Introducing the NVIDIA TITAN RTX GPU

[NoOther]

As for the special drivers and advice/support given: why charge everyone the same ridiculous amount when everyone's needs and usage will vastly vary, even if it's amongst a huge number of customers of the exact same Quadro model number?
They (nVidia) are forcing everyone to essentially pre-purchase the most expensive level of special driver and special support, regardless if most customer won't (and most of them don't) use it to that extent. Record profits due to predatory practices.

SMH, they don't charge the same amount. That is why the entry level Quadro cards are much cheaper. But to answer the question about the predatory practices, they aren't really predatory. If you don't need the extra level of support, why even buy a Quadro card? Why not just buy a Geforce card? They aren't getting record profits from their Quadro division or their Titan cards, most of their profits come from the regular Geforce line of cards.

Now, the question on why don't they sell a separate service fee for enhanced service is a viable question. Mostly because back in the day everyone expected a certain level of service when buying professional products. There is contract language involved in many transactions requiring certain levels of service when companies or ogranizations buy professional products. So, basically both AMD and Nvidia charge a significant upcharge for their professional cards. It has been less expensive to manufacture them all the same and support them all the same rather than sell it all piecemeal.

 

[tajoh111]

Please, point out where I specifically stated that it should cost $0.

RTX 2080Ti = $1250 (no signed drivers).

RTX Titan = slightly more cores and 24GB GDDR6 for $2500 (no signed drivers), now priced well into business segment territory and being marketed as a business product, so why no signed drivers?

RTX Quadro = with same core count as Titan $6300 (signed drivers).

GF 1030 = $80 (no signed drivers)
Quadro P600 = same as 1030 $170 (signed drivers)

So, you're telling us that the same $90 signed drivers and support for the Quadro version of the 1030 should add $3800 to the Quadro equivalent of the RTX Titan, despite going through the same validation and certification process to get merely added to the driver INF, just because it has more horsepower under the HSF?

How does that even make sense?

Your underestimating the performance increase from the semi professional drivers which the gtx 1080 ti line up does not get.

https://techgage.com/article/quick-...-performance-boosting-385-12-titan-xp-driver/

After the semi professional card release of the rx vega frontier edition where AMD gave their cards professional drivers without the support, Nvidia was forced to respond and add in their own support to make the cards more competitive. As a result, it is truly a semi professional card now great for certain kind of analysis and cad along with the rendering tasks it was already good at.

https://www.pcper.com/reviews/Graph...B-Air-Cooled-Review/Professional-Testing-SPEC

Before professional drivers on Titan Xp

After professional drivers.

As a result, at 1000 to 1200 dollars, Titan Xp with partial pro support provide too much value and cannibalize Quadro sales too much at 1200 dollar pricing. The WX7100 was a 799 dollar card and the Titan Xp simply outclasses it in performance and its performance is now in line with the Quadro p6000 which is a 6000 dollar card. They now beat quadros in many tasks at much lower prices which is why we won't be getting 1200 dollar geforce titans anymore. AMD forced Nvidia to make their cards better for professionals, but as a result of this added value in terms of performance, they don't make sense as consumer cards because demand is excessive from consumers and the semi pro market at the 1200 dollar price point. As a result, the geforce titan series is getting scrapped to make way for a card that bisects the difference between consumer and full professional.

At 2500 dollars, the demand should shift more to the semi pro market who are looking for a middle ground between quadro performance at the expense of signd drivers and personal support from Nvidia. Unfortunately we will see consumers buy at this price point because they see the 2500 dollar price point as attractive because it ensures more exclusivity which is more desireable(epeen). The luxury market is based on this foundation.

 

[AceGoober]

I can think of 99 things to spend $2,500 on and a video card isn't one of them. That's too damn high.

 

[triwolf]

snip

Excellent analysis, I'd also like to add that they charge $2500 because human beings are good at deception, saying - look at the "thin veneer answer" that this card gets Professional Support, so pay up, even if you will never use it. See we can tell you some BS about something you will never use and you will pay if you want the card. We are a corp and what we say goes, as long as it gets us a profit. How correct or honest it is doesn't ever come into the equation.

 

[Ski]

I can only speak for myself, but this card was clearly intended for professionals specifically for 3d rendering. Chaosgroup, Octane, Blender, and Corona are all shifting their rendering programs to GPU and one of its biggest handicaps is its limited VRAM. It's common for me when doing projects on V-Ray to utilize 32-64GB's of ram from the CPU, but since I'm still using a gtx 780 ti it's nearly impossible to render from that GPU alone. For visualizers, the only two core factors we look for in terms of GPU's is VRAM and CUDA cores.

