Looking for the latest in Nvidia RTX 50-series graphics cards? While GPUs remain elusive, leaks and rumors hint at another generational leap in performance.
The RTX 50 series of GPUs is inevitably coming. The “Blackwell” architecture was announced at GTC 2024, but for now it’s only the enterprise version.
As a result, much of the information remains speculative, as it relies on leaks provided by industry insiders Kopite7Kimi and Moore’s Law is Dead. Both of these have had relatively good accuracy rates in the past. But, until Team Green officially releases the specifications of the new GPUs, all of this should be taken with a grain of salt.
RTX 50 Series Specifications
The Blackwell architecture was officially announced by Nvidia at GTC 2024. While the chips aren’t the same as commercial gaming GPUs, the specs do at least provide some useful information. Combined with the latest leaks from Kopite7Kimi, it’s suggested that the flagship gaming GPU chip, the GB202, will use the same process node as the AI-focused GB101 chip.
Nvidia
Nvidia is also expected to deliver significant performance advancements, with the flagship GB202 gaming chip seeing significant improvements over the Ada Lovelace AD102 chips. In particular, the L1 cache will see significant performance gains.
Kopite7Kimi indicated (via (WCFTech) that the chip would offer 12 GPCs, with 8 TPCs each, for a total of 96 TPCs. Using the Ada Lovelace fabric as a model, we would get 192 SMs, which equates to 24,567 CUDA cores. This is a 33% increase over the Ada-equivalent AD102 chip.
Earlier leaks from Kopite7Kimi indicated that Nvidia would reuse the 384-bit memory bus of Ada chips for Blackwell chips, instead of moving to 512-bit. However, a report from May indicates that Nvidia will move to a 512-bit memory bus, at least on its flagship chips.
According to Moore’s Law, the performance improvement from the 40 Series to the 50 Series will not be as impressive as that of other generations.
Speculation over Blackwell GPU failure
The chips are expected to be cut into five different dies: GB202, GB203, GB205, GB206, and GB207. The GB202 is the flagship chip, and the GB207 is reserved for budget cards. Oddly, the GB204 is absent from the lineup. Historically, 204 chips have powered popular 70-suffix cards.
In September 2024, Benchlife confirmed that it was aware of five RTX 50 series card designs currently in the works, corresponding to the five chips we know of.
According to Kopite7Kimi, the GB203 chip will be “half the size of the GB202,” which would be the performance difference between the RTX 4090 and the RTX 4080. Based on that, and following the trends set by the 40 series.
It’s not yet clear whether Nvidia plans to stick with the monolithic structure for its next generation of gaming GPUs or use the MCM or multi-chip-module structure. AMD is already reportedly experimenting with chiplet configurations with its Radeon graphics cards, and many analysts expect GPU development to move in that direction to maintain improvements from successive generations.
A leak from RedGamingTech, later confirmed by Kopite7Kimi, suggested that Nvidia was developing a new Titan card using the Blackwell architecture. This would use the GB202 GPU and could offer up to 15% more performance than the RTX 5090. It is unclear at this point whether this card would make it to the consumer market.
As usual, until the specs are officially revealed, this should all be taken with a healthy dose of skepticism.
RTX 50 Memory Configuration
Once again, Kopite7Kimi is the source of information regarding the memory configuration for the RTX 50 series. This leak included the bit width of the memory bus, as well as the memory configuration and type.
As mentioned above, the GB202 is supposedly the processor that will power the flagship RTX 5090, which strengthens the likelihood that it will have the previously leaked 512-bit memory bus.
One surprise is the GB207, which is supposed to power the RTX 5060. 60-suffix cards have always been a popular option for consumers, but it looks like the GB207 will look a lot like the AD106 from the current-gen RTX 4060, without even making the jump to GGDR7 RAM.
Claimed power
By mid-July 2024, Everest on Twitter/X I noticed that an online tool called the Seasonic PSU Wattage Calculator has been updated to include data for the RTX 50 series of desktop graphics cards. This data may not be entirely accurate, but it does provide some interesting information, especially since it includes the RTX 5050.
Data published by Bench life expectancy in September 2024 suggests that the RTX 5090 will actually consume 550W of power, with the RTX 5080 being considerably less power hungry at a more modest 350W.
Kopite7Kimi disputes these figuresHowever. Instead, the RTX 5080 is pointed to as being a 400W card, with the RTX 5090 coming in at a mind-boggling 600W.
However, Kopite7Kimi did not specify whether these numbers represented total graphics power (TDP) or thermal design power (TDP), which is an important distinction when it comes to graphics cards.
Benchlife responded to Kimi’s information by saying that the 600W figure refers to the heat dissipation capacity of the special high-power cooling module for the RTX 5090.
Speculations on RTX 50 Series Features
Thanks to information provided once again by persistent leaker Kopite7Kimi, we can expect the RTX 50 series of graphics cards to feature support for Display Port 2.1, as well as HDMI 2.1. An upgrade from the Ada Lovelace series, which did not feature DisplayPort 2.1 support.
The power connector has proven to be a controversial topic. After reports of the 12VHPWR connector found on the RTX 4090 and 40 Super cards melting under certain circumstances, many consumers were hoping that a change would be made by Team Green.
Moore’s Law Is Dead recently revealed that Nvidia is planning to switch to an entirely new connector type for its next-gen cards. The leak suggests that Nvidia is working on an entirely new 16-pin connector, but this has yet to be confirmed by other sources.
If Nvidia sticks with the 12VHPWR connector, power consumption would be capped at 600W, although rumors that Nvidia plans to tout the efficiency of its new cards suggest it will be considerably lower than that limit.
RTX 50 Series Release Date Speculation
A new post on X/Twitter from Kopite has indicated that the Series 50 may not debut until CES 2025. That would put a serious dent in any hopes of seeing a 2024 release.
However, previous leaks courtesy of Moore’s Law is Dead suggest that the first RTX 50-series card could launch as early as Q4 2024. If these rumors are correct, the first of Blackwell’s gaming GPUs to hit the market will be the RTX 5090.
The anonymous source told Moore’s Law is Dead that the launch date would depend on sales of existing 40-series cards, particularly the refreshed 40 Super cards. The source goes on to say that Nvidia plans to promote the improved efficiency of the 50-series at CES 2025, with a launch to coincide with that.
Kopite7Kimi came back with a prediction that the RTX 50 series wouldn’t hit stores until 2025. Tech Outlet Red Gaming Technology The manufacturer also reiterated that the initial release date was Q4 2024, but some issue pushed the launch back. No direct reason has been identified behind this possible delay, although some commentators are pointing to a GDDR7 supply issue.
RTX 50 Series Pricing Rumors
At this point, there is very little information available on even tentative pricing for the Blackwell 50 series. The best guess for pricing is that the RTX 5090 will retail with an MSRP of $1,699 at launch.
Nvidia reportedly spent a whopping $10 billion to develop the Blackwell platform, although the company is expected to try to recoup much of that cost through sales of its dedicated AI chips.
This leaves many analysts and consumers concerned about the pricing structure Team Green will use for the 50-series. Especially since strong demand has sent the RTX 4090 well above its $1,600 launch price and well into the $2,000 range.
However, the 40 Super Refresh cards have adopted a fairly aggressive pricing strategy in an attempt to keep up with AMD. Only time will tell what position Nvidia will take.
