New details of NVIDIA's GeForce RTX 4090 & RTX 4080 graphics cards have been revealed regarding the PCB and memory capacities.
NVIDIA GeForce RTX 4080 Reportedly Features Two PCB Designs With 16 & 12 GB Capacities, RTX 4090 Gets 14-Layer PCB
According to information by MEGAsizeGPU, the NVIDIA GeForce RTX 4090 will come in one PCB flavor while the RTX 4080 will come in two flavors. The RTX 4090 will be using the same PCB for reference and AIC designs which will be equipped with up to 24 GB of GDDR6X memory and 14 layers for the PCB in total. The NVIDIA GeForce RTX 4080 has already leaked out which you can read more about here.
4080 12G GD6X 10layer PCB aic reference design
4080 16G GD6X 12layer PCB aic
reference design
4090 24G GD6X 14layer PCB aic reference designnote:4080 12G has a completely different pcb design from 4080 16G
— MEGAsizeGPU (@Zed__Wang) September 5, 2022
The NVIDIA GeForce RTX 4080, on the other hand, is reportedly going to have two entirely different PCB designs. The first one is said to feature up to 16 GB GDDR6X memory capacities and a 12-layer design for AIC cards while the second model is going to feature 12 GB GDDR6X memory support & a 10-layer PCB design for both AIC and reference models.
Now we have been hearing reports of 16 GB memory being the official model and nothing about a 12 GB variant but NVIDIA did have two RTX 3080 models, one with 10 GB and the other with 12 GB, launched previously. The RTX 3080 12 GB features a 384-bit bus while the RTX 3080 featured a 320-bit bus.
Since the RTX 4080 is rumored to feature a 256-bit bus for the 16 GB variant, the 12 GB AIC model will feature a cut-down 192-bit bus interface which is a significant reduction and surprisingly low bus width for a flagship graphics card. NVIDIA will have to compensate for the GPU bandwidth by upping the memory clocks and introducing a new memory compression technology to make up for its lost bandwidth. Maybe that's why the RTX 4080 was recently rumored to have received a 23 Gbps GDDR6X memory bump.
It would be no surprise to us if the 16/12 GB models could be nothing more than engineering samples as we have seen 16 GB and 20 GB RTX 3080 models in the wild. These graphics cards were indeed made and produced by NVIDIA but were not feasible for a retail launch.
NVIDIA GeForce RTX 4080 'Expected' Specifications
The NVIDIA GeForce RTX 4080 is expected to utilize a cut-down AD103-300 GPU configuration with 9,728 cores or 76 SMs enabled of the total 84 units whereas the previous configuration offered 80 SMs or 10,240 cores. While the full GPU comes packed with 64 MB of L2 cache and up to 224 ROPs, the RTX 4080 might end up with 48 MB of L2 cache and lower ROPs too due to its cut-down design. The card is expected to be based on the PG136/139-SKU360 PCB.
As for memory specs, the GeForce RTX 4080 is expected to rock 16 GB GDDR6X capacities that are said to be adjusted at 23 Gbps speeds across a 256-bit bus interface. This will provide up to 736 GB/s of bandwidth. This is still a tad bit slower than the 760 GB/s bandwidth offered by the RTX 3080 since it comes with a 320-bit interface but a lowly 10 GB capacity. To compensate for the lower bandwidth, NVIDIA could be integrating a next-gen memory compression suite to make up for the 256-bit interface.
For power, the TBP is now set to be rated at 340W, a 20W increase from the previous 320W spec that we got. This brings the TBP to the same ballpark as the existing RTX 3080 graphics card (up to 350W). Now it is not known whether the other RTX 40 series graphics cards will also be getting the faster GDDR6X memory treatment but we know that Micron has commenced full mass production of up to 24 Gbps GDDR6X memory modules so they have to go somewhere.
