Is NVIDIA working on a Geforce RTX 2080 Ti Super after all?

Few weeks ago: 2080 Ti Super is definitely not coming.
Now: 2080 Ti Super is probably coming.

I mean it was obvious, but some 2080Ti owners are still living in denial.

 

its not coming, the source is a driver inf that had a temporary localised string for a presently unreleased tesla part.

 

I have no more organs to sell sorry.

 

https://www.techpowerup.com/258404/nvidia-ceo-says-buying-a-gpu-without-ray-tracing-is-crazy

*trollface*

 

Looks into crystal ball:

AMD releases 5900X that matches 2080 Super performance, nVidia drops price of 2080 Super by $50-100, drops 2080Ti to $800, inserts 2080 Ti Super (full T102 with 12GB) into former $1000 price point.

 

So getting near $700 performance for $500 price tag isn't a significant performance jump or an better value or a change in price?

Your mind works in mysterious ways.

 

It's more like getting 500$ performance for 500$ instead of 700$. Honestly the RTX non Super cards were overpriced. There's absolutely no denying that imo even if you are a fan of nVidia. The main thing the Super cards do is bring the performance more in line with the price. The premium asked for ray tracing did not make any sense whatsoever at launch.

 

You can have an opinion of where you think the price to performance should be all you want, but doesn't change what reality is. GTX 1080 ti is a $700 card, RTX 2080 is a $700 card, both performed fairly similarly, but RTX 2080 typically won out in the end, and had ray tracing. Now RTX 2070 Super provides similar performance to both of those $700 cards, and has ray tracing, for $500. To say that's not a performance jump, or better value, or change in price, is ludicrous, as the word does not revolve around peoples "I feel it should have been this price", it revolves around the reality of official released prices.

 

My opinion is based purely on the price asked by nVidia for their previous generations gpu + normal inflation. nVidia asking a premium for ray tracing is not an opinion imo. I highly doubt they are giving away the dedicated RT hardware and the r&d that went into it?

 

not aiming specifically at you but everyone should see this it says pretty much what you can gather from various sources, that buying the most expensive AIB card is a waste of (even more) money
that nvidia gpus now run at their max or near max speed making AIB not very useful especially since....nvidia keep the best gpus for themselves
the best videocard unless you plan to ln2 it seems to be the ones closest to the nvidia cards (not like I have a choice they aren't sold in my country, from nvidia it would be an import so I'd have to pay full price+taxes I can buy a lightning for that price)

p.s.
my msi 1080ti gaming x always ran 1920Mhz stock, and recently I pushed all the msi afterburner limits to the max it does 1974Mhz now, manually overclocking isn't stable, the card was sold as a 1683Mhz...

 

You do not get it at all. They are made because there is market for them. And if people pay that for only marginally stronger GPU (even if that's due to better VRMs and cooling), it sends message that there is market for higher MSRP on stock nVidia/AMD cards.

 

Sure, you're absolutely right about it. Then again, if they drop 2080TI Super prices, they will most likely lower those of next year's card as well. Hopefully!

 

What about improved acoustics? That's a factor as well for some people (me for example as I prefer a silent system).

 

Lol came to obliterate ttnuagmada's ridiculously BS fanboy nonsense of a
"Going by the 5700 XT's TDP/performance, AMD would have to break 300w to match the 2080 Ti. I don't really think Nvidia is too worried"
post, with some good ol' hard facts & data that explicitly show that 7nm RDNA is either slightly MORE or less power efficient per fps than Turing depending on how far each is pushed (ala RX 5700 is slightly more efficient/fps than the RTX 2060 SUPER, whereas the RX 5700 XT is about as or slightly less efficient/fps than the RTX 2070 SUPER). Only to then find out that he'd already deleted his post after Fox2232 beat me to the punch.

Well, I already copy-pasta'd the link, so here a bit of said data is anyways. -
https://www.pcworld.com/article/3406840/amd-radeon-rx-5700-and-5700-xt-review.html

Long story short, AMD would have NO issue matching the 2080 Ti with similar power consumption figures using 7nm RDNA, and could pretty easily beat TU102 in efficiency, outright if they decided to go the HBM2/2E route.

 

I get it higher MSRP market was for titan cards and they added 2080ti also for better milking customers dat is all.

 

BS, HBM is the cause of Vega's bad power efficiency.

 

Honestly, why aren't they just doing it? We all know many people would wait for it, probably switch off Nvidia too. I want them to simply release such GPUs... but they are not. Why? I want to have a choice once I switch on to 4K graphics, but that does seem to still be an upgrade to come "next year or sometime later".

 

Tapeout costs something. ROI for mere tapeout means hefty premium on those cards as AMD does not expect high sales. And that would in return mean it would not be as competitive in terms of performance per $.

So, surely, GPU or entire card can match or even outmatch 2080 Ti. But considering that nVidia sells 3 times as many GPUs as AMD, they simply have hard time recuperating investment of low volume product.

That's why AMD is very conservative on rate at which they release new products based on new silicon.

If nVidia had just 25% of market, you would see them dealing with those issues too. Which will be interesting for intel. They seem to have intention to bring GPUs from entry to enthusiast grade. But if they do not have good sales, their GPU division will have very bad financial results.
And if intel succeeds, it will reduce competitiveness of AMD/nVidia depending on which market share intel eats.
- - - -
But above is just small part of entire market. Same silicon can be used for servers, maybe consoles. Depending on google's intention to keep game/upgrade streaming business, AMD may have required demand for bigger chips.

 

I too hope that AMD's efforts in chiplet designs pays off in the future as to make "bigger" chips more easy to come by, and circumvent the traditional issues with mGPU setups for instance. Like, they're investing in "infrastructure" of "Infinity Fabric" or what they can do. Will have to read / watch what Ms. Su has given in her speach at Hot Chip.

 

Yes, chiplets are one way to get around multiple tapeouts for different performance class cards. That can make for entire product stack at cost of one successful tapeout.