Compute-In-Memory APU Achieves GPU-Class AI Performance at a Fraction of the Energy Cost
GSI Technology (Nasdaq: GSIT) announced publication of a Cornell-led paper (ACM, presented at Micro ’25) validating its Compute-In-Memory Gemini-I APU. Cornell benchmarks found the APU delivered GPU-class throughput comparable to NVIDIA A6000 on RAG workloads while using over 98% less energy than a GPU and cutting retrieval processing time by up to 80% versus CPUs on datasets from 10GB to 200GB. GSI noted Gemini-II offers ~10x faster throughput and lower latency, and Plato targets further low-power edge capability.
GSI Technology (Nasdaq: GSIT) ha annunciato la pubblicazione di un rapporto condotto da Cornell (ACM, presentato a Micro ’25) che convalida il suo Gemini-I APU Compute-In-Memory. Le benchmark di Cornell hanno rilevato che l'APU offre una throughput di livello GPU comparabile alla NVIDIA A6000 su carichi RAG, utilizzando più del 98% in meno di energia rispetto a una GPU e - rispetto alle CPU - riduce il tempo di elaborazione del recupero fino all'80% su set di dati da 10GB a 200GB. GSI ha osservato che Gemini-II offre circa 10x più throughput e latenza inferiore, e Plato punta a ulteriori capacità edge a basso consumo.
GSI Technology (Nasdaq: GSIT) anunció la publicación de un artículo liderado por Cornell (ACM, presentado en Micro ’25) que valida su Compute-In-Memory Gemini-I APU. Las benchmarks de Cornell encontraron que el APU ofrecía un rendimiento de tipo GPU comparable a NVIDIA A6000 en cargas RAG, mientras consumía más del 98% menos de energía que una GPU y reducía el tiempo de procesamiento de recuperación hasta un 80% frente a las CPU en conjuntos de datos desde 10GB a 200GB. GSI señaló que Gemini-II ofrece ~10x más rendimiento y menor latencia, y Plato apunta a una mayor capacidad de borde de bajo consumo.
GSI Technology (Nasdaq: GSIT)은 Cornell이 주도한 논문(ACM, Micro ’25에서 발표)을 발표해 Compute-In-Memory Gemini-I APU를 검증했다고 밝혔다. Cornell 벤치마크는 APU가 RAG 워크로드에서 GPU급 처리량을 NVIDIA A6000과 비교 가능하게 제공하고 GPU 대비 98% 이상 에너지 감소를 달성하며, CPU에 비해 데이터세트가 10GB에서 200GB 사이일 때 검색 처리 시간을 최대 80%까지 단축했다고 밝혔다. GSI는 Gemini-II가 약 10배 더 빠른 처리량과 더 낮은 대기시간을 제공하며, Plato가 더 낮은 전력의 엣지 기능을 목표로 한다고 언급했다.
GSI Technology (Nasdaq: GSIT) a annoncé la publication d’un article dirigé par Cornell (ACM, présenté à Micro ’25) validant son Gemini-I APU Compute-In-Memory. Les benchmarks de Cornell ont révélé que l’APU offrait un débit de niveau GPU comparable à celui de la NVIDIA A6000 sur les charges RAG, tout en consommant plus de 98% d’énergie en moins qu’un GPU et en réduisant le temps de traitement de récupération jusqu’à 80% par rapport aux CPU sur des ensembles de données de 10 Go à 200 Go. GSI a noté que Gemini-II offre environ 10x de débit plus rapide et une latence plus faible, et Plato vise à une capacité edge à faible consommation accrue.
GSI Technology (Nasdaq: GSIT) kündigte die Veröffentlichung einer von Cornell geleiteten Studie (ACM, vorgestellt bei Micro ’25) zur Validierung ihres Compute-In-Memory Gemini-I APU an. Cornell-Benchmarks ergaben, dass der APU einen GPU-ähnlichen Durchsatz im Vergleich zur NVIDIA A6000 bei RAG-Workloads liefert, während er über 98% weniger Energie als eine GPU verbraucht und die Abfrageverarbeitungszeit im Vergleich zu CPUs auf Datensätzen von 10GB bis 200GB um bis zu 80% reduziert. GSI merkte an, dass Gemini-II rund 10x schnellerer Durchsatz und niedrigere Latenz bietet, und Plato zielt auf weitere Low-Power-Edge-Fähigkeiten ab.
