Seems likely it would be something along the lines of UE4 TAAU given how much better it looks than other 1440p games.
-
Ever wanted an RSS feed of all your favorite gaming news sites? Go check out our new Gaming Headlines feed! Read more about it here.
-
We have made minor adjustments to how the search bar works on ResetEra. You can read about the changes here.
Rumored specs for NVIDIA's upcoming "Ampere" RTX 3000 series graphics cards (RTX 3060, RTX 3070, RTX 3080 and RTX 3090)
- Thread starter Deleted member 3812
- Start date
- Rumor
- Status
You really think it looks like TAAU? I think it looks rather 1440p ish, to the point where we were counting screens of still footage and not moving ones (which is harder to do with reconstruction).
They are trying to achieve that Cinematic Feel ™
IIRC it's just a form of TSSAA, it'd get rid of jaggies well but it doesn't really make the image look more crisp than what a standard 1440P/TAA image would be like. Also their TAA has a lot of very obvious ghosting.
Me too I still enjoy my 280x but it lacks freesync and I hope it will survive until the end of the year for ampere.
No its not. If a one x is around 1060 which is like a 970, then a ps4 pro cant be near a 970.
I wouldn't even call those rumors. Someone put together some obviously fake slides last week.
On average a 980Ti is much faster. The GPU in the X1X is more or less a 580 with a bit more bandwidth. There might be games where a 580 can match a 980Ti but these are exceptions.
Correct but the 970 is not that much slower than a 980. The Pro GPU is a good bit below a 470 in raw power, which in turn is already below a 970.
I got it from Igor's Lab. (EX-tomshardware germany).
Bit unknown outside of Germany, but nice and deep enough in the hardware circles.
Last edited:
Supposedly comes from NV's expectation of Navi 21 landing above 3080 forcing them to introduce a 3080Ti/3090 card as well.
I don't see why it can't tbh. It is doubtful that AMD will take the performance crown but they can surely make a GPU which will be capable of competing in the high end.
Oh so it was a disingenuous question, just like I thought.
Turing was overpriced from the start. Everyone knew it. And now, it's getting usurped by Ampere cards that are going to knock them out of the park. As well as have to be price competitive with AMD cards and the new consoles. If you don't think that almost anyone whom spent $699+ on the first cycle of RTX isn't going to have a little bit of buyer's remorse come September then I don't know what to tell you. Other cycles like the 1080 / Ti didn't elicit that feeling, because those cards held much better long-term value. They also didn't cost as much. Whereas here on this very board we have several people looking to replace their RTX 2000 cards already. Yikes.
The new Ampere generation is shaping up to be much better than Turing, and similar to Pascal (1080 / Ti), in terms of long-term value and owner satisfaction. As shown by numerous performance leaks which you've very clearly been following. The second generation of raytracing is hittings its stride plus more games are finally incorporating it. It's all dependent on the price of course. But the gains are there, and this cycle is shaping up to be a much better time to invest for anyone who cares about "the best time to buy" an expensive piece of PC hardware to last them several years.
I have a good buddy whom I did a build for last March 2019. He bought an EVGA 2080 SC Ultra for $840 from Microcenter at the time. We built him a nice rig for around $2200. $3000 with his Alienware ultrawide. Guess how good that gpu investment is looking today in June 2020 just fifteen months later?
That's fine if you're happy buying an expensive (Turing) card and replacing it every two years. But most consumers see these types of purchasing decisions much differently. When most people buy an expensive graphics card they don't even want to THINK about having to replace it for at least the next 4 years. In this last (Turing) case, spending $800+ didn't even buy owners any extra longevity.
There's definitely a subset of people known as the "just buy it." demographic who update annually for single digit gains. They don't really care about pricing or value across models. If the obscenely priced RTX2080Ti was able to outperform the old GTX1080Ti then that was good enough for them. The same kind of people who jumped from the 9900K to the 10900K the second it hit the market.
I guess you could argue that the 1000 series was abnormally good value and that the more mediocre 2000 series was a return to form, but at the end of the day it still makes the 2000 series relatively worse than the 1000 series. The GTX1070 was a GTX980Ti with more VRAM whereas the RTX2070 was a weaker GTX1080Ti with less VRAM.
I don't understand.
