Active Topics

 


Page 3 of 4 12 3 4
Reply
Thread Tools
allnameswereout's Avatar
allnameswereout is offline allnameswereout
2008-12-14 , 20:45
Posts: 3,397 | Thanked: 1,212 times | Joined on Jul 2008 @ Netherlands
#21
Originally Posted by Benson View Post
You seem to be confusing two issues: first, even (recent) 32-bit x86 machines are all more RISC than CISC under the hood than, with translation to run the x86 instruction set; the RISClike internals are completely inaccessible, because all instructions are translated. Second, AMD64 (and Intel's clone, EM64T I think) have a 64-bit and 32-bit mode; the 64-bit mode is almost precisely a superset of the 32-bit mode, so there's no separate translation stage (and performance penalty) added for 32-bit mode. (And while the OS has little to do with performance, AFAIK 32-bit OSes can't permit 64-bit userspace code, so you have to run a 64 kernel so you can have 64 code in places where it does count (heavy math libraries). That and memory addressing are the principal reasons for 64-bit OSes, not across-the-board performance gains.)
That doesn't contradict anything I said.

However, I do assert backwards compatibility and performance are important (on desktop especially). This is what Itanium neglected, and where AMD64 succeeded.

And usually on AMD64 and Intel's AMD64 implementation everything runs in 64 bit mode on a 64 bit OS except some parts which are not native on AMD64 (these parts are getting smaller and smaller every year). Then the x86-32 compatibility mode is used.

Memory addressing can be extended on x86-32 with some hacks. But even then, most desktops don't require more than 4 GB yet, and neither do embedded systems such as a tablet or phone.
__________________
Goosfraba! All text written by allnameswereout is public domain unless stated otherwise. Thank you for sharing your output!
 
tom61 is offline tom61
2008-12-14 , 23:31
Posts: 58 | Thanked: 9 times | Joined on Jul 2007
#22
Originally Posted by Voltron View Post
if a new x86 tablet was released with equal performance and battery life to an arm based tablet which would you buy?
If you mean equal to the current OMAP2 tablets, it'd be too slow to run a full-on desktop operating system, IMO. I could barely stand a 800MHz Transmeta based ultraportable, and it looks like OMAP2 would actually be a bit slower in the same tasks. Though, x86 would give some flexibility in running non-ported applications, like say Flash 10, but it'd definitely need thin distro to work effectively.

Now the tablets coming out based on the OMAP3, an equivalent to that might might actually be fast enough to run a full-desktop environment (with some tweaks for the small screen). With 3D acceleration, you could have portable Unreal Tournament, or other older binary or Windows-only (thanks to WINE) games. It'd be something I'd consider buying if it were under $500.

also what is your opinion on x86 pocket devices?
MIDs are a bit too high priced for my tastes, and that's pretty much to only x86 pocketable devices.
 
SD69 is offline SD69
2008-12-14 , 23:39
Posts: 1,513 | Thanked: 2,248 times | Joined on Mar 2006 @ US
#23
Originally Posted by fms View Post

Well, the original poster clearly meant Nokia Internet Tablets by "tablets".
From the way he phrased it, I don't know about that. But for devices like the NIT, there are much more important aspects of the device besides CPU architecture. For me, I think if a device had poor battery life because of the CPU, then I would not buy it for that reason alone. But, if it didn't have poor battery life, then I would buy a x86 tablet or MID that had a better design.
 
allnameswereout's Avatar
allnameswereout is offline allnameswereout
2008-12-15 , 00:09
Posts: 3,397 | Thanked: 1,212 times | Joined on Jul 2008 @ Netherlands
#24
The nice thing about open source and ARM is that GPLed engines like Quake 2 will work on OMAP3. You still need a Quake 2 license to play the original Quake 2 levels or play online (very cheap now anyway) but besides that you're as free as a bird. And, Quake 1 and Quake 2 are _very_ portable. They are originally developed on IRIX (and compile without source modification on IRIX), so the chance they're ported to an architecture as ARM is high. I don't know about Quake 3. Then there are tons of games which used this game engine, some also open sourced.

If you look at current low-end x86 compatible computers, even for embedded, they suck too much juice. The uptime is simply too low on batteries. They're perfect for a lot of goals (STB, firewall, NAS -- you name it) but a portable target is one of the worst usages because there the kWh is even more important than anywhere else. This won't change for some years either...
__________________
Goosfraba! All text written by allnameswereout is public domain unless stated otherwise. Thank you for sharing your output!
 
attila77 is offline attila77
2008-12-15 , 00:18
Posts: 3,319 | Thanked: 5,610 times | Joined on Aug 2008 @ Finland
#25
Originally Posted by allnameswereout View Post
However, I do assert backwards compatibility and performance are important (on desktop especially). This is what Itanium neglected, and where AMD64 succeeded.
As somebody already noted, this does not hold for linux devices. That's why the barrier of putting linux desktop apps onto NIT style devices (if you can live with the performance) is so much easier than getting your favourite Windows desktop app on a WinMobile device. Itanium had the same problem. It wasn't ever intended as a desktop processor, it's just that when it came around, it never had enough software support in it's target segment. If Itanium's target users back then would have been Linux oriented, it would mean minimal migration costs.