- NVIDIA GeForce RTX 4080 "Expected" TBP - 340W
- NVIDIA GeForce RTX 3080 "Official" TBP - 350W
NVIDIA GeForce RTX 40 Series Official Specs:
Graphics Card Name | NVIDIA GeForce RTX 4090 | NVIDIA GeForce RTX 4090 D | NVIDIA GeForce RTX 4080 | NVIDIA GeForce RTX 4070 Ti | NVIDIA GeForce RTX 4070 | NVIDIA GeForce RTX 4060 Ti | NVIDIA GeForce RTX 4060 |
---|---|---|---|---|---|---|---|
GPU Name | Ada Lovelace AD102-300 | Ada Lovelace AD102-250 | Ada Lovelace AD103-300 | Ada Lovelace AD104-400 | Ada Lovelace AD104-250 | Ada Lovelace AD106-350 | Ada Lovelace AD107-400 |
Process Node | TSMC 4N | TSMC 4N | TSMC 4N | TSMC 4N | TSMC 4N | TSMC 4N | TSMC 4N |
Die Size | 608mm2 | 608mm2 | 378.6mm2 | 294.5mm2 | 294.5mm2 | 190.0mm2 | 146.0mm2 |
Transistors | 76 Billion | 76 Billion | 45.9 Billion | 35.8 Billion | 35.8 Billion | 22.9 Billion | TBD |
CUDA Cores | 16384 | 14592 | 9728 | 7680 | 5888 | 4352 | 3072 |
TMUs / ROPs | 512 / 176 | TBD | 320 / 112 | 240 / 80 | 184 / 64 | 136 / 48 | TBD |
Tensor / RT Cores | 512 / 128 | 456 / 128 | 304 / 76 | 240 / 60 | 184 / 46 | 136 / 34 | TBD |
L2 Cache | 72 MB | 72 MB | 64 MB | 48 MB | 36 MB | 32 MB | 24 MB |
Base Clock | 2230 MHz | 2280 MHz | 2210 MHz | 2310 MHz | 1920 MHz | 2310 MHz | 1830 MHz |
Boost Clock | 2520 MHz | 2520 MHz | 2510 MHz | 2610 MHz | 2475 MHz | 2535 MHz | 2460 MHz |
FP32 Compute | 83 TFLOPs | TBD | 49 TFLOPs | 40 TFLOPs | 29 TFLOPs | 22 TFLOPs | 15 TFLOPs |
RT TFLOPs | 191 TFLOPs | TBD | 113 TFLOPs | 82 TFLOPs | 67 TFLOPs | 51 TFLOPs | 35 TFLOPs |
Tensor-TOPs | 1321 TOPs | TBD | 780 TOPs | 641 TOPs | 466 TOPs | 353 TOPs | 242 TOPs |
Memory Capacity | 24 GB GDDR6X | 24 GB GDDR6X | 16 GB GDDR6X | 12 GB GDDR6X | 12 GB GDDR6X | 8-16 GB GDDR6 | 8 GB GDDR6 |
Memory Bus | 384-bit | 384-bit | 256-bit | 192-bit | 192-bit | 128-bit | 128-bit |
Memory Speed | 21.0 Gbps | 21.0 Gbps | 23.0 Gbps | 21.0 Gbps | 21.0 Gbps | 18.0 Gbps | 17.0 Gbps |
Bandwidth | 1008 GB/s | 1008 GB/s | 736 GB/s | 504 GB/s | 504 GB/s | 288 GB/s (554 GB/s Effective) | 272 GB/s (453 GB/s Effective) |
TBP | 450W | 425W | 320W | 285W | 200W | 160-165W | 115W |
Price (MSRP / FE) | $1599 US / 1949 EU | 12,999 RMB (China-Only) | $1199 US / 1469 EU | $799 US | $599 US | $399-$499 US | $299 US |
Price (Current) | $1599 US / 1859 EU | 12,999 RMB (China-Only) | $1199 US / 1399 EU | $799 US | $599 US | $399-$499 US | $299 US |
Launch (Availability) | 12th October 2022 | 28th December 2023 | 16th November 2022 | 5th January 2023 | 13th April 2023 | 24th May / 18th July 2023 | 29th June 2023 |
NVIDIA GeForce RTX 4090 'Expected' Specifications
The NVIDIA GeForce RTX 4090 will use 128 SMs of the 144 SMs for a total of 16,384 CUDA cores. The GPU will come packed with 96 MB of L2 cache and a total of 384 ROPs which is simply insane. The clock speeds are not confirmed yet but considering that the TSMC 4N process is being used, we are expecting clocks between the 2.0-3.0 GHz range.
