Rumors on the Nvidia RTX 20xx Series

[Crunchy Dragon]

Rumors of Nvidia's new GPU lineup have been surfacing a lot lately(no surprise to anyone, I'm sure), and this seems no exception.

 

This article comes from wccftech, so take it how you will:

Quote

NVIDIA Allegedly Branding Next-Gen GeForce As RTX 20 Series – Titan RTX 50% Faster Than 1080 Ti, RTX 2080 8 GB GDDR6 50% Faster Than 1080, RTX 2070 7 GB GDDR6 40% Faster Than 1070

NVIDIA has silently been working on their next-generation GeForce graphics cards and while we have seen the boards that would power the upcoming series, we still don’t have a confirmation regarding the naming convention of the new gaming lineup. However, if AdoredTV’s source is to be believed, then we are looking forward to an entirely new naming convention in which NVIDIA will be dropping their iconic “GTX” branding.

Quote

• NVIDIA Titan RTX – $3000 US (50% Faster Than 1080 Ti)
• NVIDIA GeForce RTX 2080 8GB GDDR6 – $500-$700 US (50% Faster Than 1080)
• NVIDIA GeForce RTX 2070 7GB GDDR6 – $300-$500 US (40% Faster Than 1070)
• NVIDIA GeForce GTX 2060 5GB GDDR6 – $200-$300 US (27% Faster Than 1060)
• NVIDIA GeForce GTX 2050 4GB GDDR5 – $100-$200 US (50% Faster Than 1050 Ti)

Say what you will about these prices, but I personally think $3000 US is pretty high for just 50% more speed than a 1080Ti, which isn't even a $2000 card. Dropping the VRAM on the 2070 and 2060 seems to be an interesting choice, I'm not sure why they would choose to do that though.

 

Quote

So let’s start with the top card, the NVIDIA Titan RTX. This flagship Titan graphics card will be a semi-consumer product featuring a price of $3000 US, which is the same as the current Titan V. The total memory capacity is not stated but we can expect around 24-32 GB of HBM2 on the GV102 GPU core. This card may be very similar to the Titan V CEO Edition but have a couple of changes in terms of clocks and core configuration. The performance is stated to be 15% faster compared to Titan V and 50% faster than the GTX 1080 Ti.

 

The flagship GeForce card would be the GeForce RTX 2080 and yes, this rumors points to NVIDIA using the 20 series label instead of the 11 series which we have been hearing about since a while now. It’s a cause of confusion but only the launch will tell what NVIDIA will opt to use. Regardless, the graphics card will feature the TU104 GPU, that’s Turing 104 for you. It will ship with 8 GB of GDDR6 memory, no issues there as we have seen boards with GDDR6 memory leak before and feature a price of around $500-$700 US.

Quote

The GTX 1080 was priced at $699 US (Founders Edition) on launch so that much is also expected. The performance of this part is expected to be 8% faster than the GTX 1080 Ti and 50% faster than the GTX 1080. Do note this is a preliminary estimate for the reference variant. Both the Titan RTX and GeForce RTX 2080 are said to launch in August so we will get to hear more about them at either SIGGRAPH or Gamescom 2018.

 

Next up, we have another GeForce RTX card, the RTX 2070. This card will feature the TU106 core and 7 GB of GDDR6 memory. This points to a 224-bit memory controller. It’s also odd since the GTX 1070 featured 8 GB of GDDR5 memory but well, we will see if there’s any truth to this in a few days. Moving on, the price is expected around $300-$500 US and performance is stated around 17% faster than the GTX 1080 and 40% faster than the GTX 1070. The card using a different core than the GTX 1080 also seems super fishy. The card is said to launch in September of 2018.

This is all very interesting. Again, the VRAM dropping by 1GB seems an odd choice for Nvidia.

