ATI Wii GPU

The ATI Wii GPU is an AMD console GPU built for Nintendo around the Hollywood chip and the Ultra-Threaded SE architecture. It was fabricated at NEC on a 90 nm process, integrating 107 million transistors on a 95 mm² die for a transistor density of 1.1M / mm². The release date is 2006-11-18, the launch MSRP was 249 USD, and the production status is end-of-life. The record contains no benchmark entries and no nearest rivals, so this analysis relies entirely on the listed specification fields.

Memory Subsystem

The memory subsystem is built around 64 MB of GDDR3 VRAM, a 64-bit memory bus, and a total bandwidth of 3.888 GB/s. The memory clock is 243 MHz, with an effective data rate of 486 Mbps. For high-resolution rendering, these figures are restrictive on two separate axes. First, 64 MB is a hard capacity ceiling; a high-resolution frame buffer must store a large amount of pixel data, and that data directly competes with other rendering data for the same limited space. Second, the 64-bit bus width and 3.888 GB/s bandwidth set a firm upper limit on how quickly that data can move in and out of memory. Even if the GPU logic were fast enough to issue more memory requests, the narrow bus and the 3.888 GB/s aggregate bandwidth would constrain the amount of texture reads, pixel writes, and intermediate data traffic per frame. The 243 MHz memory clock and the 486 Mbps effective data rate reinforce the same conclusion: this is a low-bandwidth memory design. In a fixed console context, those limitations can be worked around by controlling the rendering workload; in a high-resolution environment, the memory subsystem would become the primary bottleneck before the compute or fill-rate units are saturated.

Ray Tracing and Feature Set

The feature set is defined largely by what is not present in the specification. No ray tracing cores are listed, and no tensor cores are listed. The DirectX, OpenGL, and Vulkan API fields are all null, so the database does not record any API level support for this GPU. The architecture name is Ultra-Threaded SE, which points toward thread-level parallelism rather than dedicated ray tracing or tensor hardware. The fixed-function pipeline contains 4 texture mapping units and 4 render output units. Those units translate into a texture rate of 972.0 MTexel/s and a pixel rate of 972.0 MPixel/s. The display output field is “No outputs,” which means this is not a GPU with conventional monitor connections. The absence of RT and tensor cores, combined with null API fields, means that any hardware-accelerated ray tracing, machine learning, or modern graphics API functionality cannot be confirmed from the data. The feature set is therefore minimal and tightly bound to the console role indicated by the generation field, “Console GPU (Nintendo).”

How It Compares

The nearestRivals array in the FACT PACK is empty. There are no direct rival names, no rival scores, and no deltaPct values to draw upon. As a result, no rival-by-rival comparison can be written from the available data. The only positional field is percentileVsAllGpus, which is 50. That places this GPU at the median of the all-GPU distribution in the database. However, the average benchmark score is 0, and the benchmarks array is empty, so the 50th percentile is not supported by any measured performance entry in this record. The percentile should be read cautiously: it is the only rank available, but it is not backed by a non-zero benchmark score. The spec sheet positions the part as a low-power console GPU with a 45 W TDP, 64 MB of GDDR3 memory, and 4 TMUs / 4 ROPs, but those are specification values, not comparative performance figures. Without a populated nearestRivals list, the database cannot currently answer how this GPU stacks up against any specific alternative.

Who Should Consider It

Because the benchmark array is empty, any recommendation has to be grounded in the memory and fill-rate specifications. The 64 MB VRAM capacity and 3.888 GB/s bandwidth make this GPU a poor candidate for high-resolution rendering. A high-resolution workload would exhaust the frame buffer capacity quickly, and the memory bus would not supply data fast enough to sustain heavy texture or pixel throughput. The 972.0 MPixel/s pixel rate and 972.0 MTexel/s texture rate set a separate ceiling on how many pixels and texels can be processed per second. For high resolutions and dense detail, those rates are modest. The intended context is the fixed console hardware described by the “Console GPU (Nintendo)” generation field. Because the display output field is “No outputs,” this is not a part for a desktop PC with a separate monitor connection. Lower resolutions and reduced texture loads would lower the pressure on the 64 MB capacity and the 3.888 GB/s bandwidth, but the FACT PACK does not include benchmark scores to quantify how far settings can be raised. Users looking for a flexible desktop GPU with documented API support should not consider this part, since the API fields are null and no display outputs are listed.

Benchmark Performance

Benchmark performance cannot be analyzed in the usual way because the benchmarks array is empty. The average benchmark score is 0, and no deltaPct values are present. With no nearestRivals entries, there are no exact percentage differences to calculate or compare. The only quantitative throughput indicators in the specification are the pixel rate of 972.0 MPixel/s and the texture rate of 972.0 MTexel/s. These are theoretical fill rates for the 4 ROPs and 4 TMUs, not measured application scores. They indicate that the ROP and TMU configurations are balanced at the same numerical rate, but they do not translate into frame rates or workload performance. The percentileVsAllGpus value of 50 is the only comparative metric in the record, yet the accompanying average benchmark score of 0 leaves that percentile without a measured basis. In short, the data supports no statements about being a certain percentage faster or slower than any rival. The measured performance tier of this GPU is not currently represented in the database.

Power and Cooling

The thermal design power is 45 W, which is the only power figure in the specification. The fact pack lists no suggested PSU, no power connector requirements, and no slot width. Board dimensions are 157 mm in length, 197 mm in height, and 60 mm in width, equivalent to 6.2 by 7.8 by 2.4 inches. A 45 W TDP indicates a low heat load, so the cooling burden is modest, but the data does not define a cooler recommendation. The absence of a suggested PSU and power connector data means that system power integration cannot be described from this record. The bus interface is also null, and the display output is “No outputs,” so this is not a conventional expansion card with a standard desktop power connection. The physical dimensions provide an envelope for any cooling solution, but no slot thickness is recorded. Given the empty connector, PSU, and slot-width fields, the power and cooling requirements beyond the 45 W TDP remain unspecified in the dataset.