A home for GPU miners and builders. Mainnet is live. Building toward customer-paid AI inference. Follow for setup help, releases and progress.
United States
Joined August 2026
- Tweets457
- Following21
- Followers782
- Likes164
Pinned Tweet
Common Foundry: We unlocked nearly 2× FP8→FP32 Tensor Core throughput on an RTX 4070 Ti SUPER. A matched GEMM in a custom NVIDIA CUTLASS 4.8.1 harness ran 1.95× faster: 158.54 → 309.08 TFLOP/s. Same GPU, binary and 240 W cap. Results:
github.com/JustAResearcher/g…
Common Foundry retweeted
Unlocked And Loaded!
Thanks to the @CommonFoundry1 research & Mod, I tweaked mine a little for the dual Gpu Setup on both 4090s for windows. NOW LETS BURN SOME @MiniMax_AI Renders!
Why does Common Foundry use exact integer arithmetic?
Different implementations need to evaluate the same defined computation consistently. ForgeMatrix uses INT8 matrix operations with INT32 accumulation.
That's the reason behind the terminology.
Not sure why no one is mentioning this.
I unlocked FP8 on 40 series too! Not just BF16/FP16.
FP8 is the real prize here.
Also, I seriously doubt this guy held anything back.
He's just trying to scam people out of their money. There is a free windows version floating around.
4 days ago I posted about a FREE Linux module that turns on 2x tensor speed on the 4090. Yesterday 1usmus launched the same speedup, after holding it back for months.
HYDRA Tensor Unlock for RTX 40: about 2x FP16/BF16 tensor throughput with FP32 accumulation on the 4080 and 4090. In their own post they held it back for months and only released it because Chinese researchers found the same approach.
So the speedup existed for months. 4090 owners paid for the cards and never got it. It came out only after a free version was on GitHub.
Open came first and the other one followed. Remember the order.
And NVIDIA, the company that sold you the card with half its tensor speed switched off, still hasnt said a word.
xitter.cf/superalesha/status/211…
Which small tool would make running Common Foundry easier?
A clearer log viewer, a setup checker, an explorer feature, or something else?
Describe the task you're trying to finish. The most useful tool ideas start with a real operator problem.
Curious what Common Foundry is all about?
Check out our growing discord server to get the latest on CMFD.
discord.gg/commonfoundry
Developers: Common Foundry's source is public.
Choose one question to follow through the code: how a transaction is handled, how a block is checked, or how a node tracks chain state.
Start here:
github.com/Common-Foundry-1/…
Held it back for months?
But you couldn't even unlock the FP8 capabilities?
I dont belive you!
Common Foundry retweeted
Common Foundry is working on the fundamentals 👀l
The latest release adds independent block-work verification, wallet improvements, and AMD mining support.
Not every meaningful Web3 update is flashy. Sometimes the infrastructure is the story.
@CommonFoundry1
Explain Common Foundry to a friend in one sentence.
Ours: a live blockchain where miners perform matrix computation and nodes check proofs of that work.
What would you need explained next: the mining workload, the proofs, or how to get started?
Private beta for our Inference Marketplace are going out this week.
Also, we have beefed up our infrastructure intruding 2 new relays for faster and more consistent syncing.
Common Foundry retweeted
Nvidia ships consumer cards with tensor math throttled in software. Now people are posting real before and after numbers:
4070 Ti SUPER: 158.5 → 309.1 TFLOP/s (via @CommonFoundry1)
Worth asking: what else is locked on the card you already own?
CMFD transaction fees are burned. New CMFD is also issued through block rewards.
To understand supply, you need both numbers. Fee burning alone doesn't tell the whole story.
Inspect the published schedule:
commonfoundry.ai/docs/emissi…
Common Foundry retweeted
Common Foundry keeps building 🛠️
v1.0.16 brings independent block-work verification, faster wallet catch-up and checksummed addresses.
Less noise, more infrastructure.
$CMFD
@CommonFoundry1
Common Foundry retweeted
Common Foundry (@CommonFoundry1) reports a ~1.95x tensor speedup on an RTX 4070 Ti SUPER:
158.54 → 309.08 TFLOP/s, FP8→FP32 matched GEMM
Same GPU, same binary, same 240 W cap.
I've seen the same kind of unlock on my own 3080.
xitter.cf/CommonFoundry1/status/…
Common Foundry: We unlocked nearly 2× FP8→FP32 Tensor Core throughput on an RTX 4070 Ti SUPER. A matched GEMM in a custom NVIDIA CUTLASS 4.8.1 harness ran 1.95× faster: 158.54 → 309.08 TFLOP/s. Same GPU, binary and 240 W cap. Results:
github.com/JustAResearcher/g…
Weekend setup check: did you get a Common Foundry wallet, node or miner running?
Share one thing that worked and one thing that was confusing. GPU model and OS help others compare notes.
New here? Setup help is available in Discord:
discord.gg/XGuutqWMWP
After the 30 series was unlocked, then we unlocked the 40 series and the 20 series coming right after that.
Naturally everyone turns their attention to the 50 series.
Who will be the first to unlock its FP8 potential?
I can tell you, its not going to unlock the way 20-40 did
Nice work!!
And just like that,
We have windows gpu unlock
github.com/impacifist/Window…
A useful GPU benchmark lets someone else repeat it.
For CMFD, share your setup, test length, accepted work and power measurement alongside speed. Include the settings that changed.
A screenshot is a starting point. Reproducible results help the next operator.
Let's go!!! 20 series included too.
Nice work
Now I also confirmed that 2080ti unlock is successful with similar approach as the 30-series' method
it reach nearly same speed as 350w 3090(locked) with only 250w in mmapeak and mamf-finder
github.com/KohakuBlueleaf/Ko…
