I Asked ChatGPT to Spec a Trading Computer and Graded Its Answer

I had a trader call me last week. He was holding a parts list he got from Copilot. He'd gone online, told the AI exactly what he trades and how many screens he runs, and it handed him a clean and confident build. It listed a processor, the memory, a graphics card and a storage drive. The whole spec sheet.

The first thing out of his mouth was, "Eddie, I'm not confident in a single part on this list." He could read every line. What he couldn't tell was whether any of it matched the way he trades.

So I asked him the one thing the AI never did: what exactly do you have open on your screens when the bell rings at 9:30 a.m. Eastern time?

Copilot had sized his machine for eight charts. He does run eight charts. But he also runs two DOMs on the NQ and one on gold. A DOM, short for depth of market, is the price ladder where a lot of futures traders click buy and sell. Those three ladders change the whole machine.

What ChatGPT recommended for NinjaTrader and thinkorswim

As an experiment, I did what a lot of you have already done. I typed this into ChatGPT: "I trade using NinjaTrader and thinkorswim over four monitors. Spec me out a good trading computer." Then I graded it with a red pen.

To be fair, it wasn't a bad answer. It was a very generic answer, and you need a very specific one.

Part	ChatGPT's pick	My grade
Processor	AMD Ryzen 9 9900X, 12 cores	Miss. No benchmark floor given
Graphics card	RTX 4060, mentioned the 5060 further down	Right idea, last generation. I'd use the RTX 5060
Memory size	64 GB, said 32 GB would work	Gold star
Memory type	DDR5 6000 CL30	Miss. A gaming spec
Storage	2 TB NVMe	Gold star
Network	Wired Ethernet, 2.5 Gb motherboard port	Gold star
Power	UPS battery backup	Gold star
Sizing method	Counted charts	Miss. Should count data streams

Miss number one: no processor speed floor

This is the biggest hole in almost every AI build I've ever read. It never gives you a minimum processor score your machine needs to hit so it doesn't fall behind the market.

My floor is a CPU benchmark score of 45,000 for active day trading. That score is a measurement of horsepower for your processor, and it's the one number that cuts through all the marketing on the box.

ChatGPT picked an AMD Ryzen 9 9900X and told me it has 12 cores. Cores are the separate workers inside the chip. That's nice to know, but it never told me whether that chip clears the line. It's like a car salesman bragging about eight cylinders and never telling you how much horsepower the car has.

I don't grade chips by brand, Intel or AMD, and you shouldn't either. We grade them by the benchmark score. I wrote more about that in the 45K trading benchmark rule.

Where does your machine stand? Click the Start button and start typing "system information." When it pops up, click it. About 10 lines down, find the line that says Processor and write it down. Then type that name into my free CPU benchmark test for traders. It tells you how your chip stacks up against 45,000, and the whole thing takes about two minutes. Before you trust any build from AI, from a store or from me, find out where your current machine scores.

What ChatGPT got right

I'll give it a couple of gold stars. It got three things right, and it deserves the credit.

It didn't overspend on the graphics card

Its exact words were, "Four monitors don't require an RTX 5080." That's correct. The graphics card's main job on a trading machine is drawing your charts on your screens, and a monster gaming card doesn't make a candlestick print any faster. Its other job is supporting the AI trading modules coming out in the near future.

It picked an RTX 4060, which sadly is the last generation. That part it got wrong. That card came out back in 2023. Today I'd put the Nvidia RTX 5060 in that slot, and to be fair, ChatGPT did mention the 5060 as an option further down.

It also said to run all four monitors off the graphics card instead of mixing in the motherboard's plugs. Right again. One RTX 5060 runs four monitors: three off the DisplayPort outputs and one off the HDMI. If you're wiring that up, here's how to connect 4 monitors to a trading computer.

Where it stopped short is the future. Plenty of four-monitor traders I talk to wind up being six-monitor traders in a few years. Six to eight monitors means a second card, so build a machine with room for it.

It got the memory right

It recommended 64 GB of RAM. RAM is like the surface of your office desk. The bigger the desk, the more platforms, charts and browser tabs you can spread out without shoving things onto the floor. For NinjaTrader and thinkorswim running together on four screens, 64 GB is a smart call. It even admitted 32 GB would work. That's a pretty honest answer.

It remembered the boring stuff

It said to get a 2 TB NVMe, which is a fast kind of solid-state drive that plugs straight into the motherboard. It said to use wired Ethernet, meaning a cable plugged into your router instead of Wi-Fi. It picked a motherboard with a 2.5 Gb network port, a faster kind of wired connection, and that's what I'd want on a new trading machine too.

And it said to buy a UPS, an uninterruptible power supply. That's a battery backup that keeps your machine running when the power flickers with a position open. Most people forget these things. ChatGPT didn't.

Miss number two: a gaming part that doesn't make charts faster

Now the red pen. It specified DDR5 6000 CL30 memory. CL30 is a timing rating that gamers chase to squeeze out extra frame rates. In my experience, traders won't see their charts move any faster because of it.

