Saw this and thought hmmmm.... I know the pricing is probably wrong.

---

Yes that was part of the leaked specs. Don't expect that pricing to be accurate either.

---

Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)

I'm wondering when (if) the Volta architecture cards will hit the mainstream, by which I mean way under the Titan V, hopefully in the $1k price range? As Turing has been the chosen one this time around in the gaming realm, and at least it appears at release, will only come with 8Gb of memory on the top end cards, though it will be GDDR6, instead of the NAND like the Titan V. One other thing I came across in one article I found while trying to figure out Who's on First in this GPU shuffle, is that it was suggested that the Ampre technology would actually be the replacement for the Volta (again, aimed at the AI and scientific markets), which would either mean it would just fade away, or maybe (hopefully?) make it down to the mainstream.

At this point, who knows? Fun time to be in the PC game, though. I was initially scratching my head on why in the world the would they toss away the GTX name, as it's universally known and has been around forever, but then it hit me. The big push with this version of the cards is now Ray Tracing, hence the RT in RTX. Haven't yet actually seen it stated exactly that way though I really haven't looked all that hard, but the synapses are still able to fire pretty reliably, so that would be my captain obvious guess at this point.

---

It's official- GTX is dead; it's now RTX, starting at 2000.
And with prices to match (the current high cost of GDDR6 memory is considered to be a significant factor in that, and the fact AMD hasn't anything to put pressure on NVidia at this point in time).
11GB of RAM for extreme high end cards (RTX 2080Ti and upwards), 8GB for RTX 2080 & RTX 1070.
The GeForce RTX 2080 Ti Founders Edition will carry a suggested price of $1,199, while the RTX 2080 Founders Edition will list for $799 and the RTX 2070 will go for $599. Partner cards will apparently start at $999 for the RTX 2080 Ti, $699 for the RTX 2080, and $499 for the RTX 2070.
NVidia is taking Pre-orders now, shipping expected from (around) Sept 20.

Techreport RTX news.
Techspot RTX news.
Tom's Hardware RTX news.

---

Grant
Darwin NT

The Volta architecture is a dead-end. The Turing architecture leap-frogged it and obsoleted it. Turns out two development teams were working in parallel and the Turing team won the battle and the war for what the future product line will be. The Volta will be a one-off product, doomed to the history books.

---

Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)

Just found some prices listed for those of us here in Australia.

RTX 1070    $899
RTX 1080   $1199
RTX 1080Ti $1899

I think it'll be a long, long time till I buy one.

---

Grant
Darwin NT

I don't think ANYONE should buy any Turing product yet. Not until the review sites actually have one in hand and can test one on the current available applications and games. And only compare its performance on the same applications and games to a Pascal card. The ray tracing performance specs are meaningless. I just don't want anyone to believe the announcement hype that Turing is 10X faster than Pascal. Those specs were only comparing Turing Ray Tracing performance to Ray Tracing performance of Pascal. Pascal doesn't HAVE any ray tracing performance.

---

Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)

I just don't want anyone to believe the announcement hype that Turing is 10X faster than Pascal. Those specs were only comparing Turing Ray Tracing performance to Ray Tracing performance of Pascal. Pascal doesn't HAVE any ray tracing performance.

The claims for gaming are up to 1.5 times faster.
Yet to see any mention of general computing performance.

Bring on the reviews.

---

Grant
Darwin NT

I just don't want anyone to believe the announcement hype that Turing is 10X faster than Pascal. Those specs were only comparing Turing Ray Tracing performance to Ray Tracing performance of Pascal. Pascal doesn't HAVE any ray tracing performance.

The claims for gaming are up to 1.5 times faster.
Yet to see any mention of general computing performance.

Bring on the reviews.

The only game examples were three unknown cherry-picked games. Nothing mainstream. It will be a while before the game developers get their hands on a development SDK and can spend half a year learning to write to and utilize the new Turing hardware. Then in a couple of game generations you will finally be able to use the Turing hardware for its intended purpose of ray tracing.

Yes, no mention of GPGPU performance.

---

Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)

Moving post to a dedicated thread as I am sure it will be a hot topic in coming weeks

Yes that was part of the leaked specs. Don't expect that pricing to be accurate either.

Looking at my email and newegg has a MSI GeForce RTX 2080 Ti for $1,199.99. ASUS GeForce RTX 2080 Ti for 1239.99.

---

Looking at my email and newegg has a MSI GeForce RTX 2080 Ti for $1,199.99. ASUS GeForce RTX 2080 Ti for 1239.99.

Availability?

---

Grant
Darwin NT

Founders Edition cards from Nvidia available September 20th. No news about AIM product availability. I figure December in time for Xmas.

---

Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)

The Volta architecture is a dead-end. The Turing architecture leap-frogged it and obsoleted it. Turns out two development teams were working in parallel and the Turing team won the battle and the war for what the future product line will be. The Volta will be a one-off product, doomed to the history books.

So, the Ampre architecture is based upon the Turing, it's the next gen of it, or is it some new 3rd way yet again? I like the idea of different teams taking different directions to achieve the best possible end results, competition is often a good thing. Would be nice though if they had at least communicated during development, so it would have just been an internal struggle, instead of releasing one product and then less than a year later another one supersedes and essentially obsoletes it.

Speaking of obsolete, I saw some ramblings of the ray tracing business as the Big Thing in Turing, but the Big Thing for the Volta I thought was it's Tensor Core technology, and I didn't see anything mentioned about it in the Turing, though I might have missed it. Is Tensor a dead end as well? I know that the Volta was aimed at the AI market, and if I understand correctly, ray tracing is mostly a back end process for improved screen visuals and performance in gaming? If so, it appears they may have taken a step backwards, at least in the scientific/AI realm. But whadda I know, I'm just a guy watching it all unfold from the upper deck, nose bleed seats. ;-)

---

No, Turing has an even bigger and improved Tensor core module. It has Tensor core module, Ray Tracing module and Cuda core module.

---

Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)

Just found some prices listed for those of us here in Australia.

RTX 1070    $899
RTX 1080   $1199
RTX 1080Ti $1899

I think it'll be a long, long time till I buy one.

If these prices stand (even the US retail ones, much less the ones you are 'privileged' to pay down in Oz), I'm with you. Unless of course it is somehow optimized in some way in BOINC where I could replace 3-4 10X0 series cards in a system -thinking 1070Ti-1080Ti level, not 1050's lol - with one of those and get the same performance, I'd _maybe_ consider it, as the electricity savings over time could end up being significant. Especially if I could recoup some of the investment by selling those cards.

---

Ahh, ok. Thanks for the info! The More You Know... ;-)

---

The Volta architecture is a dead-end. The Turing architecture leap-frogged it and obsoleted it. Turns out two development teams were working in parallel and the Turing team won the battle and the war for what the future product line will be. The Volta will be a one-off product, doomed to the history books.

TitanV is a monster FP64 card and still might be preferred for professional environments when Turing is out. NV likes to gimp FP64 in consumer products to force professionals into those type of cards.

The consumer cards announced yesterday are still cut down versions of the professional cards as usual. "Big Turing" TU 100 chip has 4608 CUDA cores while the consumer version TU 102/104 chip in the RTX 2080Ti has 4352 CUDA cores. So I expect the FP64 performance to be reduced severely in the consumer cards as usual for Nvidia. They still want to push scientific compute loads to their professional card series.

[Edit] Corrected the guessed at chip name.

---

Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)