NVIDIA GeForce RTX 20-series Mobility mGPU Lineup Revealed

[btarunr]

NVIDIA is giving finishing touches to its first GeForce RTX 20-series Mobility GPUs for notebooks, based on the "Turing" architecture, with product launches expected from Q1-2019. The company could debut the series with a high-end part first, the GeForce RTX 2080 Mobility Max-Q. The rest of the lineup includes the RTX 2070 Mobility Max-Q, RTX 2060 Ti Mobility, RTX 2060 Mobility, RTX 2050 Ti Mobility, and RTX 2050 Mobility. What's interesting about this list is that NVIDIA is limiting the Max-Q design to its top-tier RTX 2080 Mobility and RTX 2070 Mobility parts.

Max-Q is an all-encompassing laptop thermal-design methodology, which allows gaming notebook designers to come up with thinner notebooks with higher performance. One of the key aspects is special Max-Q ready variants of the GPUs, which are probably binned to run the coolest, and least voltages. With a device ID 1eab, the RTX 2080 Mobility Max-Q is based on the TU104M chip, while other SKUs could be carved from the TU106M or a chip even smaller. It's being reported that with this generation, NVIDIA is playing a more active role in helping its partners engineer their Max-Q notebooks, and helping them meet NVIDIA's strict Z-height minimums.



View at TechPowerUp Main Site

 

[Vayra86]

LMAO. RTX heat in a thin and light device. I can't wait to see it perform worse than Pascal. Just add Max-Q and all the noobs think they're buying the good stuff when in fact they are stuck with overpriced performance of a GPU that is two tiers lower than its desktop equivalent. IF its not throttling, that is.

'Nvidia is playing a more active role' - that alone should be read as a big fat warning sign.

 

[Tsukiyomi91]

Sounds interesting though the "Mobility" suffix was used in ATI's mobile GPU family 10 years ago... I am stoked to see how the Turing perform in thin & light platform. Hopefully the silicon is mature enough to get a 8nm or 10nm shrink with reasonable yield rate.

 

[notb]

I haven't seen any material stating that lower parts will have RTX. Everyone is talking about a picture that doesn't include the "RTX" text at all - just model numbers.

Sure, I'd love an RTX 1050 mobile, but that's really unlikely.

 

[Vayra86]

I haven't seen any material stating that lower parts will have RTX. Everyone is talking about a picture that doesn't include the "RTX" text at all - just model numbers.

Sure, I'd love an RTX 1050 mobile, but that's really unlikely.

Good point, in fact!

 

[Tsukiyomi91]

it's been said that ANY TU10x cores that are "lower" than 104 won't get RT cores or even get the RTX naming convention, so we'll see a "GTX" 2050 or 2060 instead with Tensor & CUDA core present in those toned-down Turing chips.

 

[Ibotibo01]

Are you sure 2060 Ti?

 

[bug]

NVIDIA is limiting the Max-Q design to its top-tier

Maybe because lower tier cards don't draw that much power anyway?

 

[T4C Fantasy]

it's been said that ANY TU10x cores that are "lower" than 104 won't get RT cores or even get the RTX naming convention, so we'll see a "GTX" 2050 or 2060 instead with Tensor & CUDA core present in those toned-down Turing chips.

the thing is, that is all speculation, whether its true or not is not known.

but TU106 has rt cores and that will be 2060Ti and maybe 2060 too.
TU116 may have tensor and rt removed.

 

[Tsukiyomi91]

Ah... well, guess we have to find out by December or next year whenever NVIDIA teases us again.

 

[newtekie1]

LMAO. RTX heat in a thin and light device. I can't wait to see it perform worse than Pascal. Just add Max-Q and all the noobs think they're buying the good stuff when in fact they are stuck with overpriced performance of a GPU that is two tiers lower than its desktop equivalent. IF its not throttling, that is.

'Nvidia is playing a more active role' - that alone should be read as a big fat warning sign.

The performance/watt of Turing is better than Pascal, from what we've seen. So performance in mobile devices should be better for the RTX cards than pascal cards.

 

[R00kie]

it's been said that ANY TU10x cores that are "lower" than 104 won't get RT cores or even get the RTX naming convention, so we'll see a "GTX" 2050 or 2060 instead with Tensor & CUDA core present in those toned-down Turing chips.

TU106 in RTX2070 has them enabled.

 

[Vayra86]

The performance/watt of Turing is better than Pascal, from what we've seen. So performance in mobile devices should be better for the RTX cards than pascal cards.

Really now? And this is without RTX ON

 

[R0H1T]

I haven't seen any material stating that lower parts will have RTX. Everyone is talking about a picture that doesn't include the "RTX" text at all - just model numbers.

Sure, I'd love an RTX 1050 mobile, but that's really unlikely.

Why, you don't "have to" enable RT in games do you? If they've made RTX, with RT in mind, as many have said it'd be foolish to just include it in the high end. Without mid/low range cards having RT in silicon, it'd be a stillborn at best or worse Physx v2

 

[notb]

Really now?

Do we see a different picture?
2080: 111%
1080: 108%
2080Ti: 99%
1080Ti: 95%
?
That may not be much, but it's still better.
And then...

And this is without RTX ON

RTX and Tensor cores are designed to be way more effective in particular tasks. They'll improve performance/W further in such scenarios. Otherwise they're idle.

 

[Vayra86]

Do we see a different picture?
2080: 111%
1080: 108%
2080Ti: 99%
1080Ti: 95%
?
That may not be much, but it's still better.
And then...

RTX and Tensor cores are designed to be way more effective in particular tasks. They'll improve performance/W further in such scenarios. Otherwise they're idle.

Yeah thats a whole 2 minutes of extra battery life wooptidoo!

 

[notb]

Yeah thats a whole 2 minutes of extra battery life wooptidoo!

Well... better is better. You didn't say it has to be much better.
If you put the RTX and Tensor parts away, it's really just a Pascal refresh.

But the "not worse" result is fairly important for notebooks. This means manufacturers will be able to just replace 10-series with 20-series without much intervention to design of cooling etc.
Pascal gave us a huge jump of GPU performance in ultrabooks and thin gaming laptops. It's great this won't be reverted by the new generation.