Nvidia`s upcoming Tegra 5 playing Battlefield 3.

[ams23]

its known that mobile SOCs are always behind GPUs in adopting the bleeding edge. the fabrication capacity (or volume) required for mobile SOCs is much more than GPUs.

The fabrication capacity/volume required for one Tegra SoC is "much more" than all of NVIDIA's other GPU's? I don't think so. Anyway, there is no guarantee that mobile SoC manufacturers will always avoid the bleeding edge of process tech. For instance, Intel hopes to align Atom with the rest of their product stack in the future.

the specs of Logan GPU and Kayla are not the same.

Yes, Kayla is a dev platform for T5. NVIDIA stated that Logan will likely outperform Kayla.

 

[Lonyo]

At least we have a good few years until T1000, which is when we really need to worry.

 

[Exophase]

Yes, Kayla is a dev platform for T5. NVIDIA stated that Logan will likely outperform Kayla.

How does that make it a dev platform? There are lots of GPU + CPU solutions that Logan will likely outperform If all you want is any old CPU plus a weak Kepler GPU and Android I'm sure you could arrange that on existing laptops.

To qualify as a Logan dev platform I think you're at least going to need more than a Tegra 3 filling in for the CPUs.

 

[cmdrdredd]

I expect T4 devices to be out by the end of the year. T5 I am expecting in the middle of next year or end of next year.

Will that be too late though? That's something I'm wondering. If they will launch it at the right time with the right partners before they go with something else which may actually be slightly better in some way.

In my view for the mobile platform the CPU and the I/O must be the top performers. Graphics take a back seat unless you are going to use it dedicated to gaming which I just cannot see taking off. As I said, the hardware may be there but the software must be there to use it. The way the current Android and Apple stores are setup you basically have to force your way into the top 10 list or attach angry birds to the name to get any recognition from people browsing the app store. Especially if they expect the next games to be big titles like the 3DS gets etc. When a single game is $20 lets say and it's not even in the top 50 for most downloaded, people won't even see it and if they do they'll download 50 free games that can entertain the kids in the car instead.

 

[Stuka87]

Will that be too late though? That's something I'm wondering. If they will launch it at the right time with the right partners before they go with something else which may actually be slightly better in some way.

In my view for the mobile platform the CPU and the I/O must be the top performers. Graphics take a back seat unless you are going to use it dedicated to gaming which I just cannot see taking off. As I said, the hardware may be there but the software must be there to use it. The way the current Android and Apple stores are setup you basically have to force your way into the top 10 list or attach angry birds to the name to get any recognition from people browsing the app store. Especially if they expect the next games to be big titles like the 3DS gets etc. When a single game is $20 lets say and it's not even in the top 50 for most downloaded, people won't even see it and if they do they'll download 50 free games that can entertain the kids in the car instead.

Square Enix found out how well a high priced iOS game works out. They tried charging $15 bucks for a game only to see take rate be EXTREMELY low. The game was fine, but the price was way higher than it should have been. They would have made more selling it for $2.99.

As for the way the iOS store is setup, its very easy to go through games that are not in the top 10. If what you say is true, the top 10 would never change.

 

[cmdrdredd]

Square Enix found out how well a high priced iOS game works out. They tried charging $15 bucks for a game only to see take rate be EXTREMELY low. The game was fine, but the price was way higher than it should have been. They would have made more selling it for $2.99.

As for the way the iOS store is setup, its very easy to go through games that are not in the top 10. If what you say is true, the top 10 would never change.

Companies pay other companies to get them into the top 10 or top 50. Games would give you extra money or ammo or health packs etc. if you installed another app or game which was listed as free. The free game or app was paying the first one for each install they got because they could break into the top list that way.

Apple has recently cracked down on this practice. There were also screenshot scammers.

http://www.insidemobileapps.com/2011/04/19/apps-apple-rejected-pay-per-install/
http://gizmodo.com/5974817/apples-app-store-is-finally-cracking-down-on-the-screenshot-scam

Still what I am saying is that the games that are higher priced, but are higher quality too, will have a hard time presenting themselves against all the free and $1-$3 games out there and all over the top lists (be it top action/racing/fighting/arcade etc, top free, top paid).

You say they would have been better off selling it at $3, well maybe they felt that the time and effort were worth or should be worth more. As the mobile chips get more powerful you will be able to create bigger games. Games similar to the Vita or 3Ds. WHen you do that and want to release it on the iOS or Android stores, you still may take the same amount of time and money to develop the game as you would for a standard console game. That isn't cheap, so companies for sure would want to charge a bit more for the game because it's higher quality than fruit ninja HD and took more effort to create. That type of game would likely not do well.

One of the main problems. People view phones and tablets as media consumption and communication devices with the additional benefit of cheap games to waste a few minutes when you have a bit of time on your hands. People view the 3Ds or Vita etc as purely gaming systems.

 

[badb0y]

Will that be too late though? That's something I'm wondering. If they will launch it at the right time with the right partners before they go with something else which may actually be slightly better in some way.

