Was it worth the risk of damaging the RTX brand with a bad first impression? "RTX On" is meme now, and a point of ridicule.
Nvidia Turing Product Reviews and Previews: (Super, TI, 2080, 2070, 2060, 1660, etc)
- Thread starter Ike Turner
- Start date
Was it worth the risk of damaging the RTX brand with a bad first impression? "RTX On" is meme now, and a point of ridicule.
In my perspective, you can't operate your billion dollar business by tying it to the boom bust cycles of crypto. He hit a land mine that he could not avoid and Pascal was just a very good architecture for price/perf/watt. Hard to follow up on that.
You just gotta take your hits and keep moving. Not delivering in some is worse. It is better to try than to not try at all. It shouts that you can only play it safe, you'll only make safe moves. That's how competitors catch up and pass you, because you're only doing what you already know. RT is unknown for everyone. But now nvidia has the most experience in that field, and they will continue to grow it. A costly method of education sure, but they've only really failed if they take away nothing from this lesson.
And I think that making the value of RTX the single selling point of your new architecture knowing that that selling point was going to not only have to justify why there was no increase in the perf/$$$ for this architecture, but even worse had to actually justify why the cards delivered less perf/$$$ was a stupid choice. Better to trojan-horse that feature in with a product that was desirable with or without it and allow it to organically accrue value as it saw more and more support. Nvidia were in a great position to wait to do this because AMD had no competition and would have none (Radeon VII doesn't happen without Nvidia opening up the $699 price point for that performance level) for the foreseeable future.
Perhaps. Unfortunately we'll never know what the best method would be. It's entirely possible even with your line of thinking it could have failed as well. Not launching a product in which you are ready to launch one seems... like your leadership lost their nerve. There are shareholder meetings where Jensun must answer questions about what the are working on, it's hard to keep saying yay Turing is coming, Turing is coming. And mission abort at the goal line. It's a tough position to be in. And as much as even if he didn't want to go through with Turing knowing that this could happen, it's possible that nvidia had to because they've committed to earlier.
Shit happens. I've seen all sorts of decisions by our ELT that came out poorly. It's just a question if you can rebound or if you can turn it out in your favour later. One thing I have noticed, companies that are afraid to f* up, most certainly don't last long - see just about every major Canadian tech company that boomed big to die out a couple years later and get crushed by the competition because they thought they could rest.
Consumers don't care. If you deliver products that they want at prices they want to pay they are happy and buy those products and if consumers buy those products that will make the shareholders happy. That is a model that cannot fail. The only way it can is if someone comes along and delivers a product that is more desirable that competes with it, but AMD are not in a position to deliver that and Intel haven't gotten into the game yet.
Silent_Buddha
Legend
Going back and thinking about this, it's interesting to compare and contrast this with the GeForce 256.
In many ways quite similar.
- Change of branding. Riva to Geforce and GTX to RTX.
- Introduction of forward looking features with little to no industry support. TnL and RT respectively.
- It launched only a bit higher in price than NVs previous flagship card the Riva TNT2 Ultra (~250 USD at launch) at 279 USD.
- It launched at a lower price than previous GPUs from their biggest competitor 3Dfx. Both the Voodoo Graphics (1997) and Voodoo 2 12 MB version (1998) launched at 299 USD.
https://www.anandtech.com/show/391/25
But unlike the RTX series of cards, if you wanted to jump in and try it out, it wasn't going to be a massive investment. While the price was high for the card, it wasn't that high relative to other PC components at the time compared to the RTX cards or even flagship GTX cards.
IE - back then it was much easier to jump on new graphics technology without software support because it wasn't such a large investment relative to the rest of the PC. These days, a high end GPU represents a relatively massive investment and the RTX pushes that even further into the absurd.
As you've been saying, if there was professional support then professionals would buy it even at exorbitant prices as long as it significantly increased their workflow.
As professionals are also ones that are the least likely to jump on something without existing support in their workflow software packages this leaves RTX in currently in a weird place.
And looking back at the comparison with the Geforce 256. If they are looking at gamers to drive adoption and force software developers to support the new features, then they really needed to price it closer to 1080/1070/1060 prices for the 2080/2070/2060. At Pascal launch prices +20-50 USD, it likely would have sold in similar quantities.
