THANK YOU BASED NVIDIA

What is the purpose of the Titan SKU? Like why would someone buy a Titan over the 2080ti? The price difference is massive yet the performance isn't too far apart?

>Add tensor cores to xx80ti
>double the price
>Add 5% more cores to make it a Titan
>double the price again
The more you buy...

If this thing can hold its boost clock it should be about 20% faster than a 2080ti. That's a lot better than the gap between the 1080ti and the titan xp. Pretty extreme performance. Poor value though.

TITAN V is for FP64 compute performance, 7TFLOPs, I expect it to be discontinued anyway because Nvidia will be announcing new 7nm GPU microarchitecture at GTC 2019 in March

TITAN RTX 24GB GDDR6 is for FP16 with FP32 accumulation, INT8 and INT4 Tensor cores and Real Time Ray Tracing with RT cores, it's bad for FP64 because it's only like 500GFLOPS in performance

I wonder why Volta never got a release outside of the Titan V? They should of pushed Turing off another year so RTX could of had a better launch.

...the more money you save!

Why does it on paper look like a step down from the Titan V? Why does Nvidia compete with themselves?

Turing is Volta, only with enhanced Tensor cores that supports INT8 and INT4 ops and RT cores for ray tracing and removal of FP64 cores

double precision

I think the Titans have some left over Quadro shit in them

Volta isn't a huge arch change from pascal other than the tensor cores. They're pretty much using the intel tick tock model. Turing 7nm is the product to look out for if you're interested in big performance gains instead of features.
Volta doesn't have concurrent integer execution or most of the new frontend features. It's mostly fattened pascal with tensor cores.

What the FUCK
Guys... I just bought a 2080TI last month...

What the fuck do I do now?

Wrong, Turing does have concurrent integer execution like Volta, don't talk shit if you didn't read the Turing whitepaper

$2500 for a 24GB card is a pretty good deal actually. To get that much ram in a Pascal card you would need to buy the Quadro P6000 at about twice the price.
This card is not for gaymes, so the opinions of gaymen are meaningless.

I guess I could skip rent twice.

Titan RTX / 2080 TI
Streaming Multiprocessors (RT Cores) - 72 / 68
Cuda Cores - 4,408 / 4,352
Tensor Cores - 576 / 544
Base Clock - 1,350 / 1,350
Boost Clock - 1,770 / 1,635
Memory - 24 GB / 11 GB
Memory Bus - 384 bit / 352 bit
Memory Bandwidth - 672 GB/sec / 616 GB/sec
L2 Cache - 6,144 K / 5,500 K

Continue using it as normal, unless you really need 24GB of ram.

So this is cheaper and stronger than Titan V? I sure would be salty if I blew 3k$ on it

>24gb
>$2500
If only for the memory alone is this GPU worth it to basically every entry-level scientist that heavily uses GPU for large models. Kind of a shame that Titan V owners get shafted though.

But how can I run games in 8K on my new Samsung TV without this?

Attached: 20181203_162447.jpg (2160x2160, 1.1M)

it's between a quadro 5000 and a 6000
full rt and tensor cores, but cut down cuda
6k
>CUDA Parallel-Processing Cores 4,608
>NVIDIA Tensor Cores 576
>NVIDIA RT Cores 72
>GPU Memory 24 GB GDDR6
(incindejtally, it also costs 6k, 6.3k to be more precise)
5k
>CUDA Parallel-Processing Cores 3,072
>NVIDIA Tensor Cores 384
>NVIDIA RT Cores 48
>GPU Memory 16 GB GDDR6
(2.3k listed price at their site)

I doub't they'll sell it under 3 thosuand bucks.

I wonder if 200 extra CUDA cores and drivers can justify ~double the price increase. Nvidia might cannibalize temselves on this, especially since use cases for the RTX and consumer grade are more alike, against tensor cores and voltas, and we might start seeing custom drivers.

its got more than twice the ram of the 2080ti. Absolutely they can justify the price. You need to shell out over $5k to get that kind of RAM on another card.

I meant the new titan vs the quadro 6000

nice stains, poo-in-loo

>you need well over 5 thousand burgers to get anywhere near that kind of vram

newegg.com/Product/Product.aspx?Item=9SIAE8D8A17152
Nvidiots actually believe this.

