Rumor: NVIDIA GeForce RTX 3070 and 3080 Coming Q3 2020 + Specs

Ye ye., they did it with the last gen too, right? No.

 

And what exactly did AMD have vs the 2080ti? And before that the 1080ti?

 

I think the Ti model will command its high prices until more AAA games work well in SLI. In some cases you can almost buy 2 2080 cards for the price of a high end 2080Ti. It would be nice for Big Navi to top or even just compete but I have to agree with Solfaur.

 

Big Navi will be roughly 50% faster than the 2080 Ti, but the pricing is going to be ~$1500. Expect it to launch in October.

 

I totally trust new accounts with 1 post stating facts from the future. Hoverboard, when?

 

If anyone noticed, missing from the chart is the GA102, which is most likely the real 3080ti. The GA100 is probably the Titan.

The presumed specs of GA102 as well as the other cards (with different specs than previously posted).

GA102 - 84 SMs / 5376 CUDA cores / 12GB GDDR6 / 384-bit bus - 40% faster than RTX 2080 Ti
GA103 - 60 SMs / 3840 CUDA cores / 10GB GDDR6 / 320-bit bus - 10% faster than RTX 2080 Ti
GA104 - 48 SMs / 3072 CUDA cores / 8GB GDDR6 / 256-bit bus - 5% slower than RTX 2080 Ti

https://www.tweaktown.com/news/7215...ecs-teases-an-absolute-monster-gpu/index.html

 

more likely Tesla/Quadro for sure, and maaybe Titan as well.

 

All good until they sell out before you can get the first or second batch.
Better to keep money where it can be used.

 

There is no SLI any more, there is mGPU but it isnt the same.
With SLI you can use NVinspector and other methods to beat a lot of none SLI games into submission.
With mGPU, games must be coded to use multiple cards, there is no way force it with games that are not coded to use it.
Its a more barren world.

 

Source? Mister new user with one post.

 

I hope there's a greater increase in performance than that [GA102/103/104 (3080ti/3080/3070)], they're about 20% short of what I'd like to see in terms of what I thought could be possible/expected based on all the talk of 50% more performance at half the power (due to die shrink & new architecture).

 

And what exactly will AMD/nVidia have against Father of All Expenses (intel's huge GPU)?

Just because nVidia has biggest and most expensive GPU does not mean they did or will wipe floor with AMD's products.
Performance per $ was quite comparable or better on AMD's side in mainstream with current products.

What do you think will nVidia ask for GPU with twice the transistor count of 2080 Ti on node where transistor costs more?
And do you think such GPU will have better yields than 2080Ti on mature node?

And this assumption has exactly same problem. Bigger means = more costly.
So right questions are:
"Who will have better performing product around $250~$350?" Quite unsure.
"Who will have better performing product around $400~$500?" Likely AMD.
"Who will have better performing product around $550~$750?" Likely AMD.
"Who will have better performing product around $800~$1200?" About Equal if AMD hits TDP constraints.
"Who will have better performing product above $1250?" Likely no product from AMD will exist in this price range.

And next important question is: "How much better those next generation cards will be over current one at same price points?"

 

nVidia has already said a while back that they're no longer going to be releasing the Ti later, but up front at the start of a new lineup like they did with the 2080 Ti.

nVIdia also sells 4x more desktop GPUs than AMD does, which drives costs down quite a bit i'd imagine.

And same performance? huh? Last i checked nVidia wiped the floor with AMD's last offerings, plus they had the added selling points like raytracing, DLSS 2.0, etc.

 

I am sorry for you since you simply can't compare things of equal price and see which is better for you or other people.

What does this PC wipe floor with? Likely with nothing at 1/4 of price point that has been built by someone sensible.
Throwing more money at problem is not solution here.

It is same story with intel. Considering what AMD offers at same price points, I would not buy or advice anyone to buy intel's CPUs.
And that's ignoring all the security.
Sure, you can buy i9-9900KS or whatever you feel like. But that's not very smart use of money, nor will it get you better product overall.

