Nvidia shows signs in [2023]

Status
Not open for further replies.
Do you remember when the Game Developers Conference was about games instead of industrial applications and ai?
 
NVIDIA today announced a breakthrough that brings accelerated computing to the field of computational lithography, enabling semiconductor leaders like ASML, TSMC and Synopsys to accelerate the design and manufacturing of next-generation chips, just as current production processes are nearing the limits of what physics makes possible.
...
The new NVIDIA cuLitho software library for computational lithography is being integrated by TSMC, the world’s leading foundry, as well as electronic design automation leader Synopsys into their software, manufacturing processes and systems for the latest-generation NVIDIA Hopper™ architecture GPUs. Equipment maker ASML is working closely with NVIDIA on GPUs and cuLitho, and is planning to integrate support for GPUs into all of its computational lithography software products.

Running on GPUs, cuLitho delivers a performance leap of up to 40x beyond current lithography — the process of creating patterns on a silicon wafer — accelerating the massive computational workloads that currently consume tens of billions of CPU hours every year.

It enables 500 NVIDIA DGX H100 systems to achieve the work of 40,000 CPU systems, running all parts of the computational lithography process in parallel, helping reduce power needs and potential environmental impact.
 
The H100 NVL has a full 6144-bit memory interface (1024-bit for each HBM3 stack) and memory speed up to 5.1 Gbps. This means that the maximum throughput is 7.8GB/s, more than twice as much as the H100 SXM. Large Language Models require large buffers and higher bandwidth will certainly have an impact as well.
NVIDIA-H100-NVL-1200x754.png
 

NVIDIA-H100-NVL-1200x754.png

Blimey. I could use a couple of those. Though I really don't want to ask the price (H100 PCIe is £32k list price here).
 
NVIDIA has made a quiet yet significant update to its consumer GeForce graphics processing units (GPUs) by removing video encoding limitations. These GPUs can now encode up to five simultaneous streams, an increase from the previous limitation of three NVENC encodes.

The new change applies to certain GeForce GPUs based on Maxwell 2nd Gen, Pascal, Turing, Ampere, and Ada Lovelace GPU architectures released over the last eight years. However, there is still a limitation on the number of streams that can be encoded, depending on the resolution of the stream and the codec used.
 
The RTX 4070 pricing is to match RTX 3070 Ti at launch. Interestingly, this price has not changed much, even 1.5 years after launch (June 2021). The RTX 3070 Ti and RTX 3060 Ti SKUs were among the most popular Ampere cards.

In terms of specs, the card will feature just as many cores as the RTX 3070 (5888 CUDA), but the memory will be upgraded to GDDR6X technology, and it will have 4GB more capacity. Despite a narrower memory bus (192-bit), it will still have higher maximum bandwidth thanks to 21 Gbps modules.
 

That's a lot more reasonable. Still higher than most would like I think, but I expect it to sell well at this price given it's both cheaper than the original 3080 MSRP and will likely be as fast or a bit faster with more VRAM and all the ADA features.
 
Not so much reasonable but that it actually makes sense given the rumored specs. The $750 price rumor (without even questioning the source) just never made sense from a product stand point. The 4070ti is readily available in stock at MSRP, the rumored specs for the 4070 at $750 would not have been remotely viable from a product stand point.
 
Status
Not open for further replies.
Back
Top