Is there an RTX Titan due soon?

[AdamK47]

I didn't buy the Titan V.

 

[mohit9206]

I think I'll stick with my 1050Ti

 

[moonbogg]

$2500. I like the price. I think it's a perfect price for conducting an observational experiment. I will watch the various forums and see how many people buy it. I've always been curious as to what people's price limits were. $1200 certainly didn't do it. Not even close. $3000 Titan V sold pretty well according to some forums users, but maybe that big three grand was too much psychologically for many people, so $2500 is a great number which has yet been untested and unexplored.
Nvidia knows anything under $2000 will fly off the shelves instantly, even before reviews come out because psychologically it's just a grand with some change added on. Crossing that 2 grand barrier is entering some deeper water. As strange as it sounds, I like this quite a lot, albeit for different reasons as usual.

So, are you usual Titan buyers upgrading to this Titan? Last time you paid $1200. Now they want $2500. Ti plebs can't get away with only $700 anymore and now they get to pay $1200 this time, but you get to pay $2500. So what's it going to be? Who's in for at least one, but preferably two of these?

 

[SPBHM]

it's pretty crazy when you consider that in a couple of years it will probably perform like a midrange card,
well, it should've been expected, Nvidia has been playing with this sort of pricing for years, with the failed Titan Z, and the Titan V more recently, and now the "ti" cards costs more than the original Titan did, so yeah... only competition could stop this.

 

[Qwertilot]

Titan V & this aren't priced as gaming cards and competition wouldn't do anything.

Go and read the AnandTech review of the changes they made. Useful for some things, utterly not for gaming.

 

[moonbogg]

Titan V & this aren't priced as gaming cards and competition wouldn't do anything.

Go and read the AnandTech review of the changes they made. Useful for some things, utterly not for gaming.

That's utterly always been stated about Titans and it's utterly never been true. They game better than any Geforce card. Nvidia knows it and so do we. They are injected into the market to take advantage of the wealthy gamers and sure, maybe some pro people also. But those deep pocketed gamers are targeted just as much as anyone else is for this card.

 

[Qwertilot]

This Titan + V have both been specified as compute cards & sold and priced as such. Rich gamers a secondary market.

Titan P, in contrast, was strongly aimed at over rich gamers.

 

[mv2devnull]

Couldn't find a table. Filtered one from wiki:

GTX Titan       2013-02  $999
GTX Titan Black 2014-02  $999
GTX Titan Z     2014-03 $2999
GTX Titan X     2015-03  $999
    Titan X     2016-08 $1200
    Titan Xp    2017-04 $1200
    Titan V     2017-12 $2999
    Titan V CEO 2018-06     -
    Titan RTX   2018-12 $2499

 

[badb0y]

This Titan + V have both been specified as compute cards & sold and priced as such. Rich gamers a secondary market.

Titan P, in contrast, was strongly aimed at over rich gamers.

Ahh yes, this explains why they sent Jayztwocents a card for marketing purposes because he does extensive GPU compute testing.

 

[Qwertilot]

Their marketing department are a marketing department 🙂 Doesn't change what this product is designed or priced to do.

There is even competition scheduled for this card from that 7nm Vega. It might even be why is cheaper than Titan V. Goodness knows.

 

[LTC8K6]

TITAN RTX NVLINK™ BRIDGE

Double the effective GPU memory capacity to 48 GB and scale performance with up to 100 GB/s in total bandwidth of data transfer utilizing the NVIDIA NVLink™ technology.

Double the GPU memory capacity? Interesting difference from other RTX cards.

 

[Headfoot]

Double the GPU memory capacity? Interesting difference from other RTX cards.

* under compute workloads, just like all their other GPUs. This isn't adding anything for SLI

Given that the specs on this are basically a Quadro but the price is actually less than the equivalent Quadro (wow), I can see a lot of research institutions buying these up just like they did when the first Titan came out

 

[EXCellR8]

https://videocardz.com/press-release/nvidia-reveals-the-titan-of-turing-titan-rtx

Oh but look how shiny!!

 

[Fir]

Waiting for them to show in stock.
$5000 for two. I hope they don't charge sales tax!

