RTX Titan aka T-REX Officially Announced

Supercharged_Z06 said:
I'm going to agree with you on the "Nvidia is officially Greedy AF!!!" aspect of your post, as they have been riding the top as to the gaming GPU market with no real competition in sight for quite some time now, and as such, get to set prices to whatever they want.
Add to that they have also been riding the crypto-mining wave that blew up demand to insane levels letting the GPU market go completely bonkers when it came to GPU pricing over the last couple of years. Hopefully that wave has crested... and with AMD lining up some fairly promising parts early next year as to competition, it may rock the boat a bit concerning the mid-range gaming card market (a substantial chunk of where PC GPUs are sold), which may trigger a small slide in prices back down to at least semi-reasonable levels again.

However, your post did leave out some stuff which I feel you've overlooked. Your statement about the Titan RTX only having performance "very slightly marginally faster than the 2080 Ti" sounds a bit off because of this.

From what we know thus far, besides the massive VRAM increase, the Titan RTX is also going to have:
  • A wider memory bus (384-bit vs 352-bit), resulting in better memory throughput (672GB/sec vs 616GB/sec)
  • A higher GPU boost clock (1770MHz vs 1635MHz)
  • A better base as to power with a TDP of 280W vs 260W
  • A larger L2 Cache (6MB vs 5.5MB)
  • Better Single Precision performance (16.3 TFLOPS vs 14.2 TFLOPS)
  • Massively better Tensor Core performance (130 TFLOPS vs 57 TFLOPS) - Which should really help with DLSS and raytracing performance (when it fully materializes for gaming)
This is all going to add up. The higher TDP will also likely provide a bit more capacity as to overclocking - especially if water cooled.

Is it going to be massive? No. But, I see all of the above resulting in a solid 10% to 15% gain in overall performance over the RTX 2080Ti.

We seem to have a a slight difference of opinion as to what constitutes "very slightly marginally faster" here. I'd go so far as to drop the "very slightly marginally" bit as a double digit increase is nothing to sneeze at.

That price though, ya, it's bonkers for a gaming card. But then again, I'm getting old and have the means... I might just pick one up to play with as I don't see anything faster arriving for a good while. :D
Click to expand...

So why is the 2080ti only 57 TFLOPS now? At launch it was shown as 110TFLOPS. What is this bullshittery?

FA2D620E-E3F1-4367-9C04-E2854946B586.png


Also I’d go with a FTW 2080ti or a GALAX that can do 373 to 380W. 280W and the shitty cooler will substantially hold back a large Turing chip. Nevermind be loud. OR waterblock and shunt mod the Titan. But that’s a gaming oriented perspective.

I am more wondering what fuckery is going on with nVidia and the tensor cores.
 
parityboy said:
Again I'll ask what I asked on OCN: why isn't the Quadro RTX line being moaned about for its price, when the hardware is identical?
Click to expand...

Because you aren't paying for just the Hardware with the Quadro line of cards. You are paying for tech support and certified drivers. You are also paying for the more expensive ECC memory.
 
Dayaks said:
So why is the 2080ti only 57 TFLOPS now? At launch it was shown as 110TFLOPS. What is this bullshittery?

View attachment 126896

Also I’d go with a FTW 2080ti or a GALAX that can do 373 to 380W. 280W and the shitty cooler will substantially hold back a large Turing chip. Nevermind be loud. OR waterblock and shunt mod the Titan. But that’s a gaming oriented perspective.

I am more wondering what fuckery is going on with nVidia and the tensor cores.
Click to expand...
I don't know where he pulled that number from. The Titan has an SPU count of 8156 while the 2080 Ti has 7116, or 87.25% of the Titan. Not accounting for rounding in the marketing materials, 87.25% of 130 is about 113.
 
reaper12 said:
Because you aren't paying for just the Hardware with the Quadro line of cards. You are paying for tech support and certified drivers. You are also paying for the more expensive ECC memory.
Click to expand...

