Intel UHD Graphics 605 Mobile

The Intel UHD Graphics 605 Mobile is an IGP built by Intel on the Gemini Lake GT1.5 chip. Its architecture is Generation 9.5, the process node is 14 nm, and the foundry is listed as Intel. The database records the generation as HD Graphics-T (Goldmont Plus). The production status is end-of-life, with a release date of 2017-12-10. The record lists no series, no codename, no predecessor, and no successor. Transistor count, die size, transistor density, and physical dimensions are not supplied.

Benchmark Performance

The benchmarks array is empty. The average benchmark score is 0, and the percentile versus all GPUs is 50. Because the benchmarks array is empty, the average score and the percentile cannot be tied to measured results. The nearestRivals array is also empty, so there are no rival scores and no deltaPct values from which to compute exact percentage differences. In place of measured scores, the record supplies theoretical rates. The base clock is 200 MHz and the boost clock is 750 MHz. No game clock is listed. The memory clock is System Shared, and memory bandwidth is System Dependent. Shader throughput is defined by 144 shading units, 18 TMUs, and 3 ROPs. Pixel rate is 2.250 GPixel/s, and texture rate is 13.50 GTexel/s. FP32 throughput is 216.0 GFLOPS; FP16 throughput is 432.0 GFLOPS with a 2:1 ratio. These are the only performance-related quantities in the record.

The absence of benchmark entries means that no measured workload result, frame rate, or composite index is available. The average benchmark score of 0 is therefore not an observed performance value; it is a placeholder in an otherwise empty benchmark table. Similarly, the percentileVsAllGpus value of 50 has no supporting score behind it. Exact percentage deltas cannot be reported because no nearest rivals are provided. The theoretical rates are useful as upper bounds for hardware throughput, but they do not describe real application performance. The pixel rate of 2.250 GPixel/s and texture rate of 13.50 GTexel/s are the limits implied by the ROP and TMU counts. The FP32 rate of 216.0 GFLOPS and FP16 rate of 432.0 GFLOPS (2:1) define the compute envelope.

The memory side of the record is entirely host-dependent. Memory size is System Shared, memory type is System Shared, and memory bus width is System Shared. Bandwidth is System Dependent, so no fixed memory bandwidth figure is associated with the GPU itself. This means benchmark performance data, if it existed, would vary with the host system’s memory configuration. The clock section also lacks a game clock, leaving only the 200 MHz base and 750 MHz boost as the documented clock range. Without a game clock, the record does not indicate how the GPU behaves under sustained load.

How It Compares

The nearestRivals field is empty, so no direct per-rival paragraphs can be written. The database does not provide a comparison set for this mobile IGP. The only comparative statistic is the percentileVsAllGpus value of 50, which would normally indicate median placement among all GPUs. However, with an average benchmark score of 0 and no benchmark entries, that percentile lacks supporting measurements. No deltaPct values exist in the record, so the data cannot state whether this part is faster or slower than any other GPU.

The absence of a series, codename, predecessor, or successor means the part is not positioned in a product line by the data. No transistor count, die size, transistor density, or dimensions are available for structural comparison. The chip identity is Gemini Lake GT1.5, and the generation is HD Graphics-T (Goldmont Plus). The architecture is Generation 9.5, the process node is 14 nm, and the foundry is Intel. The bus interface is Ring Bus. The slot width is IGP. These fields define the part’s place in the database even though rival GPUs are not listed.

Because the nearestRivals array is empty, the benchmark database cannot rank this part against specific alternatives. The percentile field is the only comparative number present, but it is contradicted by the empty benchmark array. A reader should treat the percentile of 50 as an unverified entry rather than a measured median result. The structural identifiers — Gemini Lake GT1.5, HD Graphics-T (Goldmont Plus), Generation 9.5, Ring Bus, and IGP slot width — provide the context that rival comparisons normally would.

