Couldn't Gph Used A Faster Cpu For Caanoo?
- Thread starter rsuryase
- Start date
Jengo
Member
while more power is always nice, you have to factor in Battery Life and Cost. The CPU in Wiz/Caanoo is already pretty powerful, especially when you are writing software the native ARM Assembly language supported by this CPU. The only reason why it seems slow is because a lot of the software that is ported over to the Caanoo is not written in this asm code, therefore it doesn't have specific optimizations to utilize its full power. At least that is how i see it.
Exophase
Nothing good will ever come of Exophase.
It seems slow because it's not faster than Wiz, which itself wasn't a huge leap over GP2X. People expect new devices to be faster than old ones because that's the direction the industry generally moves in. And it seems slow because most other things on the market are faster. It's obvious that GPH is not trying to be competitive in performance and is hoping to be attractive enough on other merits.
Battery life and performance are not directly inversely proportional - part of the reason is that with Caanoo you're looking at a chip manufactured on an old process. Pollux is done on a 90nm process, which was standard 5 years ago but the standard now is two nodes ahead at 45nm, which is roughly 4 times more transistor dense. That means that for the same power budget you can have more functionality and cache and at a faster clock speed, or you can run at the same speeds while consuming much less. Even ignoring the process advantage later ARM designs just have better performance/Watt intrinsically.
Cost is part of it. Moving to a better SoC (and therefore faster CPU and 3D) wouldn't cost that much more; you could make a world of difference with something that cost maybe $15-20 instead of $5, but then the RAM amount would seem too low and the display resolution anemic. Moreover, something like Caanoo represents a relatively small design modification from Wiz so doesn't cost too much to design.
They wouldn't have retained backwards compatibility with another SoC, but that's moot because they didn't retain it in Caanoo in the first place. Ironically notaz was able to accomplish this himself for a variety of other platforms that are much more different from Wiz than Caanoo is, so if they used another ARM chipset + OABI compatible Linux he'd surely have been able to as well.. but notaz's software capabilities are way beyond GPH's software capabilities.
Battery life and performance are not directly inversely proportional - part of the reason is that with Caanoo you're looking at a chip manufactured on an old process. Pollux is done on a 90nm process, which was standard 5 years ago but the standard now is two nodes ahead at 45nm, which is roughly 4 times more transistor dense. That means that for the same power budget you can have more functionality and cache and at a faster clock speed, or you can run at the same speeds while consuming much less. Even ignoring the process advantage later ARM designs just have better performance/Watt intrinsically.
Cost is part of it. Moving to a better SoC (and therefore faster CPU and 3D) wouldn't cost that much more; you could make a world of difference with something that cost maybe $15-20 instead of $5, but then the RAM amount would seem too low and the display resolution anemic. Moreover, something like Caanoo represents a relatively small design modification from Wiz so doesn't cost too much to design.
They wouldn't have retained backwards compatibility with another SoC, but that's moot because they didn't retain it in Caanoo in the first place. Ironically notaz was able to accomplish this himself for a variety of other platforms that are much more different from Wiz than Caanoo is, so if they used another ARM chipset + OABI compatible Linux he'd surely have been able to as well.. but notaz's software capabilities are way beyond GPH's software capabilities.
Exophase
Nothing good will ever come of Exophase.
A little less due to not having a keyboard and having one analog stick vs a d-pad + 2 analog nubs.
traylorpark
Very Active Member
- Joined
- Sep 30, 2008
- Messages
- 593
rsuryase said:
I guess it would be nice ... but from a games perspective, I feel like much stronger hardware would just be grossly under utilized by most developers. Many devs get their start in games making on this system.
How do you imagine faster hardware benefiting users?
Last edited by a moderator:
Iorgy77
emu parasite kid
Coulda, shoulda, woulda...
rsuryase said:
I don't think anyone can say. There are many different 1Ghz ARM processors out there with widely varying capabilities.
