The primary particulars for NVIDIA’s next-generation Ampere primarily based GeForce RTX 3080 and GeForce RTX 3070 graphics playing cards have allegedly leaked out. The rumored particulars come from varied sources who’ve shared the identical collection of specs for NVIDIA’s upcoming GeForce GPUs.
NVIDIA Ampere GPU Primarily based GeForce RTX 3080 20 GB & GeForce RTX 3070 16 GB Ampere Graphics Playing cards Alleged Specs Detailed
From the main points by MyDrivers and Twitter consumer ‘KittyCorgi’, it looks as if NVIDIA is prepping up not less than two Ampere GPUs which might go on to energy the GeForce RTX 3080 and GeForce RTX 3070 graphics playing cards. The main points cowl the complete specs, together with core counts and even photographs of full block diagrams for NVIDIA’s next-generation Ampere powerhouse GPUs.
The NVIDIA GA102, GA103 & GA104 GPUs, Powered by Ampere
First up, it’s reported that there are two gaming derived GPUs primarily based on the Ampere structure, these embody the GA103 and the GA104. There’s additionally a GA102 GPU talked about however there isn’t any card talked about that might make use of it. It’s doable NVIDIA might maintain the chip underneath wraps for a extra highly effective graphics card.
Like Pascal and Volta, which stored the highest tier chips unique to HPC / AI / DNN markets, it’s also seemingly that the GA100 GPU would exist however could be developed particularly for the compute GPU market with the extra gaming-optimized chips being derived into the GA102/GA103/GA104 tiers. The unique Ampere GPU lineup leak goes far again to Might of 2019 the place a collection of NVIDIA GPUs primarily based on the Ampere structure was hinted (not confirmed), these embody (credit: kopite7kimi):
- GA100 8GPC*8TPC*2SM 6144bit
- GA102 7GPC*6TPC*2SM 384bit
- GA103 6GPC*5TPC*2SM 320bit
- GA104 6GPC*4TPC*2SM 256bit
- GA106 3GPC*5TPC*2SM 192bit
- GA107 2GPC*5TPC*2SM 128bit
Now, none of those GPUs have formally been confirmed however we now have seen references of NVIDIA’s Ampere GA104 GPU popping up in latest ECC listings and that is what these GPU configurations have been primarily based on. Like the present Turing and the earlier Pascal lineup, the GA102 and the GA104 GPU will make up the higher-end market with GA106 making up the mainstream and GA107 targetted on the entry-level section. What’s fascinating is that this leak signifies that the inexperienced crew is perhaps deciding to base the RTX **70 tier graphics card on the upper tier G**04 chips slightly than G**06.
We all know that the Turing primarily based GeForce RTX 2070 makes use of the TU106 chip and never the upper-tier TU104 chip just like the RTX 2080. On the similar time, the RTX **80 tier graphics card wouldn’t be utilizing the G**04 chip however slightly an excellent larger tier GPU referred to as the GA103 which might permit for a wider bus and better reminiscence configuration alongside extra CUDA cores. There isn’t any point out of what fabrication course of Ampere GPUs can be primarily based upon, however NVIDIA’s CEO acknowledged that almost all of orders could be dealt with by TSMC for his or her next-generation 7nm GPUs whereas Samsung would get a small variety of orders. We additionally know that NVIDIA’s Orion SOC which makes use of its next-generation GPU structure (Ampere) relies upon Samsung’s 8nm LPP course of node which makes it clear the place the smaller numbers of orders are going to. With that mentioned, let’s discuss concerning the graphics playing cards themselves.
The Ampere GPU would not solely enhance the rasterization and shading efficiency but in addition ship big uplifts over Turing in ray tracing efficiency. Since Turing was the first technology implementation of ray tracing on NVIDIA graphics playing cards, Ampere will additional optimize it and is predicted to supply extra ray-tracing onboard within the type of RT and Tensor cores. The Ampere GPUs are additionally talked about to leverage efficiency from PCIe Gen four.zero which is a given since this can be a next-generation GPU structure tuned for next-generation gaming platforms and the upper bandwidth from the Gen four.zero protocol, plus NVLINK could be a vital characteristic.
NVIDIA GeForce RTX 3080 Graphics Card – Up To 20 GB GDDR6, 320-bit Bus, GA103 GPU
As talked about above, the NVIDIA GeForce RTX 3080 would be the successor to the GeForce RTX 2080. It is going to characteristic a complete of 60 SMs with 64 CUDA cores every making 3840 CUDA cores. This is able to be a giant leap over the RTX 2080’s CUDA core rely which was set at 3072 and we aren’t even factoring within the clock speeds which NVIDIA will have the ability to churn up with a complicated 7nm course of node.
