Log in

View Full Version : XScale: Hit or Hype?


Jeff Kirvin
06-24-2002, 05:55 AM
<a href="http://www.writingonyourpalm.net/column020624.htm">http://www.writingonyourpalm.net/column020624.htm</a><br /><br />Despite some otherwise impressive features, performance on the new Toshiba e740 is a dud. Is this Toshiba's problem, or a sign that XScale-generation Pocket PCs won't live up to the hype?<br /><br />Read about it <a href="http://www.writingonyourpalm.net/column020624.htm">here</a>.

pt
06-24-2002, 07:33 AM
this was a great article. fact filled and pointed towards the questions that still need to be answered. nicely done.

Inaki C
06-24-2002, 09:24 AM
To answer your question about Xscale 'auto' speed tuning, the short answer is CPU does not tune itself. If software do nothing the core will keep runing at a constant speed.
Xscale provides two ways to change speed: the so called turbo mode and the dynamic speed setting. Turbo mode is an easy and efficient way to alternate between two speeds, by means of changing a bit in a configuration register. Dynamic speed setting means to choose a diferent speed by reprograming the core clocks. The later was possible also under StrongARM but a great deal of care was to be taken to not crash the machine. Now the procedure is much more simple and the range is larger because Xscale has the ability to change power supply dynamically.
Other than pure frequency changes a key factor in gaining performance is cache management. A *huge* improvement in speed may be achieved by careful use of Xscale powerful cache. Operating system may take benfit of it but fine cache tuning may be used in applications too.

Regarding multimedia applications, Xscale has what we could call a MMX instruction set, that is a way to work with typical multimedia code that deals with short size data in paralell operations and multiply and acc. ops.

Operating system may be tuned to take profit of speed tuning and cache management, and applications, at least the ones provided with the system, should be able to manage turbo speed seting and multimedia specific opcodes. Probably is not a good idea to let user applications to change speed at will, this feature should be managed by OS. Apps should merely suggest the OS to change speed (thread management is an issue here).

Applications made by third parties might or might not use specific Xscale features. Which is true is that the operating system itself could take a great benefit of Xscale.

Timothy Rapson
06-24-2002, 12:38 PM
Interesting info Inaki, and crisply stated Jeff.

You get right to the bottom line: that is that we are in a stew. No one has definitively stated why these are not working and so we don't really know how to fix them.

The biggest problem is that we don't have firm answers about whether the true potential of the X-Scale will ever be realized. As Jeff states, this leaves the Ipaq 3800 series, the one to buy. But, Ipaq almost certainly has planned to stop production of it to make way for the 3900 series, and they were probably planning on having the money from higher 3900 series prices.

The bottom line is that this is a huge opportunity for the Palm group. I expect that Palm itself will still mess it up. Sony is sitting real pretty. The word is that Dell is going to OEM a PDA. If they have an order in for 1.2 million PPC from Asustek, HTC, or Legend as is reported they may lose a ton of money. If they are OEMing a Palm OS 5 model designed by HandEra or Acer, they may be the big new winner in the Fall/Christmas season PDA race.

Thanks for a terrific summary article, Jeff.


RE: "So now begins the long, dark comparison shopping of the soul." Alas poor Douglas, I knew him Horatio.

captgoodhope
06-24-2002, 03:11 PM
I had an e740. With both card slots filled, the battery lasted maybe 40 minutes before the first warning. No joke.

Jason Dunn
06-24-2002, 04:36 PM
I had an e740. With both card slots filled, the battery lasted maybe 40 minutes before the first warning. No joke.

That's interesting. I had someone with an e740 do a battery test on it, and with the backlight set to the maximum and 802.11b turned off, it lasted an average of 4 hours and 15 minutes (over two tests). The iPAQ 3870 lasted 3 hours and 45 minutes. So from what I've seen, the battery life on the e740 is slightly better than the iPAQ 3870 - and I haven't heard all those doom and gloom about the iPAQ battery life (beyond the normal "I wish it lasted longer!")

captgoodhope
06-24-2002, 08:59 PM
Were there cards in the slots? I have the same cards in my ipaq with no problems.

Jason Dunn
06-24-2002, 10:22 PM
Were there cards in the slots? I have the same cards in my ipaq with no problems.

No, zero cards in the slots. Strickly a stand-alone screen power test. Go to www.scottandmichelle.net and look for his KEEPALIVE app. Remove all cards. Fully charge your unit, soft reset it, set the screen to max brightness, then run KEEPALIVE and set it to write to the log file every 10 minutes. See what kind of results you get.

DJR
06-25-2002, 11:13 PM
According to the MSNBC review: early results on the Toshiba e740 WITH WiFi ON are showing the 50% battery warning in less than an hour and shutoff battery life in under two hours (backlight on minimum setting). Hrm. Looks like you'd need to bulk it up with the extended battery pack accessory if you want it to be actually useful.