Nvidia's New Titan Xp Graphics Card Unlocks The GP102's Full Potential

[DeathBladeRSX]

Is t going to be the same amount of RAM as the Titan lineup or is there going to be more?

 

[Eximo]

It gets worse, if they don't have high end offerings then the developers won't use them when making games

Most developers don't target their games for ultra-high-end GPUs, they focus on on their recommended spec and that usually is many notches below top-end, usually in the GTX 1050-1070, RX460-480 range and equivalents. While AMD may not need a Titan Xp equivalent, it definitely needs one or two steps above the RX480 to cover the higher end of the mainstream.

I know what they target, just saying that the developers themselves will use the best cards available. Even if they turn around and do testing a lot of the underlying code will have been written and at least initially tested on the developer's machine. And they plan ahead. So the GPUs they develop and test with aren't necessarily the GPUs that will be mid-range when the game is released. Really more about that partnership. If Nvidia comes calling and says they will work with your developers and offer nice high end GPUs, while AMD comes along with their slightly worse cards, who would you pick to work with?

 

[_TheD0ct0r_]

If Nvidia hasnt <Mod Edit> up the naming of their cards already...

<You've been a member here long enough to know better than to use that language in these forums>

 

[cordes85]

buttery taste in mouth to think of SLI 2 way waterblocked of course. on micro atx baord with 2tb 960 pro, 4x 16gb g.skill trident Z RGB 3600 in a Nzxt Manta PROPS TO MP5MAFIA, to show how to convert and waterbuild and sli eith itx case with microatx board. Id have to change out the panels tho to tampered glass as thats the thing these days. have 2x 4tb samsung evo in raid 1 lol, you could pick amd or Intel, but being theres no x370 microatx boards its a 10 ten core intel to have to do

 

[Sam Hain]

This doesn't even tell you how much memory is on the card? That is the most basic stat you need for deep learning. UNINFORMATIVE.

take a wild guess 1GB to 12GB.... Here's a hint from the article, it's a GP102 chip and it's the brand-new TITAN XP.

 

[bit_user]

Yeah, because there just wasn't quite enough confusion which Titan X you were talking about. **pfft**

Many of us had already taken to calling the original Pascal Titan X as "Titan XP". Going back, people will think we were talking about Titan Xp.

Man, I swear the must be some troll in Nvidia's marketing dept, gleefully watching all the reviewers and fanatics trip over themselves having to explain exactly which Pascal-based Titan they're talking about.

 

[cryoburner]

Man, I swear the must be some troll in Nvidia's marketing dept, gleefully watching all the reviewers and fanatics trip over themselves having to explain exactly which Pascal-based Titan they're talking about.

Or maybe they just know that practically no one buys these $1200 cards anyway, so it doesn't really matter what they call them. : P

 

[Virtual_Singularity]

[video="https://www.youtube.com/watch?v=LwG7tvSCyrk"][/video]

Public Service Message: Don't be like Fry. Not for the $1200 Titan Xp, or the $800+ 1080ti. From the gimped Kep series, to Max & Pascal, many line up to feed into NV's high end hype, even after some of their AIB partners were caught passing cherry picked gpus along to influencial reviewers (to name but one scandal among others). GK110->GM207-> GM204->GM->200->GP104->GP102... Will we see another Paxwell rebrand before Volta, I wonder? I bet the Magic Answer Ball would say "Signs point to yes". Say no to the milking scheme! Break the Cycle! lol

 

[Myrmidonas]

It's a Quadro P6000 with half the RAM and little lower clocks.

 

[bit_user]

It's a Quadro P6000 with half the RAM and little lower clocks.

Except the P6000 supports ECC (okay, "soft-ECC") and costs about 4x more!!

 

[spentshells]

All these nvidia releases so close together feel like a cash in, I get the feeling Vega dx 12 performance is gonna be something to wait for.

 

[clonazepam]

Maybe they're anticipating a drop in sales when WW3 starts.

