Your PC is asleep 8 hours a night. Salad will pay you to put it to work. This kit tells you exactly how much.

Standalone Windows monitoring kit for any NVIDIA card on Salad's demand list. Logs your GPU's power, pulls your earnings four times a day, and shows you a single dashboard with the answer in dollars, net of electricity, with the hour-by-hour pattern of when you were earning and when you were getting in your own way. No accounts. No SaaS. No subscription. Runs entirely on your machine.

$29
one-time · instant download
Buy & download for $29 See what's inside
Instant Stripe checkout. The .zip lands in your inbox within 24 hours. Refundable for 14 days, no questions asked.

Where this came from

I built this for myself first. I have an RTX 4070 that mostly sits idle: at night, while I'm at work, anytime I'm AFK. Salad pays you to rent a GPU like mine to AI inference customers while you're not using it. Sounds great and maybe a little too good to be true, but I wanted to know, like actually know, not "YouTube guy says" what it was earning after I pay my power bill, and whether my own occasional use of my GPU was costing me more than I thought.

So I wrote a small monitoring artifact for Claude: nvidia-smi sampler, Salad earnings puller, a dashboard. I've been running it for over a week now. Here is what it told me, exactly, in three places I would not have figured out otherwise.

One afternoon of local LLM use cost $2.20 in foregone Salad earnings and demoted me from AI workloads to mining for the next three days. I had no idea Salad was keeping track that closely. The dashboard's heatmap showed me the exact 13-hour window where I tripped Salad off by using LM Studio, and the next day's earnings dropped from $2.51 net to $0.33 net. Salad's algorithm doesn't forgive quickly.
Salad's own container dies more often than I thought, and when I was checking Salad's dashboard it was just saying "paused until idle" the whole time at times it would say "can't connect to the kitchen" or it would start up a workload and just randomly die without me noticing. RDP into the machine, run a VM that grabs the GPU, or let a Windows Update reboot interrupt SaladBowl process any of these can stop your earnings cold without warning. The kit's heatmap shows you exactly when it happens (a gray strip where green should be) so you can restart and recover before Salad's chef algorithm demotes you to nickel workloads for three days.
Lifetime balance, day 11: $21. Net run-rate right now: ~$43/month. Not bad... but I bought this GPU used on eBay so I'm pretty happy it's paying itself back for me to mess around with LLMs. Also, knowing that number instead of guessing changes how I think about staying up late coding or training a LoRA/Whisper model. Every hour of my own use has a dollar figure on it now.

What's in the bundle

Who this is for

You have an NVIDIA card from the list below (or know yours is on Salad's demand list), you run Windows, and your machine sits idle for any meaningful stretch of the day: at night, during work hours, while you're commuting, while you're watching TV in the other room. Whether you're a gamer who sleeps eight hours, a workstation user with after-hours dead time, a builder running a homelab on the side, or just someone with a capable PC that mostly sits at the desktop screen. This is for you.

The pitch is one sentence: your hardware can earn money while you're not using it, and right now you have no idea how much. This kit answers that question with hard numbers measured on your actual box, so you can decide whether it's worth running.

Common situations buyers come in with:

Cards the kit knows about and has estimates for:

RTX 30 series
3060, 3060 Ti, 3070, 3070 Ti, 3080, 3080 Ti, 3090, 3090 Ti
RTX 40 series
4060, 4060 Ti (8/16GB), 4070, 4070 Super, 4070 Ti, 4070 Ti Super, 4080, 4080 Super, 4090
RTX 50 series
5070, 5070 Ti, 5080, 5090
Workstation
A4000, A5000, A6000

If your card isn't on this list, the kit will still log power for local-only work, but the earnings half won't apply. Check Salad's current demand at salad.com/earn/demand.

Who this is NOT for

The proof, as a picture

This is the actual 7-day heatmap from my box, with my own annotations. Green = earning. Red = either I was using it or Salad demoted me. The kit makes it obvious which is which.

Hourly busy% over the last 7 days (sample) · daily net at left
earning low idle / user no data

You can see the all-day red on 5/18 (running LM Studio cost me $2.20+ in foregone earnings) and the four-hour evening dip on 5/19 (I'd RDP'd into the box and stopped the container without realizing). The 5/17 gray strip is what a silent Salad outage looks like from the outside. Any of the usual culprits (Windows Update reboot, container crash, a VM that grabbed the GPU, a network mount that timed out) leaves this exact signature. The 5/20 row is partial: you're seeing today, mid-afternoon, while the loggers are still running. Without the dashboard you'd never know any of these happened until your weekly earnings came in low.

