Based on the performance estimates and the specifications, I don't see anything about the Titan RTX that warrants consideration as a "terror." It's just another graphics card and the only interesting thing is possibly the significantly increased memory. The performance improvement comes at a cost in TDP so all things equal, it looks like the rest of the RTX products in that it only performs better due to higher power consumption with the added benefit of ray-tracing that is marginally quick enough to work in something akin to real time.
24 GB of RAM at 672 GB/S bandwidth, full speed 32 bit accumulate on the tensor cores, 2 NVLink ports, and RT cores. What else can approach that for anywhere near $2500 dollars? It's not a terror for gaming, but for the right workloads it is. It's a good card for startups and researchers.
Maybe I'm just jaded, but I don't find anything about this product announcement or the graphics card itself terrifying. I can safely assert that I'll have lost more sleep over drinking late evening cups of coffee than over the release of the Titan RTX in the next few months.
The card isn't for you.. It doesn't do what you want it to do for a price that is reasonable for you, so it's understandable if you don't care much about it. But one man's trash is another man's treasure.
Yojimbo: But these specifications and this price for these people!
Me: I'm not scared of the Titan RTX.
Yojimbo: No Titan RTX for you! You don't care about it! You don't like the price!
Neither of my comments mentioned the price. I just pointed out that there isn't anything terrifying about this GPU. I don't understand why you're continuing to attempt to argue with me about it by highlighting unrelated specifications and the price -- neither of which worry me in the slightest. It still simply isn't fear-inducing regardless of what odd bits of information you try to incorporate into your retort or what you're attempting to project on me regarding my lack of fear. You're such a weirdo.
Well if I knew you were trying to make a stupid joke rather than have a meaningful discussion I would have safely ignored your post with a roll of my eyes.
It is possible to make a point and have fun at the same time. Those things are not mutually exclusive although a fair number of people seem to feel they should be. Life shouldn't be taken with a stiff upper lip. Loosen up. You don't get to live forever so you may as well get some giggles in on the way to the end.
"Tensor terror" - the tensor core performance boost is very exciting if you're in machine learning. I was planning on building a box with 4 2080 Tis, but now I'm considering switching to two of these linked by NVLink, if the benchmarks show that training speed scales mostly linearly with their tensor performance.
Sad is an odd term to use. nVidia will be making a significant profit on these so it's worth them building and selling them. I'm fairly sure they've figured out the maximum price they can charge for the additional performance.
(and lets face it in the greater scheme of things hardware is cheap, people are expensive. If you can get just 5% more productivity out of your employee spending $2500 on a graphics card then assuming they are earning more than $50k you've saved money inside of a year)
Sad that it get to the point that you pay that much for a GPU. The customer get rip-off and this is where I don't get it anymore. This card is not a gaming GPU, but half of them might just be used for this.
I get the whole idea of tensor cores, but from what I heard lately, FP32 or higher is the best method when it comes to image recognition.
It comes to the point that we don't know what we get for the money anymore.
You get incremental upgrades the same as you've always got for the top end of the market. If there's a market for it then people will pay for it. It's not like these are mandatory required purchases, no-one *needs* one of these, but it might be cost effective to get one.
Wow you're psychic just not a very good one. :) And whats sad about it? If you don't like it Don't buy it. And what makes you think every card from Nvidia has to fit your wallet?
> While the potential 15% performance improvement by no means justifies the greater-than 2x jump in cost, for the crazy rich out there, I do expect the Titan RTX to be a little better suited to gaming than the Titan V was.
Sadly, I'm not crazy-rich, but I am curious as to whether the additional RAM translates to more improvement in gaming performance because more textures can be loaded at once. Do any games take advantage of the additional RAM, or will they all be tuned to gaming-class cards?
Btw, For people who are not crazy-rich, but want to put a little money into gaming performance, a fast M.2 SSD can make a huge difference in load times. Exiting or entering caves in Skyrim VR only takes about 2-seconds with the game on my M.2 drive.