It's like paying extra for racing tires on the car you drive to the post office. Nothing's wrong with the part. It just tells you who the list was really written for, and it wasn't a trader. I break down more of these mismatches in gaming PC vs trading PC.

Miss number three: counting charts instead of data streams

This is the big one, and it's the one that turned off my Copilot caller. ChatGPT sized the machine by counting charts. Your machine doesn't care how many charts you have. It cares how many live data streams you're feeding it.

When my dad taught me charting with a pencil and graph paper back in 1982, a chart really was the whole workload. Today it's a small piece of it.

A 5-minute chart on the ES updates its last bar as trades come in. A DOM on the ES redraws every time an order anywhere in that ladder gets added, pulled or filled. Level two does the same thing for stocks. It shows every buy and sell order waiting at each price, not just the best one. On a busy market open, DOMs and level twos can change hundreds of times a second.

To an AI, a DOM and a chart are both just a window. It's like a moving company giving you a quote by counting bedrooms and never asking about the piano. Your DOM and level two screens are the piano. The more of them you have, and the more symbols you run, the more pianos you have.

That piano costs real money. One tick on the ES is $12.50 a contract. If your DOM is running half a second behind on a fast open and you get filled two ticks worse on three contracts, that's $75 gone on one trade. Do that a few times a week and a slow processor is costing you more than a car payment.

Here's how it ended up for my caller. He counted his three DOMs. He looked up his processor on my benchmark test while we were on the phone, and it was well short of 45,000. Once he knew that, he stopped second-guessing the list, because he finally had a way to measure any machine. He ended up with a computer sized for his three DOMs instead of just his eight charts.

The question AI should have asked first

I'll give ChatGPT a little credit. At the very end, it asked whether I run order flow or just a lot of tick charts. That's the right question. It just asked it after it had already picked every part.

That's the difference between a parts list and a trading computer. An AI chatbot answers the question you type. It doesn't know what you left out, and it's never watched a machine choke at 9:31 in the morning right after the open.

In 16 years of building these machines, the first five minutes of every call is us asking what's on your screens. How many ladders? How many level two windows? How many symbols? Which platforms run at the same time? The parts come after that, not before. If you run several platforms at once, why 16GB of RAM can't handle three platforms is worth a read.

What to do before you trust any parts list

Instead of asking ChatGPT, Claude or Gemini for the best build tonight, do this. Write down every live window you run at the open. Count the DOMs. Count the level twos. Count the symbols. Count time and sales too, not just the charts. Time and sales is a running list of every trade as it prints.

Then run my free trading benchmark test and get your number. Bring those two things to us, and we can tell you right away whether the machine you own is enough or what it would take to fix it.

If you want every part that matters in one place, in plain English and in the order that matters, get my free Complete Guide to Trading Computers. It covers the processor score, how much memory you need to fit all your platforms, and how many screens each graphics setup can drive. When you're ready, you can see our desktop trading computers.

The bottom line: AI counts your charts, but your computer cares about your data streams. And no AI parts list is worth trusting until someone who knows how these platforms work has asked you how you actually trade.

Frequently Asked Questions

What benchmark score do I need to day trade?

My floor for active day trading is a CPU benchmark score of 45,000. That score measures the horsepower of your processor, and it's the one number that cuts through the marketing on the box. Core counts and brand names don't tell you whether a chip keeps up with the market. Type your processor name into my free benchmark test and it shows how your chip stacks up against 45,000 in about two minutes.

How much RAM for a trading computer running NinjaTrader and thinkorswim?

For NinjaTrader and thinkorswim running together on four screens, 64 GB is a smart call, and 32 GB would work. Think of RAM as the surface of your office desk. The bigger the desk, the more platforms, charts and browser tabs you can spread out without shoving things onto the floor. The more live windows you run at the open, the more desk you need.

How do I check what processor is in my PC?

Click the Start button and start typing "system information." When it pops up, click it. When the page opens, look about 10 lines down for the line that says Processor and write down that name. You can then type it into my free CPU benchmark test to see how your chip compares to the 45,000 score I recommend for active day trading.

Which processor is best for trading?

I don't grade chips by brand, Intel or AMD, and you shouldn't either. I grade them by the CPU benchmark score, and my floor for active day trading is 45,000. Being told a chip has 12 cores is like hearing a car has eight cylinders without learning its horsepower. Look up any processor on my benchmark test before you trust a build from AI, a store or me.

Is a gaming PC good for trading?

Gaming parts often don't make charts faster. ChatGPT specified DDR5 6000 CL30 memory, a timing rating gamers chase for extra frame rates, and in my experience traders won't see their charts move any faster because of it. A monster gaming graphics card doesn't make a candlestick print faster either. Nothing is wrong with those parts, but they tell you the list was written for a gamer, not a trader.

What graphics card do I need for four trading monitors?

Four monitors don't require an RTX 5080. One Nvidia RTX 5060 runs four monitors, three off the DisplayPort outputs and one off the HDMI, so all four run from the graphics card instead of the motherboard's plugs. Plan ahead, though. Plenty of four-monitor traders become six-monitor traders within a few years, and six to eight monitors means a second card, so leave room for it.