 

[spdragoo]

Public Service Message: Don't be like Fry. Not for the $1200 Titan Xp, or the $800+ 1080ti. From the gimped Kep series, to Max & Pascal, many line up to feed into NV's high end hype, even after some of their AIB partners were caught passing cherry picked gpus along to influencial reviewers (to name but one scandal among others). GK110->GM207-> GM204->GM->200->GP104->GP102... Will we see another Paxwell rebrand before Volta, I wonder? I bet the Magic Answer Ball would say "Signs point to yes". Say no to the milking scheme! Break the Cycle! lol

Which reminds me...does anyone else think it's weird that GPU processor ID numbers decrease in value the more powerful they get, when CPU processor ID numbers increase? For instance, with Intel you instantly know that an i5-3570 is faster/more powerful than the i5-3330, because 3570 > 3330. And you can even tell that the i5-3570 is more powerful than the i5-2500, because you not only know it's a newer-generation chip (3 > 2), but the remaining digits are also larger (570 > 500). AMD CPUs do it as well. But when it comes to GPUs, it seems like it goes in reverse. Even AMD (which previously used island names for their chips) are doing the same thing with their newer cards: 11 > 10, but the Polaris 11 is less powerful than the Polaris 10 chip (with the Polaris 12 being even less powerful).

Is it too much to ask for logic to be applied in these situations, rather than letting marketing departments run amok?

 

[Eximo]

Internal code names and model numbers aren't really the result of the marketing team. You can think of those as version numbers so the designers have a common term.

They are marketed with numbers that represent performance levels in a given generation.

 

[bloodroses]

Nvidia marketing pitch: "Lets nerf our new Titan card, but call it a Titan anyways so we can charge $1200. Then months later we can unnerf the card and charge those same people foolish enough to buy one in the first place another $1200"

Nvidia sales board: "Great idea! Let's do one better and release a ti of our normal top gaming card, make it as fast as the titan and only half the price to make that $1200 Titan seem irrelavant. That way they'll have to spend another $1200 if they want the fastest. It's not like AMD has been able to do anything about it anyways. It's a win/win for us!"

 

[renz496]

Nvidia marketing pitch: "Lets nerf our new Titan card, but call it a Titan anyways so we can charge $1200. Then months later we can unnerf the card and charge those same people foolish enough to buy one in the first place another $1200"

Nvidia sales board: "Great idea! Let's do one better and release a ti of our normal top gaming card, make it as fast as the titan and only half the price to make that $1200 Titan seem irrelavant. That way they'll have to spend another $1200 if they want the fastest. It's not like AMD has been able to do anything about it anyways. It's a win/win for us!"

nothing wrong with that. nvidia interest is to make money. also the early titan did not use full GP102 could be because of yield where nvidia want to reserved all perfect GP102 chip for quadro. it the same situation with kepler when nvidia first transition to 28nm. we got the original titan with few of it's shader being cut of and then later we got titan black that using fully enabled GK110.

 

[kyotokid]

...not very impressed. The only changes was unlocking all the GP102 processors SMs for an extra 256 CUDA cores, expanding the memory bandwidth by 63.5 Gbps, and increasing the base clock by 1.4GFlops. Still not enough to warrant spending the additional 500$ in price over the 1080 Ti. I was expecting a memory upgrade to 16 GB GDDR5X which would make have made it a true cut above it's 11 GB sibling and more worthy of the premium price tag.

Last year Nvida doubled the memory of the Quadro M6000 to 24 GB while keeping it at the same 5,000$ price, and did the same with the P5000 over the Maxwell version (actually the top of the line Quadro has been priced at 5,000$ since the 6 GB Fermi version was introduced five years ago in spite of constant improvement).

 

[bit_user]

...not very impressed. The only changes was unlocking all the GP102 processors SMs for an extra 256 CUDA cores, expanding the memory bandwidth by 63.5 Gbps, and increasing the base clock by 1.4GFlops. Still not enough to warrant spending the additional 500$ in price over the 1080 Ti.

*sigh* it's obviously not worth the price difference. The only people buying this are those who want the very fastest GPU at any cost.

I was expecting a memory upgrade to 16 GB GDDR5X which would make have made it a true cut above it's 11 GB sibling and more worthy of the premium price tag.

And how would that work? The GP102 has 12 memory channels, meaning any full-performance memory config must be a multiple of 12. If they wanted to go above 12 GB, then they'd have to jump all the way to 24 GB. But that would risk eating into the market of their P6000 and it's high-margin price tag > $5k.

