Leak Lounge | Blackwell goes mainstream: What to expect from the upcoming Nvidia RTX 50-series gaming GPUs
It's an understatement to say that PC enthusiasts around the globe are on tenterhooks, eagerly waiting for what Nvidia has in store for the upcoming RTX 50-series gaming GPUs based on the "Blackwell" architecture. Luckily, we don't have much waiting left to do, considering that its launch is just around the corner at CES 2025 next month. Unsurprisingly, the RTX 50 series has been the epicenter of a plethora of leaks and rumors in the past few months. Here's a roundup of what to expect.
RTX 5090 and RTX 5080: The crown jewels
It is no doubt that the RTX 5090 and RTX 5080 will receive the most attention, being the highest-end GPUs of the lineup. That said, the gap between the RTX 5090 and the RTX 5080 appears to be far more massive than the gap between the RTX 5080 and the RTX 5070. This time around, the RTX 5090, which will utilize the GB202 chip, is expected to boast a whopping 21,760 CUDA cores (+33%), along with 32 GB of VRAM based on the shiny new GDDR7 spec on a 512-bit memory bus for a total bandwidth of a whopping 1.792 TB/s. This monster of a GPU is also expected to boast a massive 575-watt TBP.
The RTX 5080, on the other hand, is no slouch but sits a step down the ladder, with around 10,752 CUDA cores 'only', which is roughly a 10.5% increase compared to its predecessor. The RTX 5080 will also sport half the memory of the RTX 5090, at 16 GB, also based on the GDDR7 spec, but on a 256-bit bus allowing for a memory bandwidth of 960 GB/s. The RTX 5080 is expected to boast a TBP of around 360 watts, which is a substantial 40 watts higher than its predecessor.
RTX 5070, RTX 5060 Ti, and RTX 5060: Mainstream machines
Coming to the more mainstream and relatively affordable cards, we first have the RTX 5070 Ti, which is expected to launch alongside its more powerful siblings. This graphics card will utilize a cut-down version of the RTX 5080's GB203 chip, dubbed the GB203-300 with around 8,960 CUDA cores (+16.7%) along with 16 GB of GDDR7 VRAM on a 256-bit memory bus, as well as a 300-watt TBP. Its non-Ti sibling, on the other hand, will sport 6,144 CUDA cores (+6.1%) paired with 12 GB of GDDR7 VRAM on a slower 192-bit bus, and a 250-watt TBP.
The RTX 5060 Ti, interestingly, is expected to sport 16 GB of GDDR7 VRAM on a slower 128-bit bus, which might make it a better buy than the RTX 5070 for memory-intensive titles that will benefit from the extra 4 GB of memory. The vanilla RTX 5060, however, will have to make do with only 8 GB of GDDR7 memory, also on a 128-bit bus. For those interested in future-proofing their systems, the RTX 5060 Ti certainly seems like a far better option.
RTX 50-series Laptop GPUs: Portable powerhouses
When it comes to the RTX 50-series Laptop GPUs, the rumors are not nearly as abundant as those for their desktop counterparts. That said, we do have some nuggets surrounding the RTX 50-series Laptop GPUs, courtesy of a recent leak from a reliable source. According to the information, which contradicts a previously leaked listing, the RTX 5090 Laptop will sport only 24 GB of VRAM courtesy of 3 GB GDDR7 modules - 8 GB less than that of its desktop counterpart. The RTX 5080 Laptop might sport 16 GB of memory, and the rest of the Laptop GPUs will only get 8 GB of GDDR7 memory.
A recent report, courtesy of the YouTube channel Moore's Law is Dead, has stated that the RTX 5080 Laptop will be around 50% faster than its predecessor, which is rather optimistic to say the least, considering that the RTX 5080 Laptop is only expected to sport around 7,680 CUDA cores, which is only around 3.4% higher than its predecessor. MLID also expects an RTX 5090 Laptop variant with 32 GB of VRAM to launch down the line, although that is not certain as of now. With the recent inputs, this does seem rather unlikely. The RTX 5090 and 5080 Laptop-powered notebooks are expected to drop in February of this year, with the rest of the lineup following soon after in March.
AI-driven rendering wizardry
Apart from the raw core counts and memory configurations alone, the RTX 50-series GPUs are also expected to pack a few more tricks up their sleeves. For starters, the RTX 50-series is expected to feature a more advanced version of DLSS, likely introducing DLSS 4.0. Improvements to ray-tracing performance are also expected courtesy of improved RT cores, along with new Neural Rendering technologies that allow frames to be rendered using neural networks instead of the traditional 3D pipeline.
As mentioned previously, with CES 2025 less than a month away, we will soon find out for ourselves what the multi-trillion dollar gaming and AI giant has in store for its loyal fan base. Considering Nvidia's ever-increasing prices, now might be a good time to break open the piggy bank for those who want a piece of the Blackwell pie.
