AMD Radeon R9 M390X vs AMD Radeon RX 560X Comparison

The AMD Radeon R9 M390X and the AMD Radeon RX 560X represent two very different answers to the same question: how much gaming performance can you extract from a 75 watt power envelope. The R9 M390X is a 2015-era mobile part built on the Amethyst chip, while the RX 560X arrived in 2018 as a Polaris-based desktop card. Both are end-of-life products now, and both sit within a few percentile points of each other in the database rankings. The recorded benchmark results, however, show a clear winner in the one test where they can be compared directly, and the story behind that result is worth understanding before choosing between them.

FAQ

Q: Which GPU is faster in the shared benchmark?

A: The RX 560X wins the only head-to-head test in the database, the Geekbench Vulkan run, with a score of 20231 against 16951 for the R9 M390X. That is a 16.2 percent advantage for the Polaris card, and it is the largest single data point separating these two products.

Q: How do their overall database rankings compare?

A: The R9 M390X sits in the 66th percentile versus all GPUs in the database, while the RX 560X sits in the 62nd percentile. The R9 M390X holds an average benchmark score of 20662 against 18626 for the RX 560X, a gap of roughly 10 percent in the older card's favor when all recorded tests are averaged.

Q: Do they have the same amount of memory?

A: Yes, both ship with 4 GB of GDDR5. The difference is in the bus and bandwidth: the R9 M390X uses a 256-bit bus delivering 160.0 GB/s, while the RX 560X uses a 128-bit bus delivering 112.0 GB/s.

Q: How do their power characteristics compare?

A: Both carry a 75 W TDP and neither requires auxiliary power connectors. The key distinction is form factor: the R9 M390X is an MXM module meant for laptops, while the RX 560X is a dual-slot desktop card with a suggested 250 W system power supply.

Q: Which is newer?

A: The RX 560X, by roughly three years. The R9 M390X was released in May 2015, and the RX 560X followed in April 2018.

Q: Which rivals sit closest to each card in the database?

A: The R9 M390X scores within a fraction of a percent of the AMD Radeon RX 5600 XT, the Intel Arc A750, the Intel Arc B570, and the NVIDIA GeForce RTX 3070 Mobile on average. The RX 560X clusters near the AMD Radeon Pro 5700 XT, the AMD FirePro D500, the NVIDIA GeForce RTX 2070, and the NVIDIA Tesla K80.

Architecture Differences

These two cards belong to consecutive generations of AMD's Graphics Core Next family, and the architectural gap explains much of the benchmark picture.

The R9 M390X is built on the Amethyst chip using GCN 3.0, fabricated on TSMC's 28 nm process. It packs 5,000 million transistors onto a 366 mm² die, giving a transistor density of 13.7M per mm². The RX 560X uses the Polaris 21 chip with GCN 4.0, fabricated by GlobalFoundries on a 14 nm process. Its die is far smaller at 123 mm², carrying 3,000 million transistors at a density of 24.4M per mm².

Read those numbers together and the industry trajectory becomes obvious. The RX 560X delivers its performance from less than two-thirds the transistor count on roughly one-third the die area, a direct consequence of the node shrink. Despite the much smaller die, the newer architecture keeps pace: 2.611 TFLOPS of FP32 throughput against 2.961 TFLOPS for the older, larger chip. Both handle FP16 at a 1:1 ratio with FP32, and neither offers dedicated RT cores or tensor cores, so ray tracing and hardware-accelerated ML workloads are off the table for both.

Feature support also differs. The R9 M390X reports Vulkan 1.2.170, while the RX 560X reports Vulkan 1.3, the newer API revision. Both support DirectX 12 at feature level 12_0 and OpenGL 4.6, so baseline modern API compatibility is identical.

Head-to-Head Benchmarks

The database contains one directly comparable test between these two cards, and it is not close.

In Geekbench Vulkan, the RX 560X scores 20231 while the R9 M390X scores 16951. That is a 16.2 percent victory for the Polaris card. The head-to-head record stands at one win for the RX 560X and zero for the R9 M390X.

The reason is worth spelling out. Vulkan is exactly the kind of low-overhead API that GCN 4.0 and its driver stack were tuned for, and the RX 560X's clock structure helps: a 1175 MHz base and 1275 MHz boost, running memory at 1750 MHz (7 Gbps effective). The R9 M390X has no base or boost clock recorded in the database, with memory at 1250 MHz (5 Gbps effective).