Somebody mentioned proprietary things, like Skype, flash, etc. Well, if Nokia managed to push enough to get Skype and Flash (crappy as they are) on our beloved NIT's I don't see any reason to re-consider x86.

x86 will never be (power/performance wise) a serious competitor to architectures that could be designed from ground up. It just has too much baggage. It can be good, but the question will never be if it's better, just how much slower/inefficient it is. The only exception of course is if embedded manufacturers go bust or do zero development in the coming years - unlikely, embedded is very dynamic even in these days.
 
fms is offline fms
2008-12-15 , 06:23
Posts: 1,418 | Thanked: 1,541 times | Joined on Feb 2008
#26
Originally Posted by attila77 View Post
x86 will never be (power/performance wise) a serious competitor to architectures that could be designed from ground up. It just has too much baggage.
Well, this remains to be seen. I would not be so sure
__________________
Computer and Console Emulators for Maemo, Symbian, and more.
Pack Rat: The Maemo Package Aggregator.
Last edited by fms; 2008-12-15 at 08:37.
 
allnameswereout's Avatar
allnameswereout is offline allnameswereout
2008-12-15 , 12:43
Posts: 3,397 | Thanked: 1,212 times | Joined on Jul 2008 @ Netherlands
#27
Originally Posted by attila77 View Post
As somebody already noted, this does not hold for linux devices. That's why the barrier of putting linux desktop apps onto NIT style devices (if you can live with the performance) is so much easier than getting your favourite Windows desktop app on a WinMobile device. Itanium had the same problem. It wasn't ever intended as a desktop processor, it's just that when it came around, it never had enough software support in it's target segment. If Itanium's target users back then would have been Linux oriented, it would mean minimal migration costs.
Right.

Somebody mentioned proprietary things, like Skype, flash, etc. Well, if Nokia managed to push enough to get Skype and Flash (crappy as they are) on our beloved NIT's I don't see any reason to re-consider x86.
Yes, except when new trendy things (Fring) aren't ported (well enough) to Linux/ARM while for example they do run on Symbian/ARM. I use a proprietary application called Devicescape. It allows me to log in to thousands of hotspots (for free due to my DSL subscription) without logging on HTTPS manually (this is slow...). Too bad I cannot import it as POI...

x86 will never be (power/performance wise) a serious competitor to architectures that could be designed from ground up. It just has too much baggage. It can be good, but the question will never be if it's better, just how much slower/inefficient it is. The only exception of course is if embedded manufacturers go bust or do zero development in the coming years - unlikely, embedded is very dynamic even in these days.
We're already seeing Intel Atom (Celeron), AMD Geode GX (LX), and VIA Nano (C3, C7) boards applied on embedded fields where ARM, MIPS and POWER were normally applied. The AMD Geode GX and VIA C3/C7/Nano also support hardware cryptography, and VIA/S3 are getting more friendly towards open source thanks to open source liaison Harald Welte. They're still not the most efficient, but they're definately improving. So we're seeing them on netbooks, MIDs (tablets), STBs, phones, and firewalls & routers too. Note VIA Nano is a x86-64 compatible processor with x86-32 backwards compatibility.
__________________
Goosfraba! All text written by allnameswereout is public domain unless stated otherwise. Thank you for sharing your output!
 
Khertan's Avatar
Khertan is offline Khertan
2008-12-15 , 13:16
Posts: 1,012 | Thanked: 817 times | Joined on Jul 2007 @ France
#28
Prefer ARM proc for better consumption, as i don't want to carry a device of 3kg just to have more than 3 hours of power (x86 based device).
__________________
Benoît HERVIER, Khertan
KhtEditor - Sdist_Maemo - Khweeteur - PyPackager - KhtSimpleText - KhtNotes - KhtBMA - Wleux - and more ...
 

The Following 2 Users Say Thank You to Khertan For This Useful Post:
jolouis is offline jolouis
2008-12-15 , 15:19
Posts: 631 | Thanked: 837 times | Joined on May 2007 @ Milton, Ontario, Canada
#29
I don't consider x86 as a viable tablet solution, simply because of the reality of power management and performance. Don't get me wrong, I think x86 embedded systems are a viable solution for some scenarios (I use an AMD geode board as a dual radio AP/media box/thingy in my Jeep and another one as an AP at the house), mostly due to the fact that they're much more efficient both in size and power consumption than a full PC, but they're certainly not on the scale and power levels of the ARM boards. Quite frankly I think the software discussion is a mute point, as any handheld device is going to require at least some level of software customization in order to best suit the needs of the user, and any well written code should be portable enough to translate between x86 and ARM without a lot of difficulty... the argument for x86 here is really more of "an easy way out" than a real solution to the scenario in question, and at the end of the day the quality and performance will suffer as a result.
It's interesting times definately for this stuff, especially with the latest omap3 chips... In terms of tablets, I strongly believe Omap is king, hands down. In terms of general embedded devices, it's still more of a gray area, more dependant on the costs and features of given boards and systems along with requirements than with chipsets.
 

The Following User Says Thank You to jolouis For This Useful Post:
luca's Avatar
luca is offline luca
2008-12-15 , 21:25
Posts: 1,137 | Thanked: 402 times | Joined on Sep 2007 @ Catalunya
#30
Originally Posted by Khertan View Post
Prefer ARM proc for better consumption, as i don't want to carry a device of 3kg just to have more than 3 hours of power (x86 based device).
3 hours? I only get 2 hours max[*] from my tablet (but then, additional batteries are cheap and light)

[*]browsing, email, rss. With youtube and/or video and/or internet radio less than 1 hour.
 
Reply
Page 3 of 4 12 3 4

« Previous Thread | Next Thread »

 
Forum Jump


All times are GMT. The time now is 20:43.

Contact Us - Home - Archive - Top