Skip to content
Sign in
Technology · Global

Modders lift software limits on AI compute for NVIDIA GeForce GPUs

Modders removed software limits on certain Tensor Core operations in NVIDIA GeForce RTX 20-, 30- and 40-series GPUs. HotHardware says the changes raise throughput for relevant compute tasks by roughly 1.8 to 2 times, without hardware changes or a new VBIOS.

AI-assisted synthesis · Based on 5 outlets' reporting · Article text read from 1 of 5 outletsHow it's made

View timeline · 6 entries
5 outletsLatest reporting 2 min read
Coverage: Too few ratedL 0 · C 1 · R 0 · 4 unrated
File photo
Photo: Coolcaesar / Wikimedia Commons (CC BY-SA 4.0)

What happened

Key facts

  • HotHardware says the modifications remove software-level limits on certain Tensor Core operations (per hothardware.com, tweaktown.com).
  • The reported throughput gain reaches roughly 1.8 to 2 times for relevant AI and compute workloads (per hothardware.com, videocardz.com, tweaktown.com).
  • HotHardware says some users previously needed RTX Pro GPUs for the relevant AI and compute workloads (per hothardware.com).
  • Modders Inc's headline says a GPU unlock boosts RTX 30 BF16 AI performance by up to 80% on Linux (per modders-inc.com).

The brief

Modders have removed software-level limits on certain Tensor Core operations in NVIDIA GeForce RTX 20-, 30- and 40-series GPUs, HotHardware reports. The outlet says the changes increase throughput for relevant AI and compute workloads by roughly 1.8 to 2 times, while leaving the graphics hardware unchanged and requiring no new VBIOS.

The changes concern compute performance rather than gaming. HotHardware says they are not expected to raise in-game frame rates; the reported gains apply to workloads using the affected Tensor Core operations. The outlet describes the restrictions as artificial compute limits on consumer-oriented GeForce cards.

HotHardware says the modifications make the affected cards more capable for AI and compute tasks that users previously needed professional NVIDIA RTX GPUs to handle. It characterizes the increase as available at no extra cost, while tying that claim to use of the modifications rather than a hardware upgrade. The reported throughput gains range from about 1.8 times to 2 times.

TweakTown’s headline describes some Tensor Core throughput as nearly doubling, while TechPowerUp’s headline also reports compute performance gains of up to 2 times. Modders Inc’s headline describes a separate, more specific result: up to an 80% increase in BF16 AI performance on RTX 30 GPUs running Linux.

HotHardware identifies the RTX 20, RTX 30 and RTX 40 series as the generations involved and points readers to modification files for each. Its account does not identify a particular GPU model within each series or provide a separate performance figure for every generation.

The practical distinction in HotHardware’s account is between consumer GeForce GPUs and professional RTX GPUs: the modifications target compute tasks, not gaming performance, and do not replace or alter the underlying hardware.

The outlet directs users to the available files for the three series, while the cited performance claims remain workload-specific rather than a general measure of performance across all applications.

Why it matters

  • For users of NVIDIA GeForce RTX 20-, 30- and 40-series cards, the stated mechanism is removing software limits on certain Tensor Core operations; HotHardware reports roughly 1.8-to-2-times throughput for relevant AI and compute workloads.
  • HotHardware says the gains do not extend to in-game frame rates, distinguishing compute use from gaming performance.
  • HotHardware says users previously needed professional NVIDIA RTX GPUs for some of these compute workloads, and describes the modifications as requiring no extra cost.

What to watch next

  • Whether users apply the modification files HotHardware points to for RTX 20, RTX 30 and RTX 40 series cards.
  • Whether reported performance on specific Tensor Core workloads matches HotHardware’s roughly 1.8-to-2-times range.
  • Whether the RTX 30 Linux BF16 result described in Modders Inc’s headline reaches its stated 80% ceiling.

How it's being covered

Coverage analysisShow details
5 source articles · 5 unique outlets
Outlet political mix: Too few rated· 1 Center · 4 unrated
Where the outlets are based: Unknown 100%
Outlet political mix
Too few rated
Left
Center
Right
0 outlets
1 outlet
0 outlets
The marker shows coverage balance: each unique outlet counts once. 4 unrated outlets are not included.
Geography
Where the outlets covering this story are based.
5 unique outlets
Coverage imbalance
  • Coverage is currently concentrated among outlets based in Unknown (5 of 5 geographically-identified sources).
How we measure coverage

How outlets' accounts compare

Updat3 does not yet have enough full reports to compare how outlets framed this story.

Timeline

Developments from this story's sources, dated background we compiled and fact-checked, and our earlier coverage. Each row says which.
  1. NVIDIA announced its first CUDA-capable GPU architecture in 2006, establishing a software platform for using GPUs beyond graphics.
    Background · from our historical context
  2. Nvidia introduced Tensor Cores with its Volta architecture.
    Background · compiled and fact-checked by Updat3
  3. Nvidia announced its Turing architecture and GeForce RTX 20-series GPUs.
    Background · compiled and fact-checked by Updat3
  4. In 2020, NVIDIA also announced the Ampere-based A100 for data-center computing, while its professional RTX line served workstation users.
    Background · from our historical context
  5. Nvidia launched its A100 data-center GPU.
    Background · compiled and fact-checked by Updat3
  6. Nvidia announced its GeForce RTX 40 series, based on the Ada Lovelace architecture.
    Background · compiled and fact-checked by Updat3

How we got here

NVIDIA’s GeForce RTX 30 series, announced on September 1, 2020, brought Ampere architecture Tensor Cores to consumer graphics cards; NVIDIA described Ampere Tensor Cores as supporting new data formats, including bfloat16 (BF16).

The GeForce RTX 20 series had introduced Tensor Cores with Turing in September 2018, and the RTX 40 series followed with Ada Lovelace in September 2022. The distinction between GeForce and professional products predates these generations. NVIDIA announced its first CUDA-capable GPU architecture in 2006, establishing a software platform for using GPUs beyond graphics.

Original sources

5 articles from 5 outlets

Every story links to the reporting it was built from. How we choose sources

Spotted a factual error, a broken link or a misleading summary? Tell us.