Key Takeaways
- Chia compressed plots let you fit more effective storage into your existing drives — without buying new hardware.
- BladeBit (official, C1–C9, with C8 currently unavailable) requires no dev fee. GigaHorse (third-party) charges a small percentage of farming rewards.
- Low compression levels (C1–C5) run fine on a decent CPU. C6 also runs on a fast CPU. Only C7 and above generally need a GPU for harvesting.
- For C7 harvesting, a GTX 1060 with around 600–700 MB of VRAM is sufficient. The 8 GB VRAM requirement applies to plotting with BladeBit CUDA — not farming.
- Lookup times must stay under roughly 5–10 seconds to avoid missing rewards — always benchmark before committing your whole farm.
- Chia’s upcoming PoS 2.0 hard fork (targeted ~June 1, 2026) will make compressed plots obsolete. Plan your replot strategy now.
Chia compressed plots are plot files intentionally left incomplete during creation. Your harvester’s CPU or GPU recomputes the missing data on the fly every time the network sends a challenge. The result: smaller files on disk, more plots per drive, and more potential XCH rewards — at the cost of more compute and slightly more power draw. Since Chia 2.0 launched in 2023, compressed plotting has become the standard approach for any competitive farm.
This article is a supporting guide for our pillar page on Chialisp and next-gen Chia smart contracts. Here we dig into the specific hardware, settings, and trade-off decisions every Chia farmer needs to make when choosing a compression level.
What Are Chia Compressed Plots?
Every Chia plot is built from seven cryptographic tables. In an uncompressed plot, all seven tables are fully written to disk during the plotting process. A standard uncompressed k32 plot takes up about 101.4 GiB of space.1
A compressed plot intentionally drops one or more of those tables — or portions of them — at plot creation time. The plot file ends up smaller on disk. When the Chia network issues a challenge to see if your plot has a winning proof, your harvester’s CPU or GPU regenerates the dropped data in real time. If the proof passes, you win the reward. If it does not, the recomputed data is discarded and the cycle repeats with the next challenge.
This is a time-space trade-off at its core: you trade a fraction of your CPU or GPU’s time for the ability to store more effective space on the same drives. Technically, “compressed” is a misnomer — these are more accurately described as truncated or partial plots, since the missing data is regenerated rather than decoded from a compressed form.
How Compression Levels Work
Compression levels in Chia are labeled C1, C2, C3, and so on. Each level drops slightly more data from the plot during creation. BladeBit CUDA, the official Chia plotter, supports levels C1 through C9 (C8 is currently unavailable). GigaHorse, the third-party plotter by developer madMAx, uses its own formats that are not directly comparable to BladeBit’s level numbers — GigaHorse v1.0 runs C1–C9, while GigaHorse v2.0/3.0 uses a different format running C11–C33.3
Chia Network’s official documentation states clearly: “Chia’s plot format was designed such that higher compression levels would yield linear gains in size, at a cost of an exponential increase in required computational power.”2 In plain terms: each step up the compression ladder saves a bit more space but demands much more from your harvester hardware.
Uncompressed vs. Compressed Plot Sizes (BladeBit k32)
| Compression Level | Plot Size (GiB) | Reward Increase vs. Uncompressed | Minimum Harvester Hardware |
|---|---|---|---|
| Uncompressed (C0) | 101.4 GiB | 0% | Raspberry Pi 4 (4 GB RAM) |
| C1 | 87.5 GiB | +15.9% | Raspberry Pi 4 |
| C2 | 86.0 GiB | +17.9% | Raspberry Pi 4 |
| C3 | 84.5 GiB | +20.0% | Raspberry Pi 4 |
| C4 | 82.9 GiB | +22.3% | Desktop CPU |
| C5 | 81.3 GiB | +24.7% | Fast CPU (e.g., Intel Xeon class) |
| C6 | 79.6 GiB | +27.4% | Fast CPU (e.g., Intel Xeon class) |
| C7 | 78.0 GiB | +29.8% | NVIDIA GPU (GTX 1060 or better) |
| C9 | 75.2 GiB | +34.8% | NVIDIA GPU |
Source: Official Chia Network compression table.2 Minimum harvester specs assume a farm in the 500–1,000 TiB range. Larger farms require progressively more capable hardware at any given level. C8 is currently unavailable. Always verify current plot sizes at docs.chia.net.
