AMD Radeon R9 M390X vs AMD Radeon RX 6650 XT Comparison

# AMD Radeon R9 M390X vs AMD Radeon RX 6650 XT

The data presents a stark generational contrast: the AMD Radeon R9 M390X, a 28 nm mobile GPU from 2015, and the AMD Radeon RX 6650 XT, a 7 nm desktop part from 2022, share the same shading unit count but little else. Across the two overlapping benchmark tests, the RX 6650 XT wins decisively in both, with the R9 M390X trailing by margins of 75.1% and 79.2%. Yet the R9 M390X holds a slightly higher percentile rank (66th vs. 65th), a quirk driven by the different benchmark suites each GPU was tested with, suggesting the comparison is less about raw capability and more about architectural era.

FAQ

Q: Which GPU has more shading units, the R9 M390X or the RX 6650 XT?

A: Both GPUs feature exactly 2048 shading units. This parity in shader count makes the performance gap even more striking, as the RX 6650 XT achieves far higher scores despite matching the older chip’s core count.

Q: What is the average benchmark score difference between the two?

A: The R9 M390X posts an average benchmark score of 20662, while the RX 6650 XT averages 19765. This places the R9 M390X slightly ahead by about 4.5%, though this figure is skewed by the fact that the two GPUs were evaluated on entirely different benchmark sets.

Q: How does the R9 M390X compare to its nearest rivals?

A: The R9 M390X is within 0.6% of the AMD Radeon RX 5600 XT (which scores 20713), Intel Arc A750 (20582), Intel Arc B570 (20556), and NVIDIA GeForce RTX 3070 Mobile (20534). Its deltaPct ranges from -0.2% to +0.6%, indicating it sits almost exactly in the middle of this cluster.

Q: What about the RX 6650 XT's nearest competitors?

A: The RX 6650 XT’s nearest rivals include the NVIDIA Tesla K40m (19885, -0.6% delta), AMD FirePro D300 (19637, +0.7%), AMD FirePro W7000 (19905, -0.7%), and NVIDIA Quadro K5200 (19602, +0.8%). All four are within a single percentage point of its 19765 average.

Q: In which benchmark does the RX 6650 XT show its biggest lead?

A: The largest margin comes in the Geekbench Vulkan test, where the RX 6650 XT scores 81306 versus the R9 M390X’s 16951, a 79.2% difference. In Geekbench Metal, the lead is 75.1% (97969 vs. 24373).

Q: Do both GPUs support the same DirectX version?

A: No. The R9 M390X supports DirectX 12 (12_0), while the RX 6650 XT supports DirectX 12 Ultimate (12_2). The newer GPU also supports Vulkan 1.4 compared to the older card’s Vulkan 1.2.170.

Where Each One Wins

The R9 M390X wins in exactly zero head-to-head benchmark comparisons. The RX 6650 XT takes both available tests — Geekbench Metal and Geekbench Vulkan — with no contest. However, the R9 M390X does hold a narrow advantage in the aggregate percentile ranking (66th vs. 65th percentile across all GPUs), which reflects its stronger showing within its own benchmark set rather than any direct victory over the newer card.

Looking at the data qualitatively, the R9 M390X’s wins are contextual. Its average score of 20662 places it slightly above the RX 6650 XT’s 19765, but this is misleading because the two were tested with different applications. The R9 M390X was evaluated on Geekbench Metal and Vulkan only, while the RX 6650 XT ran a much broader suite including PassMark tests and 3DMark Steel Nomad. In terms of raw performance metrics, the RX 6650 XT wins decisively in every shared test, making it the clear choice for any workload that leverages Metal or Vulkan APIs.

For compute-heavy tasks, the RX 6650 XT’s PassMark GPU Compute score of 7532 and its FP32 throughput of 10.79 TFLOPS dwarf the R9 M390X’s 2.961 TFLOPS. The older card’s only meaningful edge is its lower 75 W TDP, which makes it suitable for power-constrained mobile platforms, whereas the RX 6650 XT demands 176 W and a 450 W suggested PSU.

Architecture Differences

The R9 M390X is built on GCN 3.0 with the Amethyst chip, fabricated on a 28 nm process at TSMC. It packs 5,000 million transistors into a 366 mm² die, yielding a transistor density of 13.7 million per square millimeter. This is a mature, power-efficient design from the mid-2010s, with a 256-bit memory bus and GDDR5 memory.

The RX 6650 XT uses RDNA 2.0 with the Navi 23 chip, also from TSMC but on a much newer 7 nm process. It integrates 11,060 million transistors into a smaller 237 mm² die, achieving a density of 46.7 million per square millimeter — roughly 3.4 times denser than the older chip. This architectural leap allows the RX 6650 XT to nearly quadruple FP32 throughput (10.79 TFLOPS vs. 2.961 TFLOPS) while also adding 32 dedicated ray tracing cores, which the R9 M390X lacks entirely.

