2070 is a TU106

[bru]

His first reference to performance was this:

"We shared some data that shows a bunch of games, you'll see the perf somewhere between 35-45% better at roughly the same generation. In most cases, if you are on high resolution and not CPU limited, Turing is going to crush it!"

Which is completely at odds with his answer when asked about the presentation slides. Then he claims that the performance over the 1080 will be 50% in most cases.

But after answering that question the interviewer also asked him this direct question, is the 2080 faster than the 1080ti. His reply "I think so. I would expect some cases where 2080 would beat 1080 Ti but I don't have the data in front of me"

I guess people are quoting that last sentence because who really believes he doesn't know the exact performance? And they are so quick to big up any product they are launching. Remember Pascal, in interviews and the launch event, they made it clear that the 1070 was going to be faster than the Titan X. He didn't have missing data for that interview. Surely if it was faster than the 1080ti, they would be doing the same thing now? What's to be gained from hiding the performance?

And you don't think that their is any chance that they deliberately don't want to tell us the exact performance, so they can shift a few more 10xx series?

Because if what this all seems to pointing too is NVidia's top 3 chips and yes I do mean chips not cards, being way ahead of the best the competition can muster. Which in its self is a freighting prospect considering the price of these RTX chips.

 

[tyler_jrb]

Bigger gap but still with in the range.

https://www.techpowerup.com/reviews/NVIDIA/GeForce_GTX_1080_Ti/30.html

1080 to Ti is 40% in most games if not more especially when both are overclocked.

1070 vs 1080Ti is a massive jump in most games 60%+.

2070 has 2304 cores.
2080Ti has 4352 cores.

If the 2070 matches the 1080Ti this puts the 2080Ti at roughly 89% faster than a 1080Ti.

 

[Nasher]

I think there is going to be a lot of disappointed people if that is the expectation.

 

[tyler_jrb]

I think there is going to be a lot of disappointed people if that is the expectation.

agreed.

 

[Unrivaledx21]

Looking at leaked architecture I am wondering why the 2080, TU104, seems to have a lot less improvements over the 2070. It looks like the 2080 is something like less than 30% increase in performance at more than 40% increase in price. Looking at the 2080 Ti in comparison it has 2x on everything at 2x the cost. Is the 2080 the worst to get? One thing to note as well is the TU104 is the only one to have 8 SMs per GPC where the others have 12 per. Any idea if that is beneficial to overclocking, or in any way at all for that matter?

 

[bru]

And that's what I found a bit suspect. Not even a footnote or passing comment. Hell, could've even gone like this:

"Hi, I'm Uncle Jensen, and this is your new RTX generation. Of course it's going to be faster than the 10 series, so there's no point in talking about it. You're here for the really good stuff"

I get the RTX technologies should receive all the focus because it's game changing stuff, but to have absolutely zero mention of the normal bits? I dunno, maybe my sarcasm is actually correct, there's little point talking about obvious performance jumps when there's a lot to talk about with RTX.

Still doesn't sit right though...

He did, he stood on stage and said plain as day "the 2070 is higher performance than the $1200 TitanXP "

Of course nobody wants to believe it and as I showed above even a 1080ti can lose to a 1070 in certain situations, So it only need to beat the TitanXP in one situation for Jensen to be correct.
But of course that is the only metric he gave us. For more we will just have to wait until the reviews.

 

[melmac]

Basically it will be within about 10% either way, mark my words pal haha.

Oh, that's what I am expecting in normal games, 2080 to be faster in some, slower in others. While games using DLSS will receive a substantial boost.

 

[melmac]

And that's what I found a bit suspect. Not even a footnote or passing comment. Hell, could've even gone like this:

"Hi, I'm Uncle Jensen, and this is your new RTX generation. Of course it's going to be faster than the 10 series, so there's no point in talking about it. You're here for the really good stuff"

I get the RTX technologies should receive all the focus because it's game changing stuff, but to have absolutely zero mention of the normal bits? I dunno, maybe my sarcasm is actually correct, there's little point talking about obvious performance jumps when there's a lot to talk about with RTX.

Still doesn't sit right though...

Could be, but, like you say, it doesn't sit right. There were a couple of slides at the launch event, but without settings used it's very hard to see what exactly is what. I presume they are best case scenario numbers. And they were at 4K.

 

[melmac]

And you don't think that their is any chance that they deliberately don't want to tell us the exact performance, so they can shift a few more 10xx series?

Because if what this all seems to pointing too is NVidia's top 3 chips and yes I do mean chips not cards, being way ahead of the best the competition can muster. Which in its self is a freighting prospect considering the price of these RTX chips.

No, because it would make no sense. All they have to do to sell more Pascal cards is drop the price a little, have sales events and game promotions.

Not sure what you mean by the second half of your post. It has nothing to do with this discussion.

 

[bru]

Er the thread is Titled "2070 is a TU106"

But yes it doesn't really connect with the currant subject under discussion.

 

[melmac]

He did, he stood on stage and said plain as day "the 2070 is higher performance than the $1200 TitanXP "

That seems to be only in relation to Ray Tracing though. Remember, he was going on about a new way to measure performance and was talking about RTX Operations and that's when he said the 2070 was faster than the Titan Xp.