There isn't a game out there capable of doing anything useful with that much RAM; by the time there is, you'll need more resources in other areas than this card provides. It's never going to be relevant in a gaming scenario.
Why do you think Nvidia only makes cards for gamers? And why do you think they should all fit your wallet? They have cards that cost WAAAAAAAAAAAAAYYY more than this one.
For those of you wondering what is the difference between this card and the US$ 6,300 Quadro RTX 6000 (which is really what it should be more compared to than the GeForce RTX 2080 Ti) is the NVENC limit. All of the Quadro RTX's (as of this comment) support unlimited concurrent NVENC sessions, while the TITAN RTX (like all other GeForce GTX/RTX/TITANs) only support 2 NVENC sessions. Also, there's a possibility that the memory in the Quadro is ECC, though I have not yet confirmed this.
Not that I'm trying to justify the price, but people should know that this particular card is NOT meant for those who will just be gaming, nor really was the TITAN V. It's more for those who want most of the capabilities of the Quadro equivalent, but are willing to sacrifice a few application-specific features. The NVENC limit is the only reason I would personally NOT buy this card, but will instead be purchasing the US$ 2,300 Quadro RTX 5000, because my video production applications can make full use of the NVENC and 2 NVDEC chips.
My point is, in certain application scenarios, the pricing may be totally worth it. Look at it from a standpoint of the Cost of Time (Investment). Let's say a cost to render a video for an hour on a 2080 Ti was US$ 1 dollar, and to render the same on a TITAN was US$ 0.85. You'd have to render a year's worth of that video for the cost difference to break even, but AFTER that year, it's profitable.
Finally, NVIDIA won't have a TITAN RTX Ti. There's simply no reason to - at that price point, a Quadro is a much better INVESTMENT.
Don't forget that since it is technically in their consumer line, 10-bit color isn't possible in OpenGL applications so if you have 10 bit color workflows in your video production environment, this Titan is even more useless to you.
Yep, if you want 10-bit in OpenGL applications, you need quadro. DirectX applications are able to do 10-bit, which includes most games and your media player as long as you set it to use the right decoder. Makes all of their consumer cards useless for anyone doing color sensitive work. Cheaper to get an AMD card for digital media creation/editing than an Nvidia.
Ok so another RTX card and the price of $2500US and it looks like Nvidia really gimped the 2080Ti Tensor performance as well when you look at the RTX Titans tensor performance on this chart. The 2080Ti costing like $1200US should have had near the same tensor performance as the RTX Titan.
Not that I wanted to reply to my own comment but no edited functions here. I wanted to add that then maybe the 2080Ti would not have had such a performance drain on it when using RTX features. I guess only time will tell after the reviews of this way over priced RTX Titan comes out and if it just sails through RTX enabled games or should I say BF 5..:)
"I wanted to add that then maybe the 2080Ti would not have had such a performance drain on it when using RTX features"
So the feature they disabled is FP16 operations with FP32 accumulate. It's pretty much only necessary for neural network training. Neural network inference doesn't require that level of precision, which is also why Turing introduced even lower precision modes (INT8/INT4). The limited FP32 accumulate should not be holding back the performance of the GeForce cards when using RT/DLSS features.
If they allowed these to push a 10 bit display in video and photo editing apps I'd jump on it. It is so frustrating they artificially limit even their titan GPUs in this respect.
"Meanwhile NVIDIA is also chasing after content creators with this card a bit." A race that they will never win as long as 10-bit color is disabled on the Titan. Such a stupid limitation because digital media creators don't need any of the other features of quadro, literally the only feature needed is 10-bit. So I would rather buy a $300 AMD card instead of a $1,000 Nvdia to get that feature. Really shooting themselves in the foot in that market segment.
