A mockup of the rumored RTX 3090 graphics card.
NVIDIA is said to be planning to build the next generation of RTX 3000 Ampere GPUs in Samsung’s 8nm process. This would be somewhat unexpected as NVIDIA’s announcement for Ampere has previously stated that they will use the 7nm TSMC process. However, the leak in question has an excellent RTX 3000 series leak history and had die settings inactive months before anyone else. That is why even though I am labeling this as a rumor simply because of the purely contrary nature of this article; Let’s say I wouldn’t be surprised if it turns out to be true.
NVIDIA plans to leverage Samsung’s 8nm process to build for RTX 3090, RTX 3080, and RTX 3070
NVIDIA has already used TSMC’s 7nm process for its high-end Ampere line, but if this rumor is true [opinion] and personally I think this is going to be true [/opinion] then things are about to get very interesting. The TSMC 7nm EUV process is more energy efficient than the Samsung 8nm process, but the latter is probably cheaper and has much more production capacity than the former. High-end TSMC processes are generally in high demand and AMD appears to have established itself there and dominates a very large supply of its wafers.
100% Samsung 8nm (developed from 10nm) https://t.co/jYqkSYfku7
– kopite7kimi (@ kopite7kimi) July 2, 2020
Going to Samsung could be the logical move here considering 1) that they have enough cash reserves to offer the process at low cost and 2) will have ample volume to meet NVIDIA’s demands. According to the tweet, Samsung’s 8nm process is based on the 10nm backbone and will be 10% more energy efficient than 10nm. 8nm will continue to be a massive leap in NVIDIA’s 12NFF performance (which is essentially an optimized 16nm version) and the company could very easily supplement the lower power efficiency by providing more power (that could explain TGP’s high ratings for the RTX. 3090 that we saw before).
Yes
– kopite7kimi (@ kopite7kimi) July 2, 2020
Initially, we had heard that NVIDIA might decide to do a staggered launch with some high performance parts (like the alleged RTX 3090) going to TSMC 7nm and lower end parts going with Samsung, but according to Kopite, all parts and segments will be based on Samsung’s 8nm process and they are 100% sure of it. This means that RTX 3090, RTX 3080, 3070, etc., all SKUs will be based on the 8nm process. Since the process is less efficient than TSMC’s NVIDIA, it would need to slightly increase TGP across the board to continue to take the massive leap in performance we’ve been seeing on leaked benchmarks (and boy, is that something to behold).
The source also previously stated that future NVIDIA Tegra processors will also be based on the same process. NVIDIA felt the brunt of its dependency on TSMC for the first time when AMD reserved its full 7nm capacity with its CPUs and GPUs. If I were in Jensen’s shoes, I’d like to also reduce dependency on the foundry and diversify some of that risk. That being said, the relocation of foundries is a * huge * decision that requires tons of R&D and EDA changes and not something to be done lightly. Much capacity has also been released at TSMC with Huawei leaving the stage and it remains to be seen if this information will be true.
At the end of the day, it really doesn’t matter if we get the RTX 3000 series on the TSMC 7nm or Samsung 8nm process, what matters is the performance and we hope they retain their performance leadership over AMD.