What does this save you?

One demotion you don't catch typically eats 1-3 days of normal earnings. Salad's chef algorithm doesn't tell you when it's downgraded your machine. You only notice when the weekly total comes in low. The dashboard's chef-risk callouts surface the pattern at every six-hour puller fire, so you can restart the container and recover before the second day of penalty rates lands.

Here's what that difference looks like by card, assuming a single ~2-day demotion you'd otherwise miss vs. catching it within twelve hours:

Card Lost per missed demotion Annual saved at 2 outages/mo
RTX 3060 12GB$1.56$28
RTX 3070 / 3070 Ti$2.26-$2.61$41-$47
RTX 4060 Ti 16GB$2.96$53
RTX 4070 / 5070$4.00-$4.35$72-$78
RTX 4080 / 5080$6.96-$8.35$125-$150
RTX 4090 / 5090$9.56-$13.05$172-$235

Realistic outage frequency for an active machine is one to three per month. Patch Tuesday reboot alone gets you to one, and that's before counting any RDP, VM, container crash, or "I started LM Studio and forgot" event. At two outages a month (typical), the kit pays for itself in:

~5 months
RTX 4070 / 5070
~3 months
RTX 4080 / 5080
~6 weeks
RTX 4090 / 5090

And that's just the demotion-prevention math. The bigger compounding effect is behavioral: once you can see what every hour of gaming or local AI work is actually costing in foregone Salad earnings, you stop running the marginal sessions. Reschedule one weekly LoRA training run from "whenever" to "Sunday 2am when you'd be idle anyway" and you've recovered another $20-40/month on a 4080-class card.

One demotion you don't catch on a 5090 costs $13, almost half the kit. Two outages a month and the kit has paid for itself by week three. After that, every catch is pure recovered revenue.

FAQ

Will this work without Salad installed?

Yes, partially. The GPU power and workload logging both work standalone. You'll get a dashboard that shows what your GPU is doing all day and what it's costing in power. The earnings side will just read $0. The kit will prompt you to install Salad if it's missing, but won't require it.

Is this a SaaS? Do you see my data?

No. Everything runs locally on your machine. Logs live in C:\SaladMonitor\. The dashboard is a single HTML file you open in your browser. There's no account, no server, no telemetry. You bought a kit; you own it.

Do I need Claude to use this?

No. The runtime (logger, earnings puller, dashboard) is standalone Windows PowerShell plus a single HTML file. The bundled SKILL.md gives you a conversational setup and a built-in opportunity-cost coach if you happen to use Claude Code, Claude Desktop, or Cowork, but the README.md walks you through the same install in five steps without Claude anywhere. Most buyers will never open SKILL.md.

Why $29? Why not free?

$29 is roughly one good week of Salad earnings on a mid-range card, about what a single missed demotion event costs you on a 4080-class GPU. The kit was iterated through months of real-world use on my own hardware until it actually answered the right questions. If it catches one outage you'd otherwise sleep through, or changes one decision you make about your hardware, it's already paid for itself.

What about AMD / Linux / Mac?

Not yet. The kit is Windows + NVIDIA only. If there's enough interest, AMD via rocm-smi is the next port. Email me if you want a heads-up.

How is this different from Salad's own dashboard?

Salad's dashboard shows you a lifetime balance and a daily earnings number, both rounded. It doesn't show power cost, opportunity cost of local use, hour-by-hour busy patterns, the gap between "GPU online" and "Salad actually paying", or what processes were on the GPU when. This kit does. It's the dashboard you'd want if you cared about the actual unit economics.

What if my Salad cookie expires?

It will, about once a month. The kit will note it in the dashboard ("earnings puller failing") and you re-paste a fresh cookie. The setup steps walk you through capturing it the first time; it's a 90-second process.

Refunds?

Yes, no questions asked, within 14 days. Reply to the receipt email. I'd rather refund a buyer who didn't get value than keep money that wasn't earned.


One more thing

The kit will probably change what you do with your GPU. Some buyers will lean in, realize their card is sitting idle 60% of the day and start treating Salad like background income. Others will see how much they actually use their GPU and decide rental isn't the right frame. Both are wins, because both are now decisions made with the actual numbers in front of you.

That's what you're paying for, not earning, Understanding.

Buy & download for $29
Instant checkout. Bundle emailed within 24 hours. Refund any time in the first 14 days.