Last year Nvida doubled the memory of the Quadro M6000 to 24 GB while keeping it at the same 5,000$ price,

Why are you comparing this to $5000 professional cards? Very different markets. And you're calling the Titan Xp overpriced?

 

[kyotokid]

...not very impressed. The only changes was unlocking all the GP102 processors SMs for an extra 256 CUDA cores, expanding the memory bandwidth by 63.5 Gbps, and increasing the base clock by 1.4GFlops. Still not enough to warrant spending the additional 500$ in price over the 1080 Ti.

*sigh* it's obviously not worth the price difference. The only people buying this are those who want the very fastest GPU at any cost.

...yeah, a small advantage over the 1080 Ti for nearly twice the price, makes me feel it's more bragging rights ("mine goes up to 12GB and 12 TFLOPS"). If I need more VRAM for rendering than 11 GB, I'd save up for a P5000.

I was expecting a memory upgrade to 16 GB GDDR5X which would make have made it a true cut above it's 11 GB sibling and more worthy of the premium price tag.

And how would that work? The GP102 has 12 memory channels, meaning any full-performance memory config must be a multiple of 12. If they wanted to go above 12 GB, then they'd have to jump all the way to 24 GB. But that would risk eating into the market of their P6000 and it's high-margin price tag > $5k.

...yeah, I see (Checked s the full specs). So there would need to be a GTX card based on the GP104 processor then as that has 8 memory channels. Though I thought it interesting that didn't seem to be the case when both the Maxwell Titan X and M6000 had 12 GB.

Last year Nvida doubled the memory of the Quadro M6000 to 24 GB while keeping it at the same 5,000$ price,

Why are you comparing this to $5000 professional cards? Very different markets. And you're calling the Titan Xp overpriced?

...just saying that it seems the pro sector has been getting a break with prices remaining the same for the last 5 years while us enthusiasts dealt with price increases for improved performance.

...and yes, compared to the 1080 Ti, the Titan Xp is overpriced given the slight degree of advantage it offers.

 

[Tech_TTT]

I dont like this .. Nvidia are pissing off their Titan X early buyers like this .. not cool.

 

[renz496]

...not very impressed. The only changes was unlocking all the GP102 processors SMs for an extra 256 CUDA cores, expanding the memory bandwidth by 63.5 Gbps, and increasing the base clock by 1.4GFlops. Still not enough to warrant spending the additional 500$ in price over the 1080 Ti. I was expecting a memory upgrade to 16 GB GDDR5X which would make have made it a true cut above it's 11 GB sibling and more worthy of the premium price tag.

Last year Nvida doubled the memory of the Quadro M6000 to 24 GB while keeping it at the same 5,000$ price, and did the same with the P5000 over the Maxwell version (actually the top of the line Quadro has been priced at 5,000$ since the 6 GB Fermi version was introduced five years ago in spite of constant improvement).

except this is not Quadro or tesla. titan is more like poor man tesla. but they cannot give everything to titan or else that will going to cut into their tesla or quadro line up.

 

[renz496]

I dont like this .. Nvidia are pissing off their Titan X early buyers like this .. not cool.

there will always be better product at the same price down the road. so you expect nvidia to price this even more expensive than last year titan just not to piss of those that already buy their last year titan?

 

[cryoburner]

there will always be better product at the same price down the road.

And within a few years, a better product will be available for a couple hundred dollars. : P

 

[411Please]

I hope everybody sees what NVIDIA did here and just ripped some dedicated customers off for a huge money grab by the TITAN X / 1080 TI / TITAN XP fiasco. I hope they lose a lot of future sales for for the TITAN cards seeing how they will just pull a fast one on TITAN GPU customers!

 

[Sam Hain]

I hope everybody sees what NVIDIA did here and just ripped some dedicated customers off for a huge money grab by the TITAN X / 1080 TI / TITAN XP fiasco. I hope they lose a lot of future sales for for the TITAN cards seeing how they will just pull a fast one on TITAN GPU customers!

The original Pascal Titan was a rip anyhoo at $1.2K, for those who bought just for gaming... Especially knowing the barely cut-down version in the way of the Ti was coming.

For those who bought the Pascal Titan for other than gaming apps, yeah... They got it in the chocolate-star with no butter, with the just released Titan XP.