Clik here to view.
Nvidia CEO Jensen Huang calls Palladium the only appliance more important to him than the refrigerator. At Cadence Design Systems annual event in Santa Clara, California, he also acknowledged that Nvidia has the largest installation of Palladium emulation systems.
Earlier, during a fireside chat, Huang said that Blackwell AI processor would not exist without Palladium. So, what’s Palladium and why is it making waves for large and powerful chip designs? It’s an emulation tool built around Cadence’s custom processors, and it’s used for pre-silicon hardware debugging.
Image may be NSFW.
Clik here to view.
Figure 1 Palladium Z3 and Protium X3 deliver fast pre-silicon verification and validation of the large, complex chip designs. Source: Cadence
At CadenceLive, held on 17 April 2024, the EDA toolmaker unveiled Palladium Z3 alongside Protium X3. “The supercharged Palladium Z3 and Protium X3 are built to deliver fast pre-silicon verification and validation of the largest and most complex devices,” said Dhiraj Goswami, corporate VP of hardware system verification R&D at Cadence.
Palladium Z3, which offers approximately 1.5 times the performance of its predecessor Palladium Z2, can scale from 16 million gates to all the way 48 billion gates. It also features specialized apps for tasks such as 4-state emulation, mixed-signal emulation, safety emulation, and fine-grained power analysis.
Next, the Protium X3 system, built around AMD Epyc processors paired to AMD Versal Premium VP1902 adaptive system-on-chips (SoCs), provides physical prototyping to accelerate bring-up times for pre-silicon software validation of complex, multi-billion gate chip designs. It’s also 1.5 times faster than its predecessor, Protium X2.
Palladium and Protium create a virtual version of a chip to start writing software while waiting for the physical chip to return from the fab. That’s how chip design emulation accelerates time to market. However, though Palladium Z3 and Protium X3 work in tandem, as explained above, they facilitate different types of workloads.
Image may be NSFW.
Clik here to view.
Figure 2 Palladium Z3 and Protium X3 feature a unified compiler and common virtual and physical interfaces. Source: Cadence
Nvidia, which used Palladium Z3 and Protium X3 predecessors in designing just-announced Blackwell AI processors, is already testing these upgraded systems in some of its AI processor designs. “The next-generation Palladium and Protium systems push the boundaries of capacity and performance to help enable a new era of generative AI computing,” said Scot Schultz, senior director for networking at Nvidia.
Today’s large chip designs serving applications like AI and high-performance computing (HPC) increasingly demand emulation solutions that offer higher performance along with faster and more predictable compile debug. New emulation systems such as Palladium fill the need with early software development, hardware-software verification, and debug tasks.
Related Content
- Hardware emulation for everyone
- On-chip debug eases MCU-based design
- Design Compilation in Hardware Emulators
- Maximizing hardware emulation’s value for networking designs
The post Palladium emulation: Nvidia’s Jensen Huang is a fan appeared first on EDN.
