What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
iTechGuides is reader-supported. When you buy through links on our site, we may earn an affiliate commission. As an Amazon Associate I earn from qualifying purchases. Learn more
Vector search design depends on where the working set lives. When vectors and the index fit in RAM, HNSW can follow graph links through fast memory accesses. When vectors are served from object storage, Infino describes a different approach: OPANN routes queries to likely regions and fetches bounded, contiguous ranges that can be requested together. Sq16 makes each vector smaller in either path. This is Infino’s architecture, not a universal rule for vector databases.
Why HNSW fits a RAM-resident working set
HNSW searches a graph rather than scoring every vector. A query starts at an upper layer and follows promising neighbor links toward candidates at the bottom. Each hop helps determine where the search goes next, so the memory addresses are dependent and difficult to plan in advance.
When the graph and vectors are resident in RAM, those hops are memory accesses. The graph lets the search examine a candidate subset instead of the full corpus. Infino describes HNSW as an excellent in-memory index, while arguing that the same dependent-access pattern is poorly suited to remote object storage. That is an assessment of access patterns in its architecture, not a claim that HNSW cannot be used with storage-backed systems.
Why Infino uses OPANN for object storage
Infino’s OPANN path changes the order of work. A routing structure first identifies likely vector regions; the system then fetches bounded, contiguous ranges, scores compact vectors, and reranks a shortlist. Because the likely ranges can be identified before their contents arrive, Infino says it can issue requests concurrently, coalesce reads, and make use of caching.
#1 Best Overall
- MEET THE NEXT GEN: Consider this a cheat code; Our Samsung 990 PRO Gen4 SSD helps you reach near max performance with lightning-fast speeds; Whether you’re a hardcore gamer or a tech guru, you’ll get power efficiency built for the final boss
- REACH THE NEXT LEVEL: Gen4 steps up with faster transfer speeds and high-performance bandwidth; With a more than 55% improvement in random performance compared to 980 PRO, it’s here for heavy computing and faster loading
- THE FASTEST SSD FROM THE WORLD'S FLASH MEMORY BRAND: The speed you need for any occasion; With read and write speeds up to 7450/6900 MB/s you’ll reach near max performance of PCIe 4.0 powering through for any use
- PLAY WITHOUT LIMITS: Give yourself some space with storage capacities from 1TB to 4TB; Sync all your saves and reign supreme in gaming, video editing, data analysis and more
- IT’S A POWER MOVE: Save the power for your performance; Get power efficiency all while experiencing up to 50% improved performance per watt over the 980 PRO; It makes every move more effective with less consumption
This matters because a remote read is not equivalent to a RAM load. With graph traversal, the result of one read may be needed before the next address is known, turning a sequence of hops into serial network requests. A planned group of ranges can instead be fetched together. Loading a cold graph into local memory before searching is another possible pattern, but it changes the cost to a bulk load and requires enough local capacity to hold the working set.
Keeping results aligned with row data
In the described implementation, vectors and the index are stored in a Parquet superfile. Infino says this lets returned IDs resolve against the same snapshot as scalar filters and BM25 results. That is an implementation-specific consistency choice; it should not be assumed for other object-storage vector systems.
Rank #2
- Ideal for high speed, low power storage
- Gen 4x4 NVMe PCle performance
- Up to 6,000MB/s read, 4,000MB/s write
- Includes Acronis cloning software
- 5-year limited warranty
What Sq16 changes
Sq16 is Infino’s 16-bit representation for vector dimensions. The article says Infino accepts fp32 input and internally represents each dimension with 16 bits, reducing the vector payload without changing the overall search architecture.
Quick wins for a faster PC:
Repair Windows errors before they cause bigger problemsFix Now →Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →Clear out junk files and repair common Windows errorsFree Scan →For scale, Infino’s August 20, 2026 article illustrates that a 1024-dimensional fp32 vector occupies 4 KiB before metadata. Ten million such vectors contain about 40 GiB of raw vector values. At 16 bits per dimension, its illustration is about 2 KiB per vector, or 20 GiB for ten million, again before index metadata. Graph links, metadata, caches, and other structures add to a system’s actual footprint.
Rank #3
- SPEED UP PROJECTS. Launch creator applications fast with uncompromising PCIe 4.0 read speeds up to 7,100MB/s,[2] (1TB and 2TB[1] models) and write speeds up to 6,700MB/s[2] (1TB[1]-4TB[1] models).
- CREATE AND STORE MORE. Make more room for your 4K videos and high-resolution images with capacities from 500GB[1] up to 4TB[1] on M.2 2280 built with our trusted 8th generation SANDISK BiCS QLC 3D CBA NAND.
- IT GOES WHERE YOU GO. With an all-new power efficient design, your drive delivers high performance with low power, giving you more time to be productive while on the go.
