PC Channel AdoredTV leaks information on new Nvidia 2000 series
AdoredTV, a YouTuber who does reviews and analyses on PC hardware and news, has allegedly obtained legitimate Nvidia documents detailing the new and not yet released Turing architecture. Although he initially expressed skepticism when he received an email from someone who claimed to be an employee at Nvidia that had information on the GeForce 2000 series (AdoredTV had recently discounted a rumor solely on the basis of there being a 2080Ti involved, believing that the 1100 series would be next), after reading through the documents he changed his mind and said he "really started to believe in this leak."
First, AdoredTV started off by revealing the document went by 2000 series naming, so the successor of the 1080 is apparently the 2080 with 8 GB of GDDR6 memory. Curiously, this was only a small snippet of the document, evident from the low resolution and the fact that the full product name was cut off; it merely read "TX 2080." Why AdoredTV did this becomes somewhat clearer when he reveals the new top of the line GPU will be called the "RTX 2080." In addition, he revealed several other GeForce 2000 series GPUs that will apparently be launching under the Turing architecture.
Launch | Card | Codename | Performance | Memory | Price |
---|---|---|---|---|---|
August | Titan RTX | Unknown | 15% faster than Titan V, 50% faster than 1080Ti | Unknown | $3K |
August | RTX 2080 | TU104 | 8% faster than GTX 1080Ti, 50% faster than GTX 1080 | 8 GB GDDR6 | $500-$700 |
September | RTX 2070 | TU106 | 17% faster than GTX 1080, 40% faster than GTX 1070 | 7 GB GDDR6 | $300-$500 |
November | GTX 2060 | TU116 | 7% slower than GTX 1070, 27% faster than GTX 1060 | 5 GB GDDR6 | $200-$300 |
November | GTX 2050 | TU117 | 18% slower than GTX 1060, 50% faster than 1050Ti | 4 GB GDDR5 | $100-$200 |
AdoredTV speculates that the RTX tag has been introduced to tie in with Nvidia's new RTX ray tracing technology and to create prestige for the new series. He also speculates that Turing either has much higher clock speeds than Pascal or has much better design than Pascal, and thus runs faster at the same clock speed; Pascal was very similar to Maxwell but in order to be better at the same clock speed significant architectural improvements would have had to be made. He discounts the possibility of these GPUs being much larger than the 1000 series because the 1080 and 2080 have a similar amount of SMs, 20 versus 23. He also said that if Turing is clocked higher, it might mean Nvidia is using TSMC's 7nm process (as 16nm might not have the headroom for higher clock speeds), and that we could see a shortage of the 2000 series at launch similar to the 1000 series.
"It's either got to be a more advanced architecture or alternatively it is higher clock speeds on 7nm, say around 2.5 GHz. It's got to be one of those two. Or this leak is just garbage. Choose one."
AdoredTV's leaked information is very interesting. According to his source, Nvidia will be leveraging their new ray tracing technology, perhaps only for high end cards, and could have made either an architecture advancement or a process advancement. If true, then AMD's current high end Vega 56 and 64 will be even further behind Nvidia's already large lead. His source also said Turing would be revealed at SIGGRAPH this year. Or, like he said, this leak could be "just garbage." This video seemingly totally contradicts a previous leak that stated Nvidia would be coming out with the 1100 series and not the 2000 series. Ultimately, it's impossible to verify this rumor (and others) are either true or false. Two groups are saying 1100 and the other is saying 2000. One of them is almost certainly wrong, and we might find out at SIGGRAPH later this year.