MSRP vs Street Price: Why the Number on the Box Is Not What You Pay

MSRP is a positioning statement made before launch, not a price you can pay. It is set by the chip designer, for a reference board, in one currency, for one market, before anybody knows what supply, competition or memory contracts will look like. Every comparison table that ranks cards on list price — including large parts of this site — is therefore ranking launch-day intent rather than the market you are shopping in.

What MSRP actually is

MSRP tells you where a card is meant to sit relative to its siblings. That is a genuinely useful signal — it is how you know a xx70 is positioned below a xx80 — and it is the only price that exists before a card ships. It is not a commitment by any retailer, and nobody is obliged to honour it.

What you pay is decided by things MSRP cannot know: how much stock exists, what the competition launched at, which board partner made your specific card, whether the same silicon is in demand for data centre products, memory contract prices, tariffs, and your local currency. Every one of those moves after MSRP is fixed, and several of them have moved violently in the last two years. GDDR7 contract pricing alone has repriced parts of the professional market by 50% since launch.

Which direction it moves, and when

Both directions, at different points in a card's life, and predictably enough to plan around:

• Above MSRP at launch. Demand exceeds supply for the first weeks to months. Premium board designs — better coolers, factory overclocks — are priced above the reference figure by design, and during a shortage they are often the only versions in stock. Buying in this window means paying the maximum the market will bear.
• At or below MSRP mid-life. Supply catches up, competitors respond, street price settles at or under list. This is the cheapest window relative to what you get, and it is where most sensible purchases happen.
• Above MSRP again at end of life. Production stops, remaining stock gets scarce, prices on discontinued cards drift up. Hunting a specific older card can cost more than it launched at — which is why "it was $329 at launch" is not an argument for what it is worth today.

The practical consequence is worth stating directly: a card's price-per-frame ranking is not a property of the card. It is a property of the card this month. Any ranking, ours included, is a snapshot of a market rather than a measurement of hardware.

MSRP is not even stable across generations

If list prices tracked cost or performance you would expect them to move in one direction per generation. They do not. Here is what NVIDIA did to its own list prices from the RTX 40 series to the RTX 50 series, tier by tier.

List price fell at 4 of the 5 tiers — the xx60 Ti 16 GB by 14% — and rose at exactly one: xx90, up 25%. That is not a cost curve. It is a company deciding that the halo product can absorb a price rise while the volume tiers cannot, and then announcing both decisions as suggested retail prices.

Once you see MSRP as a positioning decision rather than a valuation, the right way to use it becomes obvious: compare it across siblings to understand the product stack, and never treat it as the price of anything.

What one real price does to a ranking

We hold 1 observed retail price against 84 cards priced at MSRP. That ratio is the largest single limitation in every value figure on this site, and we would rather say so than let a tidy-looking table imply otherwise.

The one card we can price properly demonstrates the whole problem. NVIDIA GeForce RTX 5060 Ti 16 GB lists at $429. It is observed at $279 — 35% below list. Its measured performance did not change.

At list price it is a reasonable mid-tier option, ranking 10 of 77 on performance per dollar. At the price it actually sells for it ranks 1 — and it is the card we recommend in our sub-$300 guide for exactly that reason. If the price returns to $429, that recommendation should change, and we say so in the guide itself. A conclusion that depends on a price should carry the price it depends on.

The whole market, priced as listed

Here is every card we track with both a price and a benchmark, plotted against measured 1440p performance. Almost all of these are list prices, so read it as a map of manufacturer positioning rather than of what you would pay today.

The spread below that line is the point. If list prices were set by performance, the points would form a line. They form a cloud, because list price encodes generation, brand positioning, launch-window competition and how much margin a tier is expected to carry — none of which is performance. Cards sitting far below the line are not badly engineered; they are priced for a market that no longer exists, and nobody has repriced them because they are being sold off rather than sold.

How to read the prices on this site

We label them, which is the only honest option available to us. Prices marked MSRP are manufacturer list figures; unmarked prices are observed retail. Our methodology page states the current ratio between the two plainly, because it is the largest single limitation in our Value Scores and burying it would not make it smaller.

We are moving to live pricing through Amazon's official Product Advertising API rather than scraping retail pages — accuracy bought with someone else's terms of service is not accuracy we want, and a scraper that gets an affiliate account closed takes the whole site with it. Until that coverage is complete, treat MSRP-marked figures as relative positioning rather than a quote.

Reading any GPU price well

1. Check the listing before trusting a ranking. Any guide's price column, ours included, is a snapshot. The retailer's page is the fact.
2. Ignore "% off MSRP" entirely. A large discount off an unrealistic list price is not a deal. Frames per dollar at the price you will actually pay is the only comparison that survives scrutiny.
3. Compare total system cost. A 304 W card needing a 700 W supply is not the same purchase as a 250 W card needing 600 W, if the power supply is also on the bill.
4. Do not pay a launch premium for a mid-range card. Waiting two months for supply to normalise has historically been worth more than the specifications gained by buying early. This does not apply to halo products, which sometimes never come down.
5. Price the used market on older cards. Past its production window, the used price is the price, and the list figure is a historical artefact.

Common questions

Why is the GPU more expensive than its MSRP?

Most commonly because you are inside the launch window, or because you are looking at a premium board design rather than a reference card, or because the model is discontinued and remaining stock is scarce. All three are normal. None of them mean the retailer is overcharging in any meaningful sense — MSRP was never a ceiling.

Will GPU prices come down if I wait?

Mid-life, usually yes; at end of life, usually no. The pattern above is reliable enough to plan around: the cheapest moment to buy a card relative to what it delivers is after supply normalises and before production stops. For a card launched in the last two months, waiting is almost always correct.

Should I trust a price comparison site's price column?

Trust it for ordering, not for the number. Any static table is a snapshot, and the useful question is whether the site tells you which figures are list prices and which are observed. If it does not distinguish them, it is showing you MSRP and calling it price.

Does a lower MSRP mean a better deal?

No. As the generational chart above shows, list price is a positioning decision rather than a valuation, and the gap between list and checkout can be enormous — 35% on the one card here we can measure both ways. Two cards with the same MSRP can therefore be very different purchases. Compare performance per dollar at the checkout price, which is the only figure with a fact behind it.

Related reading: how to read a Value Score, and when to ignore it.