Friday, January 9th 2026

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

Bolt Graphics is a new computer graphics hardware startup based out of Sunnyvale, California. This company wants to shake things up in the consumer graphics hardware market with a brand-new, grounds-up graphics architecture for graphics rendering, gaming, and HPC. What makes this startup unique and refreshing is that its architecture, or at least the marketing material of it, doesn't use the most uttered two-letter acronym in human history.

Called ZEUS by Bolt Graphics, the architecture offers a contemporary raster 3D graphics, ray tracing, and path tracing hardware pipeline, using an in-house designed SIMD architecture that's controlled by a RISC-V command processor that doubles up as a CPU. The prototype graphics card features a PCI-Express 5.0 x16 host interface, and LPDDR5X as its graphics memory, with additional DDR5 SODIMM slots for the RISC-V command processor. Together, the board can have up to 384 GB of memory to deal with large data sets. The card also features a 400 Gbps or 800 Gbps enterprise network interface, and a fully-fledged BMC/IPMI circuit on board, so you can manage large clusters of the card in a rackmount render server or workstation. The most surprising aspect of the card is power, and it only features an 8-pin PCIe power input, for a total of 225 W of power draw.
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.
Add your own comment

114 Comments on New GPU Startup Bolt Graphics Promises 2.5x Path Tracing Performance of RTX 5090

#101
Unregistered
notoperableI wonder though - where did you get the impression I want "to halt any AI research"?
Why not read posts before replying to them? The entire premise of my post was that AI is here to stay, and a rebuttal to those claiming we need to halt all AI development until "it can be proven safe".

You're absolutely correct that there's no putting the genie back in the bottle. AI is a revolution far outstripping the effects of the Internet. Attempting to ban it is about as sensible as trying to ban the law of gravity.
Posted on Edit | Reply
#103
Steevo
EndymioWhy not read posts before replying to them? The entire premise of my post was that AI is here to stay, and a rebuttal to those claiming we need to halt all AI development until "it can be proven safe".

You're absolutely correct that there's no putting the genie back in the bottle. AI is a revolution far outstripping the effects of the Internet. Attempting to ban it is about as sensible as trying to ban the law of gravity.
I think you miss the point, AI is here to stay.... yes.
AI is the greatest thing since sliced bread? No
AI "outstripping the internet"? ... Not by a margin big enough for the moon to fly through.

AI is still retarded in general, its going to stay that way for YEARS. AI is in the flash animation stage of the Internets growth, AI is great at specific tasks, and in 99% of those tasks its not really AI, its just a matching algorithm that runs well on massively parallel hardware, not intelligent or no more intelligent than some advanced excel formulas running minecraft.


Banning it? Why, who said ban...? Regulated yes, companies are running huge, gargantuan deficits and sucking up investments from naive dolts who didn't live through the .com boom and bust, sure we made it out of that one, at a large cost and a lot of well, liars got rich and most paid no consequence for intentionally lying about possibilities to gain investment dollars, its happening again, the promises of AI haven't been met for the amount of funding, energy and infrastructure humanity has dumped into it, and they won't. AI for Open AI or other general purpose stuff is mediocre intelligent and still wrong sometimes. Anyone could get the same level of general knowledge by getting a free library card or learning to use google with critical thinking.

Back on topic....

This still reminds me a bit of www.tachyum.com/ this garbage..... always on the edge of a breakthrough and its big toilet paper holding it back, I mean big tech.....
Posted on Reply
#104
Unregistered
SteevoAI "outstripping the internet"? ... Not by a margin big enough for the moon to fly through.

AI is still retarded in general, its going to stay that way for YEARS.
Not even remotely correct. A year ago, a radiologist missed a significant problem on my imaging, even though he'd been instructed to look specifically for it. An AI screener found it in a few seconds, potentially saving my life. I paid the radiologist nearly $5000 and had to wait 48 hours ... the AI found it in seconds at a net cost of around 4 cents worth of electricity. That was last year's AI. I understand this year's model is significantly better. In five years, anyone who trusts a human radiologist over a ML model will need their head examined.