I'm going out to buy two of these RTX Titans: the first one I'll shove into my a...PC and the second one I'm gonna hang it around my neck from a thick 24 karat gold chain 😎
I was really excited to see they'd enabled nvlink for a Titan but disappointed that it only supports 2 cards. The V100 does 6. I'm planning on building a new neural net training machine next year with some grant money and I want 4 GPUs. 4 of these babies would be perfect but I'd have to do some serious thinking of having 2 pairs of nvlinked cards. I have no idea what that would do for performance.
Overpriced garbage. I've bought SLI every generation since the 580 came out. 2x gtx 580, 4x gtx 680, 3x gtx titan, 3x gtx titan Maxwell, 2x gtx titan pascal. And I bought 2 of the $2000 Asus 4K 144Hz monitors. I bought the latest 1tb ipad pro 12.9 and iPhone XS max 512gb outright. I'm not afraid to spend money on things that are worth it. But this RTX generation of cards is absolute garbage, and a complete rip off. And the RTX Titan makes that even worse.
This is by far the worst generational improvement Nvidia has had in at least 9 years. Yet it is the greatest generational price increase in at least 9 years. 2 years wait after the last generation came out, for a whopping 25-30% FPS increase. A much touted ray tracing feature that's only has 1/4th of the power it needs to be useful. A degradation in VRAM for gamer grade cards (Titan XP was still a gamer card, albeit a high price one, due to the price similarity with the RTX 2080ti price), going from 12GB to 11GB. And a ridiculously overpriced 24GB VRAM card that is no longer a gamer card at all.
Just absolute trash. And this is coming from a former hardcore Nvidia Fanboy. Ever since their stock went up from the AI and Mining rush, they've become 10x more greedy than normal, and care far more about gouging customers than about advancing gaming performance.
Here's hoping AMD forces them to be a bit more competitive next year, along with a node shrink. Competition is the greatest thing for consumers. Heck, due to competition between ISPs, a 150mbps plan that was $100/month 2 years ago has now been replaced with a 600mbps plan for $45 because new competition showed up in town.
I'm just incredibly disappointed. Not just by Nvidia for being so greedy, but disappointed for the limited advancement in gaming. While Apple comes out with 50% GPU performance increase every years, Nvidia gives us a price increase and just 25-30% increase after 2 years.
Youre mainly talking about high end. But even the mid range cards are completely useless this generation. Why would I buy a 2070 for that money, if a 2080 is only a few bucks away? But then again, why would I buy a 2080 with only 8 GB? And why would I dish out that much money for a Ti, while its only 10-15% faster than a non-Ti?
As you said, this isnt explainable with logic, its just pure and utter greed. But enough people seem to still buy into this crap, and so they will keep doing it.
I am actually pretty angry, because I need a new GPU, since mine is becoming too slow. But there simply arent any GPUs I would buy. And I do have the money for a 2080 ti, but that price is just perverse and I would be a hypocrite if I would buy it. Maybe they release a 2080 with 12 GB, then maybe I would grab one. But also only if the price wouldnt go up much.
Oh and I would have bought a 1080 Ti, but in their greed they stopped production of it, because it would be the MUCH better choice than a 2080. Even for years to come.
Please test turing cores, as the Geforce RTX turing cores are handicapped compared to the Quadro RTX so we need to know if the Titan RTX is also disabled or if its more like the Quadros
The reason none of you are excited about this is because you don't need that much on chip memory. People running big deep learning applications do! It's for us! The hottest NLP method right now is called BERT. It was trained on a TPU with 64 GB VRAM, and the code they published doesn't scale to multiple GPUs. So batch size has to go down when you run it on a GPU, and this has an affect on model performance. Honestly, with 2 of these linked over nvlink I could treat it as a single 48 GB gpu and nearly run BERT fine-tuning training as it was meant to be. I just wish it came with more links supported. Great looking card, but disappointed about only 2 cards being linkable.
This card seems to be strategically positioned to give a middle finger to all Nvidia's customers. Extremely expensive for gamers, lacking native 10 bit color support for professionals, and lacking some deep learning functions.