 

Check inside the spoiler for wccf's chart on memory technology changes:

Spoiler

NVIDIA GeForce GTX 1080	GDDR5X	10.0 Gbps	256-bit	320 GB/s	2016
NVIDIA Next GPU With 256-bit bus	GDDR6	12.0 Gbps	256-bit	384 GB/s	2018
NVIDIA Next GPU With 256-bit bus	GDDR6	14.0 Gbps	256-bit	448GB/s	2018
AMD Radeon RX Vega 64	HBM2	1.9 Gbps	2048-bit	483 GB/s	2017
AMD Radeon R9 Fury X	HBM1	1.0 Gbps	4096-bit	512 GB/s	2015
NVIDIA Next GPU With 256-bit bus	GDDR6	16.0 Gbps	256-bit	512 GB/s	2018
NVIDIA Titan Xp	GDDR5X	11.4 Gbps	384-bit	547 GB/s	2017
NVIDIA Next GPU With 256-bit bus	GDDR6	18.0 GB/s	256-bit	576 GB/s	2018
NVIDIA Next GPU With 256-bit bus	GDDR6	20 GB/s	384-bit	640 GB/s	2018
NVIDIA Next GPU With 384-bit bus	GDDR6	12.0 Gbps	384-bit	576 GB/s	2018
NVIDIA Titan V	HBM2	1.7 Gbps	3072-bit	652.8 GB/s	2017
NVIDIA Next GPU With 384-bit bus	GDDR6	14.0 Gbps	384-bit	672GB/s	2018
NVIDIA Tesla P100	HBM2	1.4 Gbps	4096-bit	720 GB/s	2016
NVIDIA Next GPU With 384-bit bus	GDDR6	16.0 Gbps	384-bit	768 GB/s	2018
NVIDIA Next GPU With 384-bit bus	GDDR6	18.0 GB/s	384-bit	864 GB/s	2018
NVIDIA Tesla V100	HBM2	1.7 Gbps	4096-bit	901 GB/s	2017
NVIDIA Next GPU With 384-bit bus	GDDR6	20 GB/s	384-bit	960 GB/s	2018

Quote

There also seem to be two GeForce GTX parts in the lineup which include the GTX 2060 and GTX 2050. These cards don’t use the RTX tag and could be due to performance levels not being that sufficient to handle to handle Ray tracing features like the top-tier cards. These two cards will be shipping in November and will feature price tags of $100-$200 for the 2050 and $200-$300 for GTX 2060. The GTX 2070 would feature 5 GB of GDDR6 memory, again 1 GB less than the previous GTX 1060 which had various configurations of 6/5/3 GB GDDR5 VRAM. The card will be 27% faster than the GTX 1060 at stock.

Going by their article, the RTX lineup is designed for ray tracing games, which aren't even a thing yet and likely won't be for another year or two. I don't know how games will change in the next few years, but I think it's safe to assume that ray tracing won't be fully implemented for at least another year.

 

Source:

https://wccftech.com/nvidia-geforce-gtx-rtx-20-series-titan-rtx-geforce-rtx-2080-rumors/

 

AdoredTV video mentioned by wccftech(I have not watched the video yet):

 

[ItsMitch]

i swear i seen this posted yesterday

[AntiTrust]

hmmm does the r stand for radeon? color me confused

[Crunchy Dragon]

1 minute ago, SC2Mitch said:

i swear i seen this posted yesterday

There were a couple posts yesterday with information that came from the same or similar article.

 

1 minute ago, AntiTrust said:

hmmm does the r stand for radeon? color me confused

RTX is Nvidia's ray tracing technology: https://developer.nvidia.com/rtx

[Guest]

8 hours ago, SC2Mitch said:

i swear i seen this posted yesterday

Probably another WCCFTECH. They must make this up at least every 20 seconds.

[Shreyas1]

with the amount of articles WCCFs putting out, they might even make more money than Nvidia.

[TahoeDust]

Looks like nothing for us 1080ti guys unless they want to go full blown $3000 Titan.

