Could the RTX 2000 series names be misleading?

[Danoniero]

22 hours ago, i_build_nanosuits said:

That's a 2 generation jump...Titan Volta is already a shit load faster than titan Xp in games...so how could one more generation leap deliver such aweful performance?! also, do you think they would charge 1299$ if that card would't deliver? i dont.

Of course they would charge such money for nearly no performance upgrade. Why?.

First of all AMD have nothing at the moment and Nvidia have monopoly, they can literally put any price on rtx cards. Second of all you paying extra for fancy RTX and that's all. I don't believe ceo who couldn't shut up about ray tracing wouldn't show us normal benchmark comparing rtx to gtx10xx. If the performance was lot better he would shove it in your face, but instead he only talked about ray tracing and how it's 6x faster from previous gen. Not suspicious at all. Pre order is a big risk here if you want serious performance gains. I'm not a hater, I'm all Nvidia since riva tnt and never had and card, but it's just too little information for me to justify spending £750 on rtx 2080 and shoving away my 1080

[Jrasero]

On 8/21/2018 at 5:12 AM, Dunzaus said:

Back when the Gtx 600 series where the newest thing, the top tier card was the 680.

 

Then the 700 series came and we all thought that the 780, where going to be the top tier card that generation. But then 6 months later the 780 Ti came out and suddenly that was the top tier (not counting all the Titans cards, but I don't think those are that fair to compare with this)

 

And that's how it was for the next 2 generation as well, the 900 series had the 980 Ti and there were the 1080 Ti as well. But what I always felt Nvidia really had done, was taking logic we where use to, that the ''80's'' was suppose to the best and then put Ti at the end of it and have it come out 6-12 months later, with an insanely high price tag. So how I see it, then suddenly the 780 Ti, with the logic we where used too, with which number the cards where named changed. Based on how it used to be, the 780 Ti would have been called 780, the 780 would have been 770, and so on. And the same with the 900 and 1000 series.

 

Where I am going with this, is that I think that what's going on again. I think that 6-12 months down the line, some new card with a new name or number that's not a Titan, will come out and be new ''Ti'' or ''80'' .

I think that Nvidia is messing with what we think is how things work and using it to raise the prices Even More this generation, that they already have so far, with the upcoming cards.

 

It could even make the new cards look even more powerful compared to last generation, that they actually are, at least in compare to their equivalent numbered little brothers. So really what it should be compared to (if I am right about my predictions), is the Gtx 1060 (6gb).

 

Or I am wrong about all this and Nvidia is doing us all a favor and coming out with the very best gaming card from launch day, because it's been over 2 years since last generation gaming cards, where released.

 

But I doubt it.    

I think you are right that something within the next 12 months will be released by Nvidia.   This version of RTX was really to launch a new tech feature that will hopefully be the baseline of all future cards.  The prices and traditional specs make it clear this was kind of an iPhone S type of gen in hopes of maybe funding ray tracing development, getting rid of huge amounts of Pascal cards, and just testing how far the demand for anything new is even with huge MSRP premiums.

 

There are a couple issues.  AMD should be launching 7nm next year and for a company like Nvidia that has so much hubris they will not let them have any kind of win.  I guarantee Nvidia strikes back and issues their own 7nm or a updated Turing that have "X" more performance which will probably send anyone spending $600-$1200 on this version into a crazed furry.  More than likely AMD's cards will have no shot in competing performance wise however if AMD follows their own company history they can undercut their competitor significantly by releasing cards for hundreds of dollars less than the current gen Turing cards.  

 

Oh and then you have Intel in 2020, so without a doubt Nvidia is saving a few tricks for down the line.  I believe they thought RTX would wow us and we would accept it without asking performance numbers with or without RTX enabled.  It's pretty clear even the $1200 RTX 2080 Ti can't handle ray tracing well for 2018 standards since it only managed 60 FPS w/ ray tracing on while playing the new Tomb raider 1080P at 60Hz.  Simply RTX is more of a half baked idea like the original Siri.  It will takes years for developers to catch up and really it's more of a talking point than an actual useful feature as of today or even possibly years to come.  