That's just one of several thousand fields already being impacted by AI. One of my neighbors owns a shipping firm ... a new AI-based routing module is saving him nearly a million gallons of fuel a year. In my own field, AI is identifying and screening new research candidates faster and better than people with decades of experience. The list goes on and on. Machine learning is far more than your experience with online AI chatbots.
SteevoAI is great at specific tasks, and in 99% of those tasks its not really AI, its just a matching algorithm
You're welcome to play the "it's not really intelligent" game all you wish, but the fact remains that AI is already matching and even exceeding human performance in countless areas. It may be just be doing simple pattern-matching ... but most of our modern workforce can't do even that.
Posted on Edit | Reply
#105
notoperable
EndymioNot even remotely correct. A year ago, a radiologist missed a significant problem on my imaging, even though he'd been instructed to look specifically for it. An AI screener found it in a few seconds, potentially saving my life. I paid the radiologist nearly $5000 and had to wait 48 hours ... the AI found it in seconds at a net cost of around 4 cents worth of electricity. That was last year's AI. I understand this year's model is significantly better. In five years, anyone who trusts a human radiologist over a ML model will need their head examined.

That's just one of several thousand fields already being impacted by AI. One of my neighbors owns a shipping firm ... a new AI-based routing module is saving him nearly a million gallons of fuel a year. In my own field, AI is identifying and screening new research candidates faster and better than people with decades of experience. The list goes on and on. Machine learning is far more than your experience with online AI chatbots.


You're welcome to play the "it's not really intelligent" game all you wish, but the fact remains that AI is already matching and even exceeding human performance in countless areas. It may be just be doing simple pattern-matching ... but most of our modern workforce can't do even that.
Efficient pattern-matching does not imply intelligence
Posted on Reply
#106
Steevo
EndymioNot even remotely correct. A year ago, a radiologist missed a significant problem on my imaging, even though he'd been instructed to look specifically for it. An AI screener found it in a few seconds, potentially saving my life. I paid the radiologist nearly $5000 and had to wait 48 hours ... the AI found it in seconds at a net cost of around 4 cents worth of electricity. That was last year's AI. I understand this year's model is significantly better. In five years, anyone who trusts a human radiologist over a ML model will need their head examined.

That's just one of several thousand fields already being impacted by AI. One of my neighbors owns a shipping firm ... a new AI-based routing module is saving him nearly a million gallons of fuel a year. In my own field, AI is identifying and screening new research candidates faster and better than people with decades of experience. The list goes on and on. Machine learning is far more than your experience with online AI chatbots.


You're welcome to play the "it's not really intelligent" game all you wish, but the fact remains that AI is already matching and even exceeding human performance in countless areas. It may be just be doing simple pattern-matching ... but most of our modern workforce can't do even that.
I'm glad that it found your issue. I notice you take the first part of my comment out of context due to your personal situation and then tack on the specific part where I mention that it can be very good a specific tasks such as your issue. You then fail to address the general knowledge VS intelligence. AI is NOT good at inferring, this is where the most mistakes are made, and I know there are two types of people, those who can infer information based on incomplete datasets

I have been running PhD reports through AI models for work and they are retarded, they take reports and dumb them down to drooling doodle level with misrepresented facts. The closest I have found is Microsoft Agent and it still took reports and will spit out a half unreadable summary or PowerPoint with missing data. One of our suppliers is using an advanced CUDA powered X-ray imaging system to determine product quality in almost real time, and has been since before marketing labeled it "AI". You are making very broad assumptions about people on this site that you seem to know little to nothing about and I would wager you feel superior to based off your arguments and posts, I hope I am wrong.
Posted on Reply
#107
notoperable
SteevoI'm glad that it found your issue. I notice you take the first part of my comment out of context due to your personal situation and then tack on the specific part where I mention that it can be very good a specific tasks such as your issue. You then fail to address the general knowledge VS intelligence. AI is NOT good at inferring, this is where the most mistakes are made, and I know there are two types of people, those who can infer information based on incomplete datasets

I have been running PhD reports through AI models for work and they are retarded, they take reports and dumb them down to drooling doodle level with misrepresented facts. The closest I have found is Microsoft Agent and it still took reports and will spit out a half unreadable summary or PowerPoint with missing data. One of our suppliers is using an advanced CUDA powered X-ray imaging system to determine product quality in almost real time, and has been since before marketing labeled it "AI". You are making very broad assumptions about people on this site that you seem to know little to nothing about and I would wager you feel superior to based off your arguments and posts, I hope I am wrong.
What happens nexts is usually the stage when you try to fine tune the model and you realize that it is you who actually pays for training their model and not the other way around :-)

@Endymio case is actually one of the best showcases where AI excells at, but this is not universal, not even close, its a marketing go-to headline that is being misused as golden example actually at least 2-3 years before openai shocked the world with their lingustics hijack - pattern matching and visual data processing, especially in radiology was one of the strongest example showing benefits of inserting machine learning based processing aka AI that actually makes the wait time for radiologist / neurologist shorter but did not replaced those - noone sane will sign on a AI only match without reviewing it first, that is what is either ignored or not understood, that yes, the AI didnt missed something human missed, but the very same humans that missed it, signed off the match.

