possible soc updates.


don't know what it's idle power is. Apparently single threaded power draw is 800-1200 mW. http://www.anandtech...-z3770-tested/2
 Is sunspider multithreaded at all?

If you can really get 20x (or even 5x if your benchmark is multithreaded while your applications aren't) a pandora's performance on 1.2W, I'm impressed
I'm pretty sure his 20x quote was in reference to running at full blast.

-God Ginrai
 
don't know what it's idle power is. Apparently single threaded power draw is 800-1200 mW. http://www.anandtech...-z3770-tested/2
Is sunspider multithreaded at all?

If you can really get 20x (or even 5x if your benchmark is multithreaded while your applications aren't) a pandora's performance on 1.2W, I'm impressed
I went looking around for an answer - no luck.  I saw references to sunspider being single threaded and comments that it was multi-threaded.

If the Z3770 were getting those numbers running a single thread though - I'd be shocked and amazed too.
 
don't know what it's idle power is. Apparently single threaded power draw is 800-1200 mW. http://www.anandtech...-z3770-tested/2
 
Is sunspider multithreaded at all?


If you can really get 20x (or even 5x if your benchmark is multithreaded while your applications aren't) a pandora's performance on 1.2W, I'm impressed
I'm pretty sure his 20x quote was in reference to running at full blast.

-God Ginrai
That is why it is relevant to know if Sunspider is multi-threaded or not.  If it's single thread - and max thermal is 1.2 on a single thread full out, then yeah, it could be doing 17.7 times faster on 1.2W.  I find it hard to believe that.  I'm more inclined to believe that Sunspider is multi-threaded (meaning it would be pulling 2.5W) - but have no way to know at the moment.
 
4.5 pandora's worth of computing power at 1.2W sounds good, but could be completely negated by high idling power. (I do plenty of stuff that doesn't begin to stretch my pandora @ 800MHz)

Debate adjourned until we have more data then?
 
4.5 pandora's worth of computing power at 1.2W sounds good, but could be completely negated by high idling power. (I do plenty of stuff that doesn't begin to stretch my pandora @ 800MHz)

Debate adjourned until we have more data then?
Yup.
 
More benchmark results on Z3770.  

http://www.anandtech.com/show/7263/intel-teases-baytrail-performance-with-atom-z3770-cinebench-score

"Finally, this Baytrail SKU delivers similar performance to a mid-2GHz mobile Penryn based Core 2 Duo. Let that sink in for a moment - this tablet SoC delivers better multithreaded FP performance than a 2010 Macbook Air. "

Not that the world revolves on FP performance - but another tidbit.
FP as in Floating Point? I'm not so sure what's so special about the Macbook Air for it to get the special mention. o.o

Also, in the future, could you please refrain from double posting? Any additions can easily be added to your most recent post with the Edit function.

-God Ginrai
 
Last edited by a moderator:
Grench, that benchmark is ancient... it was run in 2010, and android's webkit has had some significant improvements since. My phone (samsung infuse: cortex A8 overclocked to 1.6Ghz, 512MB ram, android 4.3) runs sunspider in 2193ms. Or 2721ms at the stock 1.2Ghz. So that makes the atom ±6x faster than the 1Ghz Pandora in javascript-code (single-threaded. afaict, Webkit doesn't seem to effectively utilize multiple cores), or 2.5-3x clock-for-clock (the atom was supposedly running at 2.4Ghz)(plus a bit for my newer android-version...)
 
OK.  Everything is new enough here that there isn't anything that puts them in the same spreadsheet.

OMAP3530 Sunspider = 9397 ms.

http://processors.wiki.ti.com/index.php/Android_Comparative_Benchmarks

Intel Z3770 Sunspider = 531.6 ms.

http://www.legitreviews.com/intel-atom-processor-z3770-bay-trail-first-look-and-performance-testing_123335/3

OK, so 20 times faster was an exaggeration.  Just not much of one.
Are you aware those kind of comparisons are basically apples-to-orages? You are comparing not only CPUs, but the quality of code output of some JIT engine, which is very unlikely to output properly paired instructions that CortexA8 needs to show it's potential. It has been known that this particular benchmark generated poor code for ARM (don't know how it is now), and we don't know what version was tested on OMAP tn that wiki. Meanwhile Intel is known to optimize benchmark suites for their CPUs as much as it can, even to point of cheating (see AnTuTu), not to mention that those benchmarks often poorly represent things that people actually use.
 
Grench, that benchmark is ancient... it was run in 2010, and android's webkit has had some significant improvements since. My phone (samsung infuse: cortex A8 overclocked to 1.6Ghz, 512MB ram, android 4.3) runs sunspider in 2193ms. Or 2721ms at the stock 1.2Ghz. So that makes the atom ±6x faster than the 1Ghz Pandora in javascript-code (single-threaded. afaict, Webkit doesn't seem to effectively utilize multiple cores), or 2.5-3x clock-for-clock (the atom was supposedly running at 2.4Ghz)(plus a bit for my newer android-version...)
In his defense, he was using benchmarks for the OMAP3530 (which is the original Pandora SoC) for comparison, not the SoC used in the 1GHz Pandora.

-God Ginrai
 
Last edited by a moderator:
Wow - lots of critique.  The underlying fact is that I pulled what I could find for what is being presented as the same benchmark over different platforms.  Yes, the results could be loaded.  No idea if they are or to what extent.  If you have the time or inclination, you're welcome to run the Sunspider benchmark on your own Pandora using current builds to see what you get.  Mine is currently at home sick with a bad internal video cable.

If you can find better (more sound) benchmarks, by all means, post them up.  That was the closest to oranges and oranges as I could find.