[Jrasero]

5 hours ago, Danoniero said:

Of course they would charge such money for nearly no performance upgrade. Why?.

First of all AMD have nothing at the moment and Nvidia have monopoly, they can literally put any price on rtx cards. Second of all you paying extra for fancy RTX and that's all. I don't believe ceo who couldn't shut up about ray tracing wouldn't show us normal benchmark comparing rtx to gtx10xx. If the performance was lot better he would shove it in your face, but instead he only talked about ray tracing and how it's 6x faster from previous gen. Not suspicious at all. Pre order is a big risk here if you want serious performance gains. I'm not a hater, I'm all Nvidia since riva tnt and never had and card, but it's just too little information for me to justify spending £750 on rtx 2080 and shoving away my 1080

I agree and unfortunately the 2080 Ti meet be the only RTX card that sees decent traditional performances increases over Pascal, albeit at $1200 MSRP.  I don't mind spending $1200 if I know that the RTX 2080 Ti will be the top dog for the next 18-24 months, but it seems this gen might be a place holder at best even though it has a whole new moniker dedicated to it 

[Mira Yurizaki]

5 hours ago, Danoniero said:

Of course they would charge such money for nearly no performance upgrade. Why?.

First of all AMD have nothing at the moment and Nvidia have monopoly, they can literally put any price on rtx cards. Second of all you paying extra for fancy RTX and that's all. I don't believe ceo who couldn't shut up about ray tracing wouldn't show us normal benchmark comparing rtx to gtx10xx. If the performance was lot better he would shove it in your face, but instead he only talked about ray tracing and how it's 6x faster from previous gen. Not suspicious at all. Pre order is a big risk here if you want serious performance gains. I'm not a hater, I'm all Nvidia since riva tnt and never had and card, but it's just too little information for me to justify spending £750 on rtx 2080 and shoving away my 1080

AMD does have something to compete against NVIDIA in the high-end: the Vega 64 series. But for some reason nobody seems to really care about it (probably because it's hard to find)

 

Also why do we need to have performance figures of current things? Why can't a presentation just talk about the biggest feature that these GPUs are bringing to the table? NVIDIA in the GeForce 10 announcement presentation didn't present any performance figures, they only showcased the features that Pascal would bring to the table (at least from what I could gather). I'm sure they haven't done so in the past save for the GeForce 900 series, which was compared against the GeForce 600 series. The comparison was also likely done to show that despite being stuck on 28nm, they could eke out more performance to alleviate concerns that performance is strictly tied to process node size.

On 8/21/2018 at 5:44 AM, syn2112 said:

actually the advantage with Vulkan is about 20-40% performance bump, but that only ties with AMD.
Nvidia didn't get any advantage with Vulkan but there's a reason for that, it's because AMD uses Asynchronous Compute and Nvidia doesn't that's where the advantage comes in.

but it's reported than RTX uses Async Compute, so i guess we will see performance improvements there as well  

and who says that the frames are wasted? and either way with a performance bump like that, you can basically run the game at higher settings because of it, what's not to like?

or for example Devs can use the hardware more efficiently and do more complex scenes and such etc etc etc

NVIDIA doesn't need to implement asynchronous compute because it's not a feature of DirectX 12 or Vulkan. If you don't believe me, look at the feature level list of DirectX 12. The only thing it requires is that the GPU supports compute functions, nothing more. The "asynchronous" part is simply the compute tasks are now independent from graphics tasks, or rather, they're not running on the same queue. All this means in GPU parlance is that the GPU can do both graphics and compute tasks at the same time. Anything more is just using the term as a buzzword to give the product some virtual advantage.

 

Maxwell 2 and Pascal can do graphics and compute tasks at the same time. The problem is with Maxwell 2, it allocated resources ahead of time with no way to re-allocate resources as necessary leaving execution bubbles (well and the drivers never actually supported multiple queues). Pascal resolved this. NVIDIA's solution to the scheduling problem is to make their GPUs excel at parallelism whereas AMD wants their GPUs to excel at concurrency (https://techdifferences.com/difference-between-concurrency-and-parallelism.html). One or the other isn't strictly better overall.