How Compressed Plots Affect Farming Performance
The single most important metric for compressed plot farming is lookup time — how long your harvester takes to respond to a Chia network challenge. If your harvester is too slow, it will miss signage points. Missed signage points mean missed rewards. Chia Network recommends keeping maximum lookup times under roughly 5 seconds, with 10 seconds being a hard upper boundary to watch for in your logs.2
This became more demanding in June 2024. Chia implemented a plot filter reduction as part of CHIP-0012, cutting the filter from 512 to 256. What this means practically: twice as many plots now pass the filter at each signage point, which doubles your harvester’s decompression workload compared to before the reduction. If you sized your hardware to handle a large farm at C6 before June 2024, that same setup may now struggle at the same compression level unless you had headroom to spare.4
CPU vs. GPU Decompression: Which Do You Need?
Whether you need a GPU comes down almost entirely to your compression level and farm size. C1 through C6 can typically be handled by a modern CPU — C1 through C3 even run on a Raspberry Pi 4 for most farm sizes, while C5 and C6 require a faster CPU such as an Intel Xeon class processor. Only at C7 and above does a GPU become the practical requirement for reliable lookups.
An important distinction: the GPU requirements for plotting and harvesting are different. To plot compressed plots using BladeBit CUDA, you need an NVIDIA GPU with CUDA capability 5.2 or higher and at least 8 GB of VRAM. To harvest (farm) C7 plots, the GPU requirement is much lower — around 600–700 MB of VRAM, meaning a GTX 1060 will work fine.2 You can plot on a high-end machine and farm the resulting plots on a lower-spec GPU harvester. The plotting hardware and farming hardware do not need to match.5
Chia Network noted at the v2.0 launch that GPU harvesting is highly efficient, with most large farms “only increasing the overall power consumption by a few percent” when using GPU decompression.4
Energy Trade-offs at Higher Compression
Compressed plots increase your farm’s power draw because your CPU or GPU is doing continuous extra work. At lower levels (C1–C3), the added power cost is minimal — often just a few extra watts per harvester. At high levels like C6–C7, your overall system power consumption will rise meaningfully compared to an uncompressed setup at the same raw storage volume, though the exact increase depends heavily on your specific hardware. The right compression level is the one where the value of your extra effective capacity — measured in additional XCH rewards — outweighs the cost of that extra electricity. For farms in regions with high electricity prices, this calculation often favors staying at C3 or C5 rather than pushing to C7.
Quick Decision: Which Compression Level Is Right for You?
| Your Situation | Recommended Level | Why |
|---|---|---|
| Small farm (<50 TB), no GPU, low power costs | C1–C3 (BladeBit) | CPU-only decompression on a Pi 4; +15.9% to +20% effective reward increase |
| Mid-size farm (50–200 TB), capable CPU, moderate power costs | C3–C6 (BladeBit) | +20% to +27.4% reward increase; C6 still Fast-CPU territory — run simulator to confirm lookup times |
| Large farm (200 TB+), dedicated GPU harvester, low power costs | C7 or C9 (BladeBit) | Maximum effective space from existing drives; GPU required; ~600–700 MB VRAM minimum for harvesting |
| High electricity costs in any farm size | C1–C3 or uncompressed | Extra power draw at higher levels may negate reward gains |
| Farm running on mixed or low-spec hardware | Uncompressed or C1 | Avoids missing rewards due to slow lookup times on older CPUs |
BladeBit vs. GigaHorse: Choosing Your Chia Compressed Plotter
Once you decide to use compressed plots, you face a second major decision: which plotting and farming software to use. There are two main options — BladeBit (the official Chia plotter) and GigaHorse (a popular third-party tool). They are not interchangeable. Plots created with one cannot be farmed with the other’s stack.
