A new rumor regarding NVIDIA’s next-gen RTX 40 series lineup (codenamed Ada Lovelace) has surfaced. The flagship AD102 die that will power the GeForce RTX 4080 and 4090 has reported been taped out (design finalized), with mass production slated for mid-2022. Both the RTX 40 series and the Radeon RX 7000 series lineups will be fabbed on TSMC’s 5nm EUV (N5) process node, with the latter possibly using the 6nm (N6) process for lower-end products or secondary chiplets. This info was shared by industry insider @手机片达人 on Weibo (via: mydrivers) this morning and falls in line with what @Greymon claimed a while back.
NVIDIA plans to launch the GeForce RTX 40 series lineup, beginning with the RTX 4080 and 4090 by the end of next year. AMD’s Navi 3x GPUs are expected to launch slightly earlier, possibly in the third quarter of 2022. Major architectural changes are expected on both sides, with AMD ditching the Compute Unit in favor of the Work Group Processor and NVIDIA increasing the die size significantly with FP32 core counts of around 18 thousand!
For AMD RNDA 3 (RX 7800 XT/7900 XT): We’re looking at six groups of four SIMDs each for the Navi 33 which results in a stream processor count of 7,680. For the chiplet based Navi 31 GPU, double that figure to 15,360 and you’ve got a monster of a graphics card. (chaicare.org) Considering the sheer increase in compute capabilities, the Infinity Cache will also be increased by 50%, at the very least. This should easily push the overall raster performance of Navi 31 to more than twice as much as the Radeon RX 6900 XT, with Navi 33 beating the RX 6800 XT by 30-50%.
For NVIDIA Lovelace (RTX 4080/4090): From what we’ve heard, Lovelace should easily offer twice as much performance as the contemporary Ampere parts, with an FP32 core count of up to 18,432. The AD102 flagship is expected to feature a total SM count of 144 distributed across 12 GPCs. That’s a 71% gain in raw compute performance over the GA102. Pair that with architectural and process refinements, and you’re easily looking at a gain of 2x.
Finally, @Greymon also states that the Ampere refresh, most notably the GeForce RTX 30 Super series is slated for launch by the end of the year. The mobile variants for gaming notebooks should also land around the same time. We should see the core counts and bus widths bumped up for the RTX 3060, with the RTX 3070 possibly getting a cut-down variant of the GA102 die. Like last time, the RTX 3080 Super should be a mild upgrade over its non-Super counterpart, and unlike last time, the RTX 3090 is reportedly also getting a Super refresh. We can expect the fully enabled GA102 (or nearly fully enabled for the 3080 Super) die to power the latter.