I still have a P5000 on my shelf. Quadro doesn't buy you any tech support at all - the responsibility is kicked down to the software provider. The idea of "certified drivers" is bunk as well. In the real world, those drivers are no more stable than the plain old GeForce drivers. ECC VRAM is similarly devoid of value unless you're trying to compute in space or next to an unshielded nuclear reactor.
 
Have to admit it looks cool, but since i only see my GPUs twice (once when installing, and once when getting rid of it) that shouldn't really matter.
 
Thunderdolt said:
The idea of "certified drivers" is bunk as well. In the real world, those drivers are no more stable than the plain old GeForce drivers.
Click to expand...
I made the mistake of buying Quadro for CAD once, never again. It was an eye opener. Certified drivers are outdated, and customized drivers usually only available for very specific software versions. Zero benefits. The only good thing that came from quadro is that all but a few of them died due to their crappy cooling solutions and we were able to exchange them for Geforce cards in almost equal value.
 
I am actually interested in 1-2 of these for my deep learning projects. I am currently using GTX 1080s and while I game the price is crazy for gaming only but I use it for development as well.
 
Thunderdolt said:
I still have a P5000 on my shelf. Quadro doesn't buy you any tech support at all - the responsibility is kicked down to the software provider. The idea of "certified drivers" is bunk as well. In the real world, those drivers are no more stable than the plain old GeForce drivers. ECC VRAM is similarly devoid of value unless you're trying to compute in space or next to an unshielded nuclear reactor.
Click to expand...

Regardless of how good or bad your experiences may be, or how good or bad the drivers, support etc is. It's still the reason that Quadro's are more expensive.
 
So how long lived do we think the first generation RTX series will be? Considering the entire line of cards launched within weeks of each other, a much more compressed schedule than previous generations.

Also due to the unimpressive Ray Tracing performance ($1,200+ to be barely playable at 1080p...), 7nm around the corner, new AMD cards, huge legacy card backlogs, supposed quality issues, and other factors.

Still interested to see the Titan's performance, hopefully the bickering about the price will stop and lead to more beneficial conversations, yes we know its expensive but I'd still want to see a super car at the tack even if I can afford it.
 
vjhawk said:
For those of you who claimed that the 2080 Ti was a Titan... you have officially been proven WRONG!

Now the new Titan is also double the price of the old Titan.

Let's see Nvidia's pricing scheme:

1080 Ti $699 -> 2080 Ti $1199
Titan XP $1200 -> Titan RTX $2499

No more excuses, Nvidia is officially Greedy AF!!!

Oh, and the cherry on top is that the Titan only has 5% more tensor cores and raytracing cores than the 2080 Ti. It's only real advantage is the 24GB of memory.

It's only going to be very slightly marginally faster than the 2080 Ti UNLESS you are using it for AI learning, etc.
Click to expand...

The titan gaming line since it did not get professional drivers was basically an early adopter tax for getting a big die early. Subsequent xx80 ti cards, were released later and cheaper with relatively the same performance.

But the number 1 reason people bought those cards was not because of the name, but it was because of the early release along with the bid die and performance relative to the xx80 series. They also liked the epeen factor because the price meant some form of exclusivity.

The GTX 2080 ti in the timeline of things is very much early release since it arrived at the same time as the RTX 2080. Usually Nvidia would attempted to milk the 2080 series more at 700 but since this series is likely to replaced in late 2019 with 7nm cards, releasing a 2080 ti means missing out on 6 months of sales for the series. As a result, they are releasing the RTX 2080 ti series early which means it has to be expensive to not cannibalize rtx 2080 sales.

Consumers might want the 2080/2080 ti to accept the price point of the 1080/1080 ti price point for for Nvidia it doesn't make sense. This is because the dies of the RTX 2080/2080 ti are much larger on a better process which makes there cost higher.

When there is no competition, does it make sense to obsolete your existing products with product that cost more to make at the same selling price? It doesn't, particularly when you have a whole crap load of existing last gen inventory that you want sold without taking a huge inventory write down.