BladeBit CUDA (Official)
BladeBit CUDA is maintained by Chia Network Inc. and ships with the Chia client starting from version 2.0. It supports compression levels C1 through C9 (C8 is currently unavailable) for k32 plots. To plot with compression, you need an NVIDIA GPU with at least 8 GB of VRAM and CUDA capability 5.2 or higher. For full RAM plotting (fastest and no SSD wear), you need at least 256 GB of system RAM. If you have less RAM but still have a compatible GPU and a large NVMe SSD, BladeBit can use the SSD as a temporary buffer instead — requiring 128 GB RAM in that configuration.8
BladeBit also supports CPU-based compressed plotting via BladeBit RAM, which requires around 416 GB of system RAM — a valid option for large farm operators with high-memory servers but no CUDA GPU available for plotting. BladeBit compressed plots are farmed directly by the standard Chia client — no special farmer software needed. There is no developer fee. For most farmers who want a clean, officially supported path, BladeBit is the straightforward recommendation.
GigaHorse (Third-Party by madMAx)
GigaHorse, developed by madMAx, offers a higher compression ceiling. GigaHorse v1.0 supports levels C1–C9 using a different format from BladeBit, while GigaHorse v2.0/3.0 uses yet another format with levels C11–C33. These version formats are not interchangeable even within GigaHorse itself. GigaHorse can also create compressed plots using a CPU rather than a GPU, which is a meaningful option for farmers without a CUDA-capable card. However, GigaHorse plots require the GigaHorse harvester and farmer software — they cannot be farmed by the standard Chia client. You also pay a developer fee: 1.562% of farming rewards when CPU-farming, and 3.125% when GPU-farming.6
GigaHorse’s compression level numbering does not map directly to BladeBit’s. A BladeBit C3 plot is closer in size and decompression load to a GigaHorse v1.0 C1 or GigaHorse v2.0/3.0 C11, depending on which version you’re comparing. Always compare actual plot sizes and measured lookup times rather than assuming level numbers are equivalent between the two tools.3
| Feature | BladeBit CUDA | GigaHorse |
|---|---|---|
| Developer / Maintainer | Chia Network Inc. (official) | madMAx (third-party) |
| Compression levels (k32) | C1–C9 (C8 currently unavailable) | v1.0: C1–C9 / v2.0–v3.0: C11–C33 |
| GPU required for plotting? | Yes for BladeBit CUDA (NVIDIA CUDA 5.2+, 8 GB VRAM); BladeBit RAM is CPU-only (416 GB RAM) | No (CPU plotting supported) |
| GPU required for harvesting? | Only at C7+; ~600–700 MB VRAM minimum | C1–C9 CPU-farmable; higher levels GPU-only |
| Farmer software required | Standard Chia client | GigaHorse farmer (custom) |
| Developer fee | None | 1.562% (CPU) / 3.125% (GPU) |
| Open source | Yes | Binaries only |
| Compatible with standard Chia client? | Yes | No (requires GigaHorse node) |
How to Choose Your Compression Level: Practical Steps
Choosing a compression level is not a “set it and forget it” decision based on one rule. It is a calculation that depends on your specific hardware, farm size, and electricity cost. Chia Network provides a simulator tool built into BladeBit’s standalone version that makes this process straightforward.
Step 1 — Run the BladeBit Simulator
Before committing to replot your entire farm at a given level, use the BladeBit simulate command. The basic form runs a quick analysis against a single compressed plot and extrapolates your farm’s maximum capacity:
bladebit simulate -f 256 <path to plot file>
The -f 256 flag reflects the current plot filter size since June 2024. For a more detailed real-time simulation — for example, to simulate running a 1 PB farm for ten minutes — you can add the --power (duration in seconds) and --size (farm size) flags:
bladebit simulate --power 600 --size 1PB -f 256 <path to plot file>
Replace the farm size with your actual planned total. A maximum simulated lookup time around 5 seconds is a safe target. If it comes out near or above 10 seconds, lower the compression level, reduce farm size on that harvester, or upgrade to a faster CPU or GPU before replotting.2
Step 2 — Benchmark a Small Batch First
Once the simulator looks promising, plot a small test batch at your chosen compression level — maybe 10–20 plots. Monitor your harvester logs in real time, paying close attention to actual lookup times and stale proof percentages. Stale proofs indicate your harvester is too slow to respond to some signage points. If stale proofs appear consistently above 1–2%, you need to back off the compression level or improve your hardware before plotting the rest of your farm.