But as it stands at current RTX pricing, sales will be lackluster at best, IMO, until there's a massive library of support on PC, and even then sales for RTX are unlikely to take off. While Pascal did respectably with 1070 and 1080, it only really massively took off when 1060 launched at a more budget friendly price. 2060 being closer to launch 1070 prices than launch 1060 prices really hurts adoption there WRT gamers.
This leaves the RTX series mostly being adopted by game developers, software researchers, and the 1% of ultra high end gamers. A fairly narrow market segment.
And while 2060 pushed prices into more reasonable territory, at lot of people in that price range likely have a 1080 or better already due to the sales that Pascal has had and all the GPUs being dumped into the 2nd hand market. Which means taking a performance hit in order to try out RT.
Regards,
SB
What most of you probably don't realize about Turing is that it was never meant to be a Pascal-like success with the general audience in the first place. It was always about bringing a hardware-enabled RT card to developers. It's all about getting devs to work on RT and probably get some feedback that they can incorporate into their next generation, which will be the "true" successor to Pascal.
Turing basically is a beta test. That way when nVidia's first 7nm generation comes out, which might happen as early as late 2019 or early 2020, there will be wide software support from day one and nVidia is gathering a ton of experience. Even if Navi will have hardware RT support, games will already be optimized for RTX and AMD will have to play catch-up. I think it is a brilliant strategy, I think.
Of course nVidia was not prepared to take a loss with Turing by producing a card that is only relevant to devs, so they added a modest performance increase, not too much (don't want to taint that future 7nm generation), but big enough that Turing sells to enthusiasts so nVidia earns the development costs back and still makes a nice profit.
They could not, however, have foreseen the burst of the crypto bubble when they came up with Turing, so sales may be lower than anticipated. Even if the stock market frowns now, it's not such a big deal. Turing was a necessary step in order to stay ahead technologically.
Turing without hardware RT but better performance may have resulted in better sales now, but strategically it would have been a loss if AMD would have been first to market with RT (no matter if "compute" RT or dedicated hardware).
Turing basically is a beta test. That way when nVidia's first 7nm generation comes out, which might happen as early as late 2019 or early 2020, there will be wide software support from day one and nVidia is gathering a ton of experience. Even if Navi will have hardware RT support, games will already be optimized for RTX and AMD will have to play catch-up. I think it is a brilliant strategy, I think.
Of course nVidia was not prepared to take a loss with Turing by producing a card that is only relevant to devs, so they added a modest performance increase, not too much (don't want to taint that future 7nm generation), but big enough that Turing sells to enthusiasts so nVidia earns the development costs back and still makes a nice profit.
They could not, however, have foreseen the burst of the crypto bubble when they came up with Turing, so sales may be lower than anticipated. Even if the stock market frowns now, it's not such a big deal. Turing was a necessary step in order to stay ahead technologically.
Turing without hardware RT but better performance may have resulted in better sales now, but strategically it would have been a loss if AMD would have been first to market with RT (no matter if "compute" RT or dedicated hardware).
nVidia is working with everybody on integration. It just takes time. It is a whole new world for those companies. ChaosGroup demonstrated their real time project at Siggraph 2018:
No. It is more that nVidia should not have invented better hardware because their doesnt exist any software for it. That never happens on the PC. Software always comes after hardware.
If that was the plan, they should have had conservative estimates and not have had to downgrade expectations.
That is, either nVidia thought Turing would sell loads and made loads, only for it to sell moderately - why did they think it would sell loads without the software to drive it?
...or...nvidia thought Turing would sell conservatively but it hasn't even managed that - why hasn't it even sold conservatively? Because the value isn't there.
I agree no company is going to get it right 100% of the time, but now we can see the response to RTX's releases, I think fair analysis shows either it could have been done better (IMO), or releasing a raytracing card was always going to cost nVidia billions and...they did it anyway?? Released alongside pro imaging software, the share price probably would have risen on account of nVidia having a monopoly on professional imaging.
If so, and these billions in wiped stock value are considered a good investment, and nVidia takes lessons learnt from RTX while AMD fails to get anything out, then it is indeed a brilliant strategy
Being first doesn't win you anything. Google wasn't the first search engine. Facebook wasn't the first social network. Boeing weren't eh first commercial jet. There are many examples. If AMD was first to market and no software implemented their raytracing, their cards would also be undesirable. If nVidia came second with a far more economical/performant solution, they'd 'win'.
I'd say the Nvidia's sales expectations (those which were missed by $0.5 bln) indicate otherwise.
