NVidia's growing opportunity in litho (serves whole chip industry) - cuLitho

At GTC 2023, Nvidia announced its new cuLitho software library for speeding up a critical bottleneck in the semiconductor manufacturing workflow. The new library speeds computational lithography, a technique used to create photomasks for chip production. Nvidia claims its new approach enables 500 DGX H100 systems wielding 4,000 Hopper GPUs to do the same amount of work as 40,000 CPU-based servers, but do so 40X faster and with 9X less power. Nvidia claims this reduces the computational lithography workload to produce a photomask from several weeks down to eight hours.

Chipmaking leaders TSMC, ASML, and Synopsys have all signed on for the new tech, with Synopys already integrating it into its software design tools. Over time, Nvidia expects the new approach to enable higher chip density and yield, better design rules, and AI-powered lithography.

Nvidia scientists created new algorithms that allow increasingly-complex computational lithography workflows to execute on GPUs in parallel, exhibiting a 40X speedup using Hopper GPUs. The new algorithms are integrated into a new cuLitho acceleration library that can be integrated into mask makers’ software (typically a foundry or a chip designer). The cuLitho acceleration library is also compatible with Ampere and Volta GPUs, though Hopper is the fastest solution.

New techniques have emerged that now allow etching features smaller than the wavelength of the light used to create them. However, the continued shrinkage of the features has led to issues with diffraction, which essentially ‘blurs’ the design that’s being printed onto the silicon. The field of computational lithography counteracts the impact of diffraction through complex mathematical operations that optimize the mask layout. However, this task is becoming increasingly compute-intensive as features shrink even further, thus enabling billions more transistors per design.

These complex problems require large clusters of computers, often numbering tens of thousands of servers (Nvidia cites 40,000), that crunch through the numbers in parallel on CPUs in a workload that can take up to weeks to process a single photomask (the amount of time varies based on chip complexity — Intel says it takes its team five days to create a single mask).

Nvidia contends that the number of servers required to design a modern mask is increasing at the same rate as Moore’s Law, thus pushing the server requirements and the amount of power needed to operate them into unsustainable territory. In fact, the incredible compute requirements for new mask tech, like Inverse Lithography Technology (ILT) which uses Inverse Curvilinear Masks (ILM), has already hampered the adoption of these more advanced techniques. Additionally, High-NA EUV and ILT are expected to increase the amount of data processing for masks by 10X in the coming years.

That’s where Nvidia’s cuLitho steps in, reducing the computational lithography workload to eight hours. The cuLitho library can be integrated into computational lithography software that leverages ILT (curvilinear shapes) or Optical Proximity Correction (OCP, which uses ‘Manhattan’ shapes) techniques, and is already integrated into Synopsys’ tools. TSMC and ASML are also adopting the tech. Given the sensitivity of these sorts of software, US export controls will govern any distribution of the software to China and other regions subject to sanctions.

Intel has long used its own proprietary software tools but is slowly shifting to adopting industry-standard tools, particularly as it begins implementing its own external IDM 2.0 foundry operations. As such, it is yet to be seen if other big fabs, like Intel and Samsung, will adopt the new software for their own internal tools. Regardless, the support from Synopsys, ASML, and TSMC assures broad uptake of the cuLitho library and Nvidia’s GPU-based solutions with leading semiconductor manufacturers over the coming years.


This will be great for all semi companies.

It does show how broad NVIDIA software in the AI world is, and how difficult it is going to be for either AMD or INTC to break into their almost monopoly on this.

This doesn’t have anything to do with AI. This is software to do the computations needed for ILT in parallel instead of multiple serial threads. nVidia has a new technology in Hopper, Thread Block Clusters, which is apparently the secret sauce that makes cuLitho so fast.

Laying out masks to make 2 nm chips is a minor part of the global demand for CPU and GPU cycles. But soon AMD, Intel, and Samsung, in addition to nVidia, will need those masks and TSMC technology to make their chips. I hope the Chinese government is smart enough to realize that invading Taiwan will cause China to fall apart faster than the invasion of Ukraine is destroying Russia. (Even if Putin reversed course right now, I don’t think he could hold Russia together. If he doesn’t change course and remains in charge through the end of this year, look for Kyiv to become the new Russian capital next year or at least for Ukrainian tanks to reach Moscow.)


Thanks for the correction. As we move very off topic, the speculation about the Ukrainian spring offensive is interesting. Personally, I think moving into Crimea is the most likely scenario, followed by moving into Russia to take out a few military bases closer to the Ukrainian border, and your thoughts on a direct assault on Moscow as a distant third.

With the more modern hardware they are getting, I think they could be in Moscow in about 4 hours… which might be quick enough to catch them off guard. It will be interesting to see how it unfolds.

As for Putin @eachus , I don’t see at all how he moves forward from here either. This business of moving nukes to Belarus on the one hand and China meeting with the Central Asian republics without Russia on the other makes me think China is not going to be much in his corner except insofar as they think limited support through some indirect channel could help drain Western arsenals without causing much blowback. (Is China feeding weapons to Iran or DPRK, so those two can feed weapons to Russia? Dunno. Wild-eyed idea.)

Ukraine can’t hold Moscow if they manage to attack it. And I don’t think they will have the international community on their side, either. While there isn’t much will to take Ukraine, I think Russians finding themselves facing Ukrainian ground attack on Moscow – or any attack on Moscow-- would find their patriotism and it would get ugly.

Will they try to retake Crimea? Perhaps. Seems like a reasonable bet. Might they do something to Russian forces on Russian soil as a way to destabilize Russia and maybe make the Crimea liberation easier? My ignorant wild-eyed guess says maybe.

What does this mean for Taiwan? China, whatever they do about Taiwan, won’t be as ham handed as Russia has been.

1 Like

Did this speculation about Ukraine invading Russia come from some news source you listen to? Seems crazy to me. I don’t see why the Ukrainians would have any desire to do that, nor would the Russians ever accept it. They didn’t want this war, they just want Russia to get off their land (and maybe Putin to be deposed). And all Western material support is for the defense of Ukraine. It would disappear in an instant if it were used for aggression into Russia.

1 Like

Ukraine has already used their “own stuff” to attack targets inside Russia. I agree they can only use NATO hardware inside Ukraine (which includes Crimea), but once they are at the border it gets easier to attack the Russian military based on the Russian side of the border.

IF Ukraine were to attempt an assault on Moscow it would be over before the west could react, but I suspect dire consequences for both Russia and Ukraine if they make such a move. I think Russians need to remove Putin, rather than having Ukraine do it.

I don’t expect Ukraine to invade Russia. However, if Russia falls apart as it did at the end of WWI, all bets are off. (Did you know that the US sent troops to Russia back then. The Forgotten Story of the American Troops Who Got Caught Up in the Russian Civil War | History| Smithsonian Magazine )

Hi bob

here’s a video talking about Zen 5/6 with a lot of possible inside information on 4nm and 3nm. There’s a lot of technical info and I was wondering if you would review this and let us know what you think. Thanks…doc

AMD Zen 5 Full Leak: Core Counts, IPC, Clocks, Sorano, Turin, and even Zen 6 Venice Details! - YouTube

edit: he speaks about Turin vs Granite Rapids and the problems possible in the future out a year or two. Its a good well done thoughtful explanation of the next 12-24 months IMHO

1 Like