzen AI MAX PRO 390 average of 18,541. VideoCardz also says it scores above the Core i9-14900 and Core i7-13700KF in the Geekbench 7 database.
Treat these as rough context, not a head-to-head test. Igor's Lab made a similar point about an earlier Medusa Point result: Geekbench averages come from different devices with varying memory, power profiles and cooling. "Beats the 9950X3D" is true only for one single-core run against a database average. It says nothing about sustained or real-world application speed.
The 2.0 GHz clock
The clock is the striking part. A logged maximum of 1,996 MHz is far below what any shipping desktop chip runs at. Early engineering samples often sit at low clocks, and TechPowerUp said the same of the first Medusa Point listing. It noted that sample spent most of its run around 1.39 GHz and barely passed 2 GHz.
There is a plausible reading: Zen 6 may deliver a lot of work per clock. But a single database run can't prove that. We don't know the power limits, thermals, firmware maturity or repeatability, and the logged clock may not reflect what the chip really ran at. This listing does not isolate an IPC figure.
The cache puzzle
Geekbench shows 16 MB of L3 here. VideoCardz notes an earlier 100-000001713-33_N sample showed 32 MB. The explanation is unresolved. Possible causes:
- AMD may be testing different configurations.
- Geekbench may be misreading early firmware. Guru3D noted that the 32 MB figure on the earlier sample is widely believed to be incorrectly reported by the benchmark.
The evidence doesn't favor either explanation, and neither should be treated as fact.
Is "_Y" really an OEM marker?
Tech Times says the "_Y" suffix marks an OEM-destined build, while "_N" marks AMD's own samples. I couldn't find primary confirmation of that convention, so treat it as the outlet's interpretation. What is documented is narrower. The earlier -33_N sample ran on AMD's Plum-MDS1 FP10 development platform, and this one is listed on a Lenovo model. That suggests AMD silicon is reaching partner hardware. It doesn't tell us how far Lenovo's validation has gone.
Don't compare Geekbench 6 and 7
Earlier Medusa Point leaks used Geekbench 6. One sample hit 3,329 single-core and 16,555 multi-core. Another reached 3,174 and 15,092 at around 2 GHz. Geekbench 7 uses an updated scoring system and workloads, so those numbers can't be lined up against 3,100 and 18,313. Reading the new score as a "decline" from 3,329 would be wrong, and so would reading it as a gain.
What this means for Windows PC buyers
- Nothing is confirmed. AMD hasn't confirmed final specs, clocks or a launch date for this part. VideoCardz says the same.
- Timing is rumor. The CES 2027 announcement and first-half 2027 laptop availability come from Tech Times, not AMD. Other claims in that piece, including the FP10 socket size, the manufacturing process and the IPC figures, weren't independently verified here. I'm not repeating them as fact.
- Don't delay a refresh on this alone. One engineering-sample run tells us nothing about battery life, thermals, drivers or software compatibility. IT teams planning refresh cycles should keep to their existing schedules and watch for retail benchmarks and OEM specs.
- Expect more leaks. Different OPN suffixes are showing up in Geekbench entries. Repeat runs on Lenovo hardware at higher clocks would be more informative than this one.
Bottom line
The verified news is modest but real. A Geekbench 7 entry ties a 10-core Medusa Point engineering sample to a Lenovo system, and it posts a very high single-core score at a very low logged clock. The stronger claims about validation stage, architecture and timing remain unconfirmed until AMD or its partners say more.