Secondly, the price while expensive, and believe you me I cannot stand Nvidia's arbitrary price increase on the RTX generation for a plethora of reasons, from a value standpoint this is way more inexpensive than having to shell out $6300 for a Quadro. It's the gamers who are the problem, most of the debate is purely immature and uninformed comments so naturally they have a limited understanding on ray tracing but also the differences between CUDA and RT cores. So it's no wonder this debate has become ugly and vitriolic from all sides. I will say however, the price point should be more in the sub $2000 range, then it would be a more reasonable price, but I'd go a little bit further and say all the GPU's are overpriced anyways.

With that said, I highly encourage you guys read this blog from Vlado, chief architect from Chaosgroup, who explores early beta testing of RT cores in comparison to the CUDA cores and even though it's still in early beta testing, the results look extremely promising. And quite frankly I'm excited as hell but more importantly Chaosgroup is the first to show RTX's real potential in the graphics industry. At the end of the day boys, I still have mixed feelings because on one hand Nvidia is the only one pushing the envelope from a tech standpoint, but their business practices in the last year has disappointed the hell out of me because their nature is becoming more corporate and greedy, so I don't know how to feel about everything. Anyways guys, check out that performance blog from Vlado because it's really fascinating.

 

[DejaWiz]

SMH, they don't charge the same amount. That is why the entry level Quadro cards are much cheaper. But to answer the question about the predatory practices, they aren't really predatory. If you don't need the extra level of support, why even buy a Quadro card? Why not just buy a GeForce card? They aren't getting record profits from their Quadro division or their Titan cards, most of their profits come from the regular GeForce line of cards.

Now, the question on why don't they sell a separate service fee for enhanced service is a viable question. Mostly because back in the day everyone expected a certain level of service when buying professional products. There is contract language involved in many transactions requiring certain levels of service when companies or organizations buy professional products. So, basically both AMD and Nvidia charge a significant upcharge for their professional cards. It has been less expensive to manufacture them all the same and support them all the same rather than sell it all piecemeal.

Sorry, I wasn't more clear with how I worded that (the last seven words of the first sentence that you quoted was my point):
What I meant by charging the same amount of "Quadro markup" to every customer of the same Quadro model number. Meaning, all Quadro M4000 or P6000 (to name only two model numbers as examples) buyers get charged the same "Quadro" markup (over the GeForce equivalents of either model), even though each customer's usage and level of need for those signed drivers and special support/advice will vastly differ. Maybe 80% of all Quadro M4000 buyers will never need much specialized support, but they are being charged the same cost as the remaining 20% of a M4000 buyers that will use quite a bit of that specialized support. That's what I mean by predatory practice.

But I do see your point about cookie-cutter pricing, rather than piecemeal support on a case-by-case basis. I would counter that with: doesn't it essentially become piecemeal support as soon as a customer reaches out to nVidia to ask for specialized support that would meet their unique needs (differing from the usage/environment of almost every other customer), which would mean nVidia probably gives them a ticket/case number, a specialist/subject matter expert assigned to that ticket/case, and a record of all correspondence between customer and SME, until the customers needs are met whether it means a custom device driver, or working with whichever devs for a given application to rectify any issues the customer is experiencing, etc?


If I am interpreting this correctly, then the record profits are indeed coming from their business/professional segment: https://nvidianews.nvidia.com/news/nvidia-announces-financial-results-for-third-quarter-fiscal-2019

 

[NoOther]

Sorry, I wasn't more clear with how I worded that (the last seven words of the first sentence that you quoted was my point):
What I meant by charging the same amount of "Quadro markup" to every customer of the same Quadro model number. Meaning, all Quadro M4000 or P6000 (to name only two model numbers as examples) buyers get charged the same "Quadro" markup (over the GeForce equivalents of either model), even though each customer's usage and level of need for those signed drivers and special support/advice will vastly differ. Maybe 80% of all Quadro M4000 buyers will never need much specialized support, but they are being charged the same cost as the remaining 20% of a M4000 buyers that will use quite a bit of that specialized support. That's what I mean by predatory practice.

Just to be clear, everyone that buys a Geforce card is being "charged" some amount for support whether they ever use support or not. The same with just about any hardware or software product. It's kind of like taxes, everyone pays into the tax system to cover benefit programs that they may never use.