GSI Technology (Nasdaq: GSIT) تُعلن عن نشر دراسة بقيادة كورنيل (ACM، مقدم في Micro ’25) تُ validates Gemini-I APU المبني على Compute-In-Memory. أشارت اختبارات كورنيل إلى أن الـAPU حقق إنتاجية على غرار GPU مقارنة بـ NVIDIA A6000 في أحمال RAG مع استهلاك طاقة يقل عن 98% مقارنة بوحدة GPU، كما أدى إلى تقليل وقت معالجة الاسترجاع حتى 80% مقارنة بمعالجات CPU على مجموعات بيانات من 10GB إلى 200GB. وأشارت GSI إلى أن Gemini-II يوفر نحو 10x من سرعة الإخراج ولها Latency أقل، و Plato يستهدف قدرات طرفية منخفضة الطاقة إضافية.
GSI Technology (纳斯达克: GSIT) 宣布发表由康奈尔大学主导的论文（ACM，在 Micro ’25 展示），验证其 Gemini-I APU 的 Compute-In-Memory。康奈尔基准测试发现，该 APU 在 RAG 工作负载下的 GPU 级吞吐量与 NVIDIA A6000 相当，同时比 GPU 少用超过 98% 的能量，并将检索处理时间相对于 CPU 在数据集从 10GB 到 200GB 的情况下缩短多达 80%。GSI 指出 Gemini-II 提供约 10x 的更快吞吐量和更低延迟，Plato 目标实现进一步的低功耗边缘能力。
- Energy >98% lower vs GPU
- Throughput comparable to NVIDIA A6000 on RAG workloads
- Processing time reduced up to 80% vs standard CPUs
- Gemini-II delivers roughly 10x faster throughput
- Benchmarks limited to RAG workloads over 10GB–200GB datasets
- Results reported on Gemini-I; Gemini-II and Plato are forward-looking
Insights
Cornell validation shows GSI's APU matches GPU throughput while slashing energy use, implying meaningful product-market leverage.
GSI Technology reports an independent, peer-reviewed evaluation (ACM; Micro ’25) showing the Gemini-I APU achieved throughput comparable to NVIDIA’s A6000 on RAG workloads while using over
Key dependencies and risks include reproducibility across broader workloads and system-level integration factors not quantified here, such as software stack maturity, end-to-end latency in deployed systems, and cooling/power realities in target environments. The Cornell study and the new analytical framework increase technical credibility, but system integrators must still validate latency, accuracy, and tooling for their specific stacks.
Watch for adoption signals and timelines: publications and conference presentation occurred on
SUNNYVALE, Calif., Oct. 20, 2025 (GLOBE NEWSWIRE) -- GSI Technology, Inc. (Nasdaq: GSIT), the inventor of the Associative Processing Unit (APU), a paradigm shift in artificial intelligence (AI) and high-performance compute (HPC) processing providing true compute-in-memory technology, announced the publication of a paper led by researchers at Cornell University. Findings confirmed that GSI Technology’s APU CIM (Compute-In-Memory) architectures can match GPU-level performance for large-scale AI applications with a dramatic reduction in energy consumption due to high-density and high-bandwidth memory associated with the CIM architecture.
Key findings include:
- GPU-class performance – The Gemini-I APU delivered comparable throughput to NVIDIA’s A6000 GPU on RAG workloads.
- Massive energy advantage – The APU delivers over
98% lower energy consumption than a GPU over various large corpora datasets, underscoring its efficiency and sustainability. - Faster and more efficient than CPUs – The APU’s unique design allows it to perform retrieval tasks several times faster than standard CPUs, shortening total processing time by up to
80% .
“Cornell’s independent validation confirms what we’ve long believed—compute-in-memory has the potential to disrupt the
Published on ACM and presented at the Micro ’25 conference, the paper by the Cornell research team titled “Characterizing and Optimizing Realistic Workloads on a Commercial Compute-in-SRAM Device,” represents one of the first comprehensive evaluations of a commercial compute-in-memory device under realistic workloads. The Cornell-led team benchmarked the GSI Gemini-I APU against established CPUs and GPUs, focusing on retrieval-augmented generation (RAG) tasks over datasets ranging from 10GB to 200GB.
The researchers’ findings point to significant opportunities for GSI Technology as customers increasingly require performance-per-watt gains across various industries, including Edge AI for power-constrained robotics, drones, and IoT devices, as well as defense and aerospace applications where the APU can deliver high performance in environments with strict energy and cooling constraints.