Memory architecture differs fundamentally as well. The R9 M390X uses 4 GB of GDDR5 on a 256-bit bus, delivering 160.0 GB/s bandwidth. The RX 6650 XT uses 8 GB of GDDR6 on a narrower 128-bit bus, yet achieves 280.3 GB/s due to much faster memory clocks (17.5 Gbps effective vs. 5 Gbps). The newer card also doubles the ROP count from 32 to 64, and its pixel rate jumps from 23.14 GPixel/s to 168.6 GPixel/s.

Specification Differences

The most glaring specification gap is in clock speeds. The R9 M390X has no listed base, boost, or game clocks, while the RX 6650 XT runs at 2055 MHz base, 2410 MHz game, and 2635 MHz boost. Memory clocks also differ: 1250 MHz (5 Gbps effective) for the old card versus 2190 MHz (17.5 Gbps effective) for the new one.

FP32 performance tells the story: 2.961 TFLOPS for the R9 M390X versus 10.79 TFLOPS for the RX 6650 XT, a 3.6x improvement. FP16 is even more lopsided — the R9 M390X offers 2.961 TFLOPS (1:1 ratio), while the RX 6650 XT delivers 21.59 TFLOPS (2:1 ratio), a 7.3x jump. Texture rate increases from 92.54 GTexel/s to 337.3 GTexel/s.

Power and physical specs diverge sharply. The R9 M390X is rated at 75 W TDP, uses an MXM Module slot, and requires no power connectors. The RX 6650 XT draws 176 W, is a dual-slot card, needs one 8-pin connector, and recommends a 450 W PSU. Bus interfaces differ too: PCIe 3.0 x16 for the older card versus PCIe 4.0 x8 for the newer. Display outputs also change from "Portable Device Dependent" to 1x HDMI 2.1 and 3x DisplayPort 1.4a.

Head-to-Head Benchmarks

The Geekbench Metal test shows a complete rout. The RX 6650 XT scores 97969, while the R9 M390X manages only 24373. This 75.1% delta means the newer card delivers roughly four times the Metal performance. Given that both GPUs have identical shader counts, this result highlights the massive efficiency gains from RDNA 2.0’s design over GCN 3.0.

Geekbench Vulkan paints an even bleaker picture for the R9 M390X. It scores 16951, while the RX 6650 XT hits 81306 — a 79.2% lead. The deltaPct of -79.2% from the R9 M390X’s perspective indicates it retains only about a fifth of the newer card’s Vulkan performance. The RX 6650 XT’s Vulkan 1.4 support versus the older card’s 1.2.170 likely contributes to this gap, as does the newer architecture’s superior compute efficiency.

Beyond these shared tests, the RX 6650 XT has a broader benchmark portfolio. Its PassMark G3D score of 17166 and GPU Compute score of 7532 provide additional context: the former suggests strong rasterization capability, while the latter confirms compute-heavy workloads are well within its reach. The R9 M390X lacks these benchmarks entirely, so no direct comparison is possible, but the FP32 and FP16 figures in the spec sheet already indicate the newer card’s dominance.

The Verdict

The data leaves no ambiguity: the AMD Radeon RX 6650 XT is the superior GPU in every measurable head-to-head metric. Its 75-79% leads in Metal and Vulkan benchmarks are decisive, and its architectural advantages — 7 nm process, 32 ray tracing cores, doubled ROPs, and 3.6x FP32 throughput — make it the clear pick for any modern workload. The R9 M390X’s only claims to relevance are its lower 75 W TDP and its slightly higher average benchmark score (20662 vs. 19765), the latter being an artifact of different test suites rather than genuine parity.

For users prioritizing raw performance, the RX 6650 XT is the obvious choice. Its 8 GB GDDR6 memory, 280.3 GB/s bandwidth, and 168.6 GPixel/s pixel rate support high-resolution gaming and compute tasks that the R9 M390X simply cannot handle. The older card’s 4 GB GDDR5 and 160.0 GB/s bandwidth, while respectable for 2015, are inadequate for today’s demands.

However, the R9 M390X has a niche: power-constrained environments. Its 75 W TDP and lack of external power connectors make it suitable for legacy MXM-based laptops where the RX 6650 XT’s 176 W TDP and dual-slot design are impossible. The percentile ranking (66th vs. 65th) further muddies the waters, suggesting that within their respective benchmark pools, the older card performs comparably well. But for any user who can accommodate the RX 6650 XT’s power and size requirements, the benchmark data shows it is overwhelmingly the better investment, with launch MSRP of 399 USD reflecting its far greater capability.