But I do see your point about cookie-cutter pricing, rather than piecemeal support on a case-by-case basis. I would counter that with: doesn't it essentially become piecemeal support as soon as a customer reaches out to nVidia to ask for specialized support that would meet their unique needs (differing from the usage/environment of almost every other customer), which would mean nVidia probably gives them a ticket/case number, a specialist/subject matter expert assigned to that ticket/case, and a record of all correspondence between customer and SME, until the customers needs are met whether it means a custom device driver, or working with whichever devs for a given application to rectify any issues the customer is experiencing, etc?

As I said above, this works much like a tax. Everyone pays into the system. The problem is neither Nvidia nor the company knows when someone will need that special support. A lot of the broader support has to do with functions for common applications like AutoCAD, SolidWorks, etc. And problems for one person may answer problems for quite a few others. So there are a lot of people that are benefiting from that enhanced support whether they personally make use of it or not. This was very much the case for me when I had to use AutoCAD and SolidWorks as well as when I was building and supporting systems for people who used them. Now in addition to that there is even more specialized support that may only benefit one person or group. But it could also be a solution for someone else down the line. For example on the behavioral simulation project I worked on, we had issues with CUDA coding and tasking a bunch of GPU across multiple systems in parallel and serial. There was a lot of information lost. Nvidia developed the NVLink to help out with those kinds of issues, but there is a lot more driver and program development that goes behind how things utilize NVLink than just the basic driver support. Many people have benefited from advancements and improvements in the NVLink, and some imrpovements only a very select few benefit from. These are just some basic examples, there are tons more.

Now if they went to a charge per support issue, then many people would scoff at that and some contracts would outright reject it. That is basically why there is the system as it is now. For instance if you look at the wording in government contracts, companies that sign the contract are automatically on the hook for warranty and support at the terms of the government at the cost they sign for. There is no real provision to not offer those services for the standard contracts.

If I am interpreting this correctly, then the record profits are indeed coming from their business/professional segment: https://nvidianews.nvidia.com/news/nvidia-announces-financial-results-for-third-quarter-fiscal-2019

I assume you are referring to this bullet: "Record revenue from Datacenter, Professional Visualization, Automotive platforms"

If so, then you are not taking away more than is there. They have record revenue from those platforms, meaning those platforms are outperforming their previous numbers. Not that those platforms are driving the lion share of profit for Nvidia.

Here is a link with some breakdowns:

https://www.fool.com/investing/2017/07/16/how-nvidia-makes-most-of-its-money.aspx

Basically the gaming specific market is about 55% of the total profit for the company. The rest of the profit is broken up into other segments. The professional and datacenter segments make up 31%. There is some gray area in that as gaming cards are also used for datacenter purposes and in professional settings. But the percentages reflect Geforce (Gaming/Datacenter) vs Telsa/Quadro (Datacenter) vs Quadro (Visualization/Professional). So basically the Datacenter market has increased quite a bit over the past year compared to previous years (now up to 21% of the profit where I think it was 11% in prior year?).

 

[DejaWiz]

Just to be clear, everyone that buys a Geforce card is being "charged" some amount for support whether they ever use support or not. The same with just about any hardware or software product. It's kind of like taxes, everyone pays into the tax system to cover benefit programs that they may never use.



As I said above, this works much like a tax. Everyone pays into the system. The problem is neither Nvidia nor the company knows when someone will need that special support. A lot of the broader support has to do with functions for common applications like AutoCAD, SolidWorks, etc. And problems for one person may answer problems for quite a few others. So there are a lot of people that are benefiting from that enhanced support whether they personally make use of it or not. This was very much the case for me when I had to use AutoCAD and SolidWorks as well as when I was building and supporting systems for people who used them. Now in addition to that there is even more specialized support that may only benefit one person or group. But it could also be a solution for someone else down the line. For example on the behavioral simulation project I worked on, we had issues with CUDA coding and tasking a bunch of GPU across multiple systems in parallel and serial. There was a lot of information lost. Nvidia developed the NVLink to help out with those kinds of issues, but there is a lot more driver and program development that goes behind how things utilize NVLink than just the basic driver support. Many people have benefited from advancements and improvements in the NVLink, and some imrpovements only a very select few benefit from. These are just some basic examples, there are tons more.

Now if they went to a charge per support issue, then many people would scoff at that and some contracts would outright reject it. That is basically why there is the system as it is now. For instance if you look at the wording in government contracts, companies that sign the contract are automatically on the hook for warranty and support at the terms of the government at the cost they sign for. There is no real provision to not offer those services for the standard contracts.