Step 3 — Calculate Your Power Trade-off
Higher compression earns more XCH by increasing your effective capacity, but it also draws more power. A C7 farm with GPU decompression will consume more electricity than the same raw storage running uncompressed, but also provides more effective space and therefore more expected rewards. At low electricity rates, this math usually favors compression. At high rates, you may find C3 or C5 is your sweet spot — or that staying uncompressed is the cleaner economic choice. Chia Network provides a Total Cost of Ownership (TCO) spreadsheet you can copy and fill in for your specific farm to calculate this more precisely.2
Step 4 — Use xch.farm Tables for Reference
The third-party resource xch.farm maintains updated tables showing maximum farm sizes at each compression level for various CPU and GPU configurations, adjusted for the current plot filter of 256. These are a useful cross-check against the BladeBit simulator results, especially if you want to quickly estimate whether a planned hardware setup will work before purchasing equipment.3
The Critical Factor Every Chia Farmer Needs to Know for 2025–2026
Chia’s upcoming Proof of Space 2.0 (PoS 2.0) hard fork will fundamentally change the compressed plot landscape. The hard fork activation is currently targeted at block height 8,800,000 — expected around June 1, 2026 — followed by a transition period ending in the first half of 2027, when old-format plots will no longer be supported.10
Under the new format, compression incentive will be severely limited. Even marginal bit-dropping — such as removing just one bit per entry to save roughly 0.5% of space — could incur more energy per TiB than the honest farmer, making it economically self-defeating for most setups. The PoS 2.0 FAQ notes that in the most extreme future scenarios with very efficient compute, bit-dropping might achieve up to 10% space savings, but only at the cost of higher energy expenditure and the ongoing risk of having to replot as minimum difficulty requirements increase over time.10
What this means for your farm right now: if you are planning to replot to compressed formats in 2025, it makes strategic sense to use lower compression levels (C1–C5) rather than committing enormous GPU hours to a full C7 or C9 replot, since you will need to replot again for PoS 2.0 anyway before the first half of 2027. The exception is if your hardware and electricity costs make high compression clearly profitable in the near term — in which case the gains between now and the transition could still be worthwhile.
Case Study: Small-Scale Farmer Replot Decision
A 100 TB farmer with a mid-range CPU harvester (no dedicated GPU) who replots from uncompressed to BladeBit C4 gains a 22.3% reward increase per the official Chia compression table — essentially turning 100 TB of raw storage into roughly 122 effective TB — with only a desktop-class CPU required for decompression. At typical XCH reward rates, that represents a meaningful gain in expected monthly earnings, with minimal added electricity cost. The same farmer pushing to C7 would require a dedicated GPU for harvesting, and would still need to replot for PoS 2.0 before 1H 2027. For that profile, C4 is the smarter near-term move.2
As Chia co-founder Bram Cohen has noted about the new proof format: the design goal is to make the honest farmer — who simply stores data and reads it — the economically dominant strategy, rather than rewarding those who invest in compute-intensive recompression rigs.
Conclusion
Chia compressed plots are one of the most practical tools available to active Chia farmers. They let you squeeze more productive capacity out of drives you already own, and at moderate compression levels — particularly C3 through C6 — the hardware demands are modest enough for most setups to handle comfortably. The decision is not about chasing the highest possible compression number. It is about finding the level where your hardware, electricity costs, and upcoming network changes all point to the same answer. Run the BladeBit simulator, test a small batch, watch your lookup times, and build from there. And keep the PoS 2.0 timeline on your radar: the plots you create today will need to be replaced before 1H 2027, so plan your replot budget and strategy with that deadline in mind. For a deeper look at how Chia’s farming economics work, see our guide on XCH block rewards and the 2024 plot filter reduction.
