A Fascinating 20GB RTX 3080 Mod With a Messy Benchmark Comparison

Rating

Video Reviewed
Rating8.3/10
I Bought A Modded RTX 3080 From Aliexpress…

Doubling an RTX 3080’s standard 10GB of video memory to 20GB immediately gives this AliExpress experiment a compelling technical mystery. The unbranded card arrives with enough unusual details to justify the curiosity: a blower-style cooler, multiple power-connector solder points, a heavy copper fin stack, and an NVLink area that raises questions about what PCB design the seller actually used. Conflicting references in the listing to an RTX 3080, 3080 Ti, and 3080M make the initial inspection even more uncertain, while the presenter keeps that uncertainty clear rather than pretending to know exactly what he bought.

The teardown is the highlight because it reveals that the extra memory is more substantial than a misleading product label. Twenty 1GB memory modules are distributed across both sides of the PCB, accompanied by what the presenter counts as an 18-phase power-delivery setup and a large vapor chamber beneath the copper heatsink. He openly admits when he cannot identify particular components or explain how the power phases are allocated, which is preferable to filling those gaps with unsupported technical conclusions. The close examination also makes the card interesting beyond the basic question of whether more VRAM improves gaming performance.

Simply getting the card running provides another useful result. NVIDIA’s regular drivers install without trouble, GPU-Z identifies a GA102 RTX 3080 and reports 20GB of memory, and the card operates normally enough to begin testing without requiring the questionable custom drivers the presenter had anticipated. That makes the modification considerably more intriguing than a novelty that merely displays an altered memory figure. The card does have an obvious practical drawback, however: under load it reaches around 82 degrees Celsius and its blower cooler becomes conspicuously loud.

The gaming tests are sensibly chosen to explore situations where 10GB might become restrictive rather than simply producing a collection of conventional benchmark numbers. Cyberpunk initially stays below that threshold before more demanding settings push memory use higher, while Doom: The Dark Ages reaches roughly 10GB at 1440p and nearly 13GB at 4K. The Last of Us provides another useful contrast, showing that high reported memory utilization does not automatically translate into a dramatic performance collapse on the standard-capacity card. These examples help demonstrate that additional VRAM only matters when the workload actually encounters a meaningful memory limitation.

Unfortunately, the comparison becomes compromised when the presenter’s EVGA 10GB RTX 3080 refuses to draw more than roughly 250 watts. He makes a substantial effort to diagnose the problem, including driver work, firmware and motherboard BIOS updates, overclocking attempts, and testing in another system, but cannot resolve it. He appropriately discloses that limitation rather than presenting the cards as perfectly equivalent. Even so, it means performance differences such as the roughly 9% deficit seen at 1440p in Cyberpunk cannot cleanly be attributed to memory capacity, weakening the experiment’s ability to quantify the modification’s benefit.

Doom: The Dark Ages nevertheless supplies the clearest demonstration of the intended point, with the 10GB card dropping to roughly five frames per second under the tested high-memory conditions while the 20GB version remains in the 30-fps range. Cyberpunk also shows a sizable difference under an extreme 4K configuration, although neither result turns this into a sensible gaming purchase. At nearly $1,000, the presenter concludes that the modified card makes little economic sense for gaming and also questions its value for local language-model workloads compared with higher-memory alternatives. The result is an entertaining and unusually revealing hardware investigation whose strongest evidence is the card itself, while its main weakness is an unavoidable but significant flaw in the control GPU used for benchmarking.

Pros

  • The teardown confirms a genuinely unusual 20GB configuration with 20 memory modules rather than relying on the seller’s listing.
  • Detailed PCB and cooling inspection makes the modification interesting beyond simple gaming benchmarks.
  • Standard NVIDIA drivers recognize the card without the custom-driver complications the presenter expected.
  • Tests deliberately target memory-heavy settings and show a particularly dramatic bottleneck in Doom: The Dark Ages.
  • The malfunctioning comparison RTX 3080 is clearly disclosed, along with the extensive attempts made to troubleshoot it.

Cons

  • The standard 10GB RTX 3080’s unexplained 250-watt power limitation prevents a clean performance comparison.
  • Several PCB components and aspects of the unusual power-delivery design remain unidentified or unexplained.
  • Some tested settings produce frame rates too low for comfortable gaming, reducing their relevance to typical use.
  • The card runs hot and the blower cooler becomes notably loud under load.
  • The discussion of its potential value for local language models is brief and not backed by actual compute-workload testing.

A genuinely bizarre graphics card turns into an engaging hardware investigation, especially once the teardown verifies that its doubled memory capacity is real and demanding games expose circumstances where 10GB can become restrictive. The faulty comparison card prevents the benchmarks from delivering a definitive measurement of the modification’s advantage, but transparent reporting of that problem and a strong physical examination still make the experiment informative.

Related Reviews