They were always going to have a big card that was going to sit well above 3080, they've always done that for years now.
Unless you are saying that without big Navi sitting above 3080, Nvidia wouldn't have made a Ti card this gen and would've just had a large gap between 3080 and Titan. But then rumours point to the 3090 being the titan equivalent in the first place and nothing above that.
Gamers Nexus also approves Igor as a very good source for leaks and they cite him quite often.
I think Microsoft itself compared the One X to a 980.
If this is to be trusted then NV will be forced to release 3090 with 3080 as they would want to keep the power crown.
Works for me, I want the highest tier day 1, no more staggered bullshit releases.
The X1X is but the Pro is a lot worse than a 1060. It doesn't help that the Pro often underperforms because of the lack of bandwidth. It's why some games show a bigger gap than what you'd expect based on Tflop difference when it comes to Pro vs X1X.
The Pro is almost literally an underclocked 470 with 36 CUs instead of 32.
Of course there are. But those people are very few and far between, and far more prevalent on a tech related thread/forum. I speak in generalizations, and one only needs to check a hardware survey to see the type of upgrade cycle that the majority of people are on.
The people that buy new cards every cycle will continue to do so,and are neither here nor there really in regards to a "best time to purchase", "best value", or "investment longevity" discussion. They are the outliers. They also buy much more frequently, are more profitable customers, and those are the folks that Nvidia is trying to target with the initial wave of (3080+) Ampere releases. But the vast majority of folks aren't upgrading their expensive core pc components every 2 yrs.
I would actually love to do a survey of RTX owners and see the collective answers for the number of raytracing games they played between Sept 2018 - Sept 2020, the estimated hours they spent playing RT games, and for them to rate their satisfaction of the enhanced experience.
The last two times I witnessed people discussing this, they kept RT off in games like Battlefield in order to get better fps.
Last edited:
Did you see also see the discussion about control rtx? Or metro rtx? I played control to completion and the visual layer that rts adds is very significant to the overall experience, same for metro exodus
It looks a lot cleaner to me than a standard 1440p->4k upscale. Went back and watched the earlier DF vids on DLSS 1.0 and there is definitely a difference.
So it's a "common knowledge" argument then.
Just like every generation of GPUs being "knocked out of the park" by the next generation of GPUs. Gotcha.
Why RTX specifically? Was there no cards at $699 previously, ever?
Yeah, I'm going on a limb here and saying that EVERYONE WILL BE FINE with that. Two years is a very long time in PC upgrades. Turing wasn't "short lived" at all. And it won't suddenly turn into trash with the launch of either Ampere or RDNA2. In fact, it will still be pretty much the only h/w currently available capable of running next gen engines with next gen features.
But I digress. You've said and I quote:
This means that you can provide us with examples of GPU h/w investments which were better than Turing during the last two years. So please do this.
For what years? They haven't been able to beat either 1080 or 2080. This means that since 2016 AMD hasn't been able to compete above x70 performance tier. If RDNA2 will be able to this will be a huge change for the market for the first time in five years.
It's possible that NV would've went with a 3080+Titan approach, as they did pretty much always since 600 series.
But it's even more possible that they simply wouldn't have hold onto the "3080Ti/3090" card and would launch it straight away with 3080 and all. Them waiting for Navi 21 means that they are planning on adjusting the pricing on this card according to Navi 21's price and performance. Hence why they seem to be expecting Navi 21 to be faster than 3080 (or 3070? we don't know much about the lineup yet).
I was talking about nvidia.
They've always had a card over the x80 series, be it a titan or a Ti for a long time now.
But yea launching the Ti equivalent at once would be something nvidia would do if big navi sits above 3080...and I think that's a good thing.
I don't know how reliable this "leak" is, but this is supposedly a "leak" of the RTX 3080 Ti spec sheet.
These specs makes little sense as NV has never ever promoted their cards in such way so...
That being said the numbers there may end up being close to truth.
If true, is this the jump that you guys have been hoping for? I'm not super well versed in specs and shit, so I don't really know.
5376 cores at 2200MHz would be over 23TF. Don't think that's all that doesn't add up.
And do Nvidia ever publish the TF numbers in their specs?
400-A1 ?
So we have binned chips again?
I dont remember exactly but I think Igor said something about them getting rid of that and now doing something different in that regard.
