Intel Iris Xe Graphics G7 96EU Mobile

Intel's Iris Xe Graphics G7 96EU Mobile is an integrated GPU built on the Tiger Lake GT2 chip, using the Generation 12.1 architecture on Intel's 10 nm process. It packs 768 shading units, 48 texture mapping units, and 24 ROPs into a 146 mm² die, with a base clock of 300 MHz and a boost clock of 1100 MHz. The part delivers 1.690 TFLOPS of FP32 compute and 3.379 TFLOPS of FP16 compute via a 2:1 ratio. It is rated at a 15 W TDP and occupies the IGP slot width, making it a purely integrated solution. Production has ended, with a release date of September 1, 2020.

How It Compares

The Iris Xe G7 96EU sits at the 50th percentile among all GPUs in the benchmark database, placing it squarely in the middle of the performance distribution. This is a notable position for an integrated part, as it means roughly half of all discrete and integrated GPUs tracked are slower. The average benchmark score in the database is 0, which indicates that this particular SKU has no recorded benchmark submissions, so direct score-based comparisons are not available. Instead, the position must be read from the percentile field and the raw specification sheet.

Against the broader field, the 1.690 TFLOPS FP32 figure is the key compute metric. A mid-pack percentile combined with sub-2 TFLOPS compute suggests the G7 96EU is competitive with older entry-level discrete GPUs but is outpaced by modern discrete parts. The pixel rate of 26.40 GPixel/s and texture rate of 52.80 GTexel/s, derived from the 1100 MHz boost clock and the 24 ROPs and 48 TMUs, reinforce that this is a part designed for 1080p-class workloads rather than high-refresh or high-resolution rendering. The 50th percentile ranking indicates that in aggregate workloads, it lands between the low end and the mid-range of the GPU spectrum.

The nearestRivals field in the database is empty for this part, meaning no direct delta percentages against named competitors are recorded. Without rival scores, the comparison must rest on the percentile and the architectural facts. The 10 nm process and Generation 12.1 architecture place it in Intel's Tiger Lake generation, which was positioned as a meaningful uplift over prior Intel integrated graphics. The 768 shading units represent a high count for an integrated part, and the 48 TMUs give it a texture throughput that exceeds typical integrated solutions.

Because the part is end-of-life, its standing is now historical. In its release window, the 50th percentile ranking would have made it one of the more capable integrated GPUs available, especially given the 15 W TDP envelope. The data shows a GPU that competes with low-end discrete parts from the same era but does not threaten mid-range or high-end discrete offerings. The empty nearestRivals list means no head-to-head deltas are published, but the percentile alone tells the story: this is a median performer.

Ray Tracing and Feature Set

The Iris Xe G7 96EU has no dedicated ray tracing cores. The rtCores field is null, and the tensorCores field is also null. This means hardware-accelerated ray tracing is not available on this part. The DirectX 12 support is at feature level 12_1, which includes some DirectX 12 features but does not mandate DXR ray tracing support at the hardware level. For workloads that rely on ray tracing, the G7 96EU would have to fall back to compute-based approaches, which are significantly slower than dedicated RT hardware.

The feature set is otherwise modern for its generation. DirectX 12 (12_1) support covers the then-current Microsoft graphics API, and OpenGL 4.6 provides broad compatibility with legacy and cross-platform applications. Vulkan 1.4 support is listed, which is a relatively recent Vulkan revision and indicates the driver stack supports the latest Vulkan feature set. This makes the part viable for modern Vulkan-based games and applications, even if it lacks RT acceleration.

The FP16 throughput of 3.379 TFLOPS at a 2:1 ratio is double the FP32 rate, which indicates support for half-precision compute that can accelerate certain workloads, particularly in machine learning inference and some shader effects. However, without tensor cores, there is no dedicated hardware for matrix operations. The 2:1 FP16 ratio is a standard feature of the architecture rather than a specialized AI acceleration path. The display outputs are listed as "Portable Device Dependent," meaning the actual connectors depend on the laptop or device implementation, not the GPU itself.

Memory Subsystem

The memory subsystem is entirely system-shared. The VRAM size is "System Shared," the type is "System Shared," and the bus width is "System Shared." There is no dedicated VRAM on this part. The bandwidth is listed as "System Dependent," which means the achievable memory bandwidth is determined by the host system's RAM configuration, including channel count, memory speed, and whether the system uses dual-channel or single-channel memory. The bus interface is Ring Bus, which connects the GPU to the rest of the Tiger Lake SoC.

