Log in

View Full Version : First Loox performance benchmark results


Jason Dunn
06-27-2002, 09:24 PM
Today is the day I wish I were Alex Kac. Why? Because he has a Loox Pocket PC for testing purposes, and I haven't laid my hands on one since CeBit. Alex ran VOBenchmark on his Loox and his iPaq 3800 series. The results? Very interesting indeed:<br /><br />• "The Loox was twice as fast in Integer and 25% faster in FP.<br />• The Loox was 4x faster in BitBlt graphic operations and 50% in StretchBlt operations.<br />• The Loox was twice as fast in Ellipse, Rectangle, and RoundedRect operations.<br />• The Loox was faster by about 20% in Memory allocation and fill.<br />• HOWEVER, the Loox was very slow on Memory Move operations - .36 in Memory Move compared to .86 for my iPaq....<br />• It was 30% faster in text operation."<br /><br />Another message from Alex on the Loox:<br /><br /><i>"Well, I just installed the final release of the Loox ROM. As promised, things are noticeable faster – I think. Its hard to tell if it just my imagination or what, but things feel a lot faster than the previous rom (10 builds ago). Little things like the ink keeping up with me perfectly at 100% zoom in notes (my iPaq doesn’t keep up with my writing – the Loox didn’t before, but now it does). IE working faster."</i>

Paragon
06-27-2002, 09:36 PM
Damn cool! I understand that the screen in the one he has isn't the one that will be in it when released... I'm very curious to know more about that.

I'm a bit slow, forgive me... memory move operation, is that when it has to move data from a storage card to RAM?

Dave

heov
06-27-2002, 09:38 PM
So, how does this compare with the Toshiba e740? Did the e740 score this well, or not. If not, then I guess Toshiba screwed up wiht xScale...

Also, as Chris DeHerra (sp) mentioned, memory move is probably due to the slower bus speed.

Jason Dunn
06-27-2002, 10:02 PM
So, how does this compare with the Toshiba e740? Did the e740 score this well, or not. If not, then I guess Toshiba screwed up wiht xScale...

I believe Alex said that the Loox was faster than the Toshiba at everything except the memory moves - same speed.

donkthemagicllama
06-28-2002, 12:00 AM
Memory move is just moving data from one location in memory to another. Not from a storage card or anything, just moving things around in RAM. The test should be dependant on memory bus speed and width.

Everyone seemed to overlook this on the Toshiba e740, since all the graphics related benchmarks were so good. But, in real life, the memory move speed is perhaps the most critical aspect of multi-media programs.

The amount of memory I/O required for example, to decompress a MPEG movie, is huge. If you're doing this sort of thing on a slow memory bus, it doesn't matter how fast you can write to the screen, it's going to be slow.

I think this is the reason XScales are underperforming. It's not due to code not being optimized for XScale, or them not taking advantage of a graphics chip (although this might help MPEG4 playback depending on architecture, it wouldn't help 3D games or the like that require memory moves to texture objects etc.) it's because they seem to perform dismally in memory move operations.

Just my $.02

jdhill
06-28-2002, 12:02 AM
Jason,

I just finished jumping all over Steve Bush at Brighthand for using information from Pocket PC Thoughts on the X-Scale issues without properly attributing all of his quotes. Now, it would appear that you may have done the same thing.

It appears that you have taken Alex Kac's statement directly from a post on Brighthand's discussion boards. If so, you need to attribute the source.

And, as I mentioned in my posts on Brighthand, it is generally considered good form to ask the author's permission to re-use a quote. It's not clear if this was done or not.

Steve Bush later changed his story to properly indicate the source and to indicate changes he had made to the original quotes. You might consider doing the same.

alex_kac
06-28-2002, 02:02 AM
Actually, I emailed the quote to Jason directly and gave him permission to publish it. He and I had talked about XScale and the Loox in particular for quite a while and so I let him know of my findings.

I also posted it on Brighthand since the discussion was so heated there.

Let me also say, I am under NDA, but I'm allowed to talk about some things on the Loox.

alex_kac
06-28-2002, 02:03 AM
&lt;dupe>

jdhill
06-28-2002, 02:17 AM
Jason,

Sorry for jumping all over you about the attribution issue. It's important, and I wanted it to be clear that I hold Pocket PC Thoughts to the same journalistic standards as I do Brighthand.

digital-doc
06-28-2002, 04:53 AM
Today is the day I wish I were Alex Kac. Why? Because he has a Loox Pocket PC for testing purposes, and I haven't laid my hands on one since CeBit. Alex ran VOBenchmark on his Loox and his iPaq 3800 series. The results? Very interesting indeed:

• "The Loox was twice as fast in Integer and 25% faster in FP.
• The Loox was 4x faster in BitBlt graphic operations and 50% in StretchBlt operations.
• The Loox was twice as fast in Ellipse, Rectangle, and RoundedRect operations.
• The Loox was faster by about 20% in Memory allocation and fill.
• HOWEVER, the Loox was very slow on Memory Move operations - .36 in Memory Move compared to .86 for my iPaq....
• It was 30% faster in text operation."

Another message from Alex on the Loox:

"Well, I just installed the final release of the Loox ROM. As promised, things are noticeable faster – I think. Its hard to tell if it just my imagination or what, but things feel a lot faster than the previous rom (10 builds ago). Little things like the ink keeping up with me perfectly at 100% zoom in notes (my iPaq doesn’t keep up with my writing – the Loox didn’t before, but now it does). IE working faster."

If Fujitsu has a "ROM," I guess there is hope because Redmond sure has no flash updates or SDK kit in the offing---specifically for X-scale.

alex_kac
06-28-2002, 04:58 AM
FlashROM

Jason Dunn
06-28-2002, 06:26 AM
Sorry for jumping all over you about the attribution issue. It's important, and I wanted it to be clear that I hold Pocket PC Thoughts to the same journalistic standards as I do Brighthand.

I appreciate you holding me accountable, but please know that I have my own particular set of standards, and they don't include taking quotes out of context or publishing information unless I believe it to be public.

Nice to see this resolved itself without me getting involved though. :D

farnold
06-28-2002, 06:53 AM
I appreciate the "standard" of this board and follow the discussions regularly. The news about the LOOX performance is really great - and it's what I expected from SIEMENS and FUJITSU. Did anyone compare performance and so forth LOOX, e740 and 3970 yet?

ctitanic
06-28-2002, 12:37 PM
So, how does this compare with the Toshiba e740? Did the e740 score this well, or not. If not, then I guess Toshiba screwed up wiht xScale...

I believe Alex said that the Loox was faster than the Toshiba at everything except the memory moves - same speed.

That's correct. And many people have been saying that one of the cause of the poor performance of e740 while playing Multimedia application is the the memory move which is in both (loox and e740) worse than the rest of ARM devices. So for those waiting for LOOX, check Toshiba's forums and you will see there what you will get with LOOX (it could be a little faster, but just a little bit, no more, and always worse than any ARM). :evil:

fmcpherson
06-28-2002, 06:09 PM
I think this is good news for E740 owners because it suggests that some things can be done in the software to gain performance improvements. All we need to do is have Toshiba work with Microsoft to to implement those improvements and for Toshiba to create a ROM upgrade for E740 owners.

The only wild card here is if the hardware designs between the E740 and LOOX are different.