Thursday, May 30th 2024
![NVIDIA](https://tpucdn.com/images/news/nvidia-v1723408558483.png)
NVIDIA to Stick to Monolithic GPU Dies for its GeForce "Blackwell" Generation
NVIDIA's GeForce "Blackwell" generation of gaming GPUs will stick to being traditional monolithic die chips. The company will not build its next generation of chips as either disaggregated devices, or multi-chip modules. Kopite7kimi, a reliable source with NVIDIA leaks, says that the largest GPU in the generation, the "GB202," is based on a physically monolithic design. The GB202 is expected to power the flagship GeForce RTX 5090 (or RTX 4090 successor), and if NVIDIA sticking to traditional chip design for this, then it's unlikely that smaller GPUs will be any different.
In contrast, AMD started building disaggregated devices with its current RDNA 3 generation, with its top two chips, the "Navi 31" and "Navi 32," being disaggregated chips. An interesting rumor suggests that team red's RDNA 4 generation will see a transition from disaggregated chips to multi-chip modules—packages that contain multiple fully-integrated GPU dies. Back to the green camp, and NVIDIA is expected to use an advanced 4 nm-class node for its GeForce "Blackwell" GPUs.
Sources:
kopite7kimi (Twitter), HXL (Twitter)
In contrast, AMD started building disaggregated devices with its current RDNA 3 generation, with its top two chips, the "Navi 31" and "Navi 32," being disaggregated chips. An interesting rumor suggests that team red's RDNA 4 generation will see a transition from disaggregated chips to multi-chip modules—packages that contain multiple fully-integrated GPU dies. Back to the green camp, and NVIDIA is expected to use an advanced 4 nm-class node for its GeForce "Blackwell" GPUs.
30 Comments on NVIDIA to Stick to Monolithic GPU Dies for its GeForce "Blackwell" Generation
The thing is, big chips are still expensive to make. Throw in the rumored 512 bit bus and GDDR7 and you have a ballpark for the upcoming pricing.
The grass is supposed to be GREENER on the other side, not covered in barbed wire. I dont understand this, how is the customer getting the shaft? If the 5090 is a step over a 4090, then its not getting shafted. Now pricing may do that, certainly. Besides, if the GB203 pulls 600w, why complain, since a full size chip would be flat out impossible to cool. I'm assuming, because if someone wants RT, buying AMD is a terrible idea and gets you, at best, ampere tier performance. That's a nice cope. In reality, the HBM equipped failure x was an embarrassment in the market, and the HBM equipped Vega 56 and 64 failed to compete with the cheaper GDDR equipped Pascal cards. HBM still exists, sure, in cards that cost more then my car and sometimes nearly as much as my house. But in consumer products? Nope. Somehow, 10 years later we still survive in the consumer world without HMB just fine.
If Nvlink can provide up to 1.8 TB/s communication, then it's still considerably slower than the communications in monolithic chips, as seen from the benchmarks of L0 and L1 caches which can go up to 7-8-9 TB/s.