Nvidia made extra sure they are wringing every last cent out of customers this time around.
If this card follows suit with the Titan X, then it will have the same media application optimizations as a Quadro. NVidia announced that at SIGGRAPH 2017.
Ryan you are the first person ANYWHERE to report on the FP16 w/FP32 performance of 2080Ti. It is a clear deciding factor for machine learning people, yet NOWHERE reported! amazing. And it is good to see the Titan RTX is more like the Quadro than the Geforce! Thanks!
We’ve updated our terms. By continuing to use the site and/or by logging into your account, you agree to the Site’s updated Terms of Use and Privacy Policy.
Facebook
Twitter
RSS
83 Comments
Back to Article
jabbadap - Monday, December 3, 2018 - link
So OC and better cooled Quadro RTX 6000. I'm actually very surprised of 24GB Vram. Hmh are int8/int4 tensors disabled or why you don't list them?And please add that RTX to the text, Quadro 6000 was dead old Fermi Quadro.
jabbadap - Monday, December 3, 2018 - link
Ach, nevermind updated those TOPS, thanks!r3loaded - Monday, December 3, 2018 - link
Unfortunately, FP64 performance is still gimped at 1/32. Super lame.jabbadap - Monday, December 3, 2018 - link
It's has all the fp64 available what tu102 chip has. There's no Turing chip with full speed fp64 compute. If you need fp64, buy Titan V.stephenbrooks - Tuesday, December 4, 2018 - link
Yep, was going to say: with this configuration, the Titan RTX is not a replacement for the Titan V, they are side-by-side alternatives.shabby - Monday, December 3, 2018 - link
Coming next month, the Titan RTX Ti!Darcey R. Epperly - Monday, December 3, 2018 - link
The Double Precision performance of this card makes it look like a joke.MrSpadge - Monday, December 3, 2018 - link
Titan V is for nVidia FP64. For everything else, Titan RTX is superior. You better know your applications before spending that amout on a GPU.Yojimbo - Monday, December 3, 2018 - link
Same for its analog signal filtering.PeachNCream - Monday, December 3, 2018 - link
Based on the performance estimates and the specifications, I don't see anything about the Titan RTX that warrants consideration as a "terror." It's just another graphics card and the only interesting thing is possibly the significantly increased memory. The performance improvement comes at a cost in TDP so all things equal, it looks like the rest of the RTX products in that it only performs better due to higher power consumption with the added benefit of ray-tracing that is marginally quick enough to work in something akin to real time.Yojimbo - Monday, December 3, 2018 - link
24 GB of RAM at 672 GB/S bandwidth, full speed 32 bit accumulate on the tensor cores, 2 NVLink ports, and RT cores. What else can approach that for anywhere near $2500 dollars? It's not a terror for gaming, but for the right workloads it is. It's a good card for startups and researchers.eva02langley - Monday, December 3, 2018 - link
Unfortunately, there is better options on the market for scientific use. These cards might be mostly used for gaming in the end.Yojimbo - Monday, December 3, 2018 - link
Better cards like what..?PeachNCream - Monday, December 3, 2018 - link
Maybe I'm just jaded, but I don't find anything about this product announcement or the graphics card itself terrifying. I can safely assert that I'll have lost more sleep over drinking late evening cups of coffee than over the release of the Titan RTX in the next few months.Yojimbo - Monday, December 3, 2018 - link
The card isn't for you.. It doesn't do what you want it to do for a price that is reasonable for you, so it's understandable if you don't care much about it. But one man's trash is another man's treasure.PeachNCream - Tuesday, December 4, 2018 - link
Me: The Titan RTX isn't scary.Yojimbo: But these specifications and this price for these people!
Me: I'm not scared of the Titan RTX.
Yojimbo: No Titan RTX for you! You don't care about it! You don't like the price!
