Nvidia GeForce: Faster RTX 3060 with GA103 full expansion appeared

0
89

Déjà vu: Once again, a GeForce RTX 3060 with modified specifications is doing the rounds, which promises more performance than the original, which was released in February 2021, with the full AD106 GPU and higher clock rates RTX 3060 (Review). But it's not a new release – and there's a catch anyway.

Another GeForce RTX 3060?!

< p class="p text-width">First, @T4CFantasy on Twitter drew attention to the GeForce RTX 3060 3840SP Edition graphics card that was added to TechPowerUp's GPU database. The model is again based on the GA103 graphics chip; In contrast to the original GeForce RTX 3060, however, the full configuration is used, which offers 30 instead of 28 streaming multiprocessors. As a result, the number of active FP32 execution units increases from 3,584 to 3,840 CUDA cores – and in this respect the GPU corresponds exactly to the mobile variants of the GeForce RTX 3060, i.e. the RTX 3060 Mobile and the RTX 3060 Max-Q.

However, because the clock rates have not been reduced from the original desktop model, but on the contrary have increased, the GeForce RTX 3060 3840SP offers around 12 percent more performance on paper than conventional RTX 3060 graphics cards and is therefore roughly at the level of one GeForce RTX 2070 Super (test), Radeon RX 6600 XT (test) or even Radeon RX 6650 XT (test), as a look at the editors' performance rating shows.

The GeForce RTX 3060 Ti (test), on the other hand, offers a total of 4,864 FP32-ALUs with 38 streaming multiprocessors and is around 15 percent ahead. In return, the power consumption of the Ti model is around 200 watts TDP, whereas the GeForce RTX 3060 3840SP is specified with the 170 watts TDP usual for the RTX 3060 – despite the wider chip and higher clock.

A possible cause can be found in the memory expansion: The GeForce RTX 3060 3840SP does not come with 12 GB GDDR6 graphics memory, but – like the mobile variants – only has half the amount of 6 GB GDDR6, which also has a lower speed of only 14 Gbps instead of 15 Gbps. Since it again stays with a 192-bit memory interface, the bandwidth only falls marginally from 360 GB/s to 336 GB/s and not, as with the GeForce RTX 3060 8 GB with a 128-bit interface, to a very low 240 GB /s.

OEM model camouflages itself as conventional RTX 3060

And now, of course, the question arises: What does Nvidia want to achieve with such an ambivalently mutated GeForce RTX 3060? In fact, it is – at least for now – not a separately sold graphics card, but an exclusive solution for OEM partners. The GeForce RTX 3060 3840SP has apparently been built into finished PCs for a long time, but is only recognized as an ordinary GeForce RTX 3060 in the graphics driver, according to TechPowerUp. However, why Nvidia is launching a faster version with half the memory for complete systems remains unanswered.

In any case, new editions and specialized models based on the GeForce RTX 3060 platform are Nvidia's particular passion. Most recently, the aforementioned GeForce RTX 3060 with a limited memory expansion and a GeForce RTX 3060 Ti with GDDR6X graphics memory were released in October 2022, which is up to 10 percent faster in 3DMark benchmarks thanks to the higher bandwidth. Special features are also the GeForce RTX 3060 Ti with a trimmed GA103 chip, which was previously used for the RTX 3080 Ti laptop GPU (test) or the GeForce RTX 3060 with a partially deactivated GA104 GPU, on which the GeForce RTX 3060 Ti is also based . The editors have compiled the specifications of all graphics cards in circulation based on the RTX 3060 platform in the table below.

Technical specifications of the RTX 3060 platform at a glance Nvidia GeForce RTX 3060 Ti Nvidia GeForce RTX 3060 Variant 3060 Ti GDDR6X 3060 Ti GA103 3060 Ti 3060 3840SP 3060 GA104 3060 3060 8GB 3060 Mobile 3060 Max-Q Release October 20202022 December 2020 February 2020 ? September 2021 February 2021 October 2022 January 2021 EIA ? 399 euros ? 329 euros ? – Architecture Ampere GPU GA104-202 GA103-200 GA104-200/202 GA106-3xx GA104-150 GA106-300 GA106-302 GA106-xxx process Samsung-Custom with 8 nm chip size 392 mm² 496 mm² 392 mm² 276 mm² 392 mm² 276 mm² SM 38 30 28 30 FP32 ALUs per SM 128 FP32 ALUs 4,864 3,840 3,584 3,840 RT cores 38, 2nd Gen 30, 2nd Gen 28, 2nd Gen 30, 2nd Gen RT cores + ALUs synchronous Yes Tensor cores 152, 3rd Gen 120, 3rd Gen 112, 3rd Gen 120, 3rd Gen Base Clock 1410MHz 1627MHz 1320MHz 900MHz 817MHz Boost Clock 1665MHz 1852MHz 1777MHz 1425MHz 1282MHz FP32 Performance 16.2 TFLOPS 14, 2 TFLOPS 12.7 TFLOPS 10.9 TFLOPS 9.8 TFLOPS FP16 Performance 16.2 TFLOPS 14.2 TFLOPS 12.7 TFLOPS 10.9 TFLOPS 9.8 TFLOPS Texture Units 152 120 112 120 ROPs 80 48 Memory 8GB GDDR6X 8GB GDDR6 6 GB GDDR6 12 GB GDDR6 8 GB GDDR6 6 GB GDDR6 Memory speed 19 Gbps 14 Gbps 15 Gbps 14 Gbps 12 Gbps Memory interface 256 bit 192 bit 128 bit 192 bit Memory bandwidth 608 GB/s 448 GB/s 336 GB/s 360 GB/s 240 GB/s 336 GB/s L2-Approx che 4 MB 3 MB TDP 225 watts 200 watts 170 watts 80 watts 60 watts Slot connection PCIe 4.0

The most popular ampere graphics cards live on

The GeForce RTX 3060 and GeForce RTX 3060 Ti are the two best-selling graphics cards of the Ampere generation, which at the beginning of 2023 still serve the mid-range segment that Nvidia has not yet occupied with GeForce RTX 4000: graphics cards based on Nvidia Ada Lovelace or AMD RDNA 3 are currently only an addition to the top, both in terms of price and performance. Although the GeForce RTX 4070 and GeForce RTX 4060 Ti will be coming soon, these graphics cards will also be positioned above the 3060 platform – and this is probably far from the end.

Future Nvidia graphics cards rumors overview

  • Nvidia graphics cards rumours: Images show alleged RTX 4090 Ti or Titan Ada
  • RTX 4070, 4060 Ti & 4090 Ti: Rumors speak of two more AD104 graphics cards
  • Nvidia graphics cards rumors: GeForce RTX 4070 Ti, 4070 and 4060 Ti become more concrete
  • Nvidia GeForce RTX 4060 Ti: Rumors speak 3070 performance for 500 USD
  • GeForce RTX 4060: Rumors circulating have no basis