While true on PC-side, didn't several console cabinets use TnL-chips before GeForce?
nVidia's stock was massively overrated because of the crypto bubble (and the KI bubble), so the stock would have taken a hit anyway. It will be much more important for nVidia earnings to still to be #1 in the GPU market in 5 years than to avoid a bad quarter in volatile times.
I think Google is a perfect example here. Google could only overtake Altavista because Altavista had become complacent and Google beat them with superior technology. That's exactly the mistake nVidia is trying to avoid here. They are investing in new technology instead of going for an incremental update of current technology.
Of course being first to market does not always guarantee that you win, we all know that. But, if done right, it allows you to hit the ground running and that's what happening here. Even if Navi comes out with RT hardware (or great software-based RT) almost everybody will have optimized their engines for RTX already. We have seen in the past that this is a really great advantage because engine developer resources are scarce. Devs may not have the resources to optimize for nVidia and AMD RT properly. Plus it sets the bar higher for Intel and their upcoming discreet GPU solution.
Why is anyone going to optimise their raytracing for RTX when the market for it is so small? Surely all devs are going to wait until there's a sizable RT market out there before writing RT engines? At which point they'll use DXR and let the drivers handle it, give or take.
I'm not sure if you have read my whole post. When they planed Turing they could not foresee the crypto bubble bursting (at the point that it did). Without that happening they would sure have sold a lot more.
Yeah, because that really works so well right now. Just write generic DX11 und DX12 code and be done with it. Performance will obviously be great! Optimizing for nVidia or AMD? Let the driver handle it!
PS: Might want to ask @sebbi how that works out for him
Not only on pc, on consoles at the start of the generation there often aint that much titles out untill the mid/end of the generation that really use the hardware.
Investments are long for the long term, MS lost what, 7 billions on xbox og/360? Their getting their foot in the market. Nvidia is a very big company.
They know that normal rasterization has its limits, and want to be the first with new tech to get further.
I dont get why ppl are upset about Nvidias new RTX line, if you dont like it dont buy it get a pascal gpu? Theres many 10XX series left that are quite cheap, in special second hand. And buy a 3000 series or whatever name it gets when its out with more developed software. Just like with new consoles, or perhaps any new devices launch, having a huge library of games is mostly never the case. Are there ppl here that arent happy with their RTX gpu investment?
I personally dont have one cause i dont really see the need yet, its abit pricey too, but i could get a RT gpu in a year or two. Im not having to be upset that way and wait till the software has evolved more, also from the community side. But i do like to see tech is moving forward, struggling or not its a beginning. RT is doing quite nice allready for what is. Turing has more then just RT too that accelerate graphics.
While i dont think Navi will have RT (arcturus or whatever does?), if it does i hope it gets optimized for both, and Intel if they have some RT ready gpu's ready at launch.
Nvidia does take a hit now with RTX, which is only a good thing imo. I feel Nvidia are abit arrogant, this will enforce them to see things over on the software side atleast.
The games are coded for DXR not RTX. So yes, the drivers will play a role here for sure.Yeah, because that really works so well right now. Just write generic DX11 und DX12 code and be done with it. Performance will obviously be great! Optimizing for nVidia or AMD? Let the driver handle it!
PS: Might want to ask @sebbi how that works out for him
Can’t run BFV RT without the Shader Model 6_3 or 6_4 which comes in a windows release with DXR
Also, the list of games atleast aint that bad considering? 11 RT games and 27 DLSS. Or am i missing something here?
https://www.forbes.com/sites/marcoc...mes-that-will-support-nvidias-rtx-technology/
https://www.kitguru.net/components/...rt-nvidias-ray-tracing-and-dlss-rtx-features/
https://www.forbes.com/sites/marcoc...mes-that-will-support-nvidias-rtx-technology/
https://www.kitguru.net/components/...rt-nvidias-ray-tracing-and-dlss-rtx-features/
Who's upset?
How many games are released with these features? How many people are going ot buy an expensive GPU now with a view to using its RT features in the future, rather than just wait until the games they want to play are out and then get an RT GPU?
This reminds me of many console announcements, especially from new players like Ouya, that list a huge number of games coming for the console. They then launch with a handful and pundits wait until the rest actually get a release, many of which may get canned when the devs realise the market just isn't large enough to be worth the investment. Is it realistic to release a product and expect people to buy it on a promise rather then on a tangible gain when the product is actually bought?