In a new tweet, Kopite7kimi has stated that NVIDIA’s GeForce RTX 4060 graphics cards may not be very fast & could be utilizing an x8 PCI interface.
NVIDIA GeForce RTX 4060 Series Reportedly Not Getting The Same Uplifts As Higher-End Ada Lovelace GPUs
The leaker says that the performance of the NVIDIA AD106 GPU, which should be powering the RTX 4060 series graphics cards, is not very strong. We don’t know which specific configuration is talked about but it is likely that he’s referring to the full-fat die than a cut-down model. An alleged 3DMark Time Spy Extreme benchmark has been shared & the GPU is said to score less than 7000 points. A score like this should put the AD106 GPU slower than the RTX 3070 Ti and on par with the RTX 3070 which doesn’t seem like the huge upgrade that the higher-end AD102 and AD103 GPUs are getting.
A Time Spy Extreme score between 6500-7000 should put the AD106 GPU (GeForce RTX 4060) up to 18% faster than the RTX 3060 Ti and 69% faster than the RTX 3060. Now, an over 50% performance boost is pretty huge but it’s not as big as the 2x jumps that we have seen rumored for the high-end cards. If there’s a 50-60% jump in synthetic benchmarks, then gaming would see a 30-40% boost at best. So it looks like we will mostly be getting RTX 3070 levels of performance in the next **60 card.
AD106 is not very strong. The typical TSE score is < 7000.
And both AD106 and AD107 are using PCIE x8.
— kopite7kimi (@kopite7kimi) September 10, 2022
Furthermore, the leaker also mentions that the AD106 and AD107 “Ada Lovelace” GPUs will utilize the PCIe x8 interface. Considering that these GPUs are mostly targeted at the mobility segment (Laptops/Notebooks), it makes sense. For desktops, this means that the GeForce RTX 4060 & 4050 will be getting an x8 link which shouldn’t be an issue for most PCs considering that next-gen systems from AMD & Intel come with Gen 5 PCIe slot support but running on an older Gen 3 platform can lead to issues related to lower bandwidth and transfer speeds.
NVIDIA GeForce RTX 4060 ‘Expected’ Specifications
Based on NVIDIA’s decision to do a mix of AD104/AD106 or go AD106 across its RTX 4060 series lineup we can either see the RTX 4060 cut-down or one with a full configuration while the Ti can use the AD104 cut-down configuration. The AD106 GPU will come packed with 32 MB of L2 cache and up to 48 ROPs.
The clock speeds are not confirmed yet but considering that the TSMC 4N process is being used, we are expecting clocks between the 2.0-3.0 GHz range. The higher than usual clock speed bump comes from the fact that NVIDIA is making a two-node jump considering the Ampere GPUs with Samsung 8nm node was in reality a 10nm process node with some optimizations. NVIDIA is skipping 7nm and going straight for a 5nm node and not even the vanilla variant but an optimized version of it. With Pascal on the TSMC 16nm node, NVIDIA delivered a huge frequency leap and we can expect a similar jump this time around too.
As for memory specs, the GeForce RTX 4060 is expected to rock 8 GB GDDR6X capacities that might come at faster 20+ Gbps speeds across a 128-bit bus interface for over 320 GB/s of bandwidth. The ‘Ti’ variant, if it ends up with an AD104 GPU, could offer up to 12 GB of memory across a 192-bit bus interface. The GeForce RTX 4060 graphics card is also said to rock a TGP of 220W which is an increase of 30% over the RTX 3060 and a 10% increase over the RTX 3060 Ti. This is a massive TGP increase and the performance needs to be really good for NVIDIA to keep its efficiency numbers up.
NVIDIA GeForce RTX 4060 Series Preliminary Specs:
|Graphics Card Name||NVIDIA GeForce RTX 4060 Ti||NVIDIA GeForce RTX 4060||NVIDIA GeForce RTX 3060 Ti||NVIDIA GeForce RTX 3060|
|GPU Name||Ada Lovelace AD104?||Ada Lovelace AD106||Ampere GA104-200||Ampere GA106-300|
|Process Node||TSCM 4N||TSCM 4N||Samsung 8nm||Samsung 8nm|
|TMUs / ROPs||TBD / 64||TBD / 48||152 / 80||112 / 64|
|Tensor / RT Colors||TBD||TBD||152 / 38||112 / 28|
|Base Clock||TBD||TBD||1410 MHz||1320 MHz|
|Boost Clock||2.2-2.6 GHz||2.2-2.6 GHz||1665 MHz||1780 MHz|
|FP32 Compute||~24 TFLOPS||~21 TFLOPS||16 TFLOPS||13 TFLOPS|
|RT TFLOPS||TBD||TBD||32 TFLOPS||25 TFLOPS|
|Tensor-TOPs||TBD||TBD||192 TOPs||101 TOPs|
|memory capacity||12GB GDDR6?||8GB GDDR6?||8 GB GDDR6||12 GB GDDR6|
|memory speed||TBD||TBD||14 Gbps||16 Gbps|
|bandwidth||>448 GB/s||>320 GB/s||448 Gbps||384 Gbps|
|Price (MSRP / FE)||~$399 US||~$329 US||$399 US||$329 US|
|Launch (Availability)||1H 2023||1H 2023||2nd December 2020||25th February 2021|
Products mentioned in this post