The architecture “Ampere” that forms the basis of the Geforce RTX 3000 series may be in short supply on store shelves, but the number of family members on the desktop side seems to be growing to include the models RTX 3050, RTX 3050 Ti and RTX 3060 with 12 GB of memory soon . Rumors have also abounded about the RTX 3000 series for laptops, where we previously reported on an Acer model that pairs AMD’s upcoming “Cezanne” processors with the RTX 3080.
The mobile version of the Geforce RTX 3080 looks to be offered with 8 or 16 GB GDDR6. In addition to lacking GDDR6X memory as its desktop siblings, Videocardz reports that mobile RTX 3080 and RTX 3070 will build on the GA104 graphics circuit – the same circuit that powers the Geforce RTX 3070 and 3060 Ti desktops.
The specifications for the Geforce RTX 3070 in portable garment are also confirmed by a Geekbench result, unearthed by the Twitter profile TUM_APISAK. Here the number of calculation units confirms, Compute Units, that the circuit is provided with 5,120 CUDA cores as its 40 units possess 128 CUDA circuits each. This leads to performance levels just over 30 percent higher than its predecessor RTX 2070 Max-Q, according to Videocardz – performance that places the circuit between stationary RTX 2080 and 2080 Super in Geekbench.
Mobile Geforce RTX 3000 series
Geforce RTX 3080 | Geforce RTX 3070 | Geforce RTX 3060 | Geforce RTX 2080 Super | |
---|---|---|---|---|
Graphics circuit | GA104-775 | GA104-770 | GA106 | TU104 |
CUDA cores | 6 144 st. | 5 120 st. | 3 072 st. | 3 072 st. |
Frequency | 1,1–1,7 GHz | 1,1–1,62 GHz | 0,9–1,7 GHz | 0,74–1,59 GHz |
Memory bus | 256 bits | 256 bits | 192 bits | 256 bits |
Memory amount | 8/16 GB | 8 GB | 6 GB | 8 GB |
Minnestyp | GDDR6 | GDDR6 | GDDR6 | GDDR6 |
TGP, Max-P | 115–150 W | 115–150 W | 80–115 W | 150–200 W |
TGP, Max-Q | 80–90 W | 80–90 W | 60–70 W | 80–90 W |
* All information is still unconfirmed.
These two mobile circuits are also expected to be joined by little brother Geforce RTX 3060, which is also based on the GA106 circuit in portable format. The circuit is characterized by, among other things, a 192-bit wide memory bus, which cuts the amount of memory to 6 GB from the big brothers’ 8 or 16 GB. The number of CUDA cores here is expected to be 3,072, compared to the 3,840 cores expected in the stationary version of the card.
It is also worth mentioning that with the transition to the architecture “Ampere”, Nvidia has also adopted a new way of counting graphics cores. The “Turing” series’ 3,072 CUDA cores, which can be seen in the table above in the top card Geforce RTX 2080 Super, would be twice as many in the “Ampere” calculation example – exactly as many as in the upcoming RTX 3080.
In terms of editions, Nvidia is expected to continue with its Max-Q and Max-P editions of the circuits, which means that each circuit will be available in at least two variants. According to Nvidia, the Max-Q variants work better for thinner laptops with lower TGP values and heat generation, while Max-P aims at high-performance computers with more large cooling solutions.
SweClocker’s users, on the other hand, have repeatedly demonstrated that gaming is best done on desktops – but Nvidia’s progress in mobile gaming is nonetheless not to be despised. Performance in class with the previous generation of desktop top cards in a portable format is impressive regardless. We will hear more on January 12 when Nvidia holds court at CES.