World’s Fastest Random Access Near Memory
HBLL-RAM
Ideal Low Density High Bandwidth Solution
HBLL-RAM is the optimal answer to the memory performance bottleneck in the AI industry.
It supports massive amounts of data transfer between memory and the processor at an exceptional speed.
HBLL-RAM accelerates the performance of algorithms in neural networks, block-chain, big data analytics, and autonomous driving.
It supports massive amounts of data transfer between memory and the processor at an exceptional speed.
HBLL-RAM accelerates the performance of algorithms in neural networks, block-chain, big data analytics, and autonomous driving.
HBLL Powers AI accelerators,
Block Chain Algorithms,
and Neural Networks.
Block Chain Algorithms,
and Neural Networks.
Features
1
Exceptional Performance
Sub-20ns random access latency
144GB/s per die random access bandwidth
1.2V core
144GB/s per die random access bandwidth
1.2V core
HBLL’s clock frequency runs at only 1GHz (2Gb/s/pin), yet its random-access bandwidth outperforms all JEDEC standard’s DRAMs on market.
A 4Gb HBLL-RAM die features an 8-Channel SRAM interface with 32 banks per channel, which optimized the array utilization and minimized the random-access latency.
A 4Gb HBLL-RAM die features an 8-Channel SRAM interface with 32 banks per channel, which optimized the array utilization and minimized the random-access latency.
2X Performance of HBM2
Random Access Bandwidth
Half Latency of HBM2
Random Access Row Latency
2
Incredible Energy Efficiency
Ultra-low Power with Energy Consumption of 5pJ/b.
HBLL-RAM runs high performance computing AI with little energy consumption.
The short-page row access design greatly reduces not only the internal power noise, but also the energy consumption of each random row access.
The short-page row access design greatly reduces not only the internal power noise, but also the energy consumption of each random row access.
One Third of LPDDR4
Energy Consumption per Bit Access
3
The Optimal LOW-DENSITY
High-Bandwidth AI Solution
High-Bandwidth AI Solution
While all AI solutions benefit from high frequency and high bandwidth, not all are in need of high-density chips.
PieceMakers provide chips with varying densities, all with exceptional bandwidths to serve a diverse group of AI customers.
PieceMakers provide chips with varying densities, all with exceptional bandwidths to serve a diverse group of AI customers.
4
Cost Effective AI Edge Solution
Save Money
Massive data storage comes with a cost, but it is not necessary for all AI solutions.
PieceMakers provide low density chips without compromising frequency and bandwidth for AI solutions at a lower cost.
Additionally, HBLL-RAM provides a single-die with single-edge-pad allocation that can be integrated with the master chip in 2D SIP form without the use of 2.5D / 3D technology.
PieceMakers provide low density chips without compromising frequency and bandwidth for AI solutions at a lower cost.
Additionally, HBLL-RAM provides a single-die with single-edge-pad allocation that can be integrated with the master chip in 2D SIP form without the use of 2.5D / 3D technology.
Key Features
1.2V core and I/O power supplies
Un-terminated LVCMOS interfaces
1GHz clock frequency，2Gb/s/pin data rate
4Gb with 8 Channels，32 banks per channel
Random access Bandwidth： 144 GBPS
Row Latency：20ns
Simple SRAM Interface
Product List
HBLL
| Part No. | Density | Bit Width | Voltage | Speed | Grade | Status |
|---|---|---|---|---|---|---|
| PLHP11440AK4 | 2.25 Gb | 288 | 1.2/1.2/3.5 V | 1 GHz | C / I | MP |
Note
C - Commercial grade
I - Industrial grade
For data sheet and support,
please contact our sales representative.
please contact our sales representative.
Contact Us
Business
sales@piecemakers.com.tw
Career
hr@piecemakers.com.tw
Or
Your name
*
Company
*
Email
*
Phone
Message
*
By submitting your personal data, you consent to PieceMakers' storage
and processing of the data for the purpose of providing you services.
and processing of the data for the purpose of providing you services.