Bolt Graphics claims 13x leap over Nvidia RTX 5090 with less wattage – and huge caveats

Add a LARGE GRAIN OF SALT: The GPU market is hotter than ever, fueled by gaming, AI, and data center demand. But real competition to Nvidia remains scarce. That's where Bolt Graphics comes in, pitching its new Zeus architecture as a disruptor. The company claims it can beat Nvidia's flagship cards in raw performance, but a closer look suggests Zeus might be more smoke than silicon, possibly repackaging server-class hardware in gamer-friendly clothing.

California-based startup Bolt Graphics recently revealed detailed specifications for its upcoming graphics cards, claiming they outperform the Nvidia RTX 5090 in path tracing workloads. While the Zeus lineup appears impressive on paper, several details raise red flags.

The company first unveiled its ambitious plans in March, boasting enormous memory pools and graphics processors that are, according to them, orders of magnitude faster than Nvidia's.

A newly released spec sheet takes these claims even further – but Bolt's vertical TFLOP numbers appear suspiciously low, and the company opting for older VRAM formats cast doubt on the gaming credentials of these GPUs.

The specifications for the company's four planned GPUs: the Zeus 1c-032, 2c-064, 2c-128, and 4c-256, appear to outperform the RTX 5080, 4090, and even 5090 in several areas, all while consuming less than half the power.

For instance, the weakest entry-level "1c-032" is said to achieve 77 gigarays in path tracing compared to the RTX 5090's 32. It also includes 128 MB of cache, and Bolt is attempting to revive socketable VRAM, enabling configurations with up to 160 GB of memory.

All of this is achieved with a power draw of just 120W, lower than even the RTX 5050. In a subtle jab at Nvidia, Bolt emphasized that Zeus uses standard 8-pin power connectors, which it notes are "known to not melt."

Click to enlarge

The higher-end Zeus models scale up the 1c's specs significantly, increasing cache, VRAM, gigarays, and 8K video stream support to levels that seem almost too good to be true.

The flagship 4c GPU, which uses slightly less power than the RTX 5090, delivers 307 gigarays, features 512 MB of cache, and supports an astounding 2.3 TB of VRAM.

Bring your ideas to life with Zeus, a new type of GPU:
- Expandable memory: 32/64/128 GB soldered + 2x/4x SO-DIMM slots for up to 384 GB memory!
- Massive improvements to path tracing performance!
- 400 GbE QSFP-DD port!
- Built-in high-performance RISC-V CPU cores capable of… pic.twitter.com/ITJyFInLGV

However, all of these numbers remain theoretical. Bolt's "pre-silicon" performance estimates are based on synthetic benchmarks, not real-world gaming tests. Additionally, production isn't expected to begin until late 2026, meaning the GPUs won't be available before 2027. By then, both AMD and Nvidia will likely have launched next-generation hardware.

If you're wondering why any of this exists, the answer might be as simple as hype. Bolt could be trying to turn heads and turn that attention into funding, by riding the current wave of GPU gold rush fever, especially as AI accelerators continue to be Silicon Valley's hottest hardware obsession.

The spec sheet also reveals areas where the Zeus cards significantly trail Nvidia's high-end RTX 5000 series. Those massive VRAM capacities rely on LPDDR5X and DDR5, which are likely much slower than AMD and Nvidia's GDDR6 and GDDR7 memory.

Furthermore, FP32 and FP16 TFLOP counts – crucial for gaming performance – fall well below the RTX 5090, and in many cases, even below the RTX 5080.

If Bolt's numbers prove accurate, the Zeus series could be compelling for certain HPC, rendering, or server workloads. But when it comes to gaming (which also needs optimized drivers, and that takes a lot of time – just ask Intel), their real-world performance will likely remain uncertain for at least the next couple of years.