The cardboard can also be mentioned to characteristic a 320-bit bus interface which implies that it might assist both 10 or 20 GB of GDDR6 reminiscence. If NVIDIA actually desires to blow the viewers away with Ampere, then 20 GB could be an ideal selection. Nonetheless, we now have to think about the excessive prices which might be related to GDDR6 reminiscence and the looming rumors of a reminiscence scarcity as a result of larger demand from the console market.
So let’s be conservative and say 10 GB of GDDR6, which might nonetheless be 2 GB greater than the present GeForce RTX 2080. If we’re taking a look at 16 Gbps modules on the upcoming RTX 3080, it will give us 640 GB/s bandwidth which might be insane over the 496 GB/s of the RTX 2080 SUPER which options 15.5 Gbps reminiscence alongside a 256-bit bus interface.
NVIDIA GeForce RTX 3070 Graphics Card – Up To 16 GB GDDR6, 256-bit Bus, GA104 GPU
The second card is the GeForce RTX 3070 and as acknowledged above, it’ll characteristic the GA104 GPU. The successor to the TU106 primarily based GeForce RTX 2070 is usually recommended to characteristic the identical CUDA core rely because the RTX 2080, at 3072, all rounded up in its 48 streaming multiprocessor models. The cardboard is predicted to characteristic a 256-bit bus interface which as soon as once more would permit for eight GB or as much as 16 GB GDDR6 reminiscence, however it’s important to think about what I simply mentioned above concerning the GDDR6 provide and value issue.
NVIDIA Ampere GPU ‘GeForce RTX 3080 & RTX 3070’ Rumored Specs Comparability:
|Graphics Card Title||NVIDIA GeForce RTX 2070||NVIDIA GeForce RTX 2080||NVIDIA GeForce RTX 3070||NVIDIA GeForce RTX 3080||NVIDIA GeForce RTX 2080 Ti|
|GPU Structure||Turing TU106||Turing TU104||Ampere GA104||Ampere GA104||Turing TU102|
|Course of||12nm FinFET NVIDIA (TSMC)||12nm FinFET NVIDIA (TSMC)||7nm TSMC?||7nm TSMC?||12nm FinFET NVIDIA (TSMC)|
|CUDA Cores||2304 CCs||3072 CCs||3072 CCs||3840 CCs||4608 CCs|
|Reminiscence||eight GB GDDR6||eight GB GDDR6||eight/16 GB GDDR6||10/20 GB GDDR6||11 GB GDDR6|
|Reminiscence Pace||14 Gbps||14 Gbps||16 Gbps?||16 Gbps?||14 Gbps|
|Reminiscence Bandwidth||448 GB/s||448 GB/s||512 GB/s||640 GB/s||640 GB/s|
What’s fascinating is that these two playing cards depart a lot room open for a GeForce RTX 3080 Ti to simply blow away the fanatic market section once more. If the studies of NVIDIA’s total Ampere structure gaining 50% larger efficiency than Turing whereas providing twice the effectivity are appropriate, than given these specs, an RTX 3070 would have the ability to ship the efficiency and even outperform the RTX 2080 Ti at a a lot cheaper price level.
In accordance with the Taipei Occasions, NVIDIA’s Ampere GPU goes to ship significantly spectacular figures in each efficiency and effectivity. The general efficiency achieve is predicted to be 50% larger than current Turing GPUs whereas providing twice the ability effectivity. It’s clearly talked about that Ampere would halve the ability consumption in comparison with current Turing GPUs. Which means a GPU with a decrease TDP determine than the RTX 2060 (125-150W) would have the ability to outperform the 2080 Ti which is significantly spectacular. One can solely count on the type of efficiency which a high-end GPU primarily based on the Ampere structure would provide as we will undoubtedly count on true 4K 60 FPS ray-traced gaming from the up to date structure.
We have been instructed NVIDIA will launch a brand new GPU within the second quarter.
AMD can also be anticipated to have Large Navi round Computex.
Data from board companions.
— VideoCardz.com (@VideoCardz) January 20, 2020
There isn’t any phrase on an actual launch plan for Ampere primarily based GeForce GPUs, however rumors have hinted at a 2H 2020 launch, nonetheless, it may very well be a lot sooner than that. NVIDIA is predicted to host their annual GTC 2020 convention in March, which might imply that we might lastly get to listen to extra concerning the next-generation inexperienced crew powerhouse chip, even when it is an HPC card announcement first. This was the case with Turing, which received to see its first public look within the type of a workstation card or Pascal which was first unveiled as an HPC oriented compute chip. We will certainly maintain you up to date in case we hear extra about NVIDIA’s next-generation GPUs.
Share Tweet Submit