The R9 M390X's other recorded result is a Geekbench Metal score of 24373, which is its strongest number in the database and the main driver of its higher 20662 average. The RX 560X's second recorded result is a Geekbench OpenCL score of 17020, which drags its average down to 18626. These tests are not directly comparable across the two cards since they were not run on both, but they explain why the averages and the head-to-head point in opposite directions: the R9 M390X has one very high score in a test the RX 560X never ran, while losing the one test they share.

Context from the rival tables reinforces how tightly clustered both cards are. The R9 M390X's average sits within about half a percent of the RX 5600 XT, Arc A750, Arc B570, and RTX 3070 Mobile. The RX 560X sits comparably close to the Radeon Pro 5700 XT, FirePro D500, RTX 2070, and Tesla K80. Neither card dominates its neighborhood; both are mid-pack performers in the database's overall distribution.

Specification Differences

The spec sheet differences are substantial despite the matching TDP and memory capacity:

• Process and die: 28 nm TSMC, 366 mm², 5,000 million transistors (R9 M390X) versus 14 nm GlobalFoundries, 123 mm², 3,000 million transistors (RX 560X).
• Shading hardware: 2048 shading units, 128 TMUs, 32 ROPs (R9 M390X) versus 1024 shading units, 64 TMUs, 16 ROPs (RX 560X). The older chip has double the rasterization and texture resources on paper.
• Throughput: 23.14 GPixel/s pixel rate and 92.54 GTexel/s texture rate (R9 M390X) versus 20.40 GPixel/s and 81.60 GTexel/s (RX 560X). Compute sits at 2.961 TFLOPS versus 2.611 TFLOPS.
• Memory subsystem: both 4 GB GDDR5, but 256-bit at 160.0 GB/s versus 128-bit at 112.0 GB/s.
• Bus and form factor: PCIe 3.0 x16 on an MXM module versus PCIe 3.0 x8 as a dual-slot desktop card measuring 170 mm (6.7 inches) long.
• Display outputs: "Portable Device Dependent" on the MXM part versus a fixed set of one DVI, one HDMI 2.0b, and one DisplayPort 1.4a on the desktop card.
• Vulkan support: 1.2.170 versus 1.3.
• Lineage: the R9 M390X succeeded the Solar System generation and was itself followed by Polaris Mobile; the RX 560X followed Polaris and was succeeded by Vega.

The suggested power supply figure of 250 W applies only to the RX 560X, which makes sense given it is a discrete desktop card while the R9 M390X lives inside a laptop chassis.

Where Each One Wins

The RX 560X wins where it matters most for a buyer choosing today: the only directly comparable benchmark. Its 16.2 percent Vulkan advantage, newer API support, and far more efficient silicon make it the stronger pick for any workload that leans on modern graphics APIs. If you are running Vulkan-native titles or software, the recorded data points one way only.

The R9 M390X wins on raw resource counts and memory bandwidth. Its 256-bit bus and 160.0 GB/s of bandwidth give it substantially more headroom than the RX 560X's 112.0 GB/s, and it holds the lead in pixel rate, texture rate, and FP32 compute. Its higher average database score of 20662, driven largely by that 24373 Metal result, suggests it can still post strong numbers in the right workload. If your software path resembles its strongest recorded test, the older chip remains competitive.

One practical note: the R9 M390X is an MXM module with device-dependent display output. You cannot buy it as a standalone upgrade in any ordinary sense. The RX 560X is a standard dual-slot desktop card with defined outputs and a defined 170 mm length, which fits far more builds.

The Verdict

Pick the RX 560X if you are choosing a graphics card for a desktop and care about real-world API performance. It won the only head-to-head benchmark by a wide 16.2 percent margin, it supports the newer Vulkan 1.3 revision, and it achieves its results from a die one-third the size with 40 percent fewer transistors. It is a more modern, more efficient design, and the data reflects that.

The R9 M390X makes sense only in its original context: as the graphics engine of a 2015-era laptop. Its superior memory bandwidth, doubled shading and texture resources, and higher average database score show the brute-force approach of the 28 nm era, and its percentile placement of 66 exceeds the RX 560X's 62. But it lost the one test the two cards share, it is an end-of-life mobile module rather than a purchasable card, and its strongest recorded result comes from a benchmark its rival never ran.

For a builder weighing these two on the merits of the recorded data, the RX 560X is the practical choice. The R9 M390X is the historical one: bigger, hungrier per transistor, and ultimately outperformed by a chip a third its size.