[KarathKasun]

On 8/21/2018 at 4:21 AM, TrigrH said:

well i didnt read what you wrote but:

 

2080 and 2080ti are using the same die

 

the 2070 is using a smaller die and doesn't support nvlink.

 

SO they are misleading if you compare them to the 10 series.

Fake news.

 

RTX 2070 is TU104-350

RTX 2080 is TU104-400A

RTX 2080 Ti is TU102-300A

 

RTX 2080 and RTX 2080 Ti are not the same die.

[KarathKasun]

On 8/21/2018 at 7:30 AM, syn2112 said:

i honestly have not seen DX12 do anything better than DX11 so far... maybe when theres DX12 based engines it will perform as promised, but i have no idea.

Vulkan didn't need "vulkan based engines" to seriously impress us, i really wish everyone would start using Vulkan, Doom is so fucking amazing because of it.

DX12 has its own "Doom", the audience for its genre (RTS) is just much smaller.  That game has been given the name... "Ashes of the Benchmark"

[Dunzaus]

34 minutes ago, KarathKasun said:

Fake news.

 

RTX 2070 is TU104-350

RTX 2080 is TU104-400A

RTX 2080 Ti is TU102-300A

 

RTX 2080 and RTX 2080 Ti are not the same die.

Yeah I just ignored him. He's a perfect example on people who only read headlines.

[TrigrH]

On 8/23/2018 at 3:56 AM, KarathKasun said:

Fake news.

 

RTX 2070 is TU104-350

RTX 2080 is TU104-400A

RTX 2080 Ti is TU102-300A

 

RTX 2080 and RTX 2080 Ti are not the same die.

Wrong.

 

RTX 2070 is TU106 full die

RTX 2080 is TU104 cut down

RTX 2080 Ti is TU102 cut down

[KarathKasun]

22 minutes ago, TrigrH said:

Wrong.

 

RTX 2070 is TU106 full die

RTX 2080 is TU104 cut down

RTX 2080 Ti is TU102 cut down

Also fake news.

 

RTX 2080 ti, RTX 2080, RTX 2070 are all referred to as "ray tracing cards".  RTX 2070 has RTX cores, TU106 (GTX 2060) is not going to have RTX cores (75% sure of this after reviewing investor conference calls). 

 

We have also seen die shots with the TU104-400A (check out GamersNexus) and TU104-350 markings.

[TrigrH]

3 minutes ago, KarathKasun said:

Also fake news.

 

RTX 2070 has RTX cores, TU106 (GTX 2060) is not going to have RTX cores (75% sure of this after reviewing investor conference calls).

 

RTX 2080 ti, RTX 2080, RTX 2070 are all referred to as "ray tracing cards".  We have also seen die shots with the TU104-400A and TU104-350 markings.

The 2070 is using the TU106 is has RT cores. It does not support NVLINK and is a smaller die. GTX 2060 is probs just a pascal refresh and not using a TU based core.

[KarathKasun]

Just now, TrigrH said:

The 2070 is using the TU106 is has RT cores. It does not support NVLINK and is a smaller die. GTX 2060 is probs just a pascal refresh and not using a TU based core.

Links?

[i_build_nanosuits]

On 21/08/2018 at 5:12 AM, Dunzaus said:

Back when the Gtx 600 series where the newest thing, the top tier card was the 680.

 

Then the 700 series came and we all thought that the 780, where going to be the top tier card that generation. But then 6 months later the 780 Ti came out and suddenly that was the top tier (not counting all the Titans cards, but I don't think those are that fair to compare with this)

 

And that's how it was for the next 2 generation as well, the 900 series had the 980 Ti and there were the 1080 Ti as well. But what I always felt Nvidia really had done, was taking logic we where use to, that the ''80's'' was suppose to the best and then put Ti at the end of it and have it come out 6-12 months later, with an insanely high price tag. So how I see it, then suddenly the 780 Ti, with the logic we where used too, with which number the cards where named changed. Based on how it used to be, the 780 Ti would have been called 780, the 780 would have been 770, and so on. And the same with the 900 and 1000 series.

