PC gaming’s eternal question: green or red? The honest 2026 answer: it depends on the tier and what you play — each brand wins on different ground. Here’s the head-to-head, tier by tier, with data from TechSpot and Tom’s Hardware.

Updated August 6, 2026. Nvidia raised prices for the third time this year (20-30%) in late July, and AMD by 10%. That has widened the gap between the two brands and made several prices in the previous version of this article obsolete. Details in the July 30 news roundup.

Before comparing: MSRP no longer works

This is 2026’s most important change and it belongs up front.

We’ve all learned to reason this way: “the 5070 Ti is $749, the 5080 is $999, so the gap is $250.” That reasoning now produces bad decisions, because real prices have decoupled from official ones by different amounts per model. In July, the RTX 5070 Ti sold as much as 38% above its MSRP while other models sat 12% up. With that spread, which card is the better buy can invert entirely.

What to do instead: open the two cards you’re comparing, at the retailer you’ll actually buy from, on the same day, and compare those two numbers. The prices below are indicative of early August 2026 — treat them as a reference point, not as truth.

The quick verdict

Battleground	Winner	Why
Pure rasterization per dollar	🔴 AMD	More raw FPS at the same price
Ray tracing and upscaling (DLSS)	🟢 NVIDIA	DLSS 4 remains the benchmark
High end (+$1000)	🟢 NVIDIA	AMD doesn’t compete there in 2026
Mid-range value	🔴 AMD	The gap has widened to ~$200
VRAM per dollar	🔴 AMD	More memory at the same price
Streaming and creation	🟢 NVIDIA	NVENC and the CUDA ecosystem
Price stability in 2026	🔴 AMD	Up 10% versus Nvidia’s 20-30%

Entry level (under $300): a technical tie with a twist

The NVIDIA case:

The NVIDIA RTX 5060 delivers better ray tracing, DLSS 4 and the most polished ecosystem. Its weaknesses: 8 GB of VRAM and a new availability problem — Nvidia cut RTX 50 series production by up to 20%, and the 5060 is hit hardest, with estimates of around six months for stock to normalise.

The AMD case:

The AMD RX 9060 XT 8GB delivers the best raw rasterization performance in the range, with the same 8 GB limitation. AMD hasn’t announced comparable production cuts, so it’s currently the easier one to find at a sane price.

The unexpected referee, with an asterisk: the Intel Arc B580 and its 12 GB of VRAM launched at $249 and was the clear pick in this tier. It now sells at $299 or more, which narrows its argument considerably. There’s an added risk too: Intel has reportedly cancelled its next-generation gaming GPU, and Arc cards depend on driver updates more than Nvidia or AMD do. Still worth considering, but no longer the obvious buy. I explain it in the GPU by budget guide.

Tier verdict: if you find any of the three near its official price, buy it and stop deliberating. With the shortage, availability outranks brand preference.

Mid-range ($300-1000): AMD takes it

This is the biggest change from the previous version of this article.

The AMD case:

The AMD RX 9070 XT sits at $630-700: 16 GB of VRAM, effortless 1440p and reasonable 4K.

The NVIDIA case:

The NVIDIA RTX 5070 Ti answers with DLSS 4 and superior ray tracing, but sits at $820-900.

Tier verdict: the gap now runs around $200 on average between two cards with broadly comparable rasterization performance, though it ranges from $120 to $270 depending on the specific models. Anywhere in that range, you’re talking about your RAM, your SSD and your power supply.

• If you play rasterised titles at 1440p, the RX 9070 XT isn’t close: it’s the clear buy.
• If you rely on heavy ray tracing, DLSS or CUDA (video, 3D, local AI), Nvidia’s advantage is real and may justify the $200. But it’s now a decision you have to justify, not the default.

One important note about that advantage: it already has an expiry date. On July 31 AMD notified another increase of at least 10% effective in August, so the “AMD raises less” argument stops existing. It was never generosity — the increase had been locked in since June, waiting for Nvidia to move first.

From here on, compare the two on features and on the price of the day, not on brand trend. Details in the August 6 news roundup.

High end (over $1000): NVIDIA walks alone

There’s no duel here: AMD chose not to compete at the high end in 2026, so the crown is green by forfeit. But pricing has run away to a point worth facing directly.

