• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

GPU Startup Bolt Graphics Promises 2.5x Path Tracing Performance of an RTX 5090

erek

erek

Fully [H]
2FA
Joined
Dec 19, 2005
Messages
17,421
Hopefully they can do better than when Bitboys had their vapourware moment

“Besides hardware, Bolt Graphics is offering a fully-fledged software that that includes Vulkan and DirectX 12 interfaces, including support for Unreal Engine and Unity. Applications for the architecture support all modern programming environments. Bolt innovated a path tracing pipeline.

Bolt Graphics has made some astoundingly tall performance claims that indicate an ASIC-like approach towards popular use-cases, using fixed-function hardware. In FP64 EM simulation benchmarks, the Zeus 4C is claimed to offer 300 times the performance of an NVIDIA B200 "Blackwell." In another FP64 math benchmark, the Zeus 1C (single silicon) is shown offering 3 times the performance of a GeForce RTX 5090. Perhaps the most fascinating claim is path tracing performance, and here, a Zeus 1C is claimed to offer 2.5 times the path tracing performance of an RTX 5090, 5 times with 2C, and 10 times with 4C.

The part about Bolt Graphics marketing that caught our eye is their claim that the Zeus 1C is "the fastest graphics processor ever," made possible due to a fundamental reimagining of conventional graphics processing workflows.”

1767957158888.png

Source: https://www.techpowerup.com/345018/...-2-5x-path-tracing-performance-of-an-rtx-5090
 
An integrated 400Gbit QSFP NIC?

Something tells me this isn't targeted at the consumer... :D

That, and chances are it is vaporware. The barriers to entry in GPU's are so utterly tremendous at this point, with the current state reflecting generational optimizations stacked on top of generational optimizations of hardware, firmware and drivers as to be extremely unlikely for any little startup to just drop in and overcome them.

The concept of both included VRAM and expandable DDR5 modules is an interesting one, but the integrated RAM is only LPDDR5, way slower than GDDR6, and if you need more RAM than what is integrated, you'll have to transit the memory bus to the slower DDR5, creating a GTX970 type of problem with RAM in two different speeds.

I don't know what they actually hope to achieve, but it seems like this product is extremely unlikely to ever hit the market, and if it does it will probably be targeted straight at datacenter AI type applications, not at consumers.

As with all of this though, I always hope I am wrong, as having more options that drive competition is always a plus.
 
Last edited:
Guaranteed vaporware. Intel, a startup in dedicated GPUs with a large budget, it’s not been able to get even close to this sort of performance.
 
erek said:
simulation benchmarks, t
Click to expand...
Here it is important to say, it is benchmark of physic simulation that we mean, but that the benchmark are simulated.

“pre-Silicon benchmarks in emulation”. Maybe some are low frequency FGPA that get extrapolated and not pure code, but the amount of things that are hard, can go wrong, cost a lot to turn it into a large amount actual chips are quite something, specially a quad chip working together,...

those cards are computers, witha a cpu, regular ram stick that run an Linux, sacrifice FP32 (regular raster) to achieve those numbers, going after a nicher market like large memory path tracing for disney renders farms make some sense (as an entry point) and some FP64 simulation, seem to be their target (thus the super fast connectivity to create farms of them).

Also way less headache than trying to run pre DX12-vulkan games drivers wise to go into those industry.
 
Zarathustra[H] said:
An integrated 400Gbit QSFP NIC?

Something tells me this isn't targeted at the consumer... :D

That, and chances are it is vaporware. The barriers to entry in GPU's are so utterly tremendous at this point, with the current state reflecting generational optimizations stacked on top of generational optimizations of hardware, firmware and drivers as to be extremely unlikely for any little startup to just drop in and overcome them.

The concept of both included VRAM and expandable DDR5 modules is an interesting one, but the integrated RAM is only LPDDR5, way slower than GDDR6, and if you need more RAM than what is integrated, you'll have to transit the memory bus to the slower DDR5, creating a GTX970 type of problem with RAM in two different speeds.