 

[melmac]

Er the thread is Titled "2070 is a TU106"

But yes it doesn't really connect with the currant subject under discussion.

I was talking about what you quoted me on, not the thread in general.

 

[bru]

That seems to be only in relation to Ray Tracing though. Remember, he was going on about a new way to measure performance and was talking about RTX Operations and that's when he said the 2070 was faster than the Titan Xp.

Really not so sure about that. watch this from 1.39 minutes and then watch it from 1.37 ( yes I know that is the later bit then the earlier bit, but trust me and just watch them.
Doesn't make sense for the second comparison to the TitanXP to just be talking about ray tracing does it.

Sorry I couldn't get it to embed with a built in start time.

 

[Jeps]

Really not so sure about that. watch this from 1.39 minutes and then watch it from 1.37 ( yes I know that is the later bit then the earlier bit, but trust me and just watch them.
Doesn't make sense for the second comparison to the TitanXP to just be talking about ray tracing does it.

Sorry I couldn't get it to embed with a built in start time.

Keep watching it. The graph he displays right after he says the 2070 its faster then the titan xp literally says RTX-OPS at the bottom.

 

[bru]

Keep watching it. The graph he displays right after he says the 2070 its faster then the titan xp literally says RTX-OPS at the bottom.

Yes it does, and yet he says at 1.37.28 45 trillion RTX ops per second, that is several times the performance of a titanX.

So he has already said the RTX 2070 with 6 Giga rays per second 5 times that of a Titan X, 45 trillion RTX ops per second that is several times the performance of a TitanX and then after the slide is put up he says the 2070 is higher performance than the $1200 TitanXP.

Now I agree he might just be saying yes it is faster than a TitanXP in three different ways but the slide at the end, as far I can see it shows the several times the performance of the TitanX in RTX ops, look like about 3 times the 1080ti which I think we could all agree is several.

Maybe I'm wrong, but why use three different metrics to say it is faster in RTX stuff than the titanX.
So I definitely reckon he is saying 5 times the giga rays, several times the RTX ops and higher performance than a TitanX.

And to try and keep this in proportion to the thread, rather than the speculation one and this is the third tier chip TU106. Wow, impressive.

 

[X-Mikuru_Beam-X]

A

Yes it does, and yet he says at 1.37.28 45 trillion RTX ops per second, that is several times the performance of a titanX.

So he has already said the RTX 2070 with 6 Giga rays per second 5 times that of a Titan X, 45 trillion RTX ops per second that is several times the performance of a TitanX and then after the slide is put up he says the 2070 is higher performance than the $1200 TitanXP.

Now I agree he might just be saying yes it is faster than a TitanXP in three different ways but the slide at the end, as far I can see it shows the several times the performance of the TitanX in RTX ops, look like about 3 times the 1080ti which I think we could all agree is several.

Maybe I'm wrong, but why use three different metrics to say it is faster in RTX stuff than the titanX.
So I definitely reckon he is saying 5 times the giga rays, several times the RTX ops and higher performance than a TitanX.

And to try and keep this in proportion to the thread, rather than the speculation one and this is the third tier chip TU106. Wow, impressive.

And yet with all those 5 x GIGARAYS I bet it’s doo doo at running raytracing enabled, especially when we know the 2080ti can only just do 1080p with raytracing on...

 

[D.P.]

But how much of that 50% increase is taken up by Tensor Cores and RTX Cores? Its gonna be faster then a 1070 but not by the amount that the increase in transistors will have you believe.

Nvidias numbers show a 50% increase in performance. Will have to wait to reviews but with Pascal Nvidia numbers were spot on.

 

[JediFragger]

especially when we know the 2080ti can only just do 1080p with raytracing on...

This is a bit of a misnomer imo, was a quick implementation to show what is possible. We cannot take those numbers seriously as their is a LOT of optimisation that can be done (as well as 'level' of RT detail as it were, a slider in settings perhaps).

Let's give it half a chance to impress, in final hardware/software

 

[bru]

Exactly, even with CEO maths, unoptimised demos and dodgy leaks, surprisingly we still have no idea of the real performance that we will see.

Roll on release day. I'm actually quite excited for this one, even though I decided they were too pricey for me.

 

[melmac]

And to try and keep this in proportion to the thread, rather than the speculation one and this is the third tier chip TU106. Wow, impressive.

It's all about RTX Operations. So of course the Tu106 is going to be faster than the Titan XP since it hasn't the hardware to do any of those operations. I am not sure how you listened to the presentation and looked at the slides and can think that he was talking about performance in Normal games.

He did some slides with normal games later and didn't mention the 2070 been faster at all. And the very slides seem to suggest that the 2080 will be only marginally faster than the 1080Ti in Normal games, how is the 2070 going to be faster than the Titan Xp?

Second if it is faster in normal games why didn't Tom Petersen just say that in his interview, because if the 2070 is faster than the Titan Xp, then the 2080 is definitely going to be a faster than the 1080ti, you don't need any data in front of you to know that.

For the 2070 to be faster than the Titan Xp, a lot of speculative theories have to come true about Nvidia sandbagging and hiding the true performance for some unknown reason. Usually the simple explanation is the right one and the simple explanation is that it's just not going to be faster in normal games. But, we shall see.