AMD Radeon Graphics 448SP Mobile

AMD Radeon Graphics 448SP Mobile is an integrated GPU entry in the database. Its chip is Cezanne-M, and its architecture is GCN 5.1. TSMC fabricated the die on a 7 nm process. The die contains 9,800 million transistors and measures 180 mm², giving a transistor density of 54.4M / mm². The generation field is "Vega II IGP (Cezanne Mobile)". The release date is 2021-04-12, and the production status is end-of-life. The predecessor is listed as Vega IGP, and the successor is Navi II IGP. The series and codename fields are null, so this entry carries no additional naming labels.

Memory Subsystem, VRAM size/type, bus width, bandwidth and what it means for high resolutions

Memory size is System Shared, memory type is System Shared, and bus width is System Shared. The memory clock entry is also System Shared. No dedicated VRAM capacity appears in the data, and no fixed memory interface width appears. Bandwidth is listed as System Dependent, which is a dependency statement rather than a fixed value. The GPU therefore has no bandwidth number of its own; effective memory throughput is tied to the host platform’s shared memory arrangement. High-resolution rendering increases memory traffic, but the data provides no bandwidth figure to quantify that increase. A fixed bus width would make memory throughput easier to bound; this entry has none. A fixed memory clock would provide a timing baseline; this entry has none. The only memory-related facts are the System Shared labels and the System Dependent bandwidth label. Consequently, the memory behavior of this GPU cannot be evaluated in isolation from the portable device that hosts it.

Ray Tracing and Feature Set

The rtCores field is null, and the tensorCores field is null. The data lists no dedicated ray tracing core count and no dedicated tensor core count. API support includes DirectX 12 (12_1), OpenGL 4.6, and Vulkan 1.3. The architecture is GCN 5.1. The recorded processing resources are 448 shading units, 28 texture mapping units, and 8 ROPs. Display outputs are Portable Device Dependent. The bus interface is IGP, and the slot width is IGP, so the feature set is delivered through an integrated implementation rather than a discrete expansion card. The game clock field is null, leaving base and boost clocks as the only populated clock fields. Without RT core or tensor core counts, any ray tracing or machine-learning acceleration would have to rely on the general-purpose shading units, though the data does not state a performance level for such workloads.

Benchmark Performance

The benchmarks array is empty, and the average benchmark score field is 0. The nearestRivals array is empty, so there are no rival names, scores, or deltaPct values available. The percentileVsAllGpus field is 50, placing the part at the midpoint of the database’s GPU distribution. That percentile is the only relative comparison figure supplied; it is not backed by any listed benchmark score.

The compute and fill data are as follows. FP32 throughput is 1.613 TFLOPS. FP16 throughput is 3.226 TFLOPS, and the data marks this as a 2:1 ratio. Pixel fill rate is 14.40 GPixel/s, and texture fill rate is 50.40 GTexel/s. Base clock is 300 MHz, boost clock is 1800 MHz, and the game clock field is null. The device has 448 shading units, 28 texture mapping units, and 8 ROPs. The FP16 figure at double the FP32 figure is consistent with the stated 2:1 ratio. These rates describe raw throughput; they are not game results. With no benchmark entries, the raw rates and the percentile are the only quantitative performance indicators in the record.

How It Compares

The nearestRivals list contains no entries. There are no nearest-rival names, no rival scores, and no deltaPct values. Therefore, no direct percentage lead or deficit can be reported for this GPU against any specific competitor. The only comparative number is the 50th percentile against all GPUs. A 50th percentile indicates a mid-database placement, but with an empty benchmark list and an average benchmark score of 0, the percentile is not supported by a measured score in the data. The absence of nearest rivals means the comparison data cannot refine that placement beyond the midpoint. The data supports a median position, but it does not support a percentage claim relative to a named rival.

Power and Cooling

TDP is 45 W. The slot width is IGP, and the bus interface is IGP, so the part does not use an expansion slot. Power connectors are listed as None. The suggested PSU field is null, so no power supply recommendation is recorded. Because no external power connectors are present, the host platform handles power delivery. The dimensions fields are null; length, height, and width are not listed. A cooler or thermal solution is not specified in the data. The 45 W TDP is the only thermal/power figure present. The production status is end-of-life, but the TDP remains 45 W. Display outputs are Portable Device Dependent, matching the integrated device configuration.

Who Should Consider It

This is an integrated GPU for portable devices. The slot width is IGP, the bus interface is IGP, and display outputs are Portable Device Dependent. The memory subsystem is System Shared, so bandwidth is System Dependent. With FP32 at 1.613 TFLOPS, pixel fill at 14.40 GPixel/s, and texture fill at 50.40 GTexel/s, the part is positioned for moderate rendering workloads. The 28 texture mapping units handle texture work, while the 8 ROPs limit final pixel output. The 448 shading units are the main parallel compute resource. FP16 throughput of 3.226 TFLOPS at 2:1 provides a higher half-precision path. Dedicated ray tracing core and tensor core counts are absent. The 50th percentile in the database supports a middle-of-the-pack classification. High-resolution rendering would rely on the shared memory path, but the data contains no fixed bandwidth number to evaluate that scenario. Users with workloads that fit within the 1.613 TFLOPS FP32 and 50.40 GTexel/s figures are the most plausible users for this end-of-life mobile IGP.