UHD 770 vs HD Graphics 530: what did we measure?
We benchmarked two Intel integrated GPUs that can be used in Linux servers: the Intel UHD 770 passed through to a test VM on a Core i5-12500 host, and the Intel HD Graphics 530 built into a Core i3-6100. Here are the OpenGL and video-encoding results, plus a test of two simultaneous 4K transcodes on the HD 530.
Results
| GPU and test environment | glmark2, 1080 × 1920 | Synthetic H.264 encode, 1080 × 1920 |
|---|---|---|
| Intel UHD 770, Debian 13 VM, i915 driver (out-of-tree module, 2026.09.14) | 2749 | about 11.7× real time |
| Intel HD Graphics 530, Core i3-6100, Debian 13 | 1038 | about 8.76× real time |
The glmark2 number is a composite score, not a game’s frame rate. Both graphics runs used glmark2 2023.01, a 1080 × 1920 portrait surface, and the build, refract, terrain, and texture scenes for five seconds each. The encoding runs used a synthetic test pattern; they do not measure Unity rendering or a real movie.
The test systems and drivers differ: the UHD 770 ran in a test VM with guest driver 2026.09.14 and kernel 7.2.11; the HD 530 ran on the Core i3-6100 node with Debian 13, kernel 6.12.113, and Intel iHD media driver 25.2.3. These results give an indication, not a universal performance ratio between the GPUs.
Two simultaneous 4K transcodes on the HD 530
We generated a synthetic 4K H.264 SDR source at 30 frames per second, then ran two simultaneous transcodes to 1080p. Both processed all 240 frames of their eight-second input at about 1.23× real time, with no GPU reset.
In this setup, VA-API handled H.264 decoding and H.264 encoding. Hardware resizing with scale_vaapi did not start on this node, so the CPU performed the resize. This validates that specific H.264 SDR path, with limited headroom. It does not validate two HEVC Main10 streams, Plex, HDR content, or a Unity game rendering at the same time.
The HD 530 node’s vainfo output exposes HEVC Main decoding, but no HEVC Main10 profile. We did not test two 4K HEVC Main10 inputs and do not claim the HD 530 is suitable for that workload.
What the numbers mean for a server
The HD 530 can suit an entry-level workload, and the test confirms it can process two synthetic 4K H.264 streams in the specific pipeline above. It is an older generation, and its i3-6100 host has two cores and four threads. The test did not include simultaneous 3D rendering, audio, or tone mapping.
The UHD 770 scored higher in these synthetic measurements, but a glmark2 score cannot directly predict game performance. For Unity, Plex, or continuous workloads, the actual project, codecs, and settings need to be tested. These results do not guarantee a minimum frame rate or support for every video profile.
The HD Graphics 530 is available on request as a no-surcharge alternative to the iGPU Start plan. The GPU is assigned to the VM, but it remains an older integrated GPU; dedicated assignment does not remove its hardware limits.
Hardware-accelerated codec support
The codecs below reflect Intel’s published media capabilities for these GPU generations. If a GPU cannot process a particular profile or pixel format, an application may fall back to CPU software processing when enabled; otherwise, the operation may fail. On Linux, the driver and profiles actually exposed through VA-API also matter, so check vainfo inside the VM.
| GPU | Hardware decoding | Hardware encoding |
|---|---|---|
| HD Graphics 530 | H.264/AVC, 8-bit HEVC/H.265 (Main), MPEG-2, VC-1, MJPEG, VP8 | H.264/AVC; 8-bit HEVC through Intel’s hybrid path, depending on driver and application |
| UHD 770 | H.264/AVC, 8/10-bit HEVC/H.265, VP9, AV1 | H.264/AVC, 8/10-bit HEVC/H.265, VP9; no hardware AV1 encoding |
The HD 530 does not hardware-decode HEVC Main10: vainfo on our node exposes HEVC Main but not Main10. HEVC 10-bit video therefore should not be assumed to decode on the GPU in this configuration. Generation-level capabilities are based on Intel’s oneVPL documentation; actual exposure depends on the hardware, kernel, and driver.
GPU contention and choosing a UHD 770 plan
We measured the UHD 770 VM on Perf 1 while other customers were placing little load on the GPU. This is an observed result under those conditions, not a constant-throughput guarantee. For sustained workloads or when consistency matters more, choose Perf 2 or Perf 3: their higher relative allocation reduces sensitivity to contention, but does not guarantee a fixed minimum throughput.