Worth mentioning is, that AI handles single color channel (shades of gray) much better then humans (radiology, ultrasound, sonar data) - but for some reasons have problems with full color spectrum imaging, and that is where you should actually try do dig for relevant answers
Posted on Reply
#108
anthraxtacide
LastDudeALiveI think there's certainly potential here. Gaming will never happen, there's too much legacy knowledge and infrastructure required to succeed in gaming. How are they going to ensure this works on games from 2006 and 2026 alike? They just don't have the manpower to do that, let alone keep up with new game releases. Just look at Intel's struggles, and they're a massive company that's had years of iGPU development.

However, the high-end rendering and HPC markets are certainly ripe for disruption. Nvidia is in the difficult situation of trying to balance a lot of different markets. AI markets want pure tensor cores, gamers want low-precision calculation for high frame rates, HPC and rendering want high-precision and care less about speed. No architecture can meet all those needs. So if they focus on supporting common rendering engines (including UE5) and make a versatile framework for writing HPC code, they can certainly make a better (on paper) product than Nvidia in those specific markets.
I agree with this assessment. Having grown up with personal computers in the 80's, there were definately moves made by IBM, Bell Labs, Cyrix, Motorola and more just to name a few to capture the professional consumer and corporate enterprise market. The standards we have today like the x86 architecture, ATX form factor and IEEE members that needed to work together to be able to benefit from each others products (patents), and that includes working with software engineers to be able to build performant applications so people would buy the hardware in the first place. Eventually new ideas come along that can indeed break the standards we have - like LastDudeAlive said "HPC markets are ripe for disruption".

Bolt GDC 2025 Live Demos

-
Here is some more data on Bolt Graphics product for those who would care to look give this a fair look before concluding that the claims they advertised are false.
- bolt-graphics.atlassian.net/wiki/spaces/EAP/pages/325124119/Developer+Guide+Path+Tracing
And here is a developer Wiki on their methodolgy they used in their benchmarks to show their performance metrics Within Their Target Applications - this is important. As we know, dedicated hardware is much more effective (Physics Processor) than multi-purpose hardware (like a CPU) .
We need more diversity in the Compute Hardware market, and we need smart people to design circuits capable of processing the data in better ways - RISC architecture has many benefits to be able to accomplish this more efficiently and faster if built right from the design phase. We've seen when nVidia stopped supporting the 32bit PhysX API - the games that relied on the programming extensions fell flat with 5090's performing worse than 10 year old GPUs. This is the difference that a dedicated hardware + software foundation can make. So yes, I am very optimistic that Bolt Graphics can do what they say *within the environment they currently are developing for*

Yes we should be critical of the claims made by start-ups, but we should not draw conclusions on the claims of a couple of news articles - Let us be open minded and enquisitive..
TRY OUT THE METHODOLGY FOR YOURSELF WITH YOUR RTX6000
^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^
> > > see if your hardware can beat their demonstration of their development kit < < <


- I certainly look forward to the day where I can install another Add-On card into my PC's - Why not have an RTX 6000 A N D a Bolt Graphics Zues PTPU (Path Tracing Processing Unit)

Then maybe can say goodbye to Fake Frame Generation.. or use it way to build better AI into games.. Can't wait to hear more about this - I hope the current market doesnt suffocate their progress - we definately could use this technology in Advanced Sciences, such as Plasma vs Electromagnetic interactions. for Fusion Energy research. Big AGI can use that data to produce better data interpolation which would improve the model to run with PTracing Processors. We all benefit as long as the big Corp-technocrats dont buy it out from under them.. but I digress..