So, if you can read post above yours, which gives you rather kind answer... it may help you a bit.
- - - -
Now to the bold part:
And that's what you support? Company that sells weaker GPU than 5700 XT at higher price while using cheaper manufacturing node and has much, much higher profits due to higher production volume?

I say: nVidia is business, and as long as their business model works, they will continue with this practice. You can call it premium product at premium price all you want.
But true meaning of it is different than lie people have in their minds.

Have you paid Premium for G-Sync? nVidia stated that board is special and has horsepower to have extra features in future. Did you get those extra features later? Or have you had to buy new monitor with Premium price again?
- - - -
Will you back your beliefs by action? Will you buy next nVidia's GPU strong enough that AMD will not have GPU to compete with it.
All the people who buy GPU of worse value just because they see some dream at top of the food chain.

 

It's obviously not always about price if you compare GPU sales over the past decade. Performance/dollar, performance/watt, overall performance, feature rich/poor, single game/overall game performance, technology leadership, product reliability, etc... are factors (weights) people use for a purchase decision. Assigned weights impact sales and hence overall GPU market share percentages, so it is not likely performance/dollar is a deciding factor for the majority of people.

The majority of people are willing to spend more for a premium product then people who won't. Market share percentages usually tell the story.

 

Again, I feel like you're cherry picking your data points to set up a false comparison - I wasn't in debate class but I'm sure that falls under some kind of logical fallacy. You picked G-Sync because it was an easy target but you ignored all the features he wrote in his post. Nvidia products tend to come with a lot of value-add. For example, I have friends that back in the 7XX days that bought a slower card, for more money, over the AMD competition because they wanted Shadow Play. With Turing Nvidia has a number of features, RT, DLSS, Mesh, VRS (A bunch of VR features), superior video encoder, on top of a bunch of software and all the hardware functions that go with it (may lead to more value in the future). Four years ago, if AMD had a similar advantage, AMD fans would argue "it's fine wine, future proofing".. now the argument has shifted to "well they are just using it to charge a premium" and for the 2080Ti with all those features on-top of +35% more general purpose.. they 100% are and should be, but at lower price points Nvidia is price competitive with those value-add features.

I hate to bring it up but driver stability is another point of contention - the recent blackscreen debacle wasn't pretty for AMD GPUs. It got a lot of coverage, whether artificially created or not, and a lot of people here and on reddit defended AMD, saying the entire thing was fake "I don't have any problems" "must be their power supply" "they must be overclocked" etc. Then AMD released a driver that had like 5+ blackscreen fixes in it - some of which weren't even specific to a game and said they were aware of more issues related to it. It doesn't look great and I've said it before but I think the community/brand that AMD has cultivated around itself makes these kinds of situations worse.

Regardless, I think the main difference is that Nvidia has the ability to command that price premium due to the brand they built around their value-add features and ecosystem. I've posted the video here before, I don't feel like finding it again, but there is a video of Jens (CEO of Nvidia for other people reading and not aware) where he says that even in the case where an AMD/Nvidia card hardware was identical that the Nvidia card would still command a higher price point because of the ecosystem they support around it. Obviously there is some tipping point, some threshold where that would flip - if Nvidia was selling all it's cards for 50% higher than it's competition there would definitely be a point where most consumers would go "lol no" but clearly they haven't hit that yet - at least as far as they are concerned. Everyone said the 2080Ti price would drop to it's MSRP because "no one is buying at that price" and yet nearly two years later the cheapest one I can find is $1150 (I didn't look that hard).

 

Fair enough, but original comment being quoted was "If those specs are true, it is going to wipe floor with RDNA 2 " which suggested focus was on the performance (not value) front. Yes, Turings top cards were a value disaster, no argument there. But before that, the 1080ti (which I also mentioned but you ignored) had pretty much ruled the roost for so long that the "wipe the floor" saying probably more applied and left an indelible impression in many peeps minds .

 

I think Nvidia knew they could maintain that price on the 2080ti because there was no competition in that performance segment.

 

The problem is that there´s a very big chance that the 3080Ti is going to have the same price... Unless AMD comes with something good or the financial crisis is worse than expected.