 

Where I am going with this, is that I think that what's going on again. I think that 6-12 months down the line, some new card with a new name or number that's not a Titan, will come out and be new ''Ti'' or ''80'' .

I think that Nvidia is messing with what we think is how things work and using it to raise the prices Even More this generation, that they already have so far, with the upcoming cards.

 

It could even make the new cards look even more powerful compared to last generation, that they actually are, at least in compare to their equivalent numbered little brothers. So really what it should be compared to (if I am right about my predictions), is the Gtx 1060 (6gb).

 

Or I am wrong about all this and Nvidia is doing us all a favor and coming out with the very best gaming card from launch day, because it's been over 2 years since last generation gaming cards, where released.

 

But I doubt it.    

Of course they will push out a titan card next...but who care...if you live to have the BEST graphics card of the moment you'll forever be disapointed...and the next titan will be so expensive anyways...so if you like what the 2080ti has to offer for the amount they ask then buy it...but if you buy it to have the top dog card...then yeah, just donT

[JoostinOnline]

On 8/21/2018 at 4:50 AM, glenalz81 said:

I think the lifespan will be short on the latest offering due to the upcoming 7nm process.

Why are people so obsessed with a minor architecture change? Fitting more transistors in a space doesn't magically mean more performance.

[glenalz81]

6 hours ago, JoostinOnline said:

Why are people so obsessed with a minor architecture change? Fitting more transistors in a space doesn't magically mean more performance.

Where in my post did i say more transistors mean more performance?

 

 

[rmarcusky]

It frustrates me when people say to wait for 7nm or wait for AMD next gen.....why not just wait till 5nm, or 4nm, or 1nm...…. if you are constantly in a wait state, they you will always be waiting. If you had a good run with your current GPU, then just upgrade already. If 7nm comes out in 3 months, then you can most likely still get a good resell from your current gen card, pay a small upgrade fee and get 7nm. SOMETHING BETTER, FASTER, SMALLER, CHEAPER, LARGER, HIGHER-RESOLUTION, MORE FEATURES, VR READY, IS ALWAYS COMING...….jump off the wait train and like Tom's hardware said, "Just Buy It!"..... I feel sorry for all those on the 'wait train' that will not be able to acquire a 2080Ti, because they will most likely remain sold out till 2019, and even then, if these cards can make mining efficient again, good luck EVER getting a card. If you are into tech, GPU's or gaming, and have the funds to get a card, DO IT, stop letting others talk you out of it.

[BuckGup]

I think people forget older GPUs don't just vanish and can equate to the same level of performance as new ones

[JoostinOnline]

21 hours ago, glenalz81 said:

Where in my post did i say more transistors mean more performance?

 

 

How is 7nm architecture worth waiting for?

[glenalz81]

2 hours ago, JoostinOnline said:

How is 7nm architecture worth waiting for?

You mean why?

 

Go read my post again.

 

All i said was i think the lifespan of the new RTX cards will be pretty short due to new architecture coming soon. That's it.

I also said a few posts later if the 2070s performance is close to the 1080ti that i'd probably get one so you're barking up the wrong tree here fella.

 

 

[JoostinOnline]

11 hours ago, glenalz81 said:

You mean why?

 

Go read my post again.

 

All i said was i think the lifespan of the new RTX cards will be pretty short due to new architecture coming soon. That's it.

I also said a few posts later if the 2070s performance is close to the 1080ti that i'd probably get one so you're barking up the wrong tree here fella.

 

 

The lifespan doesn't change just because something moderately better comes out.

 

You still seem to be linking a minor architecture change to performance.

[glenalz81]

1 hour ago, JoostinOnline said:

The lifespan doesn't change just because something moderately better comes out.

The lifespan, as a term, in this case meaning from launch until the next GPU is released that takes it's place. Not making the older version redundant in any way. Simple.

1 hour ago, JoostinOnline said:

You still seem to be linking a minor architecture change to performance.

Please for the love of god tell me how you come to this conclusion.