AMD Radeon HD 6950 vs AMD Radeon HD 8790M Comparison

Where Each One Wins

The benchmark data splits cleanly between these two AMD parts, and the split is not what the raw specifications would suggest at first glance. The AMD Radeon HD 6950 wins the only directly comparable benchmark test recorded in the database, the Geekbench OpenCL workload, with a score of 6210 against 5017 for the AMD Radeon HD 8790M. That is a 23.8% advantage for the older desktop card, a substantial margin that places it clearly ahead in general-purpose compute tasks.

The AMD Radeon HD 8790M, however, has one exclusive win that the HD 6950 cannot contest: it is the only one of the two with a recorded Geekbench Vulkan score. Its 6365 result in that test shows that the mobile part handles modern graphics APIs with a capability the desktop card simply does not possess in the database. This matters for anyone running Vulkan-based workloads, where the HD 8790M demonstrates a measurable capability while the HD 6950 has no recorded result at all.

Looking at the percentile rankings, the HD 6950 sits at the 36th percentile of all GPUs in the database, while the HD 8790M sits at the 33rd percentile. The average benchmark score tells a similar story: the HD 6950 averages 6210 across its recorded tests, while the HD 8790M averages 5691 across its two recorded tests. The desktop card holds the overall performance crown, but the mobile part counters with API compatibility and a cleaner architectural lineage.

For use-case planning, the HD 6950 is the choice for raw OpenCL throughput and any workload that can leverage its massive shading unit count and wide memory bus. The HD 8790M is the choice for portability and for anyone needing Vulkan support, since the database shows no Vulkan result for the HD 6950. The mobile card also wins on efficiency of design, being built on a smaller process node with far fewer transistors, though the database does not provide a TDP figure for the HD 8790M to confirm power draw.

Architecture Differences

The two GPUs come from different architectural generations despite sharing the AMD brand. The HD 6950 uses the Cayman chip built on TeraScale 3, while the HD 8790M uses the Mars chip built on GCN 1.0. This is a fundamental split: TeraScale 3 is the older VLIW-based design, while GCN 1.0 is the compute-oriented architecture that AMD would carry forward for years.

The manufacturing process differs significantly. The HD 6950 is fabricated on TSMC's 40 nm node, while the HD 8790M uses TSMC's 28 nm node. This explains the dramatic difference in physical characteristics. The HD 6950 packs 2,640 million transistors into a 389 mm² die, giving a transistor density of 6.8 million per square millimeter. The HD 8790M crams 950 million transistors into just 77 mm², achieving 12.3 million per square millimeter. The newer process node allows the mobile chip to be far denser, which is typical of the generational leap between 40 nm and 28 nm.

The compute resources tell the story of two very different design philosophies. The HD 6950 has 1408 shading units, 88 texture mapping units, and 32 render output units. The HD 8790M has 384 shading units, 24 TMUs, and only 8 ROPs. The desktop card has roughly 3.7 times the shading units, 3.7 times the TMUs, and 4 times the ROPs. This is a massive disparity in raw hardware, and it directly explains the OpenCL benchmark gap.

Memory configurations differ in bus width and bandwidth. The HD 6950 uses a 256-bit memory bus with 2 GB of GDDR5 running at 1250 MHz, delivering 160.0 GB/s of bandwidth. The HD 8790M uses a 128-bit bus with the same 2 GB of GDDR5 but at 1000 MHz, yielding only 64.00 GB/s. The desktop card has 2.5 times the memory bandwidth, which matters heavily for fill-rate-bound workloads.

Clock speeds tell a different story. The HD 6950 has no base or boost clock recorded in the database, only a memory clock. The HD 8790M runs at 850 MHz base and 900 MHz boost. The mobile part compensates for its smaller hardware with a higher core clock, but the sheer resource disparity of the HD 6950 overwhelms that advantage.

API support favors the newer part. The HD 8790M supports DirectX 12 (11_1), OpenGL 4.6, and Vulkan 1.2.170. The HD 6950 supports DirectX 11.2 (11_0) and OpenGL 4.4, with no Vulkan support recorded. This is the clearest architectural advantage for the GCN 1.0 part, and it explains why the HD 8790M has a Vulkan benchmark score while the HD 6950 does not.

