Tegra K1

[EssoAir]

http://arstechnica.com/gadgets/2014/01/nvidia-announces-next-generation-64-bit-tegra-k1-soc-with-192-gpu-cores/


Is there a chance that the Pyra could run Project Denver? A desktop GPU with full Open GL ES 4.4 support would be amazing. Also 64 bit so more ram connections for faster readwrite

 

[Null]

 

[EssoAir]

Why is that?

 

[Null]

Why is that?

 Because they only deal in large numbers of units and with large companies, if EvilDragon went to them and asked if they could sell him 4000 units I dare say they'd tell him to fuck off.

 

[EssoAir]

Why is that?

 Because they only deal in large numbers of units and with large companies, if EvilDragon went to them and asked if they could sell him 4000 units I dare say they'd tell him to fuck off.

What if he sent lots of cute emotes :3 :3 :3

 

[Null]

What if he sent lots of cute emotes :3 :3 :3

 

[EssoAir]

Are you just making these on the fly?

 

[Null]

Are you just making these on the fly?

 

[Annales]

That would actually be interesting to try. +1 for a fool hearted encouragement.

 

[EvilDragon]

Not a single chance.

 

[WizardStan]

So have you officially and completely decided on a SoC yet, or is this still up for debate?

And if you have, can we get more hints as to what it is? Possibly spoilers?

 

[slaeshjag]

ED said that some details will be revealed at FOSDEM. I guess we'll just have to wait for that...

 

[Hồng Thất Công]

@EssoAir:  For PYRA 1, smooth dreamcast and smooth ppsspp is enough.  Gamecube/Wii and PS2/Xbox emulations are for PYRA 2.  If PYRA 1 can run Dreamcast games and PSP games 100% like PCSX-ReArmed now on Pandora,  I am all set for the next 5 years from the PYRA 1 launch date.

Edit:  PS3, XBox360, and WiiU are for PYRA 3.  That's my timeline 

 

[Null]

ED said that some details will be revealed at FOSDEM. I guess we'll just have to wait for that...

 

[Natsu]

I read that Tegras don't live up to real life performance, anyway.

 

[fusion_power]

What should NVidia stop to sell small batches? Don't they calculate their profit per each Chip? I mean 5-10K Chips (I would set 2 batches for pyra a 5K units as realistic) are still a decent number and sill make profit.

Just in theory of course. I recently heared that Tegra support for ARM drivers is crap or vice versa so I guess Tegra would not be the best choice.

 

[WizardStan]

What should NVidia stop to sell small batches? Don't they calculate their profit per each Chip?

There's an overhead as well. They need to supply support to the people buying their chips. It's actually a lot more complex than just per-chip sales. They factor in the cost of doing business, support, lawyers, etc... and that all happens before a single chip is even sold. If they don't sell enough to make back the initial overhead then it's not worth it. I have no idea what that number is but there are probably a lot more factors that we couldn't ever think of. Doing business is surprisingly complex and you don't realize how hard it is to run one and all the little things you forget about until you actually do it.There's also a numbers thing: selling 10'000 chips to some small company that you will probably not hear from again means you have 10'000 fewer to sell to a big company that will want millions of these things and who you want to come back to you in the future when they want millions of your future product. If 10K guy doesn't get what he wants and never comes back, who cares? If big company says they want 1M and you only have 990K (or some other combination of numbers and time that means big company doesn't get exactly what it wants when it wants it because you sold 10K to a small fry) then that is bad news.

 

[bzar]

What should NVidia stop to sell small batches? Don't they calculate their profit per each Chip? I mean 5-10K Chips (I would set 2 batches for pyra a 5K units as realistic) are still a decent number and sill make profit.

Someone somewhere on this forum has answered this, but I can't find it. IIRC it's not just the chips they sell, but the support provided with it. Basically it's not worth their time to deal with small fries.

EDIT: Ninja'd by the Wizard. What he said.

 

[_wb_]

ED said that some details will be revealed at FOSDEM. I guess we'll just have to wait for that...

It's not that hard to guess though. He wants to stay with ARM for backwards compatibility, and on irc he said it would be a Cortex-A15 when drkIIRaziel asked about the CPU in the successor. We know that NVidia, Qualcomm and Samsung are not interested in small fries -- that rules out Tegra, Snapdragon and Exynos. ED doesn't want to work with Chinese SoCs because they have poor documentation and support -- that rules out HiSilicon, MediaTek, Allwinner and Rockchip. He also repeatedly said that he's looking more towards sourcing parts from the automotive sector than from the smartphone/tablet sector because of long-term availability concerns. Coincidentally, TI announced moving away from smartphones/tablets and focussing on other embedded platforms like automotive.

So by elimination, the only SoC left is TI OMAP 5. I hope it's the OMAP5432 and not the OMAP5430, but the difference is not that big (the 5432 supports more power-efficient RAM).

 

[slaeshjag]

Ssh! You're ruining the surprise for those not on IRC!

Anyway, isn't the 5432 the one that doesn't support PoP memory? That seems like an unlikely candidate, with all the bling ED wants, I can't see how everything will fit...