Heck AMD is doing it too with the RX 590 series. The RX 590 has worse price to performance than the RX 580 even when compared to the starting MSRP and worse yet, AMD didn't put any money into the RX 590 really and has been milking Polaris on it's 3rd year. Atleast Nvidia has the excuse because they put alot of R and D into this series and it is a new architecture.

What's AMD excuse for raising the price on technology 3 years old? If you take the price/performance of the RX 590 series over the RX 580 series, the pricing is actually the same as an RTX 2070/2080.

E.g

Old Price New Price Price increase Performance increase cost increase(via techpowerup)
RX 580 230 RX 590 280 21% 11%
GTX 1080 499 RTX 2070 499-599 0-20% 10-15%
GTX 1080 ti 699 RTX 2080 699-799 0-14% 8-14%

Overall it's a shitty situation but both players are participating in this small increase in price/performance. They have so much left over inventory because of the collapse of mining that they want to prop up pricing of last gen by making the new gen, not as desirable until stock is cleared out.

But atleast for Nvidia, we are getting bigger dies and a new architecture.

Charging more for the same iteration of polaris is rediculous. Before you say the shift to 14 to 12nm, remember Nvidia is going through the same with Turing and AMD 1800x to 2700x shifted similarly from 14 to 12nm but instead of being priced at 499, is priced at $329(and it also comes with a performance bump as well).

As far as RTX Titans go, it is definitely overpriced in the consumer market considering the RTX 2080 ti is already available but as a workstation card, it's a killer.

https://www.gamersnexus.net/guides/...2018-for-adobe-premiere-autocad-vray-and-more

Nvidia started giving professional drivers(not signed but with the performance of the drivers) to their Titan Xp cards which make them behave more like Quadros than 1080 ti's as shown in these charts after AMD released the frontier edition.

This made them really good deals, almost underpriced considering the flagship gaming performance and the performance close to a $5000 quadro p6000 in many work loads. This is why some people suspected the gaming Titans would end. At $1200, these things were eating into quadro sales too much and when combined with the flagship gaming performance, was disruptive within Nvidia's own product stack.

As the market had shown, people were already willing to pay 1200 dollars for a gaming card. Add the professional application performance and you have something that eats into Quadro cards which can hurt potentially your own product stack. Titan V even at 3000 dollars when judged as a professional card is already well priced as seen with any review of the card.

Take those performance optimizations and apply them to the RTX 2080 ti and you have a card that is a steal at $1200 when you combine the drivers optimization of the Titan XP along with the advancements made with turing. It will top every single one of the test in the review I linked, while having performance in gaming 90% faster than RX Vega. When your paying your engineers 40 dollars an hour or your paid 10's of thousands for rendering jobs, $1200 is insignificant and $2500 is a very easy pill to swallow.

As a result and for the intended market, $2500 is a fair asking price. If your not going to use a Titans for it's professional application performance and purely for gaming the $2500 dollar asking price is too high. But there is the RTX 2080 ti for those people who want pretty much all the gaming performance. It still a high price but considering it is sold out, it is the right price for the intended market for the time being. When supply becomes greater which means demand is not meeting supply, this is where a product truly becomes overpriced.
 
  • Like
Reactions: spine
like this
IdiotInCharge said:
So you haven't been keeping up?
Click to expand...

Yes, I have. There has been some improvement with new drivers and game optimizations however I game at 4K. Until we have a card that can turn on RTX at 4K without dropping below 60 FPS I have no use for the tech. Dropping $1K to run games at a lower resolution and refresh rate is a no go for me. But I feel in a generation or two we should see playable performance at high resolutions. These are the worst ray tracing cards that will ever be made, I don't feel like the 20** cards will stick around that long, at least I hope not.

Unless you care to elaborate
 
Zam15 said:
Yes, I have. There has been some improvement with new drivers and game optimizations however I game at 4K. Until we have a card that can turn on RTX at 4K without dropping below 60 FPS I have no use for the tech. Dropping $1K to run games at a lower resolution and refresh rate is a no go for me. But I feel in a generation or two we should see playable performance at high resolutions. These are the worst ray tracing cards that will ever be made, I don't feel like the 20** cards will stick around that long, at least I hope not.

