User La Frite David (opens in new tab) posted a table on Twitter containing alleged specs for the next series of GeForce cards. Considering Nvidia has just been the victim of a massive data leak, it’s possible this information has come from this, but stay skeptical just in case. The RTX 4000 cards are anticipated not just for being the latest generation of flagship Nvidia GPUs but also for being the first to use the company’s new Lovelace architecture, after developing using Ampere architecture since 2020. However, given the codenames in the table start with “AD”, it’s possible that the new architecture will use the name “Ada”, the first name of computer scientist Ada Lovelace, rather than her last name. Looking at the table now, we see it contains multiple RTX 4000 models, compared to previous Turing and Ampere. The codenames given in the table match up with those used for previous generations, with the AD102 model presumably being the top model 4090 and the AD107 being the lowest spec, possibly an RTX 4050. Reading the top row, we see the most powerful RTX 4000 series GPU features 144 streaming multiprocessors (SMs) vs. 84 SMs on the top RTX 3000 GPU, used in the 3080 and 3090 cards. That’s a huge generational increase. There’s no mention of CUDA cores, but we’ve heard from previous sources that the RTX 4090 could have 18,432 CUDA cores, compared to 10,752 cores in the 3090 Ti. The increase in SMs for other cards is lesser, but still notable. We should see a huge performance boost from this alone, let alone other potential refinements like higher clock speeds or greater power efficiency. This power will require more electricity though. It’s expected the RTX 40 series GPUs will require a 1,200W PSU in order to provide the 480 - 600W necessary to power the card while keeping the rest of your PC running. As a result, it may make use of new 16-pin power connectors designed for PCIe 5.0 cards, as has been recently rumored (via Tom’s Hardware (opens in new tab)). Looking at all of this together makes for potentially one of the most dramatic increases in generation-to-generation performance ever seen on Nvidia GPUs. But however the new cards perform, we’re expecting a September release based on the current rumors. If we’re lucky, we may also see the release of the RTX 4000 series help end the current GPU stock nightmare thanks to Nvidia’s efforts in securing consistent manufacturing for the components. Next: Nvidia GeForce RTX 4090’s power could leave AMD in the dust.
title: “Geforce Rtx 4000 Leak Just Tipped How Much More Powerful It Could Be” ShowToc: true date: “2022-12-03” author: “Quinn Jackson”
User La Frite David (opens in new tab) posted a table on Twitter containing alleged specs for the next series of GeForce cards. Considering Nvidia has just been the victim of a massive data leak, it’s possible this information has come from this, but stay skeptical just in case. The RTX 4000 cards are anticipated not just for being the latest generation of flagship Nvidia GPUs but also for being the first to use the company’s new Lovelace architecture, after developing using Ampere architecture since 2020. However, given the codenames in the table start with “AD”, it’s possible that the new architecture will use the name “Ada”, the first name of computer scientist Ada Lovelace, rather than her last name. Looking at the table now, we see it contains multiple RTX 4000 models, compared to previous Turing and Ampere. The codenames given in the table match up with those used for previous generations, with the AD102 model presumably being the top model 4090 and the AD107 being the lowest spec, possibly an RTX 4050. Reading the top row, we see the most powerful RTX 4000 series GPU features 144 streaming multiprocessors (SMs) vs. 84 SMs on the top RTX 3000 GPU, used in the 3080 and 3090 cards. That’s a huge generational increase. There’s no mention of CUDA cores, but we’ve heard from previous sources that the RTX 4090 could have 18,432 CUDA cores, compared to 10,752 cores in the 3090 Ti. The increase in SMs for other cards is lesser, but still notable. We should see a huge performance boost from this alone, let alone other potential refinements like higher clock speeds or greater power efficiency. This power will require more electricity though. It’s expected the RTX 40 series GPUs will require a 1,200W PSU in order to provide the 480 - 600W necessary to power the card while keeping the rest of your PC running. As a result, it may make use of new 16-pin power connectors designed for PCIe 5.0 cards, as has been recently rumored (via Tom’s Hardware (opens in new tab)). Looking at all of this together makes for potentially one of the most dramatic increases in generation-to-generation performance ever seen on Nvidia GPUs. But however the new cards perform, we’re expecting a September release based on the current rumors. If we’re lucky, we may also see the release of the RTX 4000 series help end the current GPU stock nightmare thanks to Nvidia’s efforts in securing consistent manufacturing for the components. Next: Nvidia GeForce RTX 4090’s power could leave AMD in the dust.
