NVIDIA is allegedly dropping the next-gen TITAN graphics card based on the Ada Lovelace GPU architecture and instead working on the GeForce RTX 4090 Ti.
NVIDIA Replaces Ada Lovelace TITAN In Favor of GeForce RTX 4090 Ti Graphics Card
The rumor comes from Kopite7kimi who states that NVIDIA won't be launching the TITAN Ada graphics card. Earlier, there were rumors that NVIDIA was working on a new TITAN-class card, aimed at the professional segment. The GPU was meant to feature more cores and even faster memory while doubling the capacity and also upping the power requirements over 600W. This graphics card was going to be an absolute behemoth but it looks like we won't be getting that.
We won't see Titan of Ada Lovelace. 😩
— kopite7kimi (@kopite7kimi) October 17, 2022
4090Ti is still possible in the future.
— kopite7kimi (@kopite7kimi) October 17, 2022
Kopite says that instead, NVIDIA will be preparing its flagship GeForce RTX 4090 Ti graphics card which will share the same specifications but with a 24 GB memory capacity and DRAM speeds of 24.00 Gbps for the GDDR6X VRAM.
NVIDIA GeForce RTX 4090 Ti "Rumored" Specifications
The NVIDIA GeForce RTX 4090 Ti is said to be a partially cut-die with just 2 of the 144 SMs disabled for a total of 18176 CUDA cores. The GPU will come packed with 96 MB of L2 cache and a total of 224 ROPs which is simply insane. The card will retain the TSMC 4N process node and offer a nice boost to clock speeds with the card being rated at 2.75 GHz (Default) and hitting up to 2.95 or even 3 GHz boost clocks in typical gaming workloads. That's a +190 MHz increase in the default clocks.
As for memory specs, the GeForce RTX 4090 Ti is expected to rock 24 GB GDDR6X capacities that might come at faster 24 Gbps speeds across a 384-bit bus interface. This will provide up to 1.152 TB/s of bandwidth. Now all these boosted specifications will result in higher power draw too and the flagship is expected to operate at a TBP of around 600W. The card is said to feature a TGP of 475W which is a 25W increase over the RTX 4090 but custom variants may once again push this higher up in the 600W+ range.
We have seen that the NVIDIA GeForce RTX 4090 is an extremely efficient card and can be under-volted to 70% of its power budget and deliver the same performance. If it is true that the 4090 Ti uses binned dies, that might explain the lower wattage figure as NVIDIA could have optimized the voltages further. The graphics card is said to offer around 10-20% better performance than the RTX 4090 so that should definitely hit or even surpass the 2x performance figure over the RTX 3090 Ti which will be truly impressive. However, a card of this nature won't launch soon.
Now while the Titan-Class Ada GPU has been canceled, NVIDIA already has two solutions out for the professional AI and workstation segment. The NVIDIA L40 is already announced and features 48 GB of GDDR6 memory with ECC support in a 300W TDP package while the RTX 6000 Ada comes with a similar core and memory configuration. Both of these cards come in a dual-slot design, with the RTX 6000 offering a blower-style cooler and the L40 using a passive-cooled heatsink. Both cards also use a single 16-pin connector.
NVIDIA will most likely keep the beast under wraps until AMD has a formidable opponent in the RDNA 3 Radeon RX 7000 lineup so we may not get to see the RTX 4090 Ti in action for quite some time. NVIDIA also has to focus on more mainstream parts in the coming months.
NVIDIA GeForce RTX 4090 Ti "Rumored" Specs:
| Graphics Card Name | NVIDIA GeForce RTX 4090 Ti | NVIDIA GeForce RTX 4090 | NVIDIA GeForce RTX 4080 16G | NVIDIA GeForce RTX 4080 12G | NVIDIA GeForce RTX 3090 Ti |
|---|---|---|---|---|---|
| GPU Name | Ada Lovelace AD102-400? | Ada Lovelace AD102-300 | Ada Lovelace AD103-300 | Ada Lovelace AD104-400 | Ampere GA102-225 |
| Process Node | TSMC 4N | TSMC 4N | TSMC 4N | TSMC 4N | Samsung 8nm |
| Die Size | 608mm2 | 608mm2 | 378.6mm2 | 294.5mm2 | 628.4mm2 |
| Transistors | 76 Billion | 76 Billion | 45.9 Billion | 35.8 Billion | 28 Billion |
| CUDA Cores | 18432 | 16384 | 9728 | 7680 | 10240 |
| TMUs / ROPs | TBD | 512 / 176 | 320 / 112 | 240 / 80 | 320 / 112 |
| Tensor / RT Cores | TBD | 512 / 128 | 304 / 76 | 240 / 60 | 320 / 80 |
| Base Clock | TBD | 2230 MHz | 2210 MHz | 2310 MHz | 1365 MHz |
| Boost Clock | 2710 MHz | 2520 MHz | 2510 MHz | 2610 MHz | 1665 MHz |
| FP32 Compute | 100 TFLOPs | 83 TFLOPs | 49 TFLOPs | 40 TFLOPs | 40 TFLOPs |
| RT TFLOPs | TBD | 191 TFLOPs | 113 TFLOPs | 82 TFLOPs | 78 TFLOPs |
| Tensor-TOPs | TBD | 1321 TOPs | 780 TOPs | 641 TOPs | 320 TOPs |
| Memory Capacity | 24 GB GDDR6X | 24 GB GDDR6X | 16 GB GDDR6X | 12 GB GDDR6X | 12 GB GDDR6X |
| Memory Bus | 384-bit | 384-bit | 256-bit | 192-bit | 384-bit |
| Memory Speed | 24.0 Gbps | 21.0 Gbps | 23.0 Gbps | 21.0 Gbps | 19 Gbps |
| Bandwidth | 1152 GB/s | 1008 GB/s | 736 GB/s | 504 GB/s | 912 Gbps |
| TBP | 475W | 450W | 320W | 285W | 350W |
| Price (MSRP / FE) | $2199 US? | $1599 US | $1199 US | $899 US | $1199 |
| Launch (Availability) | TBD | October 2022 | November 2022 | November 2022 | 3rd June 2021 |
News Source: Videocardz
