AMD Radeon HD 8730M vs NVIDIA GeForce 840M Comparison

The AMD Radeon HD 8730M and the NVIDIA GeForce 840M are two end-of-life mobile GPUs that, on paper, look far more alike than they sound. Both were built by TSMC on a 28 nm process, both carry 384 shading units and 8 ROPs, and both pair those shaders with 2 GB of DDR3 memory. Yet the recorded benchmark data draws a sharper line than the spec sheets suggest: the HD 8730M holds a higher average score, a higher percentile ranking against the full database, and the only head-to-head win recorded between the two. What follows is a look at where each design earns its keep, and what the numbers imply.

Where Each One Wins

The database contains a single direct head-to-head result, and it goes to the HD 8730M. In geekbench_opencl, the AMD part scores 5955 against 5764 for the GeForce 840M, a 3.3 percent advantage. That is the entire recorded win column: one win for AMD, zero for NVIDIA.

Still, the picture is worth questioning rather than accepting at face value. AMD's 3.3 percent lead sits within the noise band suggested by each chip's own rival list. The HD 8730M's nearest rivals span scores from 5929 to 5982, a spread that comfortably contains both contenders' results. In other words, the data shows AMD ahead, but by a margin small enough that these two effectively occupy the same performance tier.

Where the GeForce 840M genuinely wins is on paper characteristics that the benchmark suite does not capture. Its FP32 throughput of 863.2 GFLOPS far exceeds the HD 8730M's 537.6 GFLOPS, and its pixel rate of 8.992 GPixel/s outpaces AMD's 5.600 GPixel/s. Whether that translates into real-world wins is something the recorded data cannot confirm, since the only measured comparison favors AMD. The honest read: AMD wins the measured contest, NVIDIA wins the theoretical one.

Architecture Differences

The two chips come from opposing design philosophies of their era. The HD 8730M uses the Mars chip on AMD's GCN 1.0 architecture, part of the Solar System generation (HD 8700M), with a lineage running from the London predecessor to the Gem System successor. The GeForce 840M uses the GM108S chip on NVIDIA's Maxwell architecture, slotted into the GeForce 800M generation between the GeForce 700M and GeForce 900M.

Both were fabricated by TSMC on 28 nm, and interestingly both land on exactly 77 mm² of die area. The transistor counts diverge modestly: 950 million for AMD versus 1,020 million for NVIDIA, giving the 840M a higher density of 13.2M per mm² against AMD's 12.3M per mm². Maxwell's reputation for efficiency is visible in one recorded figure: the 840M carries a 33 W TDP, while the HD 8730M's TDP is not recorded in the database.

Clock behavior differs sharply. AMD runs a base clock of 650 MHz with a boost to 700 MHz. NVIDIA's base sits at 1029 MHz, boosting to 1124 MHz, meaning the Maxwell part clocks dramatically higher. Yet AMD counters with a wider memory bus: 128 bit against NVIDIA's 64 bit. The result is a bandwidth gap in AMD's favor, 28.80 GB/s versus 16.02 GB/s, despite NVIDIA's faster memory clocks (1001 MHz, 2 Gbps effective, against 900 MHz and 1800 Mbps effective for AMD).

Texture hardware also splits: the HD 8730M fields 24 TMUs to the 840M's 16, which explains AMD's edge in memory bandwidth feeding more texture pipelines, even though NVIDIA's higher clocks deliver a slightly higher texture rate (17.98 versus 16.80 GTexel/s). Neither chip has RT cores or tensor cores; these are pre-ray-tracing designs.

Head-to-Head Benchmarks

The measured contest is short. Geekbench OpenCL records 5955 for the Radeon HD 8730M and 5764 for the GeForce 840M, a 3.3 percent delta in AMD's favor. The database also holds a geekbench_vulkan score of 4880 for the 840M, but no corresponding AMD result exists, so no comparison can be drawn there.

Context from the percentile fields reinforces how narrow this rivalry is. The HD 8730M sits at the 34th percentile against all GPUs in the database, with an average benchmark score of 5955. The GeForce 840M sits at the 31st percentile with an average of 5322, dragged down by the lower Vulkan result. Three percentile points separate them, which the data suggests is meaningful but modest.