As for memory specs, the GeForce RTX 4090 is expected to rock 24 GB GDDR6X capacities that will be clocked at 21 Gbps speeds across a 384-bit bus interface. This will provide up to 1 TB/s of bandwidth. This is the same bandwidth as the existing RTX 3090 Ti graphics card and as far as the power consumption is concerned, the TBP is said to be rated at 450W which means that TGP may end up lower than that. The card will be powered by a single 16-pin connector which delivers up to 600W of power. It is likely that we may get 500W+ custom designs as we saw with the RTX 3090 Ti.
As for its feature set, the NVIDIA GeForce RTX 4090 and RTX 4080 graphics cards will rock all the modern NV feature sets such as the latest 4th Gen Tensor Cores, 3rd gen RT cores, the latest NVENC Encoder, and NVCDEC Decoder, and support for the latest APIs. They will pack all the modern RTX features such as DLSS, Reflex, Broadcast, Resizable-BAR, Freestyle, Ansel, Highlights, Shadowplay, and G-SYNC support too.
- NVIDIA GeForce RTX 4090 "Expected" TBP - 450W
- NVIDIA GeForce RTX 3090 "Official" TBP - 350W
NVIDIA GeForce RTX 4090 Ti & RTX 4090 'Preliminary' Specs:
Graphics Card Name | NVIDIA GeForce RTX 4090 Ti | NVIDIA GeForce RTX 4090 | NVIDIA GeForce RTX 3090 Ti | NVIDIA GeForce RTX 3090 |
---|---|---|---|---|
GPU Name | Ada Lovelace AD102-350? | Ada Lovelace AD102-300? | Ampere GA102-350 | Ampere GA102-300 |
Process Node | TSMC 4N | TSMC 4N | Samsung 8nm | Samsung 8nm |
Die Size | ~600mm2 | ~600mm2 | 628.4mm2 | 628.4mm2 |
Transistors | TBD | TBD | 28 Billion | 28 Billion |
CUDA Cores | 18432 | 16128 | 10752 | 10496 |
TMUs / ROPs | TBD / 384 | TBD / 384 | 336 / 112 | 328 / 112 |
Tensor / RT Cores | TBD / TBD | TBD / TBD | 336 / 84 | 328 / 82 |
Base Clock | TBD | TBD | 1560 MHz | 1400 MHz |
Boost Clock | ~2800 MHz | ~2600 MHz | 1860 MHz | 1700 MHz |
FP32 Compute | ~103 TFLOPs | ~90 TFLOPs | 40 TFLOPs | 36 TFLOPs |
RT TFLOPs | TBD | TBD | 74 TFLOPs | 69 TFLOPs |
Tensor-TOPs | TBD | TBD | 320 TOPs | 285 TOPs |
Memory Capacity | 24 GB GDDR6X | 24 GB GDDR6X | 24 GB GDDR6X | 24 GB GDDR6X |
Memory Bus | 384-bit | 384-bit | 384-bit | 384-bit |
Memory Speed | 24.0 Gbps | 21.0 Gbps | 21.0 Gbps | 19.5 Gbps |
Bandwidth | 1152 GB/s | 1008 GB/s | 1008 GB/s | 936 Gbps |
TGP | 600W | 450W | 450W | 350W |
Price (MSRP / FE) | $1999 US? | $1499 US? | $1999 US | $1499 US |
Launch (Availability) | 2023? | October 2022? | 29th March 2022 | 24th September 2020 |
The NVIDIA GeForce RTX 4080 and RTX 4070 graphics cards will be amongst the first graphics cards besides the RTX 4090 to launch to gamers however we can't tell whether the green team plans to launch them within this year soon after RTX 4090 or wait till early next year. The RTX 4090 is so far expected to launch around October 2022.