@9:15
He says that eg 3080 and 3080Ti (or w.e. its called) use the same chip but with different memory interface and VRAM.
Means that A and non-A chips are replaced by just naming them differently. A chip would be a 3080 Ti and a non-A chip would be 3080. The biggest card would then be the 3090.
Naming the biggest model 3090 and having a 3080Ti that isnt the biggest anymore for a lower price would make sense if they want to sell it.
The 2080Ti is a synonym for the strongest Gaming GPU (Titan excluded). With ampere they can make another xx80 Ti and many would associate it with the 2080 Ti.
...There was a time when we didn't?..
A1 is the silicon revision. You're thinking about the XXXA/XXX split which Turing had. This was basically done with with the launch of 20 Super series.
But again, all chips are always "binned". The fact that NV decided to mark some bin type on TU chips themselves doesn't mean that GPUs weren't binned previously or won't be binned in the future.
Did you happen to look into the used marked for a 1080 Ti before buying that? I know it was slim pickings after 2018 though. The 2080 SC is obviously a little better with rasterization and had the added benefit of ray tracing (and now the new image sharpening and integer scaling features that are only on the RTX cards), but $840 is still crazy for what you're getting. In November 2018 I got a really good, used EVGA 1080 Ti for $500 and it's been great since. I chose to go that route since I already had a 1080 in one of my PCs and the normal 2080 was an abysmally bad upgrade for the price. The 2080 Ti to me was just a stop-gap until the real next-gen cards actually hit, which appears to be Ampere.
Even though the 2080 Ti was way overpriced, I won't hesitate to pick up the Ampere version for up to $1,200 and it should last at least five years in my main gaming PC, and a few years more when that becomes my secondary.
Do they always intentionally stagger it just to force people to buy the release cards at launch, only to go "Aha, gotcha!" a few months later? I always assumed it was more of a manufacturing issue than anything. Get the cards out that more people are likely to buy, and then start producing the high end enthusiast version. Day one flagship is what I would want too. Otherwise I'm going to gamble and use Step-Up, buying whatever the best card is at release (get one in November that is), and then hoping the next card is out within ninety days.
I have Exodus, but have been waiting to play it. Also will grab Control after I get a new card. Would rather play at 1440p with ray tracing, without it ever going below 60fps, and higher if possible.
Yeah, no. Unless you are always chasing the bleeding edge, it's really not. More like three years is when it's time to start looking for an upgrade if you were using a mid-high end card and are no longer getting what you consider acceptable performance. I have never done a two year cycle, aside from the outlier when I went right from a launch 5850 to a launch GTX 580. Otherwise, there has never been a need for me to upgrade any sooner. This was also during a period though where most of the better GPUs would eat 1080p for breakfast and were also very competent at 1440p. Since going 4K, there's been a slight paradigm shift where GPUs were all lagging behind. Look at any random upgrade survey outside of the extreme enthusiast bubble, and most will post 3-5 years, with a select few saying two.
Last edited:
It is like we are having 3070-80-80ti rebranded into 80-80ti-90
So now a 3070 is a 3080 and will let Nvidia starts around $699
I don't know if its intentional or not. But yeah, the Ti model is always months after 80 launch. Turing was the only exception I know of.
There won't be any need to upgrade from Turing after Ampere launch either. So my point stands.
Turing is the architecture that pioneered mesh shaders, sampler feedback, dxr acceleration and variable rate shading, all the DX12 Ultimate features that are important for next gen gaming, as well as boosting machine learning performance considerably. It's far from a bad overpriced architecture if you can look beyond the raw performance in current gen titles and it might last longer than previous architectures because of that advanced featureset.
The value of Turing will be clear once more games use machine learning and DX12 Ultimate features.
I'd argue the only architecture that is overpriced and bad right now is RDNA1 because it lacks the featureset and machine learning acceleration that is so important for next generation games. Thankfully it's soon to be replaced by RDNA2.
Last edited:
Then what was the point of you even bothering to comment about two years being a very long time for an upgrade? If most people don't actually upgrade until they need to, then two years is all relative. In terms of tech advancements you might have a point, but that's not what you said.
You are also in a different argument about longevity, which the 20 series lost in a big way compared to the previous gen to gen improvements. All for the RTX, early adopter tax.
- Status