 

[alcoholbob]

Couldn't find a table. Filtered one from wiki:

GTX Titan       2013-02  $999
GTX Titan Black 2014-02  $999
GTX Titan Z     2014-03 $2999
GTX Titan X     2015-03  $999
    Titan X     2016-08 $1200
    Titan Xp    2017-04 $1200
    Titan V     2017-12 $2999
    Titan V CEO 2018-06     -
    Titan RTX   2018-12 $2499

I've seen 3 Titan V CEO Editions sold so far in the used market, one at $4k and two at $5k.

32GB HBM2 at 900GB/s, with a slight overclock can hit 1.2TB/s.

If you bought one for gaming it would be a pretty crazy gaming card, given how well Volta/Turing seems to scale with memory bandiwdth (nearly as well as with core OC).

 

[Qwertilot]

They never even wrote the drivers for gaming on Volta properly, I think? Close to Turing so it might work out all right but not really a good idea.

 

[mopardude87]

Who is buying 2 of these?

 

[moonbogg]

Who is buying 2 of these?

That feels like an unsafe question to answer, however, I'll take 4 please. JK 'cause my wife wouldn't be happy about me selling the mini-van to buy 4 GPUs.

 

[alcoholbob]

They never even wrote the drivers for gaming on Volta properly, I think? Close to Turing so it might work out all right but not really a good idea.

It's got a more brute force ability in 3d mark timespy under the same clockspeed/power limits, and is substantially faster in a few games like Sniper Elite 4.

There's a couple of games Volta underperforms at due to driving tuning, yes, for example when running Battlefield 1 it's about as fast as the 2080 Ti in the DX11 executable, but for some reason it's about 15% slower in the DX12 runtime. In general it doesn't fare worse in DX12 mode versus 2080 Ti, just in that 1 game, so that's clearly an example of a specific game driver that didn't carry over to the Titan V.

But a near doubling of memory bandwidth with the Titan V CEO Edition (with the memory overclocked) should increasing gaming performance around 25% compared to the stock Titan V, which would put it miles ahead of the 2080 Ti, even without gaming specific drivers.

 

[alcoholbob]

You have to be pretty cautious about 7nm providing an instant replacement. The big Turing chips are truly massive so even at 7nm they'll need petty big dies to go meaningfully past them.

Won't be remotely easy or cheap at this point.

It's a full die shrink, so a ~50% reduction. So a GTX 1080-sized die should be able to match the performance of 2080 Ti. So a 30 series midrange card should be able to give 2080 Ti performance at ~150W.

 

[Qwertilot]

I think some care needed there - that 7nm vega shrunk seemingly only gained 25% power efficiency and that was from 14 not 12.

Presumably you'd get more with a dedicated redesign etc but.....

 

[alcoholbob]

You know a 2080 Ti is > 90% as fast as stock at 65% power target (180W) right?

All you need is 10% more transistors and slightly slower clocks and you can easily get it under 150W.

 

[IEC]

I've seen 3 Titan V CEO Editions sold so far in the used market, one at $4k and two at $5k.

32GB HBM2 at 900GB/s, with a slight overclock can hit 1.2TB/s.

If you bought one for gaming it would be a pretty crazy gaming card, given how well Volta/Turing seems to scale with memory bandiwdth (nearly as well as with core OC).

I was curious about this, so I tried some benchmarking with Unigine Superposition on 4K optimized settings with the RTX 2080 Ti.

Despite boosting less on the core to 1890 +/- 30 with a +900 memory setting (versus 1925-1950 w/+800 memory) due to 112% power limit, my score in the benchmark went up significantly. So it appears you are indeed correct and that my gains from memory overclocking scale nearly as well (if not better).

 

[alcoholbob]

I was curious about this, so I tried some benchmarking with Unigine Superposition on 4K optimized settings with the RTX 2080 Ti.

Despite boosting less on the core to 1890 +/- 30 with a +900 memory setting (versus 1925-1950 w/+800 memory) due to 112% power limit, my score in the benchmark went up significantly. So it appears you are indeed correct and that my gains from memory overclocking scale nearly as well (if not better).

You should look at Gamernexus Titan V overclocking article, that's where found out about the huge gains from memory overclocking with the current gaming architecture.

https://www.gamersnexus.net/guides/3172-nvidia-titana-v-gpu-core-vs-hbm2-memory-overclocking