NVIDIA Announces Tegra 'Parker'

[patrickjp93]

3 minutes ago, HKZeroFive said:

I'll need a source for that.

 

The only product that used it was the Nexus 9, and that had sluggish sales figures. For a company that invested so much into Denver, only to have it in a single product which yielded poor sales (which was primarily due to the overheating and stuttering), I'd call that a failure to be honest.

And all the Shield products and one round of Google Chromebooks.

[HKZeroFive]

1 minute ago, patrickjp93 said:

And all the Shield products and one round of Google Chromebooks.

You do realise that the 32-bit Cortex-A15 Tegra K1 is different from the 64-bit Denver variant?

[patrickjp93]

5 minutes ago, HKZeroFive said:

You do realise that the 32-bit Cortex-A15 Tegra K1 is different from the 64-bit Denver variant?

Yes, and both were used.

[HKZeroFive]

1 minute ago, patrickjp93 said:

Yes, and both were used.

No, they were not. The Denver Tegra K1 was only used on the Nexus 9. Nothing else.

[Bhav]

I thought Nvidia said they were going to stop making mobile stuff and only focus on desktop / laptop GPUs?

[LAwLz]

38 minutes ago, patrickjp93 said:

Yes, and both were used.

I have never heard of nor seen the Denver version of the K1 used in any device except the Nexus 9.

The quad A15 version was not that uncommon (like you said, Nvidia's own products, some Chromebooks and such), but the Denver version is about as rare as unicorns. 

 

 

(In before "insider information" that can not be verified by anyone except by Patrick himself) 

[Tomsen]

On 8/22/2016 at 9:14 PM, LAwLz said:

I wonder if they included those four A57 cores because Denver is bad, or if they did it for some other purpose.

Perhaps for some system/security control where execution flow is critical and it needs a tons of certifications to operate a automotive vehicle, something that I don't think denver got..

 

1 hour ago, patrickjp93 said:

It was a big success. Nvidia made back its R&D 3x over.

You actually think that denver was a big success? Perhaps if they kept some of the design-wins they had in the beginning it would be decent, but far from..

[LAwLz]

Just now, Tomsen said:

Perhaps for some system/security control where execution flow is critical and it needs a tons of certifications to operate a automotive vehicle, something that I don't think denver got..

Yeah, in cars (especially self-driving) there are A TON of real-time systems. The latency from binary morphing would probably make it not be suitable. But that just makes the whole hybrid design eve more strange.

[Tomsen]

2 minutes ago, LAwLz said:

Yeah, in cars (especially self-driving) there are A TON of real-time systems. The latency from binary morphing would probably make it not be suitable. But that just makes the whole hybrid design eve more strange.

I think denver acts more like an accelerator than general purpose core in this scenario.

 

[LAwLz]

Just now, Tomsen said:

I think denver acts more like an accelerator than general purpose core in this scenario.

Sounds plausible. Again, Nvidia probably has some thought behind it but I don't get it right now.

[patrickjp93]

9 hours ago, HKZeroFive said:

No, they were not. The Denver Tegra K1 was only used on the Nexus 9. Nothing else.

Not true. It's the central chip of the Shield console.

[patrickjp93]

8 hours ago, LAwLz said:

Yeah, in cars (especially self-driving) there are A TON of real-time systems. The latency from binary morphing would probably make it not be suitable. But that just makes the whole hybrid design eve more strange.

You're assuming the binary morph takes significantly longer than the gaps in execution time for RISC code would allow.

[LAwLz]

2 hours ago, patrickjp93 said:

You're assuming the binary morph takes significantly longer than the gaps in execution time for RISC code would allow.

Yes I am assuming that, but you also have to remember that for real-time systems it's not just a race of how fast you can do something that matters. How predictable delays will be are extremely important variables too, and code morphing will most likely vary more (even if it often ends up the same or faster) than native code will (and we saw this a lot in the Tegra Denver K1).

 

 

 

2 hours ago, patrickjp93 said:

Not true. It's the central chip of the Shield console.

1) The Shield Console uses a Tegra X1 chip, not Tegra K1.

2) The Tegra X1 is a a quad core A53 + quad core A57 part. Denver is not used at all. The CPU is pure ARM design.

[HKZeroFive]

2 hours ago, patrickjp93 said:

Not true. It's the central chip of the Shield console.

You are talking about the 32-bit Cortex-A15 Tegra K1. The 64-bit Denver Tegra was not used on anything else other than the Nexus 9.

 

Wikipedia has a whole table explaining it all: https://en.wikipedia.org/wiki/Tegra#Devices_5

[patrickjp93]

4 hours ago, LAwLz said:

Yes I am assuming that, but you also have to remember that for real-time systems it's not just a race of how fast you can do something that matters. How predictable delays will be are extremely important variables too, and code morphing will most likely vary more (even if it often ends up the same or faster) than native code will (and we saw this a lot in the Tegra Denver K1).

