Kopite7kimi has several new tidbits to share on NVIDIA's next-gen GeForce RTX 50 "Blackwell" GPUs and they may sound quite a bit different than the ones we heard in his previous posts. The RTX Ada lineup also saw similar rumors before its release which used to change almost every single day but there's a good reason why these specs change too often. NVIDIA, like all other GPU manufacturers, works on various configurations, SKUs, & models of a certain product before finalizing something. We have seen variants of the Ampere lineup which were rumored, never got to see the light of day, only to pop up in certain markets. We should keep these things in mind when looking at the next-gen lineup too.
Related Story NVIDIA GeForce RTX 50 “Blackwell” GB200 Gaming GPUs To Retain Same Memory Interface As Ada, 384-bit At Max
NVIDIA GeForce RTX 50 Gaming GPUs Get Flagship GB202 & GB203 Dies, Speculation on Mono or MCM Designs
So starting with the details, the NVIDIA GeForce RTX 50 "Blackwell" B200 Gaming GPUs will kick off with the flagship GB202 and GB203. It is said that the GB202 is the full chip whereas GB203 will be half of GB203, similar to how the Ada AD103 is half of the full AD102 GPU. It remains unclear if the green team will leverage a chiplet or monolithic design. For the HPC/AI family, the GB100 is going to utilize a chiplet design (MCM) but the gaming GPU is yet to be verified.
Blackwell Flagship Now Gunning For 512-bit Memory, Up To 32 GB & 1.8 TB/s Bandwidth Possible
Interestingly, Kopite7kimi now states that the flagship GB202 "Blackwell" Gaming GPU is now going to feature a 512-bit memory interface compared to the 384-bit memory interface on the flagship Ada GPU, the AD102. This can be a game-changer for the ultra-enthusiast segment as it not only offers higher memory bandwidth but it will also increase the memory capacity by a good amount. A 512-bit interface with 16 Gb (2 GB) modules will lead to 32 GB capacities with 1792.00 GB/s of total bandwidth.
And lastly, it looks like 28 Gbps GDDR7 memory modules will be the standard in the first gen products. Once again, NVIDIA is likely going to offer some variants with lower-clocked DRAM alongside the higher-end models which feature either 28 Gbps or the full 32 Gbps modules. If Kopite7kimi is correct, then that would change the memory hierarchy across all GeForce RTX 50 "Blackwell" GPUs quite significantly.
The website gametalkz.com is an aggregator of news from open sources. The source is indicated at the beginning and at the end of the announcement. You can send a complaint on the news if you find it unreliable.
A few weeks back, PGL announced its hardware of choice for its upcoming CS2 Major Tournament which included systems with AMD Ryzen 7 7800X3D CPUs and NVIDIA GeForce RTX 4080 GPUs but it looks like things didn't go as planned as a driver crash associated with the GPU became the very reason of one team's chances of going into the playoffs being washed away.
Intel has just released its latest MLPerf v4.0 performance figures covering the Gaudi 2 Accelerators & 5th Gen Xeon "Emerald Rapids" CPUs, with the former showcasing strong performance per dollar values against NVIDIA's H100 GPU.
NVIDIA has reportedly informed its partners that the supply of its existing GeForce RTX 40 "Ada" family will be reduced significantly moving forward as the company preps for RTX 50 "Blackwell" launch in the gaming segment.