Row64 started as a low-level hardware coding experiment.
The idea was to make the fastest possible evaluation and visualization of large datasets.
We ended up using many technologies from game programming and wrote Row64 from the ground up with a low-level, bare-metal hardware programming approach. A challenge to develop, Row64 required many iterations to increase speeds and dramatically improve user experience. Several breakthroughs happened that combined the GPU, modern SSDs, and multi-threading driving innovations in low-latency memory access and data throughput.
You can think of Row64 as a real time game engine. All computation is done live and we try to work in FPS (frames-per-second) instead of seconds.
Row64 is the first ground up application based on the GPU and low level hardware. It can be anywhere from 2-1000 times faster than traditional CPU driven applications.
This means faster data work, massive datasets, and next generation visuals.
The tools will feel familiar but the underlying engine is ground-breaking and fun.
Our research has shown that Business Analysts, Data Scientists & Data Analysts all work in different ways, but have overlapping workflow patterns. Combining data expert tools and business/data analyst tools, Row64 is a first-of-its-kind interactive data analytics engine that democratizes working on massive amounts of data.
- From our testing, Row64 is the leader in quickly opening and manipulating large files. We’re strongest at :
As we blend the spreadsheet with Python dataframes, we have a unique tool for spreadsheet users to learn Python, and Python users to see their data live.
You can also think of Row64 as a Python Preset Library.
We bring the power of Python and it’s most popular libraries to simple formulas that will be easy for spreadsheet users.
Row64 also helps everyone avoid the hassle of Python installation, and automatically setup the work environment. Batteries are included with Row64.
Row64 is free for non-commercial use. It’s also free for training and education of data science tools and business analysis.
If you find it useful and want to use Row64 commercially, please clicking on the paid version. Pricing links to upgrade on the front page of Row64.com
- Row64 uses the latest in GPU technology but will run on many older machines.
- You may need to upgrade your graphics drivers to get it to work well.
- If your GPU is older you may have certain features disabled like raytraced GEO and Charts.
No… that’s not something possible. But we are trying to make Row64 run as fast as possible on older hardware.
We’re launching with the Windows version first, just because we had the most client requests for this OS.
The first year of development of Row64 was done on MacBooks so the code is very portable to mac.
- We have done many tests with Linux & Mac builds and hope to get that out in the near future.
Starting Point: Strength at the Basics
3.5 years ago when Row64 started we had the idea that performance was a gap in the data market. Our founding concept - lightning fast at the basics:
- Fast scroll of 100M’s of records
- GPU Based: Sort, Dedup, Text Search, Fuzzy / Wildcard Search
These generally targeted data investigation and market research jobs we had done in the past. Ex: crawl through a huge pile of SEC files for competitor analysis.
This area is pretty much complete, we’ve recently dialed in a ton of speed records. For example our wildcard search is up to 250X faster than CPU search.
Our plan is to test our Python recipes (which are fast to make), get feedback when we release, and build the hits on the GPU to break records.
The team that started Row64 first made a website builder called CirclePad around 2008. They learned the power of the spreadsheet to a wide audience when they integrated a spreadsheet into the website builder. So there is a long history of our team helping make software easy to use and dialing in the details of how spreadsheets work.
A standard workflow is to import a “Raw Data” sheet, do “Column Formulas” and pull findings into a “Summary Sheet”. We’re transforming this workflow with Python Pandas Dataframes as the “Raw Data Transformer”. A powerful way to bring Python superpowers to the spreadsheet user.
So our recipe language runs spreadsheet Column Formula syntax. This is a minimum bar for those users
To reduce the cost of ads/marketing/sales we did a ton of work on what Python features are hard to do in spreadsheet and will be instantly useful. That will make us more like a “Swiss Army Knife” of data problems and get used out-of-box.
Several members of our team have done corporate jobs with lots of data work. An area of expertise is merge and reconciliation of data. Spreadsheets are typically limited at this so we’re putting quite a few recipes in this area.
To communicate all this we’ve broken down our recipes into categories and then also make sure we target hits and useful features in each category:
• Record breaking speed at loading and working with large files (over 500M records)
• Search up to 50X faster than traditional CPU search & unlimited characters
• Familiar spreadsheet UI with easy & powerful pattern matching
• Record breaking speeds at fuzzy merge & fuzzy dedup
• Swiss army knife of formats:
• EXCEL, DOC, HTML, PARQUET, JSON, SQL, STATA, SAS, XML
• Simplified scraping of web site data
• Explode & Combine: Many files to one
• Subscription Analysis Simplified
• 1 Click Pareto Analysis - quick graphs and sample sets
• Quick binning of time into day/month/year
• Entity Recognition
• Sentiment Analysis
• 1 Click data summary reports
• Summaries that would take a day in a spreadsheet take minutes to seconds
• Display and filter over 50 million records at over 60 FPS
• Next generation heat maps and ray traced graphics
• Manage by City, IP Address and region
• Regression & Predictive Values
• Categorical & Multiple-imputation data
- Row64 is designed to take advantage of both local hardware and cloud based
data. We intend to have more features in this area in the future.