Neither of my comments mentioned the price. I just pointed out that there isn't anything terrifying about this GPU. I don't understand why you're continuing to attempt to argue with me about it by highlighting unrelated specifications and the price -- neither of which worry me in the slightest. It still simply isn't fear-inducing regardless of what odd bits of information you try to incorporate into your retort or what you're attempting to project on me regarding my lack of fear. You're such a weirdo.
Yojimbo - Tuesday, December 4, 2018 - link
Well if I knew you were trying to make a stupid joke rather than have a meaningful discussion I would have safely ignored your post with a roll of my eyes.PeachNCream - Tuesday, December 4, 2018 - link
It is possible to make a point and have fun at the same time. Those things are not mutually exclusive although a fair number of people seem to feel they should be. Life shouldn't be taken with a stiff upper lip. Loosen up. You don't get to live forever so you may as well get some giggles in on the way to the end.Oxford Guy - Friday, December 7, 2018 - link
"Terror" is simply the latest in lame hype.Creig - Monday, December 3, 2018 - link
I can't agree that the Titan RTX isn't a terror. Just the price tag itself is pretty scary.kordian - Monday, December 3, 2018 - link
"Tensor terror" - the tensor core performance boost is very exciting if you're in machine learning. I was planning on building a box with 4 2080 Tis, but now I'm considering switching to two of these linked by NVLink, if the benchmarks show that training speed scales mostly linearly with their tensor performance.eva02langley - Monday, December 3, 2018 - link
ROFL... 2500$, I was not far, I said up to 3000$...This is just sad.
mkaibear - Monday, December 3, 2018 - link
Sad is an odd term to use. nVidia will be making a significant profit on these so it's worth them building and selling them. I'm fairly sure they've figured out the maximum price they can charge for the additional performance.(and lets face it in the greater scheme of things hardware is cheap, people are expensive. If you can get just 5% more productivity out of your employee spending $2500 on a graphics card then assuming they are earning more than $50k you've saved money inside of a year)
eva02langley - Monday, December 3, 2018 - link
Sad that it get to the point that you pay that much for a GPU. The customer get rip-off and this is where I don't get it anymore. This card is not a gaming GPU, but half of them might just be used for this.I get the whole idea of tensor cores, but from what I heard lately, FP32 or higher is the best method when it comes to image recognition.
It comes to the point that we don't know what we get for the money anymore.
mkaibear - Monday, December 3, 2018 - link
You get incremental upgrades the same as you've always got for the top end of the market. If there's a market for it then people will pay for it. It's not like these are mandatory required purchases, no-one *needs* one of these, but it might be cost effective to get one.piiman - Sunday, December 23, 2018 - link
Wow you're psychic just not a very good one. :)piiman - Sunday, December 23, 2018 - link
Wow you're psychic just not a very good one. :) And whats sad about it? If you don't like it Don't buy it. And what makes you think every card from Nvidia has to fit your wallet?JeffFlanagan - Monday, December 3, 2018 - link
> While the potential 15% performance improvement by no means justifies the greater-than 2x jump in cost, for the crazy rich out there, I do expect the Titan RTX to be a little better suited to gaming than the Titan V was.Sadly, I'm not crazy-rich, but I am curious as to whether the additional RAM translates to more improvement in gaming performance because more textures can be loaded at once. Do any games take advantage of the additional RAM, or will they all be tuned to gaming-class cards?
Btw, For people who are not crazy-rich, but want to put a little money into gaming performance, a fast M.2 SSD can make a huge difference in load times. Exiting or entering caves in Skyrim VR only takes about 2-seconds with the game on my M.2 drive.
