** The Official Nvidia GeForce 'Pascal' Thread - for general gossip and discussions **

[bru]

This is so similar to the VR debate, the people who like it like it and the people who don't don't.

Did anyone else actually watch the NVidia press conference, Jen-Sun made no bones about the fact that there is along way to go, but the progress being made is immense, the leap and bounds being made in this field are making the whole autonomous driving idea possible. It will happen, it is just a matter of time.

 

[pmc25]

Because it offers up 16tflops of performance.

8. Less than a Fury X.

 

[pmc25]

I think this confirms everything a few of us said and NVIDIA had hinted at. Pascal's #1 priority is compute, and compute products are going to be the first chips .... that it's an automotive product is a big surprise though .... I'd have thought that the stuff to fulfil their supercomputer contracts would go out first ... i.e. the IBM powerPC joint venture and nvlink interface which they've talked about before (the only platform nvlink is on). Also surprised that it's rolled into a tegra product, which has largely speaking been a disaster in terms of commercial success.

 

[Pottsey]

Least nVidia will never be the same level of fail as PowerVR.

That makes no sense as this isn't about PowerVR and PowerVR isn't a fail like Tegra.

Because it offers up 16tflops of performance.

Only 8tflops @ FP32 and it does that 8tflops for 250watts with liquid-cooling while competing solutions do the same job but only use up 3watt without liquid-cooling. How is a 250watt liquid-cooled mobile chip anything but a massive fail

 

[Pottsey]

This is so similar to the VR debate, the people who like it like it and the people who don't don't.

Did anyone else actually watch the NVidia press conference, Jen-Sun made no bones about the fact that there is along way to go, but the progress being made is immense, the leap and bounds being made in this field are making the whole autonomous driving idea possible. It will happen, it is just a matter of time.

I admit I haven't had chance to watch most of the press conference but this new NV solution seems to be a brute force that's far behind what competitors are doing. I agree its going to happen but with so many deals sighed now with competitors I don't see how NVidia can make a profit from this.

 

[Ayahuasca]

That makes no sense as this isn't about PowerVR and PowerVR isn't a fail like Tegra.

Oh like all those other threads you derailed with PowerVR talk when it wasn't about PowerVR, you're quick to jump in and defend PowerVR and trash talk nVidia when it suits you though, never see you in any thread when nVidia is providing a positive subject.

Who cares if Tegra is a fail to be honest there's plenty of other alternatives, this is the graphics card section. If you wanna talk mobile there's a forum for that

 

[pmc25]

I admit I haven't had chance to watch most of the press conference but this new NV solution seems to be a brute force that's far behind what competitors are doing. I agree its going to happen but with so many deals sighed now with competitors I don't see how NVidia can make a profit from this.

They probably won't. Per Tegra, it's late, big, hot and power hungry, and expensive.

 

[pmc25]

Who cares if Tegra is a fail to be honest there's plenty of other alternatives, this is the graphics card section. If you wanna talk mobile there's a forum for that

This is their first Pascal product, it's a Pascal thread, and the first Pascal product is a Tegra package.

 

[Ayahuasca]

This is their first Pascal product, it's a Pascal thread, and the first Pascal product is a Tegra package.

It's still not a graphics card, can easily make a Tegra thread in the mobile forum if the topic bothers you that much.

It shows how bored people are at the moment when the debate is a mobile chip nobody is bothered about the rest of the time.

 

[Pottsey]

“you're quick to jump in and defend PowerVR and trash talk nVidia when it suits you though, never see you in any thread when nVidia is providing a positive subject.” .

I only defend PowerVR when the comments are unfair or wrong. As for not seeing me post when NVidia is providing a positive subject that’s because I only follow limited areas that are of interest to me. If NVidia do something positive in those areas I would post in them. You never see me doing that as so far as recently NVidia haven’t done anything impressive in the areas I follow. My interests are Mobile, real time Ray Tracing, IoT, Digital Radio & Automotive. The only reason I followed this thread is due to the automotive talk related to Pascal. NVidia have done positive things but I rarely post about them as they are in areas I do not follow.

EDIT: Are you forgetting about PhysX which I am a big fan off and posted many positive things about? I don't hate everything NVidia and I don't Trash talk.

“Oh like all those other threads you derailed with PowerVR talk when it wasn't about PowerVR,” .

Like when? Most of the time if you go back and read either PowerVR was on topic or like this thread others like you brought up PowerVR. This isn’t a topic for PowerVR and you cannot blame me for the PowerVR references in this thread. It’s not my fault people like you derailed this thread by bringing up PowerVR when they don’t belong in this thread.