I assume you are referring to this bullet: "Record revenue from Datacenter, Professional Visualization, Automotive platforms"

If so, then you are not taking away more than is there. They have record revenue from those platforms, meaning those platforms are outperforming their previous numbers. Not that those platforms are driving the lion share of profit for Nvidia.

Here is a link with some breakdowns:

https://www.fool.com/investing/2017/07/16/how-nvidia-makes-most-of-its-money.aspx

Basically the gaming specific market is about 55% of the total profit for the company. The rest of the profit is broken up into other segments. The professional and datacenter segments make up 31%. There is some gray area in that as gaming cards are also used for datacenter purposes and in professional settings. But the percentages reflect Geforce (Gaming/Datacenter) vs Telsa/Quadro (Datacenter) vs Quadro (Visualization/Professional). So basically the Datacenter market has increased quite a bit over the past year compared to previous years (now up to 21% of the profit where I think it was 11% in prior year?).

The "nVidia tax", much like the "Apple tax"...I dig it. Well put.


PS - taxation is theft.

 

[ChosenUno]

Please, point out where I specifically stated that it should cost $0.

RTX 2080Ti = $1250 (no signed drivers).

RTX Titan = slightly more cores and 24GB GDDR6 for $2500 (no signed drivers), now priced well into business segment territory and being marketed as a business product, so why no signed drivers?

RTX Quadro = with same core count as Titan $6300 (signed drivers).

GF 1030 = $80 (no signed drivers)
Quadro P600 = same as 1030 $170 (signed drivers)

So, you're telling us that the same $90 signed drivers and support for the Quadro version of the 1030 should add $3800 to the Quadro equivalent of the RTX Titan, despite going through the same validation and certification process to get merely added to the driver INF, just because it has more horsepower under the HSF?

How does that even make sense?

RTX Titan pricing makes a lot of sense because

1. Not every potential buyer needs the level of support and service behind the pro-level support Quadros get. Maybe they're tech enthusiasts and can troubleshoot and solve problems themselves, and would like to save a pretty penny so they buy the T-Rex.
2. The markup vs the Ti is because the T-Rex is a prosumer card, as evident by its pro-level VRAM amount. It's basically a Quadro RTX 6000. I'm pretty sure the Ti had 11GB of VRAM for a market segmentation reason. If it was 12GB it could've been a decent replacement for the M6000 cards, with 12GB VRAM also.
   1. It's not an easy and simple matter to optimize down VRAM usage. The costs involved in shaving that 1GB to fit in a 2080Ti is most likely much more than the cost difference of just buying a T-Rex and enjoy the extra 12GB of VRAM compared to existing infrastructure. You can even spend some dev time to parallelize the workload and run 2 instances per T-Rex, which would be a pretty big and worthwhile performance boost.
3. T-Rex is like a cheap demo version for the Quadro RTX 6000. Buy it, dev software on it, deploy with Quadros to get the benefit of drivers and pro support. Profit!

 

[DejaWiz]

RTX Titan pricing makes a lot of sense because

1. Not every potential buyer needs the level of support and service behind the pro-level support Quadros get. Maybe they're tech enthusiasts and can troubleshoot and solve problems themselves, and would like to save a pretty penny so they buy the T-Rex.
2. The markup vs the Ti is because the T-Rex is a prosumer card, as evident by its pro-level VRAM amount. It's basically a Quadro RTX 6000. I'm pretty sure the Ti had 11GB of VRAM for a market segmentation reason. If it was 12GB it could've been a decent replacement for the M6000 cards, with 12GB VRAM also.
   1. It's not an easy and simple matter to optimize down VRAM usage. The costs involved in shaving that 1GB to fit in a 2080Ti is most likely much more than the cost difference of just buying a T-Rex and enjoy the extra 12GB of VRAM compared to existing infrastructure. You can even spend some dev time to parallelize the workload and run 2 instances per T-Rex, which would be a pretty big and worthwhile performance boost.
3. T-Rex is like a cheap demo version for the Quadro RTX 6000. Buy it, dev software on it, deploy with Quadros to get the benefit of drivers and pro support. Profit!

That's all well and good, but it doesn't change my feelings that the first Titan was born (and advertised) to be a gaming GPU with a modest price premium over the top-tier GeForce, and now nVidia is simply trying to justify seriously ridiculous price tags under the ruse of marketing these past couple iterations as "what a bargain compared to our Quadro" business products*


*but gamers should still buy them at these artificially inflated high price tags, because it bears the Titan name.

