According to rumors using the TSMC 4N process, NVIDIA Ada Lovelace processors have a crossover advantage over AMD RDNA 3.
By NVIDIA Etta Lovelace handlers The next-generation GeForce RTX40 gaming graphics cards will have an edge over AMD’s RDNA3, as previously announced. Moore’s law is dead.
NVIDIA Ada Lovelace GPUs use the TSMC 4N processing terminal, which offers little advantage over AMD’s RDNA 3 GPUs.
From what we know so far, it’s Nvidia Anticipated Use the TSMC 5nm processor terminal for Ada Loveless GPUs running next-generation gaming graphics cards known as the GeForce RTX40 Series. According to a recent rumor, NVIDIA Ada Lovelace GPUs will be based on the TSMC 4N process terminal.
Note that I did not specify PCIe 5.0 on my Loveless chipset: https://t.co/n7sdIL1Uku
Speaking of “4”, #Nvidia Lovelace is actually 4 nanometers!
சொல்ல Tell us more about the next broken silicone … https://t.co/jjJ1GSod0j
– Morslavist law April 24, 2022
Yes, the same TSMC 4N process node just runs a coil Huber GPUs To the HPC market for data centers. As far as we know about the TSNC 4N process terminal, this is an update of the 5nm process (not to be confused with the 4nm / N4, which is a completely different terminal). The TSMC 4N process node is specifically designed for Nvidia and offers various enhancements that allow for improved power efficiency, performance and smaller density enhancements compared to the 5nm vanilla TSNC node.
The reasons why NVIDIA chose TSMC’s 4N as a candidate for the next generation of Matrix GPUs are clear. The following cards There will really be greed for power NVIDIA and company will work with the 4N process node to improve it as much as possible. AMD, on the other hand, uses a combination of TSMC It was cut with 5 nm and 6 nm process Based on upcoming MCM and monolithic GPUs RDNA Graphics Engineering3 Although they do not bring the improvements provided by 4N, they do offer MCM approach, which is expected to be very effective.
So at the end of the day, Nvidia gets a better edge, while AMD offers a better design approach. At the end of the day, these may not be important for end users who want to play their games only on the best hardware (graphics cards).
NVIDIA CUDA GPU (Rumors):
| GPU | TU102 | GA102 | M102 |
|---|---|---|---|
| Flag SKU | RTX 2080 Ti | RTX 3090 Ti | RTX 4090? |
| Architectural Engineering | Touring | Ampere | Ada Lovelace |
| Treatment | TSMC 12nm NFF | Samsung 8nm | TSMC 4N? |
| The amount of death | 754 mm | 628 mm | ~ 600 mm |
| Graphics Processing Groups (GPC) | 6 | 7 | 12 |
| Textile Processing Tools (TPC) | 36 | 42 | 72 |
| Multiprocessor flow (SM) | 72 | 84 | 144 |
| CUDA cores | 4608 | 10752 | 18432 |
| L2. Temporary storage | 6 MB | 6 MB | 96 MB |
| Theoretical TFLOPs | 16 TFLOPs | 40 TFLOPs | ~ 90 TFLOPs? |
| Memory type | GDDR6 | GDDR6X | GDDR6X |
| Memory capacity | 11GB (2080D) | 24 GB (3090 Ti) | 24 GB (4090?) |
| Memory speed | 14 Gbps | 21 Gbps | 24 Gbps? |
| Memory bandwidth | 616 GB / sec | 1,008 GB / sec | 1152 GB / V? |
| Memory bus | 384 bit | 384 bit | 384 bit |
| PCIe interface | PCIe General 3.0 | PCIe General 4.0 | PCIe General 4.0 |
| DGP | 250 watts | 350 watts | 600 watts? |
| Release | September 2018 | September 20 | 2H 2022 (TBC) |
“A social media addict. Zombie fanatic. Loves to travel. Who is obsessed with music. Bacon expert.
“Professional coffee fan. Total beer nerd. Hardcore reader. Alcohol fanatic. Evil twitter buff. Friendly tv scholar.”