“Who cares if Tegra is a fail to be honest there's plenty of other alternatives, this is the graphics card section. If you wanna talk mobile there's a forum for that”
“It shows how bored people are at the moment when the debate is a mobile chip nobody is bothered about the rest of the time.

The reason we are talking about Tegra is this is a Pascal thread and so far the only pascal products are Tegra based. The CES show is on and the only new news from NVidia press conference that I have seen is Tegra. It makes no sense to spread the pascal talk over multiple threads in different sections. Pascal talk belongs all in one place.

 

[Gregster]

Yer, the Tegra is for the mobile section and not for the GPU section. Pascal is what interests me and hopefully we get some solid news soon.

 

[Simon Belmont]

Pascal is what interests me and hopefully we get some solid news soon.

This waiting is doing my head in. Money waiting nvidia?

 

[bru]

The only reason it is being discussed in hear is because the press conference was mentioned earlier, at the time we didn't know what NVidia were going to talk about, as it happens it was the autonomous driving field, which of course isn't really relevant to the GPU section.
The whole press conference was about the entire eco system not just the in car hardware, of course the in car hardware is what interests us as it contains the first Pascal part we have seen.

As for whether Tegra has been a success or failure, does it really matter to you personally ?

 

[JediFragger]

Yer, the Tegra is for the mobile section and not for the GPU section. Pascal is what interests me and hopefully we get some solid news soon.

Maybe we'll get something built with wood screws again, being tacked together as we speak

 

[Orangey]

Probably nothing until GTC.

 

[Gregster]

Maybe we'll get something built with wood screws again, being tacked together as we speak

I would take it lol

 

[Mtom]

This is their first Pascal product, it's a Pascal thread, and the first Pascal product is a Tegra package.

On some other forums they came to the conclusion that the Drive PX2 sport two 980m maxwell not pascal gpus.





Can that be true?

 

[Pottsey]

On some other forums they came to the conclusion that the Drive PX2 sport two 980m maxwell not pascal gpus.





Can that be true?

I don’t understand how they came to that conclusion. NVidia listed the specs as Drive PX2 with Pascal GPU cores.

https://www.flickr.com/photos/21075872@N02/23887556650

 

[Mtom]

I don’t understand how they came to that conclusion. NVidia listed the specs as Drive PX2 with Pascal GPU cores.

https://www.flickr.com/photos/21075872@N02/23887556650

Well first the gpu is the same size, and on 16nm with the given specs it should be a lot smaller. It really looks like it, and 2x 980M is 8 tfolps.

http://forums.anandtech.com/showthread.php?p=37942387

http://www.overclock.net/t/1586689/pconline-drive-px2-nvidia-showed-gp104-for-the-first-time-at-ces#post_24764683

Edit:i should have been more specific. Not the whole product line itself is questioned, but the one piece they showcased today.

 

[drunkenmaster]

Nvidia said a card was a 480gtx, it was fake, with wood screws holding it together. What the final product is and what the dev kit or show piece is aren't the same thing.

This almost certainly wouldn't be the first pascal product at all and people who think it might be are nuts. You don't take a new chip and slap it in a shipping manufactured car. You get a dev kit and develop software, do exhaustive testing and maybe at a later date choose to place a large order and put it in a final production car. AS yet these are only being ordered for TESTS for the manufacturers, Nvidia have only stated that, Audi/Volvo testing vehicles with it. They are likely testing the dev kit of everyone. Same way people made multiple designs for Tegra 1-3 but the products never actually made it to market. They wouldn't be doing due diligence to not see if Nvidia's product is better, a dev kit is usually in the region of a few thousand dollars so it's not a firm commitment and certainly not an indication of willingness to place an order of millions.

Nvidia lying about what they are showing on stage would be nothing new, using existing parts as a dev kit for something not yet available is pretty standard also. Sometimes you show a working pre production model, sometimes you show a dev kit set to emulate the final product.

It could easily be a couple of maxwells, maybe using well over 250W in the dev kit but with Pascal replacements in the final product or most likely a new version of the dev kit, the power/specs drop(this is the part Nvidia has lost the massive majority of their Tegra contracts at) to match what they promised or miss out.

EDIT:- the really quite odd thing is, AMD can do 8TF at 250W currently, Nvidia aren't far behind, on 16nm, honestly 8TF should be dropping to the 125-150W range and a couple of even 20W Tegra socs(which frankly even for Nvidia would be crazy power, think 5-10W realistically for a mobile Tegra style chip), you're still no where near 250W, 20W for cooling?