 

[Dayaks]

That's all well and good, but it doesn't change my feelings that the first Titan was born (and advertised) to be a gaming GPU with a modest price premium over the top-tier GeForce, and now nVidia is simply trying to justify seriously ridiculous price tags under the ruse of marketing these past couple iterations as "what a bargain compared to our Quadro" business products*


*but gamers should still buy them at these artificially inflated high price tags, because it bears the Titan name.

The Titan V they flat out said wasn’t for gamers so there’s that... consumers could be responsible and look at price/performance.

Not directed at Deja - This getting mad over pricing is somewhat strange to me. I’ve never been one to think, “those bastards at xyz company! This luxury item is too expensive!” Never once crossed my mind to get angry. Worst case if I didn’t have the cash I’d play on high instead of ultra and wait for 2019....

 

[DejaWiz]

The Titan V they flat out said wasn’t for gamers so there’s that... consumers could be responsible and look at price/performance.

Not directed at Deja - This getting mad over pricing is somewhat strange to me. I’ve never been one to think, “those bastards at xyz company! This luxury item is too expensive!” Never once crossed my mind to get angry. Worst case if I didn’t have the cash I’d play on high instead of ultra and wait for 2019....

I'm concerned for the fact that nVidia is now price gouging and bastardizing their Titan name to try and fool everyone into believing their gouged pricing on the RTX lineup is affordable.

 

[IdiotInCharge]

I'm concerned for the fact that nVidia is now price gouging and bastardizing their Titan name to try and fool everyone into believing their gouged pricing on the RTX lineup is affordable.

...you're full of it.

You are aware that the Titan RTX provides excellent value for the performance offered. You are aware that Nvidia is a publicly traded corporation whose purpose is to make money. I certainly hope that you're aware that companies use marketing tactics to do that, based on market research.

It's hard to see your 'concern', except as a feint to continue your unfounded criticism.

 

[iamjanco]

The original 1971 version of Straw Dogs comes to mind:

 

[Aireoth]

 

[DPI]

...you're full of it.

You are aware that the Titan RTX provides excellent value for the performance offered. You are aware that Nvidia is a publicly traded corporation whose purpose is to make money. I certainly hope that you're aware that companies use marketing tactics to do that, based on market research.

It's hard to see your 'concern', except as a feint to continue your unfounded criticism.

Next you're going to tell us Santa isn't real and that hooker didn't really love me.

The weird, narcissistic internet-anger that drags it's runny nose across these nvidia threads whenever a new flagship part comes out is definitely baffling.

I'm sitting pretty with a 1080Ti while waiting for either AMD to lean into Vulkan more heavily so I can drop Nvidia, or waiting for a more powerful Nvidia card to drop to what I paid for the 1080Ti -- maybe that'll take 6-12 months? Don't care. But to sit and get angry about the existence of more expensive cards, WTF. Try indifference.

 

[IdiotInCharge]

The weird, narcissistic internet-anger that drags it's runny nose across these nvidia threads whenever a new flagship part comes out is definitely baffling.

It is kind of wild...

And yeah, I'll be waiting with my 1080Ti. I find that I'm lucky to have it- it's the most I've ever spent on a GPU (I'm typically in the xx70 category), and the first time I've owned the top-end. One of the rare points in GPU history where the top-end was actually 'worth it'.

If I hadn't picked one up, or if my requirements changed, though...

 

[JargonGR]

I honestly don't care if something fast is coming out and it is expensive or not. I judge based on my needs and spend accordingly in order to get the most value. If at a point the new Titan proves to be a productivity gain for me (i.e. making me money) I would buy it. For gaming I am not since I am not playing that much (or any ) these days. The truth is that I don't need anything faster than my 1080Ti right now since even when I game I am playing 3-5 years old games from my backlog and until I am done with them it would go to waste even at 4K. By the time I catch up with more recent titles it will be time for new cards to be announced and will then see what my options are.

 

[zkostik]

I'm a TitanXp owner myself, so I don't have a problem splashing out for a GPU. This is far outside my price range, but for those who would have bought a Quadro but didn't need the FP64 (or the TitanV's HBM), this is an incredible bargain. It's what, a 60% discount on the equivalent Quadro? I can see a computing niche jump for joy over this announcement.

The issue with non Quadro cards are driver limitations that include 3D and HDR. I suppose if you only need it as computer card for things like machine learning it isn't a bad deal.