Spunjji - Monday, December 3, 2018 - link
There isn't a game out there capable of doing anything useful with that much RAM; by the time there is, you'll need more resources in other areas than this card provides. It's never going to be relevant in a gaming scenario.piiman - Sunday, December 23, 2018 - link
My 2080ti never gets anywhere near full but this card isn't for gaming, even though you could use it for gaming.Hxx - Monday, December 3, 2018 - link
wow As a gamer I m not sure whether to laugh or cry. I guess im gonna try both.haukionkannel - Monday, December 3, 2018 - link
Multitasking?Good ;)
philehidiot - Monday, December 3, 2018 - link
I'm using hyperthreading to simultaneously laugh and cry on one cortex.Remaining cortexes are being used to mock purchasers.
piiman - Sunday, December 23, 2018 - link
Why do you think Nvidia only makes cards for gamers? And why do you think they should all fit your wallet? They have cards that cost WAAAAAAAAAAAAAYYY more than this one.cp3o - Monday, December 3, 2018 - link
For those of you wondering what is the difference between this card and the US$ 6,300 Quadro RTX 6000 (which is really what it should be more compared to than the GeForce RTX 2080 Ti) is the NVENC limit. All of the Quadro RTX's (as of this comment) support unlimited concurrent NVENC sessions, while the TITAN RTX (like all other GeForce GTX/RTX/TITANs) only support 2 NVENC sessions. Also, there's a possibility that the memory in the Quadro is ECC, though I have not yet confirmed this.Not that I'm trying to justify the price, but people should know that this particular card is NOT meant for those who will just be gaming, nor really was the TITAN V. It's more for those who want most of the capabilities of the Quadro equivalent, but are willing to sacrifice a few application-specific features. The NVENC limit is the only reason I would personally NOT buy this card, but will instead be purchasing the US$ 2,300 Quadro RTX 5000, because my video production applications can make full use of the NVENC and 2 NVDEC chips.
My point is, in certain application scenarios, the pricing may be totally worth it. Look at it from a standpoint of the Cost of Time (Investment). Let's say a cost to render a video for an hour on a 2080 Ti was US$ 1 dollar, and to render the same on a TITAN was US$ 0.85. You'd have to render a year's worth of that video for the cost difference to break even, but AFTER that year, it's profitable.
Finally, NVIDIA won't have a TITAN RTX Ti. There's simply no reason to - at that price point, a Quadro is a much better INVESTMENT.
Freakie - Tuesday, December 4, 2018 - link
Don't forget that since it is technically in their consumer line, 10-bit color isn't possible in OpenGL applications so if you have 10 bit color workflows in your video production environment, this Titan is even more useless to you.lowlymarine - Tuesday, December 4, 2018 - link
Wait, even with HDR displays everywhere these days, nVidia is STILL doing this? FFS.Freakie - Wednesday, December 5, 2018 - link
Yep, if you want 10-bit in OpenGL applications, you need quadro. DirectX applications are able to do 10-bit, which includes most games and your media player as long as you set it to use the right decoder. Makes all of their consumer cards useless for anyone doing color sensitive work. Cheaper to get an AMD card for digital media creation/editing than an Nvidia.tamalero - Monday, December 3, 2018 - link
Is this a distraction for the baffling problems with the Founder cards?Yojimbo - Monday, December 3, 2018 - link
Truly baffling...iwod - Monday, December 3, 2018 - link
I misread the Ti being a full TU102 previously, and thought $999 would be a bargain in terms of full ~750mm Die Size.I do really want to a ~750mm2 7nm GPU. Pushing technology to its limit. Along with may be HBM3? I doubt GDDR6 could keep up in the 7nm era.
rocky12345 - Monday, December 3, 2018 - link
Ok so another RTX card and the price of $2500US and it looks like Nvidia really gimped the 2080Ti Tensor performance as well when you look at the RTX Titans tensor performance on this chart. The 2080Ti costing like $1200US should have had near the same tensor performance as the RTX Titan.rocky12345 - Monday, December 3, 2018 - link
Not that I wanted to reply to my own comment but no edited functions here. I wanted to add that then maybe the 2080Ti would not have had such a performance drain on it when using RTX features. I guess only time will tell after the reviews of this way over priced RTX Titan comes out and if it just sails through RTX enabled games or should I say BF 5..:)Ryan Smith - Monday, December 3, 2018 - link
"I wanted to add that then maybe the 2080Ti would not have had such a performance drain on it when using RTX features"So the feature they disabled is FP16 operations with FP32 accumulate. It's pretty much only necessary for neural network training. Neural network inference doesn't require that level of precision, which is also why Turing introduced even lower precision modes (INT8/INT4). The limited FP32 accumulate should not be holding back the performance of the GeForce cards when using RT/DLSS features.
