Nvidia’s next-gen RTX 60 GPUs might not be released until 2028, prominent leaker claims — gaming takes a back seat as company focuses on delivering data center

Nvidia's next-gen RTX 60 GPUs might not be released until 2028, prominent leaker claims — gaming takes a back seat as company focuses on delivering data center

Sayem Ahmed Social Links Navigation Subscription Editor Sayem Ahmed is the Subscription Editor at Tom's Hardware. He covers a broad range of deep dives into hardware, both new and old, including the CPUs, GPUs, and everything else that uses a semiconductor. He has worked as a professional tech journalist since 2015 and has written for Gamespot, IGN, and Dexerto.

usertests Leaker Moore's Law is Dead has claimed that the RTX 6090, or something from the 60 series, is "technically aiming for H1 2027 as of NOW". CP98gJdcTCo Logically, not releasing any consumer or technically consumer-oriented products in 2027 seems to make moore sense. So if you were hoping to see an epic MLID faceplant or backtracking, this may be a good chance. Even if there's no graphics card coming in 2027, we can at least enjoy some more mud slinging between these two leakers. Reply

thesyndrome usertests said: Leaker Moore's Law is Dead has claimed that the RTX 6090, or something from the 60 series, is "technically aiming for H1 2027 as of NOW". CP98gJdcTCo Logically, not releasing any consumer or technically consumer-oriented products in 2027 seems to make moore sense. So if you were hoping to see an epic MLID faceplant or backtracking, this may be a good chance. Even if there's no graphics card coming in 2027, we can at least enjoy some more mud slinging between these two leakers. I take MLiD with a HEAVY grain of salt these days, because I was following every leak leading up to Blackwell, and people like him an redgamingtech would constantly talk about how Blackwell was going to be "the new Pascal" for like a year, then it finally launched and it was just a VERY SLIGHTLY better Ada Lovelace with a bunch of artificially exclusive (as we've now learned) AI upscaling and frame generation features. Reply

usertests thesyndrome said: I take MLiD with a HEAVY grain of salt these days, because I was following every leak leading up to Blackwell, and people like him an redgamingtech would constantly talk about how Blackwell was going to be "the new Pascal" for like a year, then it finally launched and it was just a VERY SLIGHTLY better Ada Lovelace with a bunch of artificially exclusive (as we've now learned) AI upscaling and frame generation features. You should treat any leaks as entertainment and conversation starters. Even the "good" leakers like Jaykihn issue corrections often. I think he has had worse predictions/leaks under his belt than Blackwell. I'd have to recheck what he was saying in the run up to release (always a chore sifting through old videos), but a lot of people were blindsided by Blackwell's meager uplifts. Ultimately, the big clue was that Blackwell used the same node as Lovelace. Not the end of the world, but never a good sign if you're looking for big performance and efficiency gains. There's almost no chance that 60 series doesn't use TSMC 3nm, which will be an improvement. Soft predicting (relaying what a "source" has said) a launch in the first half of 2027 is going against the grain given market conditions, and will be proven right or wrong within the next 6-9 months. Reply

txfeinbergs This is excellent news as far as I am concerned. That means I will have the best card available for at least 3 years before something better is available. I wish new cards only came out every 5 years. At least that way game publishers wouldn't be able to get lazy and just assume more powerful cards will fix their crappy optimization. Reply

thestryker I can't imagine we're going to see any consumer video cards with a new architecture before the end of next year. I have seen the rumor that a singular GPU from AMD may launch next year, but unless that ends up being faster than a 5090 I don't see nvidia caring enough to release competition. Intel's next architecture will launch next year, but we already know there aren't any in dGPU form. The only add in board design doesn't have ROPs as it has been designed purely for datacenter. There is one advantage of the memory supply shortages and that's more emphasis being placed on upgrading technology. I believe the GDDR7 specification calls for up to 64Gb IC and the highest we've seen so far is 24Gb. If 32Gb goes into production next year that could drive down the costs on 24Gb which have risen more than 16Gb. In theory that could unlock volume which makes sense to use in consumer level video cards. I'm also curious if Intel's work on Crescent Island will have any sort of impact on the consumer side once LPDDR6 is in high volume. When looking at LPDDR6 it would take 6 maximum bus width packages at 14400 to roughly match the bandwidth of 8 GDDR7 packages at 32000. While it wouldn't particularly be important for the consumer side using such a design would allow for high VRAM capacity variance. This would allow for the potential of something like 12GB on a consumer SKU and 96GB on an enterprise class SKU without fundamentally changing anything anything. The thing I'm finding myself most curious about though is RZL-AX and the RXL-S with 12 Xe3P cores (should it still happen). It sounds like both of these were intended to be NVL parts but have been pushed back. If RZL isn't pushed back it's possible both would appear before the end of 2027. Reply

usertests thestryker said: There is one advantage of the memory supply shortages and that's more emphasis being placed on upgrading technology. I believe the GDDR7 specification calls for up to 64Gb IC and the highest we've seen so far is 24Gb. If 32Gb goes into production next year that could drive down the costs on 24Gb which have risen more than 16Gb. In theory that could unlock volume which makes sense to use in consumer level video cards. https://videocardz.com/newz/4gb-and-6gb-gddr7-chips-reportedly-planned-for-the-next-two-years Maybe you already saw this featured article. Memory demand could already be winding down by the time 32 Gb or 48 Gb GDDR7 reach the market. So I'm optimistic about it. An interesting aspect is that the PS6 or Helix could be using 24 Gb or 32 Gb. It's unlikely they will use the latter, but not impossible. This decision may have less profound effects than previous memory decisions (like Xbox 360 going with 512 MB instead of 256 MB), but will still affect gaming for a long time. An extra 10-12 GB could make the use of LLMs in gaming more credible. Reply

usertests thestryker said: This would allow for the potential of something like 12GB on a consumer SKU and 96GB on an enterprise class SKU without fundamentally changing anything anything. Low-end RDNA5 dGPU may use LPDDR5X, with relatively large L2 cache, shared chiplet with mobile. If so, a good first test of this strategy. thestryker said: The thing I'm finding myself most curious about though is RZL-AX and the RXL-S with 12 Xe3P cores (should it still happen). Intel Razor Lake-AX reportedly has 16 or 32 Xe3 graphics cores It used to be 48, now we're looking at 16/32, still more than 12. Razor Lake-AX should be thought of as a Strix Halo competitor, while Razor Lake-S w/ 12 Xe3P is more like AMD's socketed 128-bit desktop APUs, except much faster than 8700G and friends. That's from extrapolating from Panther Lake, and due in part to having a big L2 cache. Reply

Key considerations

  • Investor positioning can change fast
  • Volatility remains possible near catalysts
  • Macro rates and liquidity can dominate flows

Reference reading

More on this site

Informational only. No financial advice. Do your own research.

Leave a Comment