The rival lists tell a curious story of clustering. The HD 8730M's neighborhood includes the NVIDIA Quadro K620M (5977 average, a dead-even 0 percent delta), the AMD Radeon HD 8750M (5970, AMD 0.3 percent ahead), the Intel UHD Graphics 730 (5929, AMD 0.4 percent ahead), and the NVIDIA Quadro K4000 (5982, AMD 0.5 percent behind). Every one of those contests is essentially a tie. The 840M's rivals are similarly bunched: the GeForce 930A (5317, 0.1 percent), the GeForce GTX 980M (5308, 0.3 percent), the Radeon R7 M445 (5358, 0.7 percent behind), and the GeForce 940M (5284, 0.7 percent ahead).

One oddity worth flagging: the GeForce GTX 980M appearing in the 840M's rival list at nearly identical average score is a quirk of averaging across different benchmark mixes, and it underscores why single cross-family score comparisons deserve scrutiny. The OpenCL-specific head-to-head remains the cleanest measured signal, and it favors AMD.

Specification Differences

The shared ground first: both are 28 nm TSMC parts with 384 shading units, 8 ROPs, 2 GB of DDR3 memory, a PCIe 3.0 x8 bus interface, OpenGL 4.6 support, and end-of-life production status. Their dies are identical in area at 77 mm².

Where they diverge:

• Architecture: GCN 1.0 (Mars) versus Maxwell (GM108S)
• Release date: the HD 8730M arrived on 2013-03-31; the GeForce 840M followed on 2014-03-11
• Clocks: 650 MHz base / 700 MHz boost for AMD against 1029 MHz base / 1124 MHz boost for NVIDIA
• Memory bus: 128 bit for AMD, 64 bit for NVIDIA; bandwidth 28.80 GB/s versus 16.02 GB/s
• TMUs: 24 for AMD, 16 for NVIDIA
• Theoretical rates: AMD leads nowhere except bandwidth; NVIDIA posts 8.992 GPixel/s pixel rate and 863.2 GFLOPS FP32 against AMD's 5.600 GPixel/s and 537.6 GFLOPS
• Transistors: 950 million (AMD) versus 1,020 million (NVIDIA)
• TDP: 33 W recorded for the 840M; not recorded for the HD 8730M
• API support: NVIDIA holds newer interfaces, DirectX 12 (11_0) with Vulkan 1.4, while AMD lists DirectX 12 (11_1) with Vulkan 1.2.170, a slightly higher feature level on DirectX but an older Vulkan version
• Form factor: the 840M is listed as IGP with no power connectors and portable-device-dependent display outputs; no corresponding fields are recorded for the HD 8730M

FAQ

Q: Which GPU is faster in the recorded benchmarks?

A: The AMD Radeon HD 8730M. It won the only direct head-to-head, geekbench_opencl, 5955 to 5764, a 3.3 percent margin, and holds a higher average score (5955 versus 5322) and percentile (34 versus 31).

Q: Is the performance gap large?

A: No. The 3.3 percent OpenCL delta is smaller than the spreads within each card's own rival clusters, so the data treats them as near-equals.

Q: Which has more memory bandwidth?

A: The HD 8730M, by a wide margin: 28.80 GB/s on a 128 bit bus versus 16.02 GB/s on a 64 bit bus, both with 2 GB of DDR3.

Q: Which has higher theoretical compute throughput?

A: The GeForce 840M. Its 863.2 GFLOPS FP32 and 8.992 GPixel/s pixel rate substantially exceed the HD 8730M's 537.6 GFLOPS and 5.600 GPixel/s, thanks to much higher clocks.

Q: Do they support modern graphics APIs?

A: Both list DirectX 12 and OpenGL 4.6. AMD records feature level 11_1 against NVIDIA's 11_0, while NVIDIA carries the newer Vulkan 1.4 versus AMD's Vulkan 1.2.170.

Q: Are either of these still in production?

A: No. Both are marked end-of-life in the database.

The Verdict

The measured data favors the AMD Radeon HD 8730M, and it is the only card here with a recorded head-to-head victory. Buyers or upgraders comparing these two strictly on benchmark evidence should lean AMD: it wins the OpenCL contest, posts the higher average score, and sits three percentile points higher against the full database.

The case for the GeForce 840M rests on unmeasured strengths. Its theoretical FP32 throughput is dramatically higher, its pixel rate is higher, its Vulkan support is newer at version 1.4, and its 33 W TDP is the only power figure recorded between the two. If a workload rewards raw shader clocks and newer API support, Maxwell's on-paper advantages are real, but the database cannot confirm them in measured results.

The most defensible conclusion is a split decision: the HD 8730M for anyone who follows recorded benchmark scores, the 840M for those who weight theoretical throughput and newer Vulkan support. Given that both sit in the low 30th percentile range against all GPUs, neither was ever a performance champion, and the data suggests the choice matters less than either spec sheet implies.