Claim Says Hidden RTX AI Compute Was Unlocked
A user claims hidden Tensor Core AI compute across NVIDIA RTX 20, 30, and 40 cards was unlocked, but the claim remains unverified by NVIDIA.
GPU
All AI Pulse updates tagged "GPU".
20 signalsMarvell photonics module puts memory 50 meters from GPUs
Marvell's Dave Lazovsky shows a tiny photonics module that moves HBM and DDR memory up to 50 meters from a GPU, versus the usual 1.5 mm, freeing memory placement for the first time.
Aethir Rings Nasdaq Bell as $AGPU Debuts
AethirCloud celebrated its journey from decentralized GPU cloud to ringing the Nasdaq bell with $AGPU, saying real compute demand is only getting started.
Frontier labs say unpaid entry still too costly
Financially independent people asked about joining frontier labs without salary. The labs' response: each person's GPU and token consumption far exceeds a salary, and wasted resources are unaffordable.
3-year GPU commits can cost more than ownership
The poster says on-demand GPUs are scarce, so teams sign long offtakes; competitive rates require 3+ year commitments, while shorter terms raise hourly prices.
NVIDIA GPUs Act as High-Yield Assets, H100 Returns 76%
Ornn's data shows H100 delivered about 76% total return in one year and B300 about 21% in under three months. Rental income and strong secondary-market residual value give GPUs both utility and yield.
Finland.
Finland. Thailand. Australia. India and more. RTX GPUs at the edge, from $0.10/hr. Deploy on Theta EdgeCloud.
BC250 Board Gets GitHub Firmware Flash
A $40 BC250 board with 16GB GDDR6 rewrites CPU/GPU split and opens 15.5GB unified memory.
Chinese Dev Buys Failed Startup's H100 Server for $40K
A Chinese developer bought a failed startup's H100 server for $40K and now bills $80K a month.
HPE DL380a Chassis Holds 4 RTX Cards
A single HPE DL380a chassis now holds 4 RTX PRO 6000 cards, designed for frontier-class local models.
Scale AI workloads with GPU power
Amazon EC2 G7 instances are now available, powered by NVIDIA GPUs.
A $351 mini PC runs 26-billion-parameter AI models
Running at 20 tokens/sec with Hermes agent.
On this episode of the Odd Lots podcast
Carmen Li discusses her company's quest to build a compute market.
Local AI stops being a toy
He holds a 128GB AI mini PC, showcasing the shift in local AI.
GEMMA 4 12B WITH MULTI TOKEN PREDICTION ON GMKTEC EVO-X2
Video shows 14+ tokens per second at nearly 30k context.
Switch between TPUs and GPUs easily
You can switch between TPUs and GPUs without rewriting code, now with native PyTorch support.
AMD CEO showcases mini PC running 235B model
The mini PC showcased by AMD CEO runs a 235B model, replacing a $440/month AI stack.
267 tokens per second on a single RTX 5080
Ollama runs Llama 3.2 1b at impressive speeds.
Developer loaded 670GB of Kimi K2.6
Cut AI costs from $4,000 to $700/month while increasing processing speed.
Nvidia GPU farm makes $18,000 a month
Ten cards running 24/7, paid off in seven months, renting compute to AI companies.