ServerBudget

Cheap VPS Benchmark Per Dollar: What $1 a Month Actually Buys

By Marcus Reid · Published Nov 11, 2026 · Last updated Nov 11, 2026

Short answer: With the public pages of the three comparison sites locked in for cheap vps hosting, you can compute exactly how much advertised hardware one dollar buys, but you cannot compute how much measured speed it buys. Dividing each published plan by its monthly price gives a hard, recomputable result: the $10 entry tier buys 0.100 vCPU, 0.100 GB of RAM and 2.00 GB of NVMe storage per dollar, while the $11 entry tier at another vendor buys 0.091 vCPU, 0.182 GB of RAM, 5.45 GB of storage and 0.182 TB of transfer per dollar. The cheapest tiers win on almost every axis: one vendor's NVMe per dollar halves from 2.00 GB to 1.00 GB between its $10 and $80 plans, and the other vendor's vCPU per dollar falls from 0.118 to 0.045 while its TB per dollar falls from 0.182 to 0.056 between $17 and $179. Not one of these three sites publishes an IOPS figure, a CPU score or a latency measurement beside a price, and the only third-party report cited in this comparison supplies rankings without numbers. So the honest answer is that spec-per-dollar is computable today and performance-per-dollar is not. Below is the full table, the arithmetic, the public report's test design, and the public probe command that lets you produce the missing half yourself. Every figure in this article is transcribed from a public page listed under Sources.

The verdict, in one line

QuestionAnswer from published data
How much CPU, RAM and disk does $1/month buy?Computable, and listed tier by tier below
How much measured speed does $1/month buy?Not computable: no priced plan publishes a benchmark result
Does the cheapest tier give the most hardware per dollar?Yes on CPU, disk and transfer at both vendors with tiered pricing
Can you compare the two vendors' storage columns directly?No: one publishes NVMe capacity, the other publishes capacity without a disk type
What would close the gap?A public probe run on each tier, divided by the same monthly price

1. The one number you can actually verify

A plan page gives you two kinds of facts: a monthly price and a list of resources. Neither is a performance measurement, but both are published, dated and checkable. Dividing one by the other produces a ratio that anyone can recompute from the source page. That is the entire method used in this article.

spec_per_dollar = published_resource_amount / published_monthly_price_usd

    examples:
      vCPU per $  = 1 vCore / $10 = 0.100
      RAM  per $  = 1 GB    / $10 = 0.100
      NVMe per $  = 20 GB   / $10 = 2.00
      TB   per $  = 2 TB    / $11 = 0.182  (rounded to three decimals)

Every division in the tables below is derived arithmetic, not a vendor quote. Prices are the monthly "starting at" or list prices printed on the plan pages, read on 2026-11-11. No tax, promotion, renewal price or add-on is folded in, because those are separate published figures belonging to a different comparison.

2. Table A: four flat-priced tiers, recomputed per dollar

One vendor publishes four KVM tiers with a 1 Gbps port, NVMe capacity and unmetered traffic that is footnoted to a fair-use policy. [S1]

PlanMonthlyvCPURAMNVMevCPU per $RAM GB per $NVMe GB per $
KVM1-US$1011 GB20 GB0.1000.1002.00
KVM2-US$2022 GB25 GB0.1000.1001.25
KVM4-US$4044 GB40 GB0.1000.1001.00
KVM8-US$8088 GB80 GB0.1000.1001.00

Two findings fall straight out of the arithmetic. First, vCPU per dollar and RAM per dollar are exactly 0.100 on all four tiers: the price scales linearly with those two resources, so moving up a tier buys more capacity but not a better ratio. Second, NVMe per dollar is not linear. It drops from 2.00 GB per dollar on the $10 tier to 1.00 GB per dollar on the $40 and $80 tiers, so the cheapest plan gives twice the storage per dollar of the most expensive one. [S1]

The traffic column is deliberately absent from this table. "Unmetered" is a policy statement, not a quantity, and the footnote on the same page points to a fair-use policy rather than a TB allowance. [S1] A ratio needs a number in the numerator, and there is none here.