I don't know what they actually hope to achieve, but it seems like this product is extremely unlikely to ever hit the market, and if it does it will probably be targeted straight at datacenter AI type applications, not at consumers.

As with all of this though, I always hope I am wrong, as having more options that drive competition is always a plus.
Click to expand...
Shit will be my next pfsense box with that NIC
 
To be fair, if all it did was path tracing, then what they said could be true. Modern graphics cards do a lot more than just that though. If it was so easy, Intel would have cleaned ship -- I'll remain skeptical.
 
It'd be kinda neat if they made this into something like the original PhysX cards, but dedicated to RT + Path tracing. Could be an interesting way to bring improved Raytracing to systems without great support through their GPU (ie: RX 7000 series).
 
DF-1 said:
so... DDR5 (not GDDR/HBME)
PCIE connectors on BOTH sides...???
and an RJ-45??

this is a picture an AI made when asked to make a revolutionary GPU.
Click to expand...
This GPU has 2 mommy boards. It has 2x the performance but what they didn't tell you is that requires it is plugged into 2 PCs.
 
Yeah. Making a computation unit is hard, but not insurmountable.

Making a GPU which does all the things - well, that's a bit of a sticky wicket.
 
I'll believe it when I see it. The Bitboys comparison is apt: It's easy to make a claim and show something on a simulator. Call me when it is a real product. In part because even IF the claims end up being real, if it takes too long it could be unimpressive. Like lets say they do release a card that is 2.5x the power of a 5090... 10 years from now. I'd bet at that point nVidia will have something that is considerably more than 2.5x as fast, and thus it'll be unimpressive.

As a more recent example: When ChatGPT first became a thing, remember all those claims from startups about the "AI accelerators" that were going to totally trounce nVidia? How many of those have we actually seen out there? Google and Amazon have made their own, not sure how good they are (Amazon in particular is cagey about benchmarks) but for everyone else it appears to be all-nVidia all the time because all the other claims have thus far failed to deliver. Every year they do, the less impressive that claim looks because nVidia produces better and better hardware.

I just really have no fucks to give about announcements that aren't backed up by hardware. If your shit doesn't exist at least in testing, then I am just not impressed because not only do I not know if it'll make it to market, but even if it does it could be way too late for me to care.
 

Bolt Graphics Completes Tape-Out of Test Chip for Its High-Performance Zeus GPU, A Major Milestone in Reducing Computing Costs By 17x​

News provided by
Bolt Graphics, Inc.
Apr 22, 2026, 10:19 ET

Company now targets production in Q4 2027 to supply chips for high-performance compute (HPC), rendering, and next-generation workloads

SUNNYVALE, Calif., April 22, 2026 /PRNewswire/ -- Bolt Graphics today announced the successful tape-out of its test chip, marking a key milestone in the development of its Zeus GPU. Zeus is a next-generation compute platform designed to reduce the total cost of compute by up to 17 times across high-performance computing (HPC), rendering, and emerging compute-intensive applications.

The Zeus platform integrates a custom GPU architecture with a full software stack to create a unified system designed to operate across multiple compute markets. The platform uses established semiconductor processes, with the test chip successfully designed into TSMC 12 FFC. The Zeus scalable architecture also addresses advanced nodes, including 5 nm.


Performance-Per-Dollar

companies optimized existing compute architectures for peak performance rather than cost efficiency, causing infrastructure cost to become a primary constraint. As the majority of workloads remain dependent on these architectures, large segments of the available market remain economically unviable.

Bolt Graphics takes a fundamentally different approach to delivering compute to the market. The company achieves superior performance while optimizing for performance-per-dollar rather than maximizing for just peak performance. By focusing on cost efficiency at the system level, the Zeus platform delivers step-function improvements in compute economics. With cost savings up to 17 times compared to incumbent architectures, Zeus unlocks new classes of workloads previously constrained by cost.


https://x.com/IanCutress/status/2046992442486899036?s=20
 
It says HPC.... so it's for datacenters for cloud-hosted computing.

If your custom solution can't beat a generic one, then you have failed at step 1, and you need to stop what you are doing.
 
You must log in or register to reply here.
Back
Top