- UNCOMPROMISED RELIABILITY. With up to 1,200 TBW[3] (4TB[1] model) endurance rating, your drive is designed for creators.
- KEEP YOUR DRIVE UPDATED. Monitor your SSD’s performance and check for updates with the downloadable SANDISK Dashboard application.[5]
The broader tradeoff is not unique to Infino. OpenSearch documentation says 16-bit vectors can require half the memory of 32-bit vectors in the best case, and explains that quantization can affect recall depending on the technique and vector distribution. Its documented mechanisms and memory estimates are specific to OpenSearch; they do not establish how Infino’s OPANN internals work.
What Infino’s published measurements show—and do not show
In an Infino-reported Cohere run on 9.4 million vectors at 768 dimensions with top-10 results, its post-drain OPANN + Sq16 path reported recall@10 of 0.995, warm p50 latency of 9.37 ms, and warm p99 latency of 10.25 ms. These are vendor-published figures from an article dated August 20, 2026, not independently reproduced results.
Rank #4
- HUGE SPEED BOOST: Get random read/write speeds that are 40%/55% faster than 980 PRO; Experience up to 1400K/1550K IOPS, while sequential read/write speeds up to 7,450/6,900 MB/s reach near the max performance of PCIe 4.0*
- BREAKTHROUGH POWER EFFICIENCY: Use less power and get more performance; Enjoy up to 50% improved performance per watt over 980 PRO, plus optimal power efficiency with max PCIe 4.0 performance**
- SMART THERMAL CONTROL: Samsung's own nickel-coated controller delivers effective thermal control; With its slim size, 990 PRO is a perfect fit for desktops and laptops that meet the PCI-SIG D8 standard***
- THE CHAMPION MAKER: Up to 65% improvement in random performance enables faster loads for an ultimate gaming experience on PS5 and DirectStorage PC games****
- SAMSUNG MAGICIAN SOFTWARE: Get the most out of your SSD with Samsung Magician's advanced yet intuitive optimization tools; Monitor drive health, protect valuable data, and receive important updates for your 990 PRO
The cold-start figures from the same described path provide essential context: the cold open median was 661 ms, and the first cold query took 4.60 seconds while fetching 2.55 GiB of index ranges. A warm p50 or p99 alone does not describe the cost of opening or querying a cold index.
The Tool Desk
Outbyte PC Repair FREEClear out junk files and repair common Windows errorsFree Scan →Outbyte Driver Updater FREEScan for outdated or missing drivers - takes under a minuteDriver Scan →These figures do not constitute a controlled comparison against HNSW or another system. OpenSearch’s documentation discusses memory, quantization, recall, on-demand loading, and rescoring as design tradeoffs, but does not rank OPANN against HNSW. AWS separately describes S3 Vectors as purpose-built object storage for storing and querying vectors; its product claims should not be read as evidence that it uses Infino’s OPANN, Parquet superfiles, or Sq16 design.
Best Value
- This product has been replaced by our latest generation. Please search for the SANDISK Optimus GX 7100 NVMe SSD
- HIGH-OCTANE GAMING. Experience speeds up to 7,250MB/s read and 6,900MB/s write (1-2TB models), with up to 35% faster performance than previous generation.
- PURPOSE-BUILT. Designed for serious on-the-go gamers, with a PCIe Gen4 interface and SANDISK’s next generation TLC 3D NAND.
- MORE TIME TO CLEAR THAT CHECKPOINT. Built with laptops and handheld gaming devices in mind, with up to 100% more power efficiency over the previous generation.
- DO MORE WITH DASHBOARD. Ensure your drive is optimized for prime performance with the downloadable WD_BLACK Dashboard (Windows only).
How to decide which access pattern suits a workload
The useful question is not simply whether a system supports a particular index. Evaluate where the working set sits and what the complete query path costs in both hot and cold states.
- Working-set fit: Can the vectors and graph stay resident in RAM, including metadata and other application needs?
- Remote I/O shape: Does the query make dependent reads one at a time, or can it identify and batch ranges?
- Cold and warm behavior: Measure index open time and first-query latency alongside warm p50 and p99.
- Recall at the target k: Test the workload’s required recall, and determine whether a compact candidate set is rescored with higher-precision vectors.
- Total footprint and build cost: Include graph links, metadata, caches, loading overhead, and index construction—not just raw vector bytes.
- Data consistency: Check how vector IDs line up with filters and associated row data, particularly when data changes.
If the working set is hot and fits in memory, HNSW’s graph traversal can make sense. If the serving path depends on object storage, the relevant design question is whether the system can avoid turning each query into a chain of dependent remote reads. Infino’s OPANN + Sq16 combination is one implementation response to that constraint.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