They would most likely have to design a completely new system and have a totally new group of partners to obtain parts for it. By keeping mostly the same specs from the Wiz to the Caanoo they greatly simplified the development and production of the device, which would have made it significantly cheaper. Not to mention the fact that it would already be a tried and tested platform.
For a small company producing products in a niche market that doesn't rely solely on hardware capabilities, it seems pretty logical for them to not take many risks by starting from scratch.
I think the more interesting question is why did they make just enough changes to make Wiz software incompatible? If it already supported everything that the Wiz supported and it had some kind of internal "wrapper" to make the analog stick emulate the directional pad of the Wiz, I think it would be getting more attention.
Still, none of this really matters. Its available, it has a lot of talented developers working hard to support it, and it seems like a solid system so far. On top of that its reasonably priced. I'm happy with mine.
Last edited by a moderator:
Exophase
Nothing good will ever come of Exophase.
If it means anything:
Odroid-S Premium: $299
Odroid Premium: $239
Odroid Lite: $199
The difference between the Odroid Premium and Lite is that the former comes with a second battery and an 8GB SD card. So in theory an Odroid-S Lite would cost $259. The differences between Odroid-S and Odroid are:
- Odroid-S has a 1GHz S5PC110 (with SGX540), Odroid has an 833MHz S5PC100 (with Samsung in-house graphics that are much lower performance)
- Odroid-S has 400MHz/512MB of RAM partitioned as 384MB of LPDDR1 and 128MB of "OneDRAM" and some unknown amount of "OneNAND", Odroid has 400MHz of less expensive 512MB DDR2 and no NAND
- Odroid-S has a magnetometer
I think that's about it.
So if they made an Odroid with S5PC110 and otherwise kept things the same it might cost somewhere between $199 and $259. And it could cost less if they used less RAM.
Of course, these devices are hardly what I'd consider mass production, given how hard it is to get one and how they're usually sold out. But going larger volume should lower prices, not raise them, assuming there's a market for it.
What's interesting about the current crop of HardKernel platforms is that they all share the same CPU daughterboard that contains the S5PC110, OneNAND + LPDDR1 + OneDRAM in a PoP, and PMIC (power chip), and they sell this board standalone for $120. They also include full schematic of the connector. So really you could take this board and do your own handheld around it; at the very least it'd be great for prototyping a new concept. This S5PC110 setup is better than the other stuff an average person can currently buy (OMAP3 and its Cortex-A8 based successors, i.MX51L).
Odroid-S Premium: $299
Odroid Premium: $239
Odroid Lite: $199
The difference between the Odroid Premium and Lite is that the former comes with a second battery and an 8GB SD card. So in theory an Odroid-S Lite would cost $259. The differences between Odroid-S and Odroid are:
- Odroid-S has a 1GHz S5PC110 (with SGX540), Odroid has an 833MHz S5PC100 (with Samsung in-house graphics that are much lower performance)
- Odroid-S has 400MHz/512MB of RAM partitioned as 384MB of LPDDR1 and 128MB of "OneDRAM" and some unknown amount of "OneNAND", Odroid has 400MHz of less expensive 512MB DDR2 and no NAND
- Odroid-S has a magnetometer
I think that's about it.
So if they made an Odroid with S5PC110 and otherwise kept things the same it might cost somewhere between $199 and $259. And it could cost less if they used less RAM.
Of course, these devices are hardly what I'd consider mass production, given how hard it is to get one and how they're usually sold out. But going larger volume should lower prices, not raise them, assuming there's a market for it.
What's interesting about the current crop of HardKernel platforms is that they all share the same CPU daughterboard that contains the S5PC110, OneNAND + LPDDR1 + OneDRAM in a PoP, and PMIC (power chip), and they sell this board standalone for $120. They also include full schematic of the connector. So really you could take this board and do your own handheld around it; at the very least it'd be great for prototyping a new concept. This S5PC110 setup is better than the other stuff an average person can currently buy (OMAP3 and its Cortex-A8 based successors, i.MX51L).