[Phentos]

RTX tech is neat and all that, but I'd rather wait for the following generation where the real benefits would start to be seen.

[Morgan Everett]

10 hours ago, Tedny said:

8Gb for 4k gaming..... that very very bad 

P.S. Waiting for next next gen 

Not at all: that’s more than enough, in general.

[dfsdfgfkjsefoiqzemnd]

6 hours ago, TahoeDust said:

Looks like nothing for us 1080ti guys unless they want to go full blown $3000 Titan.

At that price I'd rather add another 1080Ti to my rig and deal with any SLI issues that might occur.

[Bananasplit_00]

If the 2050 costs the same as the 1050 or 1050TI, with about the same TDP that would be so awesome... 

[chckovsky]

Found this on the SIGGRAPH site listing showcases: https://s2018.siggraph.org/press/news-releases/siggraph-2018-spotlight-creators-pioneers-games-industry-real-time-live-showcase-future-production-live-stage/

 

Quote

The “Reflections” Ray-Tracing Demo Presented in Real Time and Captured Live Using Virtual Production Techniques

Gavin Moran, Epic Games, and Mohen Leo, ILMxLAB

Epic Games, NVIDIA, and ILMxLAB present 2018’s GDC demo, “Reflections,” set in the “Star Wars” universe. The team will also record a character performance live using virtual production/virtual reality directly into Unreal Engine Sequencer, and then play the demo with real-time ray tracing live at 24fps.

 

Doesn't look that impressive to me, really.

[Taf the Ghost]

11 minutes ago, Bananasplit_00 said:

If the 2050 costs the same as the 1050 or 1050TI, with about the same TDP that would be so awesome... 

Assuming they don't just rebrand the current 1030/1050 range, since it's manufactured at Samsung.

[John Ellmaker]

Maybe you’ll be able to slap some optane into the cards

[Amazonsucks]

22 hours ago, Crunchy Dragon said:

Rumors of Nvidia's new GPU lineup have been surfacing a lot lately(no surprise to anyone, I'm sure), and this seems no exception.

 

This article comes from wccftech, so take it how you will:

Say what you will about these prices, but I personally think $3000 US is pretty high for just 50% more speed than a 1080Ti, which isn't even a $2000 card. Dropping the VRAM on the 2070 and 2060 seems to be an interesting choice, I'm not sure why they would choose to do that though.

 

This is all very interesting. Again, the VRAM dropping by 1GB seems an odd choice for Nvidia.

 

Check inside the spoiler for wccf's chart on memory technology changes:

  Reveal hidden contents

NVIDIA GeForce GTX 1080	GDDR5X	10.0 Gbps	256-bit	320 GB/s	2016
NVIDIA Next GPU With 256-bit bus	GDDR6	12.0 Gbps	256-bit	384 GB/s	2018
NVIDIA Next GPU With 256-bit bus	GDDR6	14.0 Gbps	256-bit	448GB/s	2018
AMD Radeon RX Vega 64	HBM2	1.9 Gbps	2048-bit	483 GB/s	2017
AMD Radeon R9 Fury X	HBM1	1.0 Gbps	4096-bit	512 GB/s	2015
NVIDIA Next GPU With 256-bit bus	GDDR6	16.0 Gbps	256-bit	512 GB/s	2018
NVIDIA Titan Xp	GDDR5X	11.4 Gbps	384-bit	547 GB/s	2017
NVIDIA Next GPU With 256-bit bus	GDDR6	18.0 GB/s	256-bit	576 GB/s	2018
NVIDIA Next GPU With 256-bit bus	GDDR6	20 GB/s	384-bit	640 GB/s	2018
NVIDIA Next GPU With 384-bit bus	GDDR6	12.0 Gbps	384-bit	576 GB/s	2018
NVIDIA Titan V	HBM2	1.7 Gbps	3072-bit	652.8 GB/s	2017
NVIDIA Next GPU With 384-bit bus	GDDR6	14.0 Gbps	384-bit	672GB/s	2018
NVIDIA Tesla P100	HBM2	1.4 Gbps	4096-bit	720 GB/s	2016
NVIDIA Next GPU With 384-bit bus	GDDR6	16.0 Gbps	384-bit	768 GB/s	2018
NVIDIA Next GPU With 384-bit bus	GDDR6	18.0 GB/s	384-bit	864 GB/s	2018
NVIDIA Tesla V100	HBM2	1.7 Gbps	4096-bit	901 GB/s	2017
NVIDIA Next GPU With 384-bit bus	GDDR6	20 GB/s	384-bit	960 GB/s	2018

