ReRAM steps into the spotlight
The world of memory is under pressure. Between AI workloads, data centers, and PC hardware demand, traditional NAND flash is in short supply and prices are rising. That is why a new type of non volatile memory called ReRAM is getting fresh attention.
ReRAM stands for resistive random access memory. Despite the name, it is not RAM in the same sense as DDR5 in your gaming rig. It is a form of non volatile memory like NAND flash used in SSDs, USB drives, and storage chips. The big news is that startup Weebit Nano has signed a licensing deal with Texas Instruments to integrate its ReRAM technology into advanced process nodes for embedded processors.
This does not mean ReRAM SSDs will suddenly appear on shelves, but it does mean the tech is moving out of pure research and into real products. With ongoing memory shortages and scaling challenges for flash, that matters for the future of PCs and gaming hardware.
What makes ReRAM different from flash?
ReRAM is designed to do the same basic job as NAND flash but with some important advantages. It stores data by changing the resistance of a material instead of trapping electrons in a floating gate as flash does. That switch in approach brings a few key benefits.
- Non volatile and persistent ReRAM, like NAND, keeps data even when power is turned off. That makes it useful both for storage and for firmware that has to survive reboots.
- Much higher endurance Weebit Nano claims its ReRAM can handle between 100,000 and 1,000,000 write cycles, compared with roughly 10,000 program and erase cycles for typical flash. For systems that write data frequently this can dramatically extend lifetime and reliability.
- High temperature retention The company describes its technology as having excellent retention at high temperatures, which is attractive in harsh environments, industrial systems, automotive electronics, and hot compact devices.
- Low power and cost effective ReRAM is pitched as a low power alternative to embedded flash, and as a back end of line module it can be integrated without changing the core transistor stack. That helps control manufacturing complexity and cost.
The most exciting idea for PC enthusiasts is that ReRAM blurs the line between system memory and storage. In theory you could have one type of memory that is fast enough to act like RAM and persistent enough to act like storage. That concept is sometimes called universal memory.
We are not there yet. Current ReRAM implementations are more focused on embedded use inside chips than on replacing your DDR5 or NVMe drive outright. But the path is being laid for more flexible memory designs in future PCs.
Why chip makers care and what it means for future PCs
Weebit’s ReRAM is built as a back end of line memory module. That is important because it can be added to existing chip manufacturing flows without redesigning the core logic transistors. For a chip manufacturer, less disruption means less risk and lower costs.
Compared to embedded flash, ReRAM integration can reduce added wafer cost by almost 15 percent according to the company. It also scales better to smaller process nodes. While flash often has to sit off chip or on a separate die, ReRAM can live right next to logic even at nodes like 22 nanometer and below. That is particularly interesting for AI accelerators and compact SoCs where you want memory as close as possible to compute.
Today, many systems use an external flash chip to hold firmware and data. On boot, that data is copied into SRAM or DRAM so the processor can access it quickly. This creates overhead in power, performance, and even security, because the data has to move and can be intercepted or tampered with along the way.
Weebit Nano’s CEO describes an alternative: replace on chip SRAM with ReRAM and you get a one chip solution that boots instantly, keeps its contents when powered off, and reduces data exposure. For embedded systems this is huge. For gaming PCs and laptops this could one day mean near instant boots, lower standby power, and simpler motherboard designs.
There is also interest in using ReRAM for neuromorphic computing, where memory cells behave a bit like synapses in a brain inspired architecture. Each ReRAM bit can naturally represent a weight or connection strength, which lines up well with how neural networks work. While claims about artificial general intelligence are probably overhyped here, it is clear that ReRAM fits nicely with AI accelerators that want dense, low power, non volatile memory close to the compute cores.
The race for universal memory
ReRAM is not the only player in the next gen memory race. Other projects like ULTRARAM and various superlattice based memories are also chasing the goal of a single technology that can replace both RAM and storage. The history here is littered with ambitious ideas that struggled to reach mass adoption, from Intel and Micron’s 3D XPoint to earlier phase change memories.
So while this new Texas Instruments deal and partnerships with foundries like SkyWater, DB HiTek, and Onsemi push ReRAM closer to reality, it is still early days. Full scale production for mainstream PCs and gaming hardware is some way off. For now, NAND flash and DRAM will continue to power your SSDs and gaming rigs, and the current memory price squeeze is not going away overnight.
Still, the direction is promising. As more chips adopt ReRAM internally, the ecosystem and manufacturing experience will mature. That sets the stage for future PC platforms where the divide between RAM and storage starts to blur, boot times shrink toward zero, and memory endurance and efficiency get a serious upgrade.
If you care about where PC performance is headed over the next decade, keep an eye on ReRAM and its rivals. The next big leap in gaming performance might not come from a faster GPU but from a smarter and more flexible memory system feeding data to everything else.
Original article and image: https://www.pcgamer.com/hardware/memory/reram-has-reentered-the-chat-with-high-profile-manufacturing-deal-positioning-it-as-the-successor-to-flash-memory/