The NVIDIA RTX 5080 was spotted at $2,099 at major retailers in late July — above the RTX 5090’s $1,999 launch MSRP. And the RTX 5090 has passed $4,300 at some retailers.

The honest note: before spending $2,000 on a 5080, remember the mid-range RX 9070 XT gives you a very large share of that experience for around $650. With the stack scrambled, the high end only makes sense if you have a high-refresh 4K monitor and the price on the day is reasonable.

Watch one side effect: a 5080 at $2,099 makes the 5090 “look” reasonable. It isn’t — it remains an absolute niche product. The tier below running away doesn’t turn the tier above into a good purchase.

Should you wait for the RTX 50 Super?

No. And this time for a numerical reason, not a hunch.

The RTX 50 Super cards are built but held back over memory costs: a 3GB GDDR7 chip costs $60-70 against roughly $20 for a 2GB part. An 18GB RTX 5070 Super needs six 3GB chips — $360-420 in memory alone, against $120 for the standard 5070. By the time that reaches a shelf, it’s $300-400 on the final price.

In other words: a 5070 Super would land $300-400 above a standard 5070, dropping squarely into RTX 5070 Ti price territory without matching its performance. At that price the product competes against Nvidia’s own lineup and stops making commercial sense — which is why it’s parked with no date. The practical consequence is that current cards have become more valuable, not less: with no refresh coming, they hold their position in catalogue far longer than usual. I break it down in the July 23 news roundup.

Each brand’s pros and cons

🟢 NVIDIA

• ✅ DLSS 4: the best AI upscaling on the market
• ✅ Benchmark ray tracing
• ✅ NVENC for streaming + CUDA for AI/creation
• ✅ The only real high-end option
• ❌ You pay a brand tax: fewer FPS per dollar in rasterization
• ❌ Stingy with VRAM in lower tiers
• ❌ Three price increases in 2026 (January, May and July), the last one 20-30%
• ❌ Production cut of up to 20% across the RTX 50 series

🔴 AMD Radeon

• ✅ Better raw performance per dollar at entry and mid tiers
• ✅ More VRAM at the same price (ages better)
• ✅ FSR has closed much of the gap with DLSS
• ✅ Excellent Linux drivers
• ✅ Up 10% versus Nvidia’s 20-30%
• ❌ Ray tracing still a step behind
• ❌ No high-end option in 2026
• ❌ Less mature creation/AI ecosystem

Frequently asked questions

NVIDIA or AMD in 2026? It depends on the tier. AMD wins the mid-range clearly: the RX 9070 XT delivers rasterisation performance NVIDIA doesn’t match for the money. At the high end NVIDIA walks alone, because AMD chose not to compete there this year.

Why can’t I trust the MSRP? Because in 2026 it stopped reflecting what you pay. Street prices have detached from official pricing, so the useful comparison is between real prices, not launch figures.

Is it worth waiting for the RTX 50 Super? No. They’re on indefinite hold over GDDR7 memory costs with no date. Waiting on a product whose viability depends on memory getting cheaper isn’t a plan.

When does paying for NVIDIA pay off? If you rely on heavy ray tracing, on DLSS, or if your work uses CUDA — editing, 3D or local AI. There the advantage is real. For pure rasterisation at 1440p, the price gap weighs more.

What if I only play esports? In the entry tier the two are very evenly matched, so buy on price and availability on the day. At those resolutions and loads, neither will hold you back.

Bottom line: decide in 10 seconds

1. Mid-range and rasterised games? → RX 9070 XT. At a $200 gap, it isn’t close.
2. Ray tracing, streaming or AI apps? → NVIDIA, knowing you pay a ~$200 premium.
3. 4K with no limits? → NVIDIA, no alternative — but check the price on the day first.
4. Entry level? → Whichever you find near its official price. Availability rules.
5. Were you waiting for a price drop or the Supers? → Stop waiting. There’s no low-price window in sight.

Picked a side? Specific models with prices are in our GPU by budget guide, and check that your power supply and case can handle the card.

---

Sources: TechSpot – The Best GPUs, Early 2026, Tom’s Hardware – GPU Price Tracking 2026. Prices verified July 11, 2026; reviewed August 6, 2026 after the late-July Nvidia and AMD increases.