NVIDIA GeForce4 488 Go

The NVIDIA GeForce4 488 Go is an end-of-life mobile graphics processor in the GeForce4 Go generation, built around the NV18 chip on the Celsius architecture. It was fabricated by TSMC on a 150 nm process, with 29 million transistors on a 65 mm² die and a transistor density of 446.2K per mm². The GPU uses an AGP 8x bus interface, lists no auxiliary power connectors, and reports display outputs as Portable Device Dependent. Its release date is 2002-02-28. In the database ranking, the GeForce4 488 Go is placed at the 50th percentile of all GPUs, although the benchmark array is empty and the average benchmark score field is 0. The product line is bracketed by the GeForce2 Go as predecessor and the GeForce FX Go 5 as successor.

Memory Subsystem

The memory subsystem consists of 64 MB of DDR memory on a 128-bit bus, with bandwidth listed at 8.800 GB/s. The memory clock is 275 MHz, with an effective transfer rate of 550 Mbps. The 550 Mbps effective figure is exactly double the 275 MHz memory clock, reflecting the DDR transfer behavior. This is the only populated clock in the specification; the base and boost clock fields are not listed. With the memory clock and effective rate both present, the record provides a clear picture of the memory side, but no core clock figures are available to place the memory speed in a wider performance context.

The 64 MB capacity is the most important constraint for high-resolution work. A frame buffer of this size must hold color data, depth information, and any textures or render targets used by the scene. As resolution increases, the memory required for those buffers grows, and 64 MB leaves little headroom for large scene data. The 128-bit bus and 8.800 GB/s bandwidth are moderate figures, but they are not necessarily the first bottleneck; the capacity limit is more immediate. The GPU cannot compensate for a shortage of storage by relying purely on bandwidth, because the data has to have somewhere to reside in memory.

At higher resolutions, the 8.800 GB/s bandwidth defines how quickly pixel data can move through the frame buffer, but the 64 MB total memory means the active scene complexity must fit within a small footprint. The display outputs are Portable Device Dependent, so the actual resolution available to a given system is also tied to the portable device rather than solely to the GPU. Overall, the memory subsystem points toward lower-resolution rendering in portable systems, with the 64 MB capacity acting as a hard ceiling on frame-buffer and texture storage.

Ray Tracing and Feature Set

The feature set contains no RT cores and no tensor cores. Hardware-accelerated ray tracing is therefore not a feature that can be enabled on this part. No Vulkan version is listed; the API entries are DirectX 7.0 and OpenGL 1.5. These are the only exposed graphics APIs in the data, and they define the available software-level feature surface. Because neither an RT-core count nor a tensor-core count is present, there is no basis for treating this GPU as capable of hardware ray tracing or tensor-accelerated workloads.

The rendering pipeline is represented by 4 texture mapping units and 2 raster operation units. The texture rate is 1.200 GTexel/s, and the pixel rate is 600.0 MPixel/s. The record lists no shading-unit count, no fp32 throughput, and no fp16 throughput, so compute capability cannot be quantified from the available fields. The power connector field is None, indicating that the board does not use auxiliary power connectors. The bus interface is AGP 8x, and display outputs remain Portable Device Dependent.

The API support is limited to DirectX 7.0 and OpenGL 1.5, with no Vulkan entry. For ray tracing specifically, the absence of RT cores and the absence of any ray-tracing-capable API entry mean the GPU cannot be considered a ray-tracing part. The feature set is instead built around fixed rasterization and texture-sampling resources: 4 TMUs, 2 ROPs, a pixel rate of 600.0 MPixel/s, and a texture rate of 1.200 GTexel/s.

Benchmark Performance

The benchmarks array is empty, and the average benchmark score is recorded as 0. This 0 is a placeholder accompanying an empty benchmark list, not a measured result from a workload. No nearest rivals are listed, so there are no rival scores and no percentage-delta comparisons to report. The only quantitative placement is the percentileVsAllGpus value of 50, which places the GeForce4 488 Go at the midpoint of the database’s GPU ranking.

With no workload scores, the listed throughput figures become the primary performance indicators. The pixel rate is 600.0 MPixel/s, the texture rate is 1.200 GTexel/s, and memory bandwidth is 8.800 GB/s. These figures define a hardware envelope: the GPU can output up to 600.0 MPixel/s, sample up to 1.200 GTexel/s, and transfer up to 8.800 GB/s. However, without benchmark results, translating that envelope into actual game performance, settings scaling, or resolution scaling is not possible.

The 50th percentile is a rank position, not a performance score. It indicates that the GeForce4 488 Go sits at the median of the tracked GPU distribution, but the empty benchmark table means this placement is not accompanied by any workload-specific evidence. The absence of nearestRivals entries also prevents any comparison to specific competing parts. The benchmark section of this record is therefore incomplete: it offers a percentile position and fill-rate figures, but no measured scores.

Who Should Consider It

Given the available data, the GeForce4 488 Go is suited to systems that operate at lower resolutions and with modest texture workloads. The 64 MB DDR memory, 128-bit bus, and 8.800 GB/s bandwidth are the main memory-side constraints, and the 600.0 MPixel/s pixel rate with 1.200 GTexel/s texture rate places a firm ceiling on fill-heavy rendering. Applications that need large frame buffers or complex scene data will be limited by the 64 MB capacity. Lower-resolution scenes with simpler geometry and smaller texture sets are a better match.

Because no RT cores, no tensor cores, and no Vulkan entry are present, this GPU should not be selected for ray-traced rendering or for content that depends on a newer low-level API. The exposed API level is DirectX 7.0 and OpenGL 1.5, which further narrows the software environment. The part belongs to the GeForce4 Go generation and the Celsius architecture, and it sits between the GeForce2 Go and GeForce FX Go 5 in the product line. As an end-of-life product, it is more relevant to legacy portable systems than to current hardware.

No TDP is listed, so thermal and power requirements cannot be quantified. The only power-related fact is that the power connector field is None, which indicates that no auxiliary power connectors are present. Physical dimensions and slot width are also absent from the record, so integration details are limited. The functional profile, though, is clear: the GeForce4 488 Go is a low-capacity, fill-rate-limited mobile GPU for lower-resolution workloads, not for high-resolution, high-fidelity rendering.

How It Compares

The nearestRivals field is empty in the provided data. As a result, there are no rival names, no rival scores, and no percentage-delta comparisons available in this record. The only comparative position is the 50th percentile versus all GPUs, indicating a midpoint rank in the database. Without nearestRivals entries, a per-rival breakdown cannot be constructed.

The product line context is provided by predecessor and successor fields: the GeForce2 Go precedes the GeForce4 488 Go, and the GeForce FX Go 5 follows it. Neither of those products has benchmark scores in this record, so no numerical comparison to adjacent generations is possible. They serve only as product-line markers.

Because no nearest rivals are listed, there are no per-rival paragraphs to write. The empty nearestRivals array is the complete comparative picture beyond the 50th percentile position. Any future comparison would require added entries with rival names and scores, but none are present in this dataset.