Power and Cooling

The TDP is 5 W. The slot width is IGP, meaning the part is not an expansion card format. The power connector field is null, so no connector requirements are recorded. The suggested PSU field is null, so no power supply recommendation is provided. Memory is System Shared, and memory bandwidth is System Dependent, so the host platform’s memory subsystem is part of the graphics memory interface. Display outputs are Portable Device Dependent. The database does not list a dedicated cooling solution, and dimensions are not supplied.

With a 5 W TDP and no power connectors, the power delivery design is minimal. The absence of a suggested PSU reinforces that this is not a discrete add-in card with its own power requirements. The memory being System Shared means the GPU does not carry dedicated video memory; the host system memory is the memory interface. Memory bandwidth being System Dependent means the system memory speed and channel configuration will influence graphics throughput. The record contains no cooler size, fan specification, or thermal solution, so the only thermal figure is the 5 W TDP.

Who Should Consider It

The record describes a low-TDP IGP for portable devices. The display outputs are Portable Device Dependent, so the target hardware is a portable system with the GPU’s display connections integrated into the device. Memory is System Shared, and memory bandwidth is System Dependent, so overall graphics performance will be tied to the host memory. The compute resources are 144 shading units, 18 TMUs, and 3 ROPs. Pixel rate is 2.250 GPixel/s, and texture rate is 13.50 GTexel/s. The FP32 rate is 216.0 GFLOPS, and FP16 is 432.0 GFLOPS (2:1). These figures indicate a part sized for light graphics rather than high-resolution, high-detail rendering.

No benchmark scores are present to support settings-specific recommendations. The data cannot confirm whether this GPU is adequate for a particular game, resolution, or graphics preset. What the data does show is a small shader configuration, a low pixel rate, a low texture rate, and a 5 W TDP. The 3 ROPs limit fill-rate-bound work, while the 18 TMUs limit texture-heavy workloads. The 216.0 GFLOPS FP32 rate is the general shader throughput, and the 432.0 GFLOPS FP16 rate with a 2:1 ratio offers double-rate half-precision compute.

The production status is end-of-life, and the release date is 2017-12-10, so the part appears in existing platforms rather than new high-end builds. It should be considered in the context of a 5 W IGP with no listed power connectors and no suggested PSU. The System Shared memory and System Dependent bandwidth mean that host memory configuration is a major factor. Because no benchmark scores exist, any resolution or settings recommendation would have to be derived from the theoretical rates alone. Those rates do not suggest high-resolution, high-detail rendering as a plausible use case.

Ray Tracing and Feature Set

The RT core count is null, and the tensor core count is null. The database therefore does not document dedicated ray tracing or tensor acceleration hardware. The API list is DirectX 12 (12_1), OpenGL 4.6, and Vulkan 1.3. These API versions form the software feature set for compatibility. Vulkan 1.3 is the newest API entry listed, and DirectX 12 is listed with feature level 12_1. OpenGL 4.6 is also present. API support alone does not confirm the presence of hardware ray tracing, especially when the RT core field is null.

The architecture is Generation 9.5, built on a 14 nm process at Intel. The chip is Gemini Lake GT1.5, and the generation is HD Graphics-T (Goldmont Plus). The bus interface is Ring Bus. Memory size, type, and bus width are all System Shared, with bandwidth recorded as System Dependent. Display outputs are Portable Device Dependent. The FP16 throughput is 432.0 GFLOPS with a 2:1 ratio, indicating half-precision compute is available at twice the FP32 rate. No tensor core count is listed, so tensor-style acceleration is not part of this record.

Because the RT core count is null, hardware ray tracing cannot be confirmed. Because the tensor core count is null, tensor acceleration cannot be confirmed. The feature set is defined instead by the API list and the 144 shading units, 18 TMUs, 3 ROPs, and the theoretical rates already noted. The 5 W TDP and IGP slot width place the part in a low-power integrated context, and the absence of game clock and benchmark entries leaves the feature list as the only detailed specification.