Attached: 4392e0ab68.png (1607x504, 305K)

Probably driver differences. There are openGL extensions supported on quadro only and nvlink on quadro is vastly superior to what is offered in the geforce tier. Then there is also the licensing shennanigans. You are right though this does hurt the top end quadro appeal alot.

It's already listed for $2500. What do you mean you doubt they'll sell it for under $3000?

That card us a fucking dinosaur.

Oh I really missed that.
Last line in the videocardz article.
Quite baffling, actually.

It is a lot of money, but leatherjacketman made me feel like it's a bargain. I feel conned.

which would make the Quadro 6000 ~triple the price.

holy fuck, only bill gates can afford it

>buying this gen of nvidia
retard

Yeah, it really feels like the Quadrio 6000 is awkwardly priced with basically almost an equivalent amount of cuda cores. I imagine the Titan won't get the same number of supported technologies though. I can see the NVLink bridge not allowing vRAM pooling.

This Titan card it will feature nvlink.
>TITAN RTX provides multi-precision Turing Tensor Cores for breakthrough performance from FP32, FP16, INT8 and INT4, allowing faster training and inference of neural networks. It offers twice the memory capacity of previous generation TITAN GPUs, along with NVLink to allow researchers to experiment with larger neural networks and data sets.

The Ray Tracing all the things is a new market segmentation, and this line will really be a bit lost inbetween the gaming segment and the upcoming Voltas (expecting them to still be full tensor cores).

geforce RTX has NVlink as well, however it is cut down in the driver to act like SLI, not an actual unified frame buffer like the Quadro. I'm not sure where the Titan will fit in those options of NVlink.

thinking nvidia will ever go lower now that they established those price points because amd won't deliver anything new before 2020 and you don't really believe its high end so 1,4k for a 80ti is now standard

You're looking from the bottom and upwards. Look from the top. Why would anyone part with over 6 thousand dollars for a Quadro 6000 now. Imagine buying 10 quadros vs 10 titans, that price difference alone is enough incentive to make in house drivers, plugins/addons - saving about 38k in the process.

>2080Ti is the new Titan
I remember how you were all fooled.
Every specs of 2080Ti screamed it was a cut down GPU.
Prepare for 2090 down the line.

Unless they fuck you with drivers and licenses for the drivers. If you are putting cards in a data center the driver license requires quadros. Nvidia fucking sucks.

>you don't need FP64 goy
>you need tensor cores
>ECC? No goy, why could you possibly want that on a card aimed at mission critical work?
>just buy it!

Attached: 1455687281173.png (439x290, 141K)

The impotent amdpajeet rage and bitching is hilarious. They even call for the gov to step up and break nvidia. The whole "rebellion" marketing apparently worked. Rise up fellow amdrones.

Attached: mfwtur.jpg (144x145, 10K)

3162 American Dollars in Norway.

Attached: Untitled.jpg (858x966, 189K)

That beings said, I haven't seen a 2080Ti for that price.

Usually 200USD or so more than that.

that's barely an upgrade

When you know that they are related family wise and probably they are planing a dynasty to control the pc hardware market.

Attached: Worried.png (520x678, 301K)

>Definitely not paid to shill.

Attached: (63)_Nvidia_Screwed_Everyone_Over_with_the_20-Series_-_YouTube_-_2018-12-03_01.08.34.jpg (1058x600, 65K)

Well you know not too bad. I called $10 000 but it's only $2500. I think that's actually cheaper than their previous Titans

The first TItan was aimed at workstations and actually cheap compared to Quadro cards.

Didn't later Ttitan cards have gimped double precision performance in order to not compete with themselves.

Not cut down, just not topmost bin.

Very poor design. If it's not blower then you can't put it side by side without ruining thermal performance. As such it's useless for scientific computing (it doesn't have half speed FP64 anyways) and deep learning if you're going for a 4x GPU workstation build or 8x GPU rack build

You're saying this new Titan is not from the same wafer?
I'm laughing at you.

Nvidia for pc gayming and Amd for console. Why does it look like they are not competing on purpose?

Nvidia = Good GPUs
AMD = Good CPUs/getting there now
Its as if they are trying to control their own market

Ayymd has more compute power I thought? They slaughtered nvidia in mining?