CoryS - Monday, December 3, 2018 - link
If they allowed these to push a 10 bit display in video and photo editing apps I'd jump on it. It is so frustrating they artificially limit even their titan GPUs in this respect.Freakie - Tuesday, December 4, 2018 - link
"Meanwhile NVIDIA is also chasing after content creators with this card a bit." A race that they will never win as long as 10-bit color is disabled on the Titan. Such a stupid limitation because digital media creators don't need any of the other features of quadro, literally the only feature needed is 10-bit. So I would rather buy a $300 AMD card instead of a $1,000 Nvdia to get that feature. Really shooting themselves in the foot in that market segment.Gastec - Monday, December 3, 2018 - link
I'm going out to buy two of these RTX Titans: the first one I'll shove into my a...PC and the second one I'm gonna hang it around my neck from a thick 24 karat gold chain 😎Lord of the Bored - Tuesday, December 4, 2018 - link
At last, someone with a good use for one of these things!Gmn17 - Monday, December 3, 2018 - link
Too bad FP64 is gimped otherwise I’d buy 1Gmn17 - Tuesday, December 4, 2018 - link
Will Nvidia now drop the prices on Titan V and Titan Xp?Ryan Smith - Tuesday, December 4, 2018 - link
Likely not.alpha754293 - Tuesday, December 4, 2018 - link
"Strictly speaking the Quadro 6000 should be superior here (if only due to drivers & support), however it’s also a good deal more expensive."Why would a Fermi card be superior than the Turing card?
(You know that there IS an actual, separate, and much older Quadro 6000 model, right?)
(Source: https://en.wikipedia.org/wiki/List_of_Nvidia_graph...
DeepLearner - Tuesday, December 4, 2018 - link
I was really excited to see they'd enabled nvlink for a Titan but disappointed that it only supports 2 cards. The V100 does 6. I'm planning on building a new neural net training machine next year with some grant money and I want 4 GPUs. 4 of these babies would be perfect but I'd have to do some serious thinking of having 2 pairs of nvlinked cards. I have no idea what that would do for performance.Socius - Wednesday, December 5, 2018 - link
Overpriced garbage. I've bought SLI every generation since the 580 came out. 2x gtx 580, 4x gtx 680, 3x gtx titan, 3x gtx titan Maxwell, 2x gtx titan pascal. And I bought 2 of the $2000 Asus 4K 144Hz monitors. I bought the latest 1tb ipad pro 12.9 and iPhone XS max 512gb outright. I'm not afraid to spend money on things that are worth it. But this RTX generation of cards is absolute garbage, and a complete rip off. And the RTX Titan makes that even worse.This is by far the worst generational improvement Nvidia has had in at least 9 years. Yet it is the greatest generational price increase in at least 9 years. 2 years wait after the last generation came out, for a whopping 25-30% FPS increase. A much touted ray tracing feature that's only has 1/4th of the power it needs to be useful. A degradation in VRAM for gamer grade cards (Titan XP was still a gamer card, albeit a high price one, due to the price similarity with the RTX 2080ti price), going from 12GB to 11GB. And a ridiculously overpriced 24GB VRAM card that is no longer a gamer card at all.
Just absolute trash. And this is coming from a former hardcore Nvidia Fanboy. Ever since their stock went up from the AI and Mining rush, they've become 10x more greedy than normal, and care far more about gouging customers than about advancing gaming performance.
