Past cadence doesn't always indicate what they do these days, and claiming they're been pushed back because some random supposed leakers saying so is not even remotely credible at all. Radeon vii came at an odd launch time for amd, that being in February, rdna 4 could be something similar.
Some of those leakers have a great track record - they are not MLID. Vega20 was launched in November 2018 for commercial cards,so it hit the same cadence as previous large AMD dGPUs. It was never really a consumer dGPU,just re-purposed later on. Large RDNA2 came 21 months later.
If past cadence wasn't followed that is by definition a delay. One of the reasons Blackwell is delayed is because GDDR7 production was pushed back and those leakers did cover it:
The competition among memory manufacturers to advance GDDR7 graphics memory into mass production is intensifying. Samsung and Micron aim to launch GDDR7 chips in Q4 2024, while SK Hynix has announced at Computex 2024 that it plans to begin mass production in the first quarter of 2025.
www.guru3d.com
Samsung and Micron have launched GDDR7 late this quarter. SK Hynix is Q1 2025. There was also a flaw in the commercial Blackwell chips which needed fixing:
Also looking at the cadence(first new release):
1.)RTX4090 - October 12th, 2022
2.)RTX3090 - September 24th, 2020
3.)RTX2080TI - December 18th, 2018
4.)Titan V - December 7th, 2017
4.)Pascal Titan - August 2, 2016
4.)GTX1080 - May 27th 2016
5.)GTX980 - September 18, 2014
Time between generations:
1.)Ampere to Ada Lovelace - 24 to 25 months. Was delayed one month.
2.)Turing to Ampere - 21 months
3.)Volta to Turing - 12 months
4.)Pascal to Volta - 14 months
5.)Pascal to Turing - 27 months. Smaller Volta cancelled.
6.)Maxwell to Pascal - 20 to 23 months
In the last decade,Nvidia refreshed ranges within 24 months.So far Ada Lovelace to Blackwell if launched next month,will be 27 months.But next month might be a reveal at CES,not a launch so it could be even longer. That will be the longest wait for a decade. Even the Pascal to Turing timing had one Volta card launch. So sure you can go with it not being delayed. I don't agree.
AMD,OTH is using an old TSMC 4N process and GDDR6. Instead of exploiting the Nvidia delay,as usual they are launching as close as possible to Nvidia IMHO,so they can price as high as possible.So AMD as usual will delay and delay,right until Nvidia launches Blackwell,which will steal it's thunder next month. I wouldn't be surprised if the RTX5070 ends up appearing soon afterwards.