Something to stay up for: Rumors and leaks concerning Nvidia’s next-generation graphics chips proceed to movement as the firm’s plans seemingly stay in flux. Although the RTX 50 collection flagship product will nearly definitely launch this yr, the standing of its companions stays unclear. Meanwhile, the lineup of AI chips that follows is already taking form.
Established leaker @kopite7kimi lately claimed that Nvidia would possibly start delivery the RTX 5080 before the extra highly effective 5090. New GPU generations usually debut with the flagship product, so this technique could be considerably uncommon.
The projected launch dates of the upcoming RTX 50 collection graphics playing cards, codenamed Blackwell, have lately fluctuated between late 2024 and early 2025. Most predictions place the 5090 and 5080 before the finish of this yr, with mid-range and mainstream variants following in 2025. However, YouTuber Moore’s Law is Dead up to date his info to recommend that solely the top-end GPU would launched in 2024, prompting Kopite to reply, stating that the 5080 ought to debut first.
It’s not ture. RTX 5080 must be launched first.
– kopite7kimi (@kopite7kimi) May 7, 2024
One potential purpose for the switch-up is that the RTX 5080 would possibly outperform the strongest Nvidia GPU presently accessible – the 4090 – thus making it the most potent graphics card on the market till the 5090 begins delivery. Most studies point out that rival vendor AMD will introduce the RDNA 4 lineup this yr, however it should solely encompass mid-range merchandise. However, sources lately knowledgeable Benchlife that the new Radeon collection could be delayed till 2025. Intel plans to launch its first-ever enthusiast-class GPU in 2024, however there’s nonetheless no indication of the way it compares to Blackwell.
Benchlife additionally studies that coolers for the RTX 50 collection are presently present process testing, revealing 4 designs ranging between 250w and 600w. Trial wattage scores do not essentially mirror last merchandise, so prospects should not count on the 5090 to require anyplace close to 600w.
For occasion, early studies surrounding the 40 collection sparked concern that its flagship would possibly draw over 800W, however the 4090 ended up solely utilizing 450W. On the reverse finish of the scale, the preliminary 250W assessments for Blackwell could characterize an RTX 5070 or 5070 Ti.
Looking additional forward, analyst Ming-Chi Kuo studies that the successor to Nvidia’s Blackwell AI GPUs, codenamed “Rubin,” will start mass manufacturing in the fourth quarter of 2025, with rack options following in early 2026.
The lineup’s inaugural chip, the R100, will advance to TSMC’s 3nm N3 node, proceed utilizing CoWoS-L packaging, incorporate a 4x reticle design, and have HBM4 reminiscence. The GR200 Grace CPU may even make the most of N3, transferring the total system past Blackwell’s 5nm node.