Throughput depends on the algorithm in question and the choices made for the GPU architecture. Raw compute power is a useless measurement.

Remember when the first Kepler Titan was released and everyone was outraged over the $1000 price tag?
Oh how far we've come.

Jow Forums fukken loves monopolies and being price gouged as anons use materialism to escape the reality they ARE worthless people. Normally you'd think the iphone would be a better choice than a gpu but that involves actually communicating with other meatbags and fuck that noise.

The Titan Z was the future that Nvidia wanted and look where we are now.

It's as if Intel a shit.

Technically the original Titan was far better than the cards of the time. It's just the 780 Ti was released shortly after with the same performance without as much vram. The Titan black was also released which was the full uncut die but wasn't voltage unlocked like the previous Titan.

The 780ti only existed because Nvidia didn't think hawaii could match fat kepler. To be fair it was never designed to do so which is why its clocked so far beyond its ideal voltage vs clock point precisely to get the 290x to be the fastest gpu on the planet.

To this day i'm not sure if any AIB vega card can top MSI's 390x - that thing is clocked to 100mhz out of the box and is easily running 1.3v. I swear it has a 400w bios.

Attached: money-down-the-toilet.png (400x400, 171K)

>TU102-400 GPU

wait, does that mean there's a potential TU-100?

>amdlet /v/edditor
People are buying them for machine learning and make money off it.

>tfw running an Nvidia GPU with an AyyMD CPU
I kind of feel like being in the winning side. I refuse Intel and their scamware, they are the Apple of hardware.

Attached: 1462238810888.jpg (800x604, 60K)

FUCK YES! Gonna SLI these bad boy(s).

Good deal actually.

If they're designed for machine learning they should be blower cards or passive cooling cards. You can't put the dual-fan reference design cards side by side and expect them to not throttle especially when they're 280W each.

Jelly amdbabies?

Attached: ama.png (829x464, 793K)

Titan V has half speed FP64 and Titan RTX doesn't.

>no ecc
>gaming drivers only

Useless for professionals and too expensive for goymers.

Titans are for low cost ML. Most people are content with 1. The market you're talking about uses teslas.

>Playing trash games
Lul
Bf3 was last good bf game.

I'll buy a Navi card when they come out to play csgo at 720p.

Machine learning doesn't need ECC and quadro drivers.

Holy shit reflections in a game engine, it's like I'm really in 2010

Screen space reflections are nothing new and doesn't need raytracing cores.

Meant for

When you are in the middle of a fight where you are trying to prevent some mudslime from cutting your throat you don't stop to admire how nice your surroundings are.

People who want their personal ML workstations are stuck with Titan V if they don't want to buy the horrendously overpriced DGX station.

>Driven by the new NVIDIA Turing™ architecture, TITAN RTX — dubbed T-Rex
>dubbed T-Rex
Cringe

>The x80ti cards have always been a much better deal.
Until now

>Nvidia isn't really price gouging, guys, they're just shifting the product stack, the 2080ti is just the new Titan with a different name
>I know there's no evidence of this and that the VRAM is too low for what Nvidia would go for with a Titan and that Jensen would never leave money on the table like that, but it's definitely what they're doing and you guys are being totally unfair by paying attention to history and not making up Nvidia's excuses for them
>The Founders Edition is totally cool and we can definitely take partner MSRPs at face value, Nvidia isn't planning launch day price hikes, that would only happen because of scalpers and crypto miners
>Nvidia did nothing wrong
>Fuck Radeon for their Vega launch though, can you believe a company would launch a product that costs more than its performance merits and with the knowledge that its price is going to go above MSRP after launch? Not recommended

Attached: JayzTwoTons.jpg (527x646, 38K)

Finally a card that can run Total War Warhammer at 2k 144hz

RTX are expensive because AMD is incompetent.

AMD is competent on the CPU front and Intel's response is to jack up prices even in the face of cheap CPUs.

V has come to

>The x80ti cards have always been a much better deal
Literally the only time this was true was with the 980Ti. Nowadays, buying anything above $400 is beyond retarded.

What in the everloving fuck.
Appleshitters mean nothing to me, but Nvidiafaggots make me 1000% mad.

Nvidia are literally the Apple of GPUs, but they're somehow even more malicious.