Next-gen NVIDIA Ada Lovelace and AMD RDNA 3 GPUs reportedly more than twice as fast as Ampere and RDNA 2
Information regarding the next generation GPUs from NVIDIA and AMD has been announced by the Leaky Reliable, Kopite7kimi. Based on his latest tweets, it looks like AMD and NVIDIA will deliver double-digit performance gains with their next-gen GeForce RTX and Radeon RX graphics cards.
AMD and NVIDIA aim to deliver double-digit performance gains with next-gen Ada Love “ GeForce RTX ” and RDNA 3 “ Radeon RX ” GPU-powered graphics cards
In his tweet, Kopite7kimi reports that he received advice from his sources and although the Leaker has had an absolutely flawless record with leaks, we will still advise treating this information with a grain of salt since these products are still in the pipeline. more than a year. launch and performance estimates could change drastically once we get closer to that time point. That said, the release does mention performance estimates and some specs for the next-gen AMD and NVIDIA GPUs.
-5nm (no matter TSMC or SEC)
-AD102 in transition or GH202 in revolution?
-RDNA3, GFX11, how is AMD achieving its performance target? Double the SIMD in CU?
– kopite7kimi (@ kopite7kimi) May 22, 2021
Starting with the details, it is reported that next-gen GPUs will likely be based on a 5nm process node and use either TSMC or Samsung’s design. Key architectures that have been discussed include Ada Lovelace (AD102), Hopper (GH202), and RDNA 3 (Navi 3X) from NVIDIA. Based on the details, it looks like if Ada Lovelace will be a small architectural transition from Ampere, it will bring a massive performance boost while Hopper will be the revolution NVIDIA has been waiting for, both in terms of design (power / package). . ) and performance (raster / ray tracing).
Kopite7kimit believes that with the next generation cards we are likely to get up to 100 TFLOPs of FP32 computing power in a single graphics card. Sounds a bit too optimistic, but given that GPUs today rely on faster and more efficient compute models, 100 TFLOPs don’t seem so far-fetched and we might see a sparse model for consumer-specific chips to increase. considerably their computer performance. .
When it comes to performance, the NVIDIA Ampere GA102 architecture is used as the baseline. Compared to that, NVIDIA Ada Lovelace (AD102) chip would offer a 2.2x performance boost, while AMD’s RDNA 3 (Navi 3X) would offer a more than 2.5x performance increase.
There have been several rumors that the upcoming RDNA 3 GPUs will outperform anything NVIDIA has to offer in terms of rasterization performance. It looks like AMD will take the lead in offering the first MCM-powered GPUs as part of its Radeon RX line of graphics cards. But at the same time, NVIDIA is expected to move quickly to its Hopper MCM line of GPUs, which will deliver a performance improvement of over 3 times over Ampere GPUs. Here is the expected performance increase:
based on drafts, by my calculation only, the performance scale I guess:
NAVI31 (4 GCD)> 2.5X
I’m optimistic. But Jensen and Lisa need some time to sort out the issues, especially the bandwidth. Boring days, wait and see.
– kopite7kimi (@ kopite7kimi) May 22, 2021
One of the main bottlenecks for next-gen AMD and NVIDIA GPUs is stated to be memory bandwidth. GDDR6X won’t be able to cut it, so we might see a return to HBM or a new GDDR memory standard available around 2022-2023. GPU manufacturers are likely to invest more in near-GPU caches such as the Infinity Cache featured on the RDNA 2 graphics card to bypass bandwidth bottlenecks and reduce costs. Power consumption is also expected to increase dramatically and this generation was just a glimpse of what to expect from the next-gen line of graphics cards.
NVIDIA and AMD GPU road maps:
NVIDIA / AMD GPU Roadmap (Rumored)
|AMD (RDNA 1)||AMD (RDNA 2)||N / A||AMD (RDNA 3)||N / A||AMD (RDNA 4)?|
|NVIDIA (Turing Refresh)||NVIDIA Ampere||N / A||NVIDIA Ada Lovelace||N / A||NVIDIA Hopper|
Kopite7kimi also hinted at some spec details of NVIDIA Ada Lovelace chips some time ago which you can read more here and check out the specs in the table below:
NVIDIA Lovelace AD102 Rumored GPU Specifications
|Treat||5 nm||Samsung 8nm||TSMC 12nm NFF|
|Graphics Processing Clusters (GPC)||12||7||6|
|Texture Processing Clusters (TPC)||72||42||36|
|Streaming multiprocessors (SM)||144||84||72|
|Release||2022 (to be confirmed)||Sep 19||March 2017|
As for AMD’s Navi 31 GPU, KittyYYuko, which is also as reliable as Kopite7kimi in the GPU leak scene, has reportedly delivered over 2.5x performance already. As for Hopper, he would have been delayed in favor of Ada Lovelace some time ago, but again, it’s possible that we could see the AD10 * and GH20 * GPUs being introduced close to each other. , but NVIDIA’s roadmap cites that their “ The Ampere Next Next GPU won’t be here until 2024.