Multiple Intel Panther Lake Benchmark Scores Leaked: Surprises and the Expected
Multiple unreleased Intel Panther Lake Core Ultra Series 3 processors have surfaced in the PassMark database, including the Ultra 7 366H, Ultra X7 358H, Ultra 7 365, and Ultra 5 332.
While raw numbers are useful, their real value is in revealing Intel’s core layout philosophy, cache configuration, and die-binning strategy for this new generation of mobile chips.
⚙️ Core Configuration and Binning Strategy #
High-End Models (Ultra 7 366H & Ultra X7 358H) #
Both the 366H and 358H share a 16-core hybrid layout:
- 4 Cougar Cove P-Cores
- 8 Darkmont E-Cores
- 4 Skymont LP-E Cores
They feature 18MB L3 and 12MB L2 cache.
The key differentiator is how Intel bins the dies:
| Model | P-Core Max Boost | iGPU Units | Primary Binning Focus |
|---|---|---|---|
| Ultra X7 358H | Up to 4.8 GHz | Full 12 Xe3 | GPU quality / full iGPU enablement |
| Ultra 7 366H | Up to 5.0 GHz | 4 Xe3 | High-frequency CPU performance |
This split reflects wafer realities: dies that hit high CPU frequencies rarely sustain a full 12-unit GPU at target power. Intel segments SKUs to recover maximum usable yield.
Mid-Range and Entry-Level Models #
| Model | Core Count | P | E | LP-E | L3 Cache | L2 Cache | Segment |
|---|---|---|---|---|---|---|---|
| Ultra 7 365 | 8 | 4 | 0 | 4 | 12 MB | 12 MB | Mainstream thin-and-light |
| Ultra 5 332 | 6 | 2 | 0 | 2 | N/A | 6 MB | Entry-level / budget |
- The Ultra 7 365 removes E-cores to optimize cost and power for mainstream laptops.
- The Ultra 5 332 aggressively scales down P-cores, LP-E cores, and cache to maximize wafer utilization.
📊 Performance Analysis (PassMark ST & MT) #
Ultra 7 366H — Unexpected IPC Jump #
-
ST: 4,217 — nearly matches the Ultra 9 285H despite ~400 MHz lower boost.
→ Indicates front-end changes, improved scheduling, better branch prediction, and stronger execution-unit utilization. -
MT: 34,386 — on par with 285H.
→ The 4P + 8E configuration delivers full throughput under sufficient power and cooling.
Ultra X7 358H — E-Core Scaling Gains #
- MT: 32,288 — higher than the Ultra 7 255H (30,889), despite lower P-core frequency.
→ Demonstrates better E-core scaling and improved cross-cluster scheduling in Panther Lake.
Mid-Range & Entry-Level Metrics #
-
Ultra 7 365 (22,160) — exceeds Ryzen AI Z2 Extreme and Intel Ultra 5 226V.
→ Strong IPC and efficient LP-E involvement offset lower physical core count. -
Ultra 5 332 — firmly positioned for the low-cost tier.
Related Leak: OneXPlayer X1 i (Ultra 5 338H) #
A separate handheld leak reported:
- ST: 2,428 — slightly behind Ryzen AI 9 HX 370
- MT: 13,265 — ahead of AMD Strix Point
LP-E participation improves MT output in aggressive power limits typical of handheld gaming devices.
💡 Intel’s Mobile Strategy Comes Into Focus #
Intel’s segmentation across the Ultra 3/5/7/X7/X9 families reflects a clear strategy:
-
Standardized P-Core base
Every SKU uses 4 Cougar Cove P-cores. -
Price scaling via E-Core and LP-E counts
More clusters → higher performance tiers. -
Secondary segmentation via iGPU units
High-end models retain 12 Xe3, while CPU-centric SKUs disable GPU blocks to maximize yield.
While full performance will depend on final firmware, power policies, and OEM thermal design, the leaked data points to clear architectural themes:
- Greater dependence on E-core throughput
- Expanded LP-E scheduling participation
- Notable IPC improvements to counter lower mobile frequencies
Panther Lake marks a strategic repositioning of Intel’s mobile lineup, using architectural efficiency and binning depth to broaden coverage across performance and price segments.