If the latest rumor from Greymon55 on Twitter is to be believed, NVIDIA’s upcoming Ada Lovelace AD102 GPU could feature a clock speed of up to 2.2 GHz. There have already been some rumors about NVIDIA’s new GPUs, especially the AD102 SKU. It will be the flagship of the upcoming graphics cards, both for gamers and the workstation users. It will also be the successor to the current GA102 GPU and as such, we can definitely expect some interesting specs.
Based on previous rumors, there have been repeated hints that NVIDIA would use TSMC’s N5 (5nm) process node for the Ada Lovelace GPUs. This includes the AD102 SKU, which will rely on a completely monolithic design. His current tweet, which also writes about the specific GPU configurations, goes on to say that the AD102 GPU could have a clock speed of up to 2.2 GHz (or higher).
n31:5nm+6nm 256bit gddr6
n32 5nm+6nm 192bit gddr6
n33 6nm 128bit gddr6 perf>6900xt
ad102 5nm 384bit gddr6x
ad103 ?
ad104 ?
ad106 ?— Greymon55 (@greymon55) September 19, 2021
The AD102 GPU could have 18432 CUDA cores, following the preliminary specs from Kopite’s leak, but this could still change at any time. That would be almost twice as many cores as Ampere, which again would be a huge step up from Turing. The now rumored 2.2 GHz clock speed could enable up to 81 TFLOPs (FP32) of computing power, which would end up being more than double the performance of a current RTX 3090, which still offers 36 TFLOPs of FP32 computing power.
The 125% jump in performance looks huge at first, of course, but it’s also important to remember that NVIDIA has already made a big jump in FP32s in the current generation with Ampere. The GA102 GPU (RTX 3090) offers 36 TFLOPs, while the Turing TU102 GPU (RTX 2080 Ti) still offered 13 TFLOPs. That’s actually an increase of over 150% in FP32 flops, but the RTX 3090’s real-world gaming performance is only about 50-60% faster than the RTX 2080 Ti on average.
We must never hide the fact that FLOPS is not synonymous with actual GPU gaming performance these days. Besides, we don’t know (yet) if 2.2 GHz is the average boost or just the peak, whereas the former could mean that there might be even higher computing potential for AD102. Apart from that, the leaker also claims that the NVIDIA GeForce RTX 40 flagship could retain a 384-bit bus interface analogous to the RTX 3090
By the way, it is also interesting that the leaker explicitly mentions “only” GDDR6X, which could mean that NVIDIA could only switch to a new memory standard after Ada Lovelace, including the current problems with cooling. But we’re still in for a surprise.
Source: Twitter
46 Antworten
Kommentar
Lade neue Kommentare
Urgestein
Veteran
Urgestein
Urgestein
Mitglied
Mitglied
Urgestein
Urgestein
Veteran
Veteran
Urgestein
Mitglied
Mitglied
Veteran
Urgestein
Veteran
Urgestein
Veteran
Alle Kommentare lesen unter igor´sLAB Community →