 

 

 

 

1) The Shield Console uses a Tegra X1 chip, not Tegra K1.

2) The Tegra X1 is a a quad core A53 + quad core A57 part. Denver is not used at all. The CPU is pure ARM design.

No it won't. Embedded systems are extremely small in scope and size. It's fully measurable.

 

No, the new edition of the console uses an X1.

[79wjd]

2 minutes ago, patrickjp93 said:

No, the new edition of the console uses an X1.

https://shield.nvidia.com/store/android-tv

 

I don't know why you're so against providing a source. 

[patrickjp93]

3 hours ago, HKZeroFive said:

You are talking about the 32-bit Cortex-A15 Tegra K1. The 64-bit Denver Tegra was not used on anything else other than the Nexus 9.

 

Wikipedia has a whole table explaining it all: https://en.wikipedia.org/wiki/Tegra#Devices_5

That table is incomplete and disagrees with Nvidia's own list of products, but okay.

[patrickjp93]

2 minutes ago, djdwosk97 said:

https://shield.nvidia.com/store/android-tv

 

I don't know why you're so against providing a source. 

I'm not. It's just cumbersome on a phone. I'm in Brisbane now.

[HKZeroFive]

3 minutes ago, patrickjp93 said:

That table is incomplete and disagrees with Nvidia's own list of products, but okay.

Incomplete? Perhaps. Disagrees? No. Every single product in that table sports a NVIDIA Tegra K1 processor.

 

The table is correct in outlining that the Denver Tegra K1 was exclusively used on the Nexus 9. Nothing else. Find me a source that suggests otherwise.

[79wjd]

5 minutes ago, HKZeroFive said:

Incomplete? Perhaps. Disagrees? No. Every single product in that table sports a NVIDIA Tegra K1 processor.

 

The table is correct in outlining that the Denver Tegra K1 was exclusively used on the Nexus 9. Nothing else. Find me a source that suggests otherwise.

Wait, I thought the X1 was the newer/better of the two, not the K1.

 

Nvm, I'm stupid. 

13 minutes ago, patrickjp93 said:

I'm not. It's just cumbersome on a phone. I'm in Brisbane now.

I can't find anything that shows an Nvidia Shield using a Denver K1.  

https://en.wikipedia.org/wiki/Project_Denver

[HKZeroFive]

9 minutes ago, djdwosk97 said:

Wait, I thought the X1 was the newer/better of the two, not the K1

You're starting to confuse me.

 

The whole argument I'm having with Patrick at the moment is whether or not the Nexus 9 was the only product released so far to be sporting Denver. There are two types of Tegra K1s. The first is the 32-bit Cortex-A15 and the second being the updated 64-bit Denver variant. He thinks that the latter was not just exclusive to the Nexus 9, but rather also existed in Chromebooks and NVIDIA's Shield products. I'm arguing that he's incorrect because the products mentioned use the 32-bit variant, not Denver (which is only used in the Nexus 9).

 

You're correct in saying that the X1 is newer/better than both K1s. It is, after all, based on Maxwell whilst the K1s are based on Kepler.

[79wjd]

Just now, HKZeroFive said:

You're starting to confuse me.

 

The whole argument I'm having with Patrick at the moment is whether or not the Nexus 9 was the only product released so far to be sporting Denver. There are two types of Tegra K1s. The first is the 32-bit Cortex-A15 and the second being the updated 64-bit Denver variant. He thinks that the latter was not just exclusive to the Nexus 9, but rather also in Chromebooks and NVIDIA's Shield products. I'm arguing that he's incorrect because the products mentioned use the 32-bit variant, not Denver (which is only used in the Nexus 9).

 

You're correct in saying that the X1 is better than both K1s. It is, after all, based on Maxwell whilst the K1s are based on Kepler.

Yeah, I figured it out. I thought the X1 was denver. 

[LAwLz]

20 hours ago, patrickjp93 said:

No it won't. Embedded systems are extremely small in scope and size. It's fully measurable.

Measurable and "predictable and consistent" are not the same things.

Cars have a lot of real time systems which will gladly trade speed for consistency.

 

 

20 hours ago, patrickjp93 said:

No, the new edition of the console uses an X1.

Which "old version" of the Shield Console are you referring to then? Please link it, because I can not find any console at all on Nvidia's website that uses the Denver version of the K1. I can't even find a Shield product on Nvidia's website that uses the Denver version of the K1. I've done quite a bit of Googling (thanks for wasting my time) and I am 99.999% sure that the only product with the Denver K1 is the Nexus 9. That's it.

 

 

Bahaha, now Patrick is salty and tagging me in other threads going "look at what's in front of your eyes!".

Sorry Patrick but I can't just look at what's in front of my eyes, because you have provided 0 evidence that I can look at.