3. Table B: eight Compute tiers, recomputed per dollar

A second vendor publishes a 20-plan Cloud VPS table split into Compute, Memory and CPU groups, with vCPU, memory, storage, bandwidth, a monthly starting price and an hourly price. The eight Compute rows, recomputed, are: [S2]

MonthlyvCPURAMStorageBandwidthvCPU per $RAM GB per $Storage GB per $TB per $
$1112 GB60 GB2 TB0.0910.1825.450.182
$1724 GB80 GB3 TB0.1180.2354.710.176
$2226 GB120 GB4 TB0.0910.2735.450.182
$4548 GB240 GB5 TB0.0890.1785.330.111
$67412 GB320 GB6 TB0.0600.1794.780.090
$90616 GB440 GB7 TB0.0670.1784.890.078
$123624 GB540 GB8 TB0.0490.1954.390.065
$179832 GB640 GB10 TB0.0450.1793.580.056

Here the ratio is not flat. vCPU per dollar peaks at 0.118 on the $17 plan and falls to 0.045 on the $179 plan, a 2.6-fold spread. TB per dollar falls from 0.182 to 0.056, a 3.3-fold spread. RAM per dollar, by contrast, stays in a narrow 0.178 to 0.195 band on seven of the eight rows, with the $22 plan as the outlier at 0.273. [S2]

So two independently published price lists agree on one point: on the advertised resources, the cheapest tiers buy the most per dollar. Neither list says anything about how fast those resources run.

4. What the tables say, and what they cannot say

Three limits matter before anyone uses these numbers as a buying decision.

There is also a definitional gap. One of the comparison sites defines a fast VPS as one with a global response time under 180 ms in its own server speed checker, and lists SSD or NVMe storage, nearby data centers, free caching and HTTP/3 as supporting features. [S3] That is the only numeric speed threshold in the whole comparison set, and it is a threshold for a category, not a published measurement for any priced plan. Its roundup lists 14 starting prices from $1 to $100 per month but attaches no response-time reading to any of them. [S3]

The figures here cover benchmark results per dollar; for the part that follows, the cheap VPS decision matrix page walks through the same figures in the order you need them before ordering.

If you came here for benchmark results per dollar, the VPS downgrade and rollback page covers the rollback rules providers print after an upgrade.

Reading about benchmark results per dollar raises a separate question, and the cheap VPS entry price page shows what the lowest advertised tier actually includes.

5. Why performance per dollar is still uncomputable from published data

One of the vendor pages cites a third-party analyst to claim its VPS is faster than three named competitors. [S4] Following that citation leads to a public write-up of the underlying report, and that write-up is unusually specific about how the test was run while giving no numeric results at all. The checkable design details are: [S4]

Element of the test designWhat the public write-up states
DurationEntry-level VMs observed for 48 hours, with median results reported
Providers comparedFour cloud providers, including the vendor citing the report
VM specs testedAn AWS C5.Large with SSD EBS in US East, a 4 GB Compute Optimized instance in US East (NYC), a 2 GB SSD VPS in US Central (Zone C), and a Compute Optimized 4 GB instance in US Central (Dallas-Fort Worth)
Operating systemUbuntu 16.04 LTS
Test categoriesFive: vCPU performance, random read/write IOPS, Apache web server performance, MySQL mixed workload, PHP performance
Database loadA database of roughly 5 million rows, reviewed with 24 simultaneous threads
Web server methodApache and Apache Benchmark run on the same VM (localhost) to isolate VM performance

That table is a methodology, not a result. The write-up reports which provider led in each category in words, and states the median results are in the report; it prints no IOPS value, no requests-per-second figure, no CPU score and no percentage margin on the page. [S4] So the one third-party performance claim in the comparison set cannot be turned into a number, and it compares four configuration types that are not the same size. A 2 GB SSD VPS against a 4 GB Compute Optimized instance is a test of four products, not a controlled measurement of one variable.

This is the exact hole the pricing pages leave open: there is no published numerator for "performance", so the denominator (price) has nothing to divide. Anyone claiming a cheapest-VPS performance winner from these pages is supplying the missing number themselves.

