If the rumors about Nvidia’s next-gen lineup are true, well-heeled gamers will have GPUs worth drooling over (except please don’t – Ed) All the rumors about outrageous power consumption and triple-slot coolers have done little to diminish peoples’ appetite for the upcoming RTX 40-series. Now there’s more fuel to add to the fire. One of the most notorious leakers has dropped a benchmark result for the RTX 4090. It’s just one benchmark, and it’s a synthetic test, but if it’s true, or even close to true, Nvidia truly has a beast of a card on its hands. The only bad news is another well-regarded leaker is saying the RTX 4090 will be the only Ada Lovelace GPU to drop this year.
Notable GPU tipster kopite7kimi has revealed the RTX 4090’s score for Time Spy Extreme. This is a 4K, DirectX 12 benchmark designed to stress the latest CPUs and GPUs. They note on Twitter the RTX 4090 can achieve a score of 19,000, which would be a new world record by a large margin. Currently the world record is 14,977, obtained by an RTX 3090 Ti running at 2,850Mhz, presumably on LN2. The RTX 4090 is 26 percent faster than even that setup, likely while running on air (that’s an assumption, but a safe one). As far as how it compares to a regular RTX 3090 Ti, PC Gamer’s review showed the Asus TUF notching a score of 11,249. That would give the 4090 a surprising 68 percent advantage in this one test.
It would also give it an hard-to-believe 85 percent advantage over the Nvidia Founder’s Edition RTX 3090. That’s a significant performance gain from one generation to the next. Compared to the RTX 3080, it’s more than twice as fast. For reference, the RTX 2080 Ti only scored 5,741 in this test.
One thing to keep in mind is the amount of power these cards may consume. One straightforward way for Nvidia to boost performance without needing a tremendous boost in performance per clock from Ampere to Lovelace would be to build bigger, more power-hungry GPUs. Rumors have consistently pointed to a power consumption increase for top-end cards from both Nvidia and AMD this year. This is not to imply Lovelace won’t be more efficient — just that Nvidia need not rely solely on IPC (Instructions per Clock) improvements.
This new information lines up with previous rumors that Nvidia is pulling out all the stops this time around. Nvidia may not be sure what AMD’s chiplet-based GPU will bring to the table. Raising TDPs would help the company leave nothing to chance. That’s our assumption though, not based on any evidence, just what has been rumored so far.
The only fly in the ointment is the other reputable leaker, Greymon55, is also reporting that only the RTX 4090 will ship this year. The rest of the stack will be coming in 2023. This would, again, line up with reports that Nvidia is trying to push back the 40-series launch. This would give its partners time to clear out existing 30-series inventory. Those cards will become difficult to sell once the 4000 series stack arrives. Previously, Nvidia has typically launched at least two GPUs from the same family at the same time, so only launching the RTX 4090 would be a departure from that strategy.
As always, we’re curious to see more than one benchmark, especially in a synthetic test. Not only would we like some actual in-game numbers, but some leaks from the AMD would be appreciated as well. We’ll also be following up on these rumors once Ada Lovelace launches to gauge how accurate they are, or were.