BTC $77,884.17 -3.27%
ETH $2,444.23 -3.06%
BNB $694.11 -2.39%
XRP $1.39 -4.62%
SOL $104.70 -3.12%
TRX $0.3396 +0.48%
DOGE $0.0850 -4.57%
ADA $0.2030 -5.90%
BCH $247.52 -8.64%
LINK $11.47 -3.58%
HYPE $80.86 -5.73%
AAVE $122.54 -4.99%
SUI $0.7510 -4.07%
XLM $0.1786 -4.98%
ZEC $793.59 -3.59%
BTC $77,884.17 -3.27%
ETH $2,444.23 -3.06%
BNB $694.11 -2.39%
XRP $1.39 -4.62%
SOL $104.70 -3.12%
TRX $0.3396 +0.48%
DOGE $0.0850 -4.57%
ADA $0.2030 -5.90%
BCH $247.52 -8.64%
LINK $11.47 -3.58%
HYPE $80.86 -5.73%
AAVE $122.54 -4.99%
SUI $0.7510 -4.07%
XLM $0.1786 -4.98%
ZEC $793.59 -3.59%

flex

All
Article
Flash

hot_img Cerebras releases the fourth generation AI inference system CS-4: performance doubled, power consumption doubled, more flexible deployment

Cerebras released its fourth-generation AI inference system CS-4 this week, based on the same 5nm WSE-3 wafer, achieving double the performance by doubling the clock frequency and power consumption. A single CS-4 cabinet accommodates 3 wafers (CS-3 has 2), featuring a modular "backpack" design that simplifies manufacturing and deployment, with a TDP of approximately 125 to 135kW. The CS-4 can provide an inference speed of nearly 4000 tokens/second/user, about twice that of the CS-3, and supports decomposed inference with heterogeneous systems such as AMD and AWS Trainium.Cerebras claims that the CS-4 offers about 2000 times the on-chip memory bandwidth of NVIDIA's Rubin (43PB/s), but the 44GB SRAM capacity remains unchanged, and long-context inference still requires multi-wafer stacking. For example, with the DeepSeek V4 Pro (1.6T parameters), approximately 20 systems are needed for a 1M context window, and about 40 systems are required for 256 concurrent users, corresponding to a CAPEX exceeding 20 million USD. Cerebras is collaborating with clients such as OpenAI and plans to achieve approximately double performance improvements each year, aiming for a 20-fold throughput increase by 2027. The "backpack" cabinet design of the CS-4 will continue into the next-generation "Nexus" platform.

The Solana Foundation launches a new privacy framework for institutions: enterprise-level adoption requires flexible privacy controls

According to CoinDesk, the Solana Foundation released a report titled "Privacy on Solana: A Comprehensive Approach for Modern Enterprises," which suggests that enterprise-level adoption requires flexible privacy controls and positions privacy as a customizable feature rather than a trade-off.The report argues that the next phase of crypto adoption will depend more on allowing enterprises to control the subjects and content of information disclosure, rather than solely relying on transparency. The Solana Foundation proposes that privacy encompasses four different modes: pseudonymity, confidentiality, anonymity, and complete privacy systems. Pseudonymity hides identity while transaction data is visible; confidentiality allows participants to be known but encrypts sensitive information; anonymity hides participant identities while transaction data is visible; and complete privacy systems obscure both identity and transaction data through technologies such as zero-knowledge proofs and multi-party computation.The report emphasizes that there is no single privacy model suitable for all scenarios, and enterprises can mix different tools according to their needs. The report notes that Solana's high throughput and low latency enable advanced privacy technologies to operate at near-network speeds, making applications such as encrypted order books or private credit risk calculations possible. The Solana Foundation also proposed mechanisms such as "audit keys," allowing designated parties to decrypt transactions when necessary, thus achieving coexistence between privacy and regulation.
app_icon
ChainCatcher Building the Web3 world with innovations.