GeForce RTX 4080, 4070 And 4060 Alleged Launch Dates Put A Damper On Waiting To Upgrade
There's Intel currently releasing its first discrete GPUs in over two decades, along with what may be its last-ever desktop CPUs with a monolithic design later this year. NVIDIA's going to drop a bomb called Ada Lovelace on the industry, with a top-end GPU that may suck down over 600 watts to do its duty. Then you have AMD, ever the scrappy underdog, about to release a one-two punch of its Ryzen 7000-series processors (with its associated Socket AM5 platform) and its Radeon RX 7000-series GPUs.
We have quite a bit of information on these products thanks to myriad leaks, but most of them aren't even acknowledged to exist yet beyond the concept stage, much less being formally announced. Naturally, that means we don't have release dates for any of this stuff. To that end, here comes our favorite Digimon-themed device data distributor, Greymon55, with a roundup of release windows for some upcoming hardware.
In a few tweets and replies, Greymon55 puts AMD's Raphael down for an announcement in August with availability in September. Apparently the red team's RDNA 3 GPUs will come a bit later, with an announcement in October and availability in November. Meanwhile, Intel's 13th-generation Core processors will apparently come along between the two, announcing in September and releasing in October.
For NVIDIA's part, it seems like the best guess is that the company will launch the largest Lovelace die, known as "AD102," at some point this year, but smaller GPUs based on the same technology will have to wait until next year. That's a real bummer for folks who don't want to pay $1,500 or more for a graphics card. Scuttlebutt has it that AD102 will only be used in the flagship GPU, likely to be named GeForce RTX 4090 or similar.
If the leaks are accurate, AD102 is an absolutely massive graphics processor with some 18432 CUDA cores and a 384-bit memory bus. Recent rumors have stated that it will clock all the way up to 3 GHz, which would endow the card with single-precision floating-point (FP32) throughput of over 100 TFLOPs. It seems like like it wasn't all that long ago that the GTX 480 broke 1 TLFOPs.