Unless you care to elaborate
Click to expand...

[H] is doing a review on the latest build with RT that was supposed to increase it by 50% which should make 60Hz 4k possible.

RTX also includes DLSS which I have always been more interested in. Just need more implementation... Final Fantasy got it yesterday so maybe we’ll have some reviews soon.
 
Dayaks said:
[H] is doing a review on the latest build with RT that was supposed to increase it by 50% which should make 60Hz 4k possible.

RTX also includes DLSS which I have always been more interested in. Just need more implementation... Final Fantasy got it yesterday so maybe we’ll have some reviews soon.
Click to expand...

You can test what performance boost you get with DLSS right now if you want. Set a custom res of 1800p, enable TAA and see what FPS you are getting.
 
reaper12 said:
You can test what performance boost you get with DLSS right now if you want. Set a custom res of 1800p, enable TAA and see what FPS you are getting.
Click to expand...

That’s one game and it’s really DLSS2x I want. I want super scaling and to put those tensor cores to work.

Lets be honest... we all would have preferred the 50% more cuda cores rather than this rtx horseshit. Would have been twice the performance of a 1080ti.
 
Dayaks said:
[H] is doing a review on the latest build with RT that was supposed to increase it by 50% which should make 60Hz 4k possible.

RTX also includes DLSS which I have always been more interested in. Just need more implementation... Final Fantasy got it yesterday so maybe we’ll have some reviews soon.
Click to expand...
Through optimizations or taking stuff away, both?
 
Dayaks said:
[H] is doing a review on the latest build with RT that was supposed to increase it by 50% which should make 60Hz 4k possible.

RTX also includes DLSS which I have always been more interested in. Just need more implementation... Final Fantasy got it yesterday so maybe we’ll have some reviews soon.
Click to expand...

Is that what that 15 gb patch was for?
 
Hagrid said:
Through optimizations or taking stuff away, both?
Click to expand...

Optimizations. Like not rendering shit you can’t see lol.

They’ve only been working on RT for a few months. I am sure they will find more down the road too.
 
Hagrid said:
Through optimizations or taking stuff away, both?
Click to expand...

With DLSS enabled, internal rendering resolution is actually reduced significantly. A requested resolution of 4k is actually rendered at 2560 x 1440, has AA applied, is then upscaled to 4k using AI, and then has the GUI/HUD/Menus drawn at native resolution on top of that.

That's how DLSS can perform better than no AA at all.
 
Last edited:
This thing coming by end of year or not?

Also wondering about the cooling solution. Seems a little under sized for the wattage.
 
Probleminfected said:
This thing coming by end of year or not?

Also wondering about the cooling solution. Seems a little under sized for the wattage.
Click to expand...

Technically, all they have to do is have it available to order by December 31st. So there are still 2 weeks left...
 
From the Jay Z Two Cents benchmarks, it's about 10% faster in games. But it's slower than the 2080TI in RTX for Battlefield V. He thinks it's probably just an immature driver. Having already purchased two 2080TIs, I'll be passing on this. Had it come out earlier than the 2080, or at least the same time, I might have considered this.
 
Nytegard said:
From the Jay Z Two Cents benchmarks, it's about 10% faster in games. But it's slower than the 2080TI in RTX for Battlefield V. He thinks it's probably just an immature driver. Having already purchased two 2080TIs, I'll be passing on this. Had it come out earlier than the 2080, or at least the same time, I might have considered this.
Click to expand...

The 2080 Ti's were difficult enough to justify (but doable).. with die shrinks at some point on the horizon, I'd have a tough time with a $2500 GPU.
 
brentsg said:
Hang in there man, I'm sure you'll land on your feet. Just try to enjoy the holidays and hit the ground running in January.
Click to expand...

Ditto - chin up and just enjoy the holidays. Don't let it get you down. Think of it as a new start. Companies are always looking to hire in January. New projects kicking off, new business plans being set into motion, etc. Spend a little time after Christmas brushing up the resume and then after New Year's, hit the ground running.
 
You must log in or register to reply here.
Back
Top