We’re just 4 days away from Nvidia’s GeForce Special Event where they’re going to announce the next-gen lineup of RTX 30 Series graphics cards. A lot of rumors have been circulating already leading up to the launch, and you’re probably sick of them, especially since we’re only 4 days away from the truth.
However, I thought it would be interesting to put this article together as a sort of ‘predictions’ type so that we can all debate on the supposedly leaked specs. If they are indeed true, then how does this make you feel about next-gen? Are you excited by the enormous performance benefits? Or are you disappointed by a less-than-stellar jump in horsepower? Let’s take a look!
Note that these alleged specs are all based off of the Custom AIB non-overclocked models, which makes sense considering that the RTX 3090 Founders Edition has already been confirmed by Micron to have 12GB of video memory. Plus, all custom boards will apparently be using dual 8-pin power connectors, whereas the Founders Editions will be using the new 12-pin power connector.
|Card||GPU||Process Node||CUDA Cores||Boost Clock||Memory||Memory Clock||Bus Width||TDP||Launch Date|
|RTX 3090||GA102-300||7nm||5248||1695MHz||24GB GDDR6X||19.5 Gbps||384-bit||350W||Mid-September 2020|
|RTX 3080||GA102-200||7nm||4352||1710MHz||10GB GDDR6X||19 Gbps||320-bit||320W||Mid-September 2020|
|RTX 3070||GA104-300||?nm||TBA||TBA||8GB GDDR6||16 Gbps||256-bit||220W||End of September 2020|
According to the leaks, all 3 GPUs are set to launch in September, with the RTX 3090 and RTX 3080 both launching mid-September, whilst the RTX 3070 launches towards the end of the month. Apparently, Nvidia will be launching the 24GB model of the 3090, and the 10GB model of the 3080, with a custom 20GB version of the 3080 planned to launch later from AIB partners as well.
Apparently, all Ampere-based RTX cards will also support PCI Express 4.0, as well as connectors for HDMI 2.1 and DisplayPort 1.4a. Plus, the next-gen Ampere architecture also introduces 2nd-Gen Ray Tracing Cores and 3rd-Gen Tensor Cores.
The RTX 3090 itself will apparently feature the GA102-300 GPU, which will come with 5248 CUDA cores, 24GB of GDDR6X memory and a Bus Width of 384-bit. This should give us 936GB/s of theoretical bandwidth thanks to a memory clock of 19.5Gbps. The card will also feature the highest TGP of the lineup with a staggering 350W power draw.
Moving onto the RTX 3080, this card will be using the GA102-200 GPU die and allegedly has 4352 CUDA Cores with 10GB of GDDR6X memory and a maximum theoretical bandwidth of 760GB/s from a bus width of 320-bit and a memory clock speed of 19Gbps. This card will supposedly have 320W TGP, which is a touch lower than the 3090 but may still require a whopping 850W PSU.
Finally, the RTX 3070 is apparently estimated to have 8GB of GDDR6 memory (not the new 'X' version), with a memory speed of 16Gbps and a slightly more reasonable 220W TGP. No other specs were mentioned in the leak unfortunately.
So what do you think? Based on these specs which one will you choose? Are you planning to upgrade to Ampere? Or will you wait for the next generation afterwards? Let us know!