Here's hoping AMD forces them to be a bit more competitive next year, along with a node shrink. Competition is the greatest thing for consumers. Heck, due to competition between ISPs, a 150mbps plan that was $100/month 2 years ago has now been replaced with a 600mbps plan for $45 because new competition showed up in town.
I'm just incredibly disappointed. Not just by Nvidia for being so greedy, but disappointed for the limited advancement in gaming. While Apple comes out with 50% GPU performance increase every years, Nvidia gives us a price increase and just 25-30% increase after 2 years.
cbns - Thursday, December 6, 2018 - link
couldn't have said it better myself.if anyone from nvidia is reading this, take my big middle finger you nasty, greedy, monopoly monsters
Beaver M. - Friday, December 7, 2018 - link
Youre mainly talking about high end. But even the mid range cards are completely useless this generation.Why would I buy a 2070 for that money, if a 2080 is only a few bucks away? But then again, why would I buy a 2080 with only 8 GB? And why would I dish out that much money for a Ti, while its only 10-15% faster than a non-Ti?
As you said, this isnt explainable with logic, its just pure and utter greed. But enough people seem to still buy into this crap, and so they will keep doing it.
I am actually pretty angry, because I need a new GPU, since mine is becoming too slow. But there simply arent any GPUs I would buy. And I do have the money for a 2080 ti, but that price is just perverse and I would be a hypocrite if I would buy it.
Maybe they release a 2080 with 12 GB, then maybe I would grab one. But also only if the price wouldnt go up much.
Beaver M. - Friday, December 7, 2018 - link
Oh and I would have bought a 1080 Ti, but in their greed they stopped production of it, because it would be the MUCH better choice than a 2080. Even for years to come.Luke212 - Monday, December 10, 2018 - link
Please test turing cores, as the Geforce RTX turing cores are handicapped compared to the Quadro RTX so we need to know if the Titan RTX is also disabled or if its more like the QuadrosDeepLearner - Monday, December 10, 2018 - link
The reason none of you are excited about this is because you don't need that much on chip memory. People running big deep learning applications do! It's for us! The hottest NLP method right now is called BERT. It was trained on a TPU with 64 GB VRAM, and the code they published doesn't scale to multiple GPUs. So batch size has to go down when you run it on a GPU, and this has an affect on model performance. Honestly, with 2 of these linked over nvlink I could treat it as a single 48 GB gpu and nearly run BERT fine-tuning training as it was meant to be. I just wish it came with more links supported. Great looking card, but disappointed about only 2 cards being linkable.evernessince - Wednesday, December 12, 2018 - link
Only problem with that is certain features designed for deep learning are disabled on this card. So yeah...DeepLearner - Wednesday, December 12, 2018 - link
Yeah, if you need double precision save your money lol. I don't for any of my current work, thankfullypeevee - Tuesday, December 11, 2018 - link
There is something wrong with the tensor performance number for RTX2080Ti in your table. +5.9% more Tensor cores could not get +128% of performance.peevee - Tuesday, December 11, 2018 - link
Ah, got it, they artificially limited FP32 accumulation on 2080. Probably even in the drivers...evernessince - Wednesday, December 12, 2018 - link
This card seems to be strategically positioned to give a middle finger to all Nvidia's customers. Extremely expensive for gamers, lacking native 10 bit color support for professionals, and lacking some deep learning functions.Nvidia made extra sure they are wringing every last cent out of customers this time around.
Draven31 - Friday, December 14, 2018 - link
If this card follows suit with the Titan X, then it will have the same media application optimizations as a Quadro. NVidia announced that at SIGGRAPH 2017.Luke212 - Sunday, December 23, 2018 - link
Ryan you are the first person ANYWHERE to report on the FP16 w/FP32 performance of 2080Ti. It is a clear deciding factor for machine learning people, yet NOWHERE reported! amazing. And it is good to see the Titan RTX is more like the Quadro than the Geforce! Thanks!BuzBuz - Thursday, January 3, 2019 - link
Why not Two 2080 Ti's in SLI for the same price as a Titan?