BGR's Aaron Greenbaum published the comparison on October 8, 2026. Here is what holds up, what needs qualifying, and what it means if you're pricing a PC.
The two price tags
2006: GeForce 7950 GX2. BGR says Nvidia's 2006 GeForce 7 lineup covered a wide range of prices:
- 7900 GS: $199 to $249
- 7950 GT: $299 to $349
- 7900 GTX: $450 to $500
- 7950 GX2: $535 to $650
Contemporary launch coverage generally put the GX2's suggested price at $600 to $650, so the low end of BGR's range probably reflects street pricing or specific boards. Treat "about $600 to $650" as the launch reference and anything lower as a retail variation.
2025–2026: GeForce RTX 5090. Nvidia's launch announcement set the $1,999 MSRP for the RTX 5090. BGR says it now retails for anywhere between $9,000 and $14,000, depending on the model. Nvidia's own launch release lists the RTX 5080 at $999 and the RTX 5070 at $549.
The 2026 street price is real, but it isn't a single number
Other outlets corroborate the surge, though the figures vary:
- Yahoo Tech, reporting on the week of September 13, 2026, said stores such as Newegg were selling RTX 5090s for around $9,000.
- TweakTown described third-party listings mostly between about $6,000 and $9,300, with some above $10,000.
- A report citing Tom's Hardware said first-party inventory had almost completely dried up, leaving marketplace listings between $6,395 and $9,500.
- That same report noted one Micro Center location still had a card on the shelf at roughly $4,299.
- A Tech Insider analysis put the U.S. median at $4,699.99 in August 2026.
So "$9,000 to $14,000" describes the most extreme listings, not what every buyer pays. Median figures and the odd in-store card sit far lower, though still well above $1,999. Several reports tie the price spike to limited availability rather than any official Nvidia price increase. One analysis also reports that Nvidia told board partners in May 2026 to expect an estimated $300 increase, citing rising GDDR7 memory costs. Treat that last point with caution. It comes from a secondary source, not Nvidia.
The inflation math
BGR used the Bureau of Labor Statistics calculator to turn the 2006 price into roughly $888 to $1,079 in today's dollars. That range is consistent with adjusting $600 to $650. Two things to keep in mind:
- The BLS calculator uses the average Consumer Price Index for a given calendar year, and for the current year it uses the latest monthly value. The exact endpoint behind BGR's range isn't stated, so treat it as approximate.
- BGR says that range is about what an RTX 5070 costs today, or an RTX 5080 at launch. That holds only at launch MSRPs ($549 and $999). At the street prices the same article cites for the 5090, no one should assume current 5070 or 5080 pricing without checking retailers.
The comparison that matters most is the headline one. Adjusted for inflation, the GX2 is about $900 to $1,100 in today's money. Even at launch, the RTX 5090 was roughly double that. At today's reported street prices, it's several times higher.
Why "4x the clock and 32x the memory" is shaky
BGR frames the generational leap as four times the clock speed (500 MHz versus 2.01 GHz) and 32 times the memory (1 GB versus 32 GB). It's a fun line, but the comparison is flawed:
- The GX2 was two GPUs on one board. Its 1 GB is a total split across two chips, with 512 MB each. The RTX 5090's 32 GB belongs to one GPU. The ratio is therefore 64 times per GPU, not 32.
- Clock speed doesn't measure performance across architectures. The 2.01 GHz is a base-clock figure, and a Blackwell GPU does far more per clock than a 2006 chip. Nvidia's launch material, for example, claims up to 3,352 AI TOPS, a number with no 2006 equivalent.
- Nvidia's product page confirms the 5090's headline spec. It says the card carries 32 GB of GDDR7, but it says nothing about the 2006 board, so the 2006 specs rest on period documentation.
None of this undermines BGR's main point that GPUs are vastly more capable. It just means the multipliers are rhetorical, not engineering ratios.
What this means for PC builders
- MSRP and market price are now very different things. Anyone budgeting a high-end build should check live listings, not launch announcements.
- Prices vary widely by seller and region. Third-party marketplace listings, first-party stock and in-store inventory differ by thousands of dollars.
- Cheaper cards are the practical route. BGR notes Nvidia's lower-tier GPUs are generally under $1,000. That's a claim about launch pricing, so verify it against current listings before you buy.
- Compare flagships with flagships. The GX2 was a top-end 2006 product, not an average GPU. Its inflation-adjusted price says nothing about how much a mainstream card cost then or now.
Is the BGR piece fair?
It leans on opinion in places. The claim that GPU cost has "barely outstripped" the value delivered isn't backed by a defined benchmark or workload. The suggestion that GPU value would have outpaced cost without component shortages is speculation. Component shortages do appear in other outlets' reporting, but the value claim itself isn't proven.
The strongest takeaway is the simple one. A 2006 flagship cost roughly what a mid-to-upper RTX card costs at launch today. The 2026 RTX 5090 is the outlier, because its price has been pushed up by shortages and heavy demand, not by Nvidia's list price.