NVIDIA may soon be offering their next ultimate flagship graphics card, the GeForce RTX 2080 Ti SUPER. The card has been going through the rumor mill for a while now with a launch reportedly planned for early 2020 and now we have a new rumor which covers its specifications and they are as meaty as you could expect from a flagship NVIDIA GeForce graphics card.
NVIDIA GeForce RTX 2080 Ti SUPER Reportedly Rocks Full TU102 GPU With 4608 CUDA Cores & 16 Gbps GDDR6 Memory
The latest information on the alleged NVIDIA GeForce RTX 2080 Ti SUPER graphics card once again comes from Twitter user kopite7kimi whose previously leaked information for the NVIDIA GeForce RTX SUPER and GeForce GTX 16 series lineup turned out to be accurate. The latest rumor from the user talks about the specifications of the card and what we can expect from the graphics card in terms of performance.
It is stated in the tweet that the NVIDIA GeForce RTX 2080 Ti SUPER will feature 4608 cores and 16 Gbps GDDR6 memory. So let's start with the core configuration, 4608 CUDA Cores mean that the full Turing TU102 GPU would be featured on the RTX 2080 Ti SUPER, offering 576 tensor cores, 72 RT cores, 288 texture units, and 96 ROPs. The existing GeForce RTX 2080 Ti graphics card has a rated clock speed of 1350 MHz (base) and 1635 MHz (OC). It is also based on a cut-down die while the TITAN RTX which uses the full TU102 die, features the same 1350 MHz base clock but a boost clock of 1770 MHz.
We may see the GeForce RTX 2080 Ti SUPER being clocked somewhere around the 1700 MHz range, but whether it would be clocked higher than the TITAN RTX is a decision that NVIDIA would have a hard time making considering that if the clock speed is the same as the TITAN RTX, the only thing that makes the TITAN RTX any good would be its higher memory capacity and if not, then the TITAN RTX would stay at the top of things when comparing reference variants. The RTX 2080 Ti SUPER will get custom treatment with higher clocks but as an out of the box reference solution, the TITAN RTX would still be the fastest consumer (prosumer*) variant of the Turing family.
Coming to the memory specifications, the only thing that is rumored is that 16 Gbps GDDR6 dies would be used on the NVIDIA GeForce RTX 2080 Ti SUPER. There's no mention of the memory bus or the memory capacity which will be featured on the card. Regardless, the memory will be the main differentiating factor for both cards. The TITAN RTX features 24 GB of GDDR6 memory clocked at 14 Gbps along a 384-bit bus.
The RTX 2080 Ti SUPER could go two routes, either 12 GB of GDDR6 memory running along a 384-bit bus at 16 Gbps or 11 GB of GDDR6 memory running alongside a 352-bit bus interface. The latter seems more plausible since the top tier GeForce cards have been featuring cut-down memory buses for a while now, but if 384-bit is to happen, then that would be a super upgrade for sure. The GeForce RTX 2080 SUPER didn't support the full 16 Gbps dies because that would have required a PCB revision which would cost AIBs some extra money, but with a card like the RTX 2080 Ti SUPER, AIBs may be willing to make the necessary upgrades.