This has significant implications for high-resolution workloads. Because the GPU shares memory with the CPU, the effective bandwidth available to the GPU is a fraction of the system's total memory bandwidth, and it is subject to contention with CPU workloads. At high resolutions, where frame buffers grow and texture bandwidth demands increase, the system-dependent bandwidth becomes a bottleneck. The GPU's compute capabilities — 1.690 TFLOPS FP32 — are modest, but the memory subsystem is the more limiting factor in practice.

The "System Dependent" bandwidth figure means two systems with the same Iris Xe G7 96EU GPU can perform very differently depending on their memory configurations. A laptop with dual-channel, high-speed DDR4 will provide substantially more bandwidth to the GPU than a single-channel configuration. This is a critical caveat for any performance expectation. The memory clock is also listed as "System Shared," meaning there is no dedicated memory clock; the GPU operates at the system memory frequency.

For 1080p gaming, the shared memory subsystem can be adequate for lighter titles, but at 1440p or 4K, the bandwidth constraints and the lack of dedicated VRAM will cause significant performance degradation. The pixel rate of 26.40 GPixel/s, which is a fixed property of the GPU clock and ROP count, is independent of memory bandwidth, but the actual achieved fill rate in games is often limited by memory bandwidth rather than the ROP throughput. The texture rate of 52.80 GTexel/s similarly represents the theoretical peak, not the achievable rate under memory constraints.

FAQ

Q: What architecture does the Iris Xe G7 96EU use?

A: It uses the Generation 12.1 architecture, built on Intel's 10 nm process, with the Tiger Lake GT2 chip and a die size of 146 mm².

Q: Does this GPU support ray tracing?

A: No. The rtCores field is null, meaning there are no dedicated ray tracing cores. The DirectX 12 support is at feature level 12_1, which does not include hardware-accelerated ray tracing.

Q: How much dedicated VRAM does it have?

A: It has no dedicated VRAM. The memory size, type, and bus width are all listed as "System Shared," meaning it uses the host system's RAM.

Q: What is the power draw of this GPU?

A: The TDP is rated at 15 W, and the slot width is IGP, confirming it is an integrated part.

Q: What API support does it offer?

A: It supports DirectX 12 (12_1), OpenGL 4.6, and Vulkan 1.4.

Q: Is this GPU still in production?

A: No. The production status is "End-of-life," and it was released on September 1, 2020.

Q: What is the FP32 compute performance?

A: It delivers 1.690 TFLOPS of FP32 compute and 3.379 TFLOPS of FP16 compute at a 2:1 ratio.

Who Should Consider It

The Iris Xe G7 96EU is a 15 W integrated GPU with a 50th percentile ranking among all GPUs in the database. This makes it suitable for users who need a capable integrated solution for everyday computing, light gaming, and media playback in thin-and-light laptops. The 768 shading units and 48 TMUs provide enough throughput for 1080p gaming at medium to low settings in less demanding titles, but the system-dependent memory bandwidth means performance will vary heavily based on the host system's RAM configuration.

Users who prioritize low power consumption and portability over raw performance will find the 15 W TDP appealing. The part is end-of-life, so it is only relevant in existing devices or used markets. For high-resolution gaming, the shared memory subsystem is a clear limitation: at 1440p and above, the system-dependent bandwidth and lack of dedicated VRAM will bottleneck the GPU's theoretical 26.40 GPixel/s pixel rate and 52.80 GTexel/s texture rate. The absence of ray tracing cores and tensor cores further limits its appeal for modern RT-heavy titles and AI-accelerated workloads.

The 50th percentile position means it outperforms roughly half of all GPUs in the database, which is respectable for an integrated part. It is best suited for users who accept 1080p as the target resolution, play lighter or older games, and do not require hardware ray tracing. For those needs, the Iris Xe G7 96EU offers a balanced feature set with DirectX 12 (12_1), OpenGL 4.6, and Vulkan 1.4 support. However, for anyone planning to run demanding modern titles at high settings or high resolutions, the data clearly shows a part that will struggle due to memory constraints rather than raw compute limits.