Head-to-Head Benchmarks

The database contains one direct head-to-head comparison between these two cards: the Geekbench OpenCL test. The HD 6950 scores 6210, while the HD 8790M scores 5017. The delta is 23.8% in favor of the desktop card. That is a commanding lead, and it aligns with the resource disparity between the two parts.

To put that score in context, the HD 6950's nearest rivals in the database are the AMD FirePro W600 at 6223, the NVIDIA GeForce RTX 4070 Ti SUPER AD102 at 6270, the NVIDIA GeForce RTX 5070 Ti SUPER at 6270, and the NVIDIA Quadro K620 at 6282. The HD 6950 trails the FirePro W600 by just 0.2%, trails both RTX 4070 Ti SUPER AD102 and RTX 5070 Ti SUPER by 0.9%, and trails the Quadro K620 by 1.1%. This is remarkable: an older desktop card from 2010 sits within 1.1% of modern high-end NVIDIA parts in this specific OpenCL workload. The recorded data shows the Cayman chip's compute potential remains relevant.

The HD 8790M's nearest rivals include the Intel Iris Pro Graphics P6300 at 5712, the NVIDIA GeForce GTX 670MX at 5721, the NVIDIA GeForce GTX 550 Ti at 5731, and the NVIDIA Quadro M500M at 5604. The HD 8790M's average score of 5691 puts it 0.4% behind the Iris Pro P6300, 0.5% behind the GTX 670MX, 0.7% behind the GTX 550 Ti, and 1.6% ahead of the Quadro M500M. The mobile part sits comfortably in the midrange of its peer group, trading blows with integrated and older discrete solutions.

The Vulkan result for the HD 8790M, 6365, has no direct counterpart in the database for the HD 6950. This means the mobile part's Vulkan score exceeds its own OpenCL score by 26.9%, suggesting the GCN architecture scales well with modern APIs. For the HD 6950, the absence of a Vulkan result is a functional limitation, not just a missing data point.

When interpreting the average benchmark scores, the HD 6950's 6210 average reflects a single OpenCL result, while the HD 8790M's 5691 average blends its OpenCL and Vulkan scores. Even with the Vulkan score pulling its average up, the HD 8790M still trails the HD 6950 by 8.4% on average. The desktop card is simply the faster part in compute-heavy workloads.

FAQ

Q: Which card is faster in OpenCL compute workloads?

A: The AMD Radeon HD 6950 scores 6210 in Geekbench OpenCL, which is 23.8% higher than the HD 8790M's 5017. The desktop card wins this workload decisively.

Q: Does the HD 8790M support Vulkan?

A: Yes, the HD 8790M has a Geekbench Vulkan score of 6365 and supports Vulkan 1.2.170 according to its API list. The HD 6950 has no Vulkan score and no Vulkan support listed.

Q: How does the HD 6950 compare to its nearest rivals?

A: The HD 6950 trails the AMD FirePro W600 by 0.2%, the NVIDIA GeForce RTX 4070 Ti SUPER AD102 by 0.9%, the NVIDIA GeForce RTX 5070 Ti SUPER by 0.9%, and the NVIDIA Quadro K620 by 1.1% in average score.

Q: What is the HD 8790M's closest competitor?

A: The HD 8790M's nearest rival is the Intel Iris Pro Graphics P6300, which scores 5712, just 0.4% ahead of the HD 8790M's average of 5691. The NVIDIA GeForce GTX 670MX and GTX 550 Ti are also close at 0.5% and 0.7% ahead respectively.

Q: Which card has more shading units?

A: The HD 6950 has 1408 shading units, while the HD 8790M has 384. The HD 6950 also has 88 TMUs against 24, and 32 ROPs against 8.

Q: Do these cards have the same memory bandwidth?

A: No. The HD 6950 has a 256-bit bus delivering 160.0 GB/s, while the HD 8790M has a 128-bit bus delivering 64.00 GB/s. The HD 6950 has 2.5 times the bandwidth.

Specification Differences

The two cards differ in nearly every measurable specification. The chip design is Cayman for the HD 6950 and Mars for the HD 8790M. The architecture is TeraScale 3 for the desktop card and GCN 1.0 for the mobile card. The process node is 40 nm for the HD 6950 and 28 nm for the HD 8790M, both fabricated by TSMC.

Transistor counts diverge sharply: 2,640 million for the HD 6950 against 950 million for the HD 8790M. Die size follows, with the HD 6950 at 389 mm² and the HD 8790M at 77 mm². Transistor density reverses the trend, with the HD 6950 at 6.8 million per mm² and the HD 8790M at 12.3 million per mm².