Going by their article, the RTX lineup is designed for ray tracing games, which aren't even a thing yet and likely won't be for another year or two. I don't know how games will change in the next few years, but I think it's safe to assume that ray tracing won't be fully implemented for at least another year.

 

Source:

https://wccftech.com/nvidia-geforce-gtx-rtx-20-series-titan-rtx-geforce-rtx-2080-rumors/

 

AdoredTV video mentioned by wccftech(I have not watched the video yet):

 

The current Titan is already $3000 and saying "50% faster than 1080ti" is a huge misnomer. The current Titan is many many times more powerful for AI workloads than a 1080ti because it has tensor cores, and many times the DP cores of the GP102.

 

The $3000 Titan V is the cheapest V100 product by far at half the price or less of the Quadro and Tesla variants.

 

 

Reducing the RAM is likely due to it being expensive and GDDR6.

[S w a t s o n]

 

Based on the event description on Nvidia's site I would expect some RTX demos, maybe a new DGX (DGX Pod?). If we're lucky we might get a post-volta roadmap or announcement.

 

Starts at 7PM EST, or like 45 mins from posting time

 

Edit: WTF it's been removed, did Nvidia troll us?

Edit2: This seems to work still, https://www.ustream.tv/NVIDIA

 

[orbitalbuzzsaw]

at this point there's been so many conflicting rumors that...

 

 - content is better suited as a status update -

 

/s

[porina]

On 8/12/2018 at 4:32 PM, Crunchy Dragon said:

Say what you will about these prices, but I personally think $3000 US is pretty high for just 50% more speed than a 1080Ti, which isn't even a $2000 card. Dropping the VRAM on the 2070 and 2060 seems to be an interesting choice, I'm not sure why they would choose to do that though.

How does the Titan V sit relative to 1080Ti? Titan RTX could be going for similar spot.

 

As for ram... maybe they improved internal compression to need less, and at the same time, help offset costs while ram pricing isn't shifting.

[orbitalbuzzsaw]

1 minute ago, porina said:

How does the Titan V sit relative to 1080Ti? Titan RTX could be going for similar spot.

 

As for ram... maybe they improved internal compression to need less, and at the same time, help offset costs while ram pricing isn't shifting.

I believe it's about 25% faster than the 1080 Ti

[corsairian]

They forgot to do the 2060 TI 6gb with 3 extra cores and a soda.

[JediFragger]

19 minutes ago, OrbitalBuzzsaw said:

I believe it's about 25% faster than the 1080 Ti

Is that stock or is that with the V overclocked? Am guessing stock as I gather they clock quite well.

[orbitalbuzzsaw]

2 minutes ago, JediFragger said:

Is that stock or is that with the V overclocked? Am guessing stock as I gather they clock quite well.

Both stock IIRC

[S w a t s o n]

Why the fuck did nvidia choose ustream, it's basically unwatchable

Edit:

https://www.facebook.com/NVIDIA/videos/10155983076983253/

This seems to play solidly

[JediFragger]

Just now, OrbitalBuzzsaw said:

Both stock IIRC

Ok thanks.

[Cedimedi]

Quadro RTX 6000 announced

 

the NVIDIA Quadro GV100 was 9k, i wonder how much this costs lol