Mr. Shu continued, “This tremendous work by Cornell highlights CIM advantages using the Gemini-I silicon. Our recently released second-generation APU silicon, Gemini-II, can deliver roughly 10x faster throughput and even lower latency for memory-intensive AI workloads, while further improving energy efficiency. Looking ahead, Plato represents the next step forward, offering even greater compute capability at lower power for embedded edge applications. The APU’s unique combination of speed, efficiency, and programmability positions us to unlock high-growth opportunities across edge AI, data centers, defense, and other markets where energy efficiency is a critical strategic advantage.”
The Cornell study also introduced a new analytical framework for general-purpose compute-in-memory devices, providing optimization principles that strengthen the APU’s position as a scalable platform for developers and system integrators. A copy of the publication can be found on the GSI website at https://gsitechnology.com/characterizing-and-optimizing-realistic-workloads-on-a-commercial-compute-in-sram-device/.
ABOUT GSI TECHNOLOGY
Founded in 1995, GSI Technology, Inc. is a leading provider of semiconductor memory solutions. GSI's resources are focused on bringing new products to market that leverage existing core strengths, including radiation-hardened memory products for extreme environments and Gemini-I, the associative processing unit designed to deliver performance advantages for diverse artificial intelligence applications. GSI Technology is headquartered in Sunnyvale, California, and has sales offices in the Americas, Europe, and Asia. For more information, please visit www.gsitechnology.com.
About ACM
ACM publishes more than 50 scholarly peer-reviewed journals in dozens of computing and information technology disciplines. Available in print and online, ACM's high-impact, peer-reviewed journals constitute a vast and comprehensive archive of computing innovation, covering emerging and established computing research for both practical and theoretical applications. ACM journal editors are thought leaders in their fields, and ACM's emphasis on rapid publication ensures minimal delay in communicating exciting new ideas and discoveries.
Forward-Looking Statements
The statements contained in this press release that are not purely historical are forward-looking statements within the meaning of Section 21E of the Securities Exchange Act of 1934, as amended, including statements regarding GSI Technology’s expectations, beliefs, intentions, or strategies regarding the future. All forward-looking statements included in this press release are based upon information available to GSI Technology as of the date hereof, and GSI Technology assumes no obligation to update any such forward-looking statements. Forward-looking statements involve a variety of risks and uncertainties, which could cause actual results to differ materially from those projected. These risks include those associated with the normal quarterly and fiscal year-end closing process. Examples of risks that could affect our current expectations regarding future revenues and gross margins include those associated with fluctuations in GSI Technology’s operating results; GSI Technology’s historical dependence on sales to a limited number of customers and fluctuations in the mix of customers and products in any period; global public health crises that reduce economic activity; the rapidly evolving markets for GSI Technology’s products and uncertainty regarding the development of these markets; the need to develop and introduce new products to offset the historical decline in the average unit selling price of GSI Technology’s products; the challenges of rapid growth followed by periods of contraction; intensive competition; the continued availability of government funding opportunities; delays or unanticipated costs that may be encountered in the development of new products based on our in-place associative computing technology and the establishment of new markets and customer and partner relationships for the sale of such products; and delays or unexpected challenges related to the establishment of customer relationships and orders for GSI Technology’s radiation-hardened and tolerant SRAM products. Many of these risks are currently amplified by and will continue to be amplified by, or in the future may be amplified by, economic and geopolitical conditions, such as changing interest rates, worldwide inflationary pressures, policy unpredictability, the imposition of tariffs and other trade barriers, military conflicts and declines in the global economic environment. Further information regarding these and other risks relating to GSI Technology’s business is contained in the Company’s filings with the Securities and Exchange Commission, including those factors discussed under the caption “Risk Factors” in such filings.
Source: GSI Technology, Inc.
Contacts:
Investor Relations
Hayden IR
Kim Rogers
541-904-5075
Kim@HaydenIR.com
Media Relations
Finn Partners for GSI Technology
Ricca Silverio
(415) 348-2724
gsi@finnpartners.com
Company
GSI Technology, Inc.
Douglas M. Schirle
Chief Financial Officer
408-331-9802
FAQ
What did Cornell find when benchmarking GSI's Gemini-I APU versus NVIDIA A6000 for RAG in October 2025?
How much energy savings did the Cornell study report for GSIT's APU compared with GPUs?
What performance improvement did GSIT claim versus CPUs in the October 20, 2025 announcement?
What gains did GSI announce for its Gemini-II APU after the Cornell paper?
On what dataset sizes did Cornell evaluate GSIT's APU in the published study?
