New RTX 4080 Leak Drops Core Count Below 10,000, 40-Series Titan Rumors
GPU leaker @ Copy Te 7 Kimi (opens in new tab) is another update on the upcoming RTX 4080 spec, and we’re doing it again today. His latest rumors include the RTX 4080’s core count dropping from an estimated 10,240 to 10,752 range to he’s down to 9728. This equates to his 10% reduction in core count.
Thankfully, this change doesn’t appear to change the 4080’s estimated Time Spy Extreme score of 15,000. Performance doesn’t seem to change, at least for now. Other specs of his RTX 4080 that Kopite showed, such as 420W power limit, 16GB of his 21Gbps GDDR6X memory and AD103 die are also unchanged.
Apply a healthy dose of skepticism to these rumors and “leaks” regardless of source. may not. One important piece of information missing is the RTX 40 series core clocks. There’s been speculation that Nvidia will be boosting the core clocks significantly over the existing Ampere line based on the higher power levels, but there’s no solid data yet.
I’m not a chatty person, but I need to update a few things. please do not worry. Possible RTX 4080, PG136/139-SKU360AD103-300-A19728FP32256bit 16G 21Gbps GDDR6Xtotal power ~420WTSE ~15000 Now you have the latest updates for the 4090, 4080 and 4070.August 5, 2022
RTX 40 Series Titan Rumors and L2 Cache Update
another tweet from @kopite7 you has hinted that the Titan brand could return to the GeForce lineup in some way. This RTX 40 series Titan “has a chance” to use his AD102 chip in full fat, but Kopite has not confirmed if that will actually happen. The yield of 100% functional chips is likely to be low, probably on large GPUs, so it’s more likely that it will be a near-perfect AD102.
An interesting question is whether or not we can actually get the new Titan cards. Nvidia Will Again Have Titan “Hybrid” SKUs With Larger Memory And Enhanced Drivers That Blur The Line Between Nvidia’s A-Series Workstation GPUs And The Company’s GeForce Gaming GPU Lineup. Or do you get the RTX 4090 Ti skipping the pro driver enhancements instead? Kopite7kimi seems to offer an educated guess for now.
In the same tweet, Kopite also shared an L2 cache update for the AD103 and AD104 40-series dies. The fully unlocked AD103 die apparently has 64MB of L2 cache, while the AD104 die has 48MB. This is the same one an earlier leak suggested earlier in the year.
However, Kopite also suggests that the RTX 4090’s L2 cache size has been reduced, or that the AD102 may not have a proportionately larger cache. Alternatively, it could be a move to differentiate the potential RTX 4090 Ti (or Titan) from the “lower” RTX 4090s.
1. TITAN has a chance to use full fat chips. I will update when I can confirm. 2. Full fat AD103 has 64M L2 and AD104 has 48M. Only the RTX 4090 should have the L2 cut. (99%😁)August 5, 2022