NVIDIA GeForce RTX 20 SUPER Lineup Specifications:
| Graphics Card Name | NVIDIA GeForce RTX 2060 | NVIDIA GeForce RTX 2060 SUPER | NVIDIA GeForce RTX 2070 | NVIDIA GeForce RTX 2070 SUPER | NVIDIA GeForce RTX 2080 | NVIDIA GeForce RTX 2080 SUPER | NVIDIA GeForce RTX 2080 Ti |
|---|---|---|---|---|---|---|---|
| GPU Architecture | Turing GPU (TU106) | Turing GPU (TU106) | Turing GPU (TU106) | Turing GPU (TU104) | Turing GPU (TU104) | Turing GPU (TU104) | Turing GPU (TU102) |
| Process | 12nm FNN | 12nm FNN | 12nm FNN | 12nm FNN | 12nm FNN | 12nm FNN | 12nm FNN |
| Die Size | 445mm2 | 445mm2 | 445mm2 | 545mm2 | 545mm2 | 545mm2 | 754mm2 |
| Transistors | 10.6 Billion | 10.6 Billion | 10.6 Billion | 13.6 Billion | 13.6 Billion | 13.6 Billion | 18.6 Billion |
| CUDA Cores | 1920 Cores | 2176 Cores | 2304 Cores | 2560 Cores | 2944 Cores | 3072 Cores | 4352 Cores |
| TMUs/ROPs | 120/48 | 136/64 | 144/64 | 184/64 | 192/64 | 192/64 | 288/96 |
| GigaRays | 5 Giga Rays/s | 6 Giga Rays/s | 6 Giga Rays/s | 7 Giga Rays/s | 8 Giga Rays/s | 8 Giga Rays/s | 10 Giga Rays/s |
| Cache | 4 MB L2 Cache | 4 MB L2 Cache | 4 MB L2 Cache | 4 MB L2 Cache | 4 MB L2 Cache | 4 MB L2 Cache | 6 MB L2 Cache |
| Base Clock | 1365 MHz | 1470 MHz | 1410 MHz | 1605 MHz | 1515 MHz | 1650 MHz | 1350 MHz |
| Boost Clock | 1680 MHz | 1650 MHz | 1620 MHz 1710 MHz OC | 1770 MHz | 1710 MHz 1800 MHz OC | 1815 MHz | 1545 MHz 1635 MHz OC |
| Compute | 6.5 TFLOPs | 7.5 TFLOPs | 7.5 TFLOPs | 9.1 TFLOPs | 10.1 TFLOPs | 11.1 TFLOPs | 13.4 TFLOPs |
| Memory | Up To 6 GB GDDR6 | Up To 8 GB GDDR6 | Up To 8 GB GDDR6 | Up To 8 GB GDDR6 | Up To 8 GB GDDR6 | Up To 8 GB GDDR6 | Up To 11 GB GDDR6 |
| Memory Speed | 14.00 Gbps | 14.00 Gbps | 14.00 Gbps | 14.00 Gbps | 14.00 Gbps | 15.50 Gbps | 14.00 Gbps |
| Memory Interface | 192-bit | 256-bit | 256-bit | 256-bit | 256-bit | 256-bit | 352-bit |
| Memory Bandwidth | 336 GB/s | 448 GB/s | 448 GB/s | 448 GB/s | 448 GB/s | 496 GB/s | 616 GB/s |
| Power Connectors | 8 Pin | 8 Pin | 8 Pin | 8+6 Pin | 8+8 Pin | 8+8 Pin | 8+8 Pin |
| TDP | 160W | 175W | 185W (Founders) 175W (Reference) | 215W | 225W (Founders) 215W (Reference) | 250W | 260W (Founders) 250W (Reference) |
| Starting Price | $349 US | $399 US | $499 US | $499 US | $699 US | $699 US | $999 US |
| Price (Founders Edition) | $349 US | $399 US | $599 US | $499 US | $699 US | $699 US | $1,199 US |
| Launch | January 2019 | July 2019 | October 2018 | July 2019 | September 2018 | July 2019 | September 2018 |
Pricing of the card will be a major consideration for buyers before they go out and buy the new card. I believe that NVIDIA will do what they did with the SUPER lineup. Remove the Founders Edition tax and stick with one MSRP which in the case of the RTX 2080 Ti was $999 US. NVIDIA can afford that since Turing has been in production for over a year now and there's no graphics unit supply constraint affecting the Turing cards at all. As for the launch, NVIDIA could unveil the card at CES 2020 if they want to since their entire 2019 SUPER lineup will have launched by then. Then again, NVIDIA may just do a special game-specific event for their flagship product launch prior to the next-gen lineup which is expected to debut much later in 2020.