Give Them a Fair Go ..
Posted on Reply
#109
Alain De Loin
but this is a typical scenario, they burned through money, deadlines are looming, so they release a few worthless slides, no confirmation from independent tests, and none because there is no physical equipment, they continue to work on simulations. If their solutions are to some extent true and work, they will most likely be included in the RTX 7090. Because that's all it's about, making some noise and making as much money as possible before their securities end up on the junk market and some Wolf of Wallstreet makes a small fortune out of them by selling them to car mechanics.
Posted on Reply
#110
anthraxtacide
Given their intial seed cash of 138K was over 10 years ago in March of 2014, they must be producing a hell of a lot of worthless slides to convince investors!! I bet they were awarded the Patents on worthless Slide Reproduction Technology!! That must be why they've built a development framework and now have a fake PCB with a metal badge glued to it too!! Damn those Investors and Grant Adjudicators must be sooo stupid!! I bet the CGI Artists were actually working for nVidia trying to work out how to get their games to run like Doom..

I do agree with you that at this stage in their Slide Evolution, they have very blank slides to print. They're very lucky to have those Patents to allow them to print worthless slides about how to trace rays for the next couple of years. Those foolish investors I bet they can't wait to profit from the next worthless slide. Let's hope Bolt Graphics doesnt print the mathematical IP they developed to print the most realistic, incredible power efficient slides the industry has ever seen.

.. nVidia must be working on the RTX 70 series right now!
.. Surely they cant allow a Worthless Slide startup company to post their slides in Oracles Data Centres.. o_Oo_Oo_O



Nooo their code is going to share sooo many Worthless Slides !!

Bolt Graphics are so terrible with all their noisy slides and making so much noise in a handful of technology reports.. I wish they would all shut up, I want to hear more about AI and AI the of AI and all the AI that will AI us to AI Utopia. Burn those slides! They are too loud - the Dept of Energy might hear them and give them lots of tax money...
Posted on Reply
#111
notoperable
It would be all nice and dandy if the public get served a planet-wide beta test that might (or might not) be a serious health problem.
@Endymio have seen the recent trend of inconvenient medial studies - aka turd - bubbling up? It seems the LLMs are actually a health hazard.
And its not me that says it is so - but its some serious study results coming into focus, and, the simple fact that our brains can't differentiate,
handling AI LLMs based interface by default as human (meaning LLMs are simulacrums, tricking your brain into serious signaling problem).
This lands your unconscious part of the brain directly in conflict with conscious signalling - one parts says its human the other part says its not.
Medical studies though are hard to ignore, and the results imply exposure without clear or any signaling (that is now default) might by a serious
health problem - and we now see bigtech scuttle away from it and i bet in next 6 months all LLMs human facing interfaces will largly expand their
smallprint. Guess why?
Posted on Reply
#112
Unregistered
notoperable....? It seems the LLMs are actually a health hazard.
And its not me that says it is so - but its some serious study results coming into focus, and, the simple fact that our brains can't differentiate,
handling AI LLMs based interface by default as human (meaning LLMs are simulacrums, tricking your brain into serious signaling problem).
This lands your unconscious part of the brain directly in conflict with conscious signalling - one parts says its human the other part says its not.
Medical studies though are hard to ignore, and the results imply exposure without clear or any signaling (that is now default) might by a serious
health problem...
If you're referring to studies like this, you've managed to garble both their objectives and results into something entirely different.
#113
notoperable
EndymioIf you're referring to studies like this, you've managed to garble both their objectives and results into something entirely different.
If that's a study for you, then we have a different interpretation of what a peer reviewed study is. A study cant be garbled neither can be the results - when reviewed. A study's goal is to be objective empirical process to solidify any answers to the thesis it proposes, that being said- i think we said all there was to say and that will be it.
Posted on Reply
#114
Unregistered
notoperableIf that's a study for you, then we have a different interpretation of what a peer reviewed study is. A study cant be garbled neither can be the results - when reviewed. A study's goal is to be objective empirical process [sic] to solidify any answers to the thesis it proposes....
Was that a joke? If a study concludes that cell phones don't cause brain cancer and you come to this forum and tell us that it concludes computer use causes liver cancer, then yes, you've garbled the results. And if that isn't the study to which you were originally referring that "LLMs are actually a serious health hazard", then by all means, post your evidence and we'll have a look at it. I strongly suspect it doesn't exist.
Add your own comment
Nov 18th, 2026 04:53 CST change timezone

New Forum Posts

Popular Reviews

TPU on YouTube

Controversial News Posts