The HD 6950 has no base or boost clock recorded, while the HD 8790M runs at 850 MHz base and 900 MHz boost. Memory clocks differ as well: 1250 MHz with 5 Gbps effective for the HD 6950, against 1000 MHz with 4 Gbps effective for the HD 8790M. Both have 2 GB of GDDR5, but the bus width is 256-bit for the desktop card and 128-bit for the mobile card. Bandwidth is 160.0 GB/s versus 64.00 GB/s.

Compute unit counts favor the desktop card heavily: 1408 shading units versus 384, 88 TMUs versus 24, and 32 ROPs versus 8. Pixel rate is 25.60 GPixel/s for the HD 6950 and 7.200 GPixel/s for the HD 8790M. Texture rate is 70.40 GTexel/s versus 21.60 GTexel/s. FP32 performance is 2.253 TFLOPS against 691.2 GFLOPS.

The HD 6950 has a TDP of 200 W, uses dual-slot cooling, requires 2x 6-pin power connectors, and suggests a 550 W power supply. The HD 8790M has no TDP recorded, uses an MXM Module form factor, requires no power connectors, and has no suggested PSU. The HD 6950 uses PCIe 2.0 x16, while the HD 8790M uses MXM-A (3.0).

Display outputs differ completely: the HD 6950 has 2x DVI, 1x HDMI 1.4a, and 2x mini-DisplayPort 1.2, while the HD 8790M's outputs are portable device dependent. API support favors the mobile part: DirectX 12 (11_1) versus 11.2 (11_0), OpenGL 4.6 versus 4.4, and Vulkan 1.2.170 versus none.

The HD 6950 measures 286 mm in length, 126 mm in height, and 42 mm in width. The HD 8790M has no dimensions recorded. Release dates are 2010-12-13 for the HD 6950 and 2013-03-31 for the HD 8790M. The HD 6950 has a launch MSRP of 299 USD, while the HD 8790M has none recorded. The HD 6950's predecessor is Evergreen and successor is Southern Islands; the HD 8790M's predecessor is London and successor is Gem System.

The Verdict

The data points to a clear split based on workload. For raw compute performance in OpenCL, the AMD Radeon HD 6950 is the definitive winner. Its 6210 score against 5017 for the HD 8790M is a 23.8% margin, and its percentile ranking of 36 against 33 confirms it sits higher in the overall GPU distribution. Anyone running OpenCL workloads, legacy DirectX 11 applications, or tasks that benefit from 160.0 GB/s of memory bandwidth should pick the HD 6950. The desktop card's 1408 shading units, 88 TMUs, and 32 ROPs provide the hardware muscle that the mobile part simply cannot match.

For modern API support and portability, the AMD Radeon HD 8790M is the only choice. Its Vulkan 1.2.170 support and DirectX 12 (11_1) capability put it in a different software generation than the HD 6950, which maxes out at DirectX 11.2 and OpenGL 4.4 with no Vulkan. The HD 8790M's 6365 Vulkan score demonstrates that GCN 1.0 scales well with contemporary APIs, and its 28 nm process with 12.3 million transistors per mm² shows a far more efficient design. The MXM Module form factor means it fits in portable systems where a 286 mm dual-slot card never will.

The HD 6950's competitive position relative to its nearest rivals is striking. Sitting within 1.1% of the NVIDIA Quadro K620, 0.9% of the RTX 4070 Ti SUPER AD102, and 0.9% of the RTX 5070 Ti SUPER in average score means the old Cayman chip still trades blows with parts from much later generations. The HD 8790M, by contrast, sits in a tighter pack, within 0.7% of the GTX 550 Ti and 0.5% of the GTX 670MX, and only 1.6% ahead of the Quadro M500M.

The verdict is straightforward: choose the HD 6950 for desktop compute tasks where raw throughput and memory bandwidth dominate, and choose the HD 8790M for mobile systems or any workload that requires Vulkan. The desktop card is the performance king in OpenCL; the mobile card is the compatibility king in modern APIs. Neither card is a general-purpose winner across all tests, because they target different form factors and software generations. The database records one head-to-head win for the HD 6950 and zero for the HD 8790M, but the HD 8790M's Vulkan capability is a functional advantage that no benchmark comparison can deny.