In my view for the mobile platform the CPU and the I/O must be the top performers. Graphics take a back seat unless you are going to use it dedicated to gaming which I just cannot see taking off. As I said, the hardware may be there but the software must be there to use it. The way the current Android and Apple stores are setup you basically have to force your way into the top 10 list or attach angry birds to the name to get any recognition from people browsing the app store. Especially if they expect the next games to be big titles like the 3DS gets etc. When a single game is $20 lets say and it's not even in the top 50 for most downloaded, people won't even see it and if they do they'll download 50 free games that can entertain the kids in the car instead.

Well, whether it's too late or not is still up in the air at this time. If we take what nVidia says at the truth they have had more design wins for T4 then T3 so I guess that's something. I really don't know how well it will pan out throughout the year, it's still too early to call right now.

 

[ams23]

How does that make it a dev platform? There are lots of GPU + CPU solutions that Logan will likely outperform If all you want is any old CPU plus a weak Kepler GPU and Android I'm sure you could arrange that on existing laptops.

To qualify as a Logan dev platform I think you're at least going to need more than a Tegra 3 filling in for the CPUs.

It appears that the goal is to provide an ARM development platform for OpenGL 4.3 and CUDA 5: https://developer.nvidia.com/content/kayla-platform . So my mistake, Kayla is not meant to be a dev platform specifically for Logan (although the GPU probably does approximate the performance of Logan while having the same feature set as Logan), but rather for OGL and CUDA on ARM. Yes, the CPU in Kayla is probably much slower than what will be used in Logan, but better than nothing I suppose, and still 4+1 CPU cores like T4. Apparently T3 was used in this platform due to it's PCIe bus support for use with the Kepler GPU.

 

[Arzachel]

In my view for the mobile platform the CPU and the I/O must be the top performers. Graphics take a back seat unless you are going to use it dedicated to gaming which I just cannot see taking off.

This is is a line of thought that I can't really understand. What the hell would an average person use anything more than a reasonably clocked dual core A15 in their phone/tablet (and their laptops for that matter)? Is everybody secretly reencoding movies on their phones? Rendering CAD landscapes? Doing fluid dynamics calculations? What is the reason to push for more CPU grunt? Unless you're really into javascript benchmarks, you'll mostly be bottlenecked by the connection. All the UI is going to be hardware accelerated, at least focusing on the GPU makes some sense.

 

[ultimatebob]

That really wasn't a fair comparison. They should have showed the iPad playing Infinity Blade or Rage instead.

That lame game they were showing looked like it had graphics about on par with the original Half Life... a game that was released in 1998.

 

[Stuka87]

That really wasn't a fair comparison. They should have showed the iPad playing Infinity Blade or Rage instead.

That lame game they were showing looked like it had graphics about on par with the original Half Life... a game that was released in 1998.

They definitely pick and chose this test. That are a lot of games that look a lot better.

 

[cmdrdredd]

This is is a line of thought that I can't really understand. What the hell would an average person use anything more than a reasonably clocked dual core A15 in their phone/tablet (and their laptops for that matter)? Is everybody secretly reencoding movies on their phones? Rendering CAD landscapes? Doing fluid dynamics calculations? What is the reason to push for more CPU grunt? Unless you're really into javascript benchmarks, you'll mostly be bottlenecked by the connection. All the UI is going to be hardware accelerated, at least focusing on the GPU makes some sense.

Everything runs on the CPU. The os, the apps running in the background, the browser tabs. Try turning power saver on and forcing only 1 core on tegra 3 and then tell me CPU doesn't matter. It is much more important for pleasant user experience. Tegra 3 is better than exnos in the GPU department but slower for multitasking since the CPU has less power. The whole experience of using a tablet or phone depends on the CPU and I/O performance.

 

[Arkadrel]

The whole experience of using a tablet or phone depends on the CPU and I/O performance.

Somethings do better on the GPU

http://www.youtube.com/watch?v=jJSSWBykAMU

1 cpu core = 0.69 amps = 14 fps
2 cpu core = 0.83 amps = 24 fps

Opencl gpu = 0.65 amps = 42 fps

Doing Image processing on the phone of say a picture you just took with the camra of the phone.

Some tasks even on a phone, can be done faster and useing less power, than useing the cpu. So not every aspect of your experiance is dependt on the CPU.

at least focusing on the GPU makes some sense.

^ I agree with the sentiment.
The workloads you do on a phone usually arnt that demanding.
Most people probably consider their phones to be "good enough" and would rather they have longer battery life.

When their not doing something like gameing on them.


Edit:
This video is cool:
http://www.youtube.com/watch?v=CaS0kpozyMM

shows OpenCL editing too, better demo (than one above)

 

[SickBeast]

It sounds like nVidia is putting other projects on hold and really focussing on Tegra 5. They want to bury Apple. I have to say that I'm extremely impressed at how well Battlefield 3 ran on the Tegra 5 test platform. I can only imagine how games that are coded specifically for it will look.

 

[sontin]

How does that make it a dev platform? There are lots of GPU + CPU solutions that Logan will likely outperform If all you want is any old CPU plus a weak Kepler GPU and Android I'm sure you could arrange that on existing laptops.

It's not about Android it's about ARM and Cuda. nVidia run Ubuntu on Kayla at GTC. So they don't need Android.

To qualify as a Logan dev platform I think you're at least going to need more than a Tegra 3 filling in for the CPUs.

And how is this possible without PCIe? Tegra 3 supports it, Tegra 4 not. And right now their is no other ARM CPU out there which have PCIe support.