6. The missing half: a public, recomputable probe

The tooling to produce that numerator is public and its parameters are documented. A widely used Linux benchmark script runs three families of tests: disk performance with fio, network performance with iperf3, and CPU/memory performance with Geekbench. [S5] The documented specifics are: [S5]

# Run the full suite (public command from the project README)
    curl -sL yabs.sh | bash

    # fio: four random read/write disk tests at 4k, 64k, 512k and 1m block sizes,
    #      each with a 50/50 read-write split; reports MB/s and IOPS
    # iperf3: 8 parallel threads, both download and upload, IPv4 and IPv6,
    #      ~20s per location (10s each direction); reports send/recv speed and ping
    # Geekbench 6 by default (4, 5 and 7 available via flags);
    #      reports single-core and multi-core scores

    # JSON output, useful for repeat runs and for dividing by price
    curl -sL yabs.sh | bash -s -- -j

The project also documents flags to skip individual test families (-f for disk, -i for network, -g for Geekbench), to reduce the number of iperf locations (-r), and to write results to a file (-w <filename>). [S5] These are the public, checkable parameters of the tools named in the comparison set's own gap analysis.

Disclosure: this article does not report any self-run benchmark result. No probe was executed on these plans for this piece, so there is no measured IOPS, Geekbench score, throughput or latency figure in it, and none is estimated. What is provided is the reproducible command and the arithmetic that turns a future measurement into a per-dollar figure:

performance_per_dollar = measured_result / published_monthly_price_usd

      IOPS per $        = fio 4k total IOPS      / monthly price
      MB/s per $        = fio 1m total MB/s      / monthly price
      Geekbench per $   = Geekbench multi score  / monthly price
      Mbit/s per $      = iperf3 recv speed      / monthly price

      Run the same command on each tier, at the same hour, and publish
      the raw JSON next to the price. Only then does the comparison exist.

That last step is the part no vendor page in this comparison set has done. Until raw results sit next to prices, "benchmark per dollar" is a worksheet, not a leaderboard.

7. Five-minute worksheet

Use this before you treat any advertised spec as a performance claim:

  1. Write down the monthly price and the four allocations: vCPU, RAM, storage, transfer.
  2. Divide each by the price. Check whether the ratio improves or degrades as you move up a tier.
  3. Mark which columns are allocations and which are measured outcomes. If there are no measured outcomes, say so explicitly.
  4. Note the disk type, the traffic policy and whether transfer is inbound-only or outbound. Do not divide across mismatched definitions.
  5. Ask the provider for a raw fio / iperf3 / Geekbench result on the exact plan, or run the public command yourself and keep the JSON.

Bottom line: on the published plan pages, one dollar buys between 0.045 and 0.118 vCPU, between 0.100 and 0.273 GB of RAM, between 1.00 and 5.45 GB of storage, and between 0.056 and 0.182 TB of transfer, depending on vendor and tier - and the cheapest tiers buy the most of it. [S1][S2] What one dollar buys in measured performance is not published anywhere in the comparison set: the only numeric speed threshold defines a category rather than a plan, and the only third-party report cited gives a detailed test design with no results on the page. [S3][S4] Run the public probe, divide by the same price, and the number that is missing today becomes one you can verify.

Frequently Asked Questions

Which cheap VPS gives the most performance per dollar?

That cannot be answered from published data. The three comparison sites publish allocations and prices but no measured benchmark result beside a price, so there is no numerator to divide. What can be computed is spec per dollar, and on that measure the entry tiers lead.

Is the cheapest plan always the best value per dollar?

On advertised resources, yes in both tiered price lists examined here: NVMe per dollar falls from 2.00 GB to 1.00 GB across one vendor's four tiers, and vCPU per dollar falls from 0.118 to 0.045 across the other's Compute tiers. That is a ratio of allocations, not a claim about speed, support or renewal price.

Can I compute a real benchmark-per-dollar figure myself?

Yes, with a public tool and one division. Run the documented fio, iperf3 and Geekbench suite on the exact plan you are considering, keep the raw JSON, and divide the measured result by the published monthly price. Repeat on each tier under the same conditions and publish the raw output next to the price.

Sources

All pages read on 2026-11-11.

Every figure above is transcribed from the listed public pages; all divisions and ratios are labeled as derived arithmetic. No sentence in this article is reproduced from the three compared sites; only topic, structure and angle are borrowed. No self-run benchmark result, traffic estimate, search-volume estimate, analytics figure or ranking data is used.

Marcus Reid
Systems Architect & VPS Infrastructure Analyst

Marcus has been working with virtualized server infrastructure since 2015, deploying and benchmarking VPS and cloud instances across DigitalOcean, Vultr, Hetzner, OVHcloud, Contabo, RackNerd, Hostinger, and the major hyperscalers. His work focuses on the gap between what providers advertise and what actually happens when you deploy a real workload. He verifies every price and specification against official provider pages before publication.