Chia Compressed Plots FAQs
What are chia compressed plots and how do they differ from regular plots?
Chia compressed plots are plot files intentionally created without some of their data, making them smaller on disk. Unlike regular (uncompressed) plots, chia compressed plots require your harvester’s CPU or GPU to recompute the missing data each time the network sends a farming challenge — trading compute time for reduced storage footprint.
What hardware do I need to farm chia compressed plots at high compression levels?
For BladeBit compression levels C7 and above, you need an NVIDIA GPU connected to your harvester — but the VRAM requirement for farming is much lower than for plotting: around 600–700 MB, meaning a GTX 1060 will work. Levels C1–C6 can typically be handled by a CPU, though you should always run the BladeBit simulator to confirm your specific setup keeps lookup times under 5–10 seconds.
Is GigaHorse better than BladeBit for chia compressed plots?
GigaHorse offers higher compression ceilings and CPU-based plotting without a GPU, but it requires its own custom farmer software and charges a developer fee of 1.562–3.125% of rewards. BladeBit has no fee, works with the standard Chia client, and is fully open source — making it the simpler choice for most farmers, especially given the PoS 2.0 transition ahead.
Will chia compressed plots still work after the Chia PoS 2.0 hard fork?
Current-format compressed plots (BladeBit and GigaHorse) will remain valid through the transition period after the hard fork, which is targeted at June 2026, but will be fully phased out in the first half of 2027. All farmers will eventually need to replot using the new proof format, which is designed to make high-level compression economically nonviable going forward.
How do I find the right compression level for my chia compressed plots?
The most reliable method is to first run the BladeBit standalone simulator with bladebit simulate -f 256 [path to a compressed plot] for a quick capacity estimate, then use the extended simulation mode (--power 600 --size [your farm size]) for a real-time 10-minute benchmark. Follow that with a small real-world test batch before replotting your entire farm at that level.
Chia Compressed Plots Citations
- Chia Network – Plotting Basics | docs.chia.net
- Chia Network – Choosing a Compression Level (Official Table) | docs.chia.net
- XCH.farm – Chia Compressed Plots Reference Tables, Hardware Capacity & GigaHorse Details | xch.farm
- Chia Network – Version 2.0.0 Release Notes (Plot Filter Reduction, GPU Efficiency) | chia.net
- Chia Network – Plotting Hardware Requirements (CUDA, VRAM, RAM) | docs.chia.net
- madMAx – Chia GigaHorse GitHub Repository (Compression Levels, Dev Fee Structure) | github.com
- Chia Network – Plotting Software Overview (BladeBit CUDA Requirements) | docs.chia.net
- Chia Network – Plotting Software Overview (BladeBit CUDA Requirements) | docs.chia.net
- Chia Network – Approaching the Next Generation of Proof of Space (PoS 2.0 Hard Fork Timeline) | chia.net
- Chia Network – New Proof of Space 2.0 FAQ (Compression Limits Under New Format) | docs.chia.net
SEO Deliverables
Slug: chia-compressed-plots-tradeoffs-settings
Meta Title (54 chars): Chia Compressed Plots: Settings That Save You Money
Meta Description (154 chars): Learn exactly which chia compressed plots level matches your hardware. Avoid missed rewards, wasted GPU hours, and costly replots before the PoS 2.0 deadline.
Target Keywords: chia compressed plots, bladebit compression levels, gigahorse vs bladebit, chia plot filter 256, chia proof of space 2.0 replot
DALL-E Image Prompts:
- A clean data center shelf filled with rows of hard drives glowing green, with holographic percentage labels floating above each drive showing compression ratios, dark background, futuristic and technical, landscape orientation, photorealistic style
- A split-image illustration showing a Raspberry Pi 4 on the left side labeled “C1–C3” and a sleek NVIDIA GPU on the right labeled “C7+”, connected by a glowing Chia logo in the center, dark blue tech background, landscape orientation, modern flat design
- An infographic-style hero image of a Chia farm decision tree: icons for farm size, electricity bolt, and GPU branching into colored paths leading to compression levels C1 through C9, on a clean dark background with Chia green accents, landscape orientation