Here is another article - better written than the other ones.  Not a lot of new information, but well organized.  Still using the Intel prototype tablets in their lab.  Apparently they had a media day.

http://us.hardware.info/reviews/4792/intel-atom-z3770-lbay-trailr-review-strong-competition-for-arm

Still nothing about idle power requirements.
 
@God Ginrai

Don't know why you say the Pandora should only be a gaming handheld, it has always been a Linux UMPC & Gaming handheld.  I made the point that it's not pushed enough as a UMPC and that most people only see it as a gaming handheld which is a problem. The fact that you see it only as a gaming handheld seems to emphasize that issue.

As for Bay Trail, we should see very soon what it's like, there are a number of tablets 8" - 10" coming out with it in the months to Christmas.

From everything Ive read, it should be priced aggressively with ARM offerings, offer better CPU performance than ARM offerings (S800 & T4), The HD4000 it uses looks to deliver around iPad 4 graphics performance - so not class leading , And finally good battery life.

But we shall see when reviews of the tablets come out.  If it's decent,  I think a Bay trail Pandora which will be able to run another OS along with Linux and Android would only help it's sales , support and boost it's cred as a UMPC
 
Last edited by a moderator:
Those mediatek SOCs make me want to make humping actions..... the power usage is going to be nuts tho
 
MediaTek is out of question, no Linux driver support.


(... and that video had one of the weirdest soundtracks ever for a tech-comparison :) )
 
[citation needed]

(Not that I'm accusing you of untruths, or anything - I'd just like to see your sources :) )
 OK.  Everything is new enough here that there isn't anything that puts them in the same spreadsheet.

OMAP3530 Sunspider = 9397 ms.

http://processors.wiki.ti.com/index.php/Android_Comparative_Benchmarks

Intel Z3770 Sunspider = 531.6 ms.

http://www.legitreviews.com/intel-atom-processor-z3770-bay-trail-first-look-and-performance-testing_123335/3

OK, so 20 times faster was an exaggeration.  Just not much of one.
Sunspider is a bullshit benchmark, and moreover performance of Javascript engines has improved tremendously over the last few years. If you ran the test using the same browser version you'd get a much different result. 1GHz Pandora would probably score under 3000, maybe under 2000. Again, it's a bullshit benchmark so it doesn't mean much.

Also, for those asking, it's a strictly single threaded test.
 
I just want to mention some points that aren't mentioned any more.

1: Pandora compatibility

There's so much optimized stuff for Pandora.

Don't throw it away.

2: WINE and Quemu.

Pandora successor might get a quad core CPU with 1.5-2 GHz.

Shouldn't this be enough to at least play some closed source games?
 
[citation needed]


(Not that I'm accusing you of untruths, or anything - I'd just like to see your sources :) )
 
OK.  Everything is new enough here that there isn't anything that puts them in the same spreadsheet.


OMAP3530 Sunspider = 9397 ms.

http://processors.wiki.ti.com/index.php/Android_Comparative_Benchmarks


Intel Z3770 Sunspider = 531.6 ms.

http://www.legitreviews.com/intel-atom-processor-z3770-bay-trail-first-look-and-performance-testing_123335/3


OK, so 20 times faster was an exaggeration.  Just not much of one.
Sunspider is a bullshit benchmark, and moreover performance of Javascript engines has improved tremendously over the last few years. If you ran the test using the same browser version you'd get a much different result. 1GHz Pandora would probably score under 3000, maybe under 2000. Again, it's a bullshit benchmark so it doesn't mean much.


Also, for those asking, it's a strictly single threaded test.
I was asked for benchmarks.  I posted the only comparison I could find that was even using benchmarks of the same name.

Please, take a look through the benchmarks in this article.  There are piles of benchmarks in it and it is the best written one I could find.

http://us.hardware.info/reviews/4792/intel-atom-z3770-lbay-trailr-review-strong-competition-for-arm

If you spot any benchmarks in there that you consider fair, that could also be done on a Pandora to get apples to apples, that would be very helpful.

*********************************

I just want to mention some points that aren't mentioned any more.

1: Pandora compatibility

There's so much optimized stuff for Pandora.

Don't throw it away.

2: WINE and Quemu.

Pandora successor might get a quad core CPU with 1.5-2 GHz.

Shouldn't this be enough to at least play some closed source games?
Consider the above.  Which is easier:

1.  Emulate a 5 year old ARM environment on X86.

2.  Emulate a 5 year old X86 environment on ARM.

?

*******************************

I hate how this forum blends multiple posts together into ONE message - those were meant to be two separate messages to two different sub-strings within this thread.  Sorry.  Row of **********s added for separation of conversations.
 
Last edited by a moderator:
[citation needed]


(Not that I'm accusing you of untruths, or anything - I'd just like to see your sources :) )
 
OK.  Everything is new enough here that there isn't anything that puts them in the same spreadsheet.


OMAP3530 Sunspider = 9397 ms.

http://processors.wiki.ti.com/index.php/Android_Comparative_Benchmarks


Intel Z3770 Sunspider = 531.6 ms.

http://www.legitreviews.com/intel-atom-processor-z3770-bay-trail-first-look-and-performance-testing_123335/3


OK, so 20 times faster was an exaggeration.  Just not much of one.
Sunspider is a bullshit benchmark, and moreover performance of Javascript engines has improved tremendously over the last few years. If you ran the test using the same browser version you'd get a much different result. 1GHz Pandora would probably score under 3000, maybe under 2000. Again, it's a bullshit benchmark so it doesn't mean much.


Also, for those asking, it's a strictly single threaded test.
I just re-read your last line there and it sunk in.  

That means that in a single threaded test that the Z3770 performed roughly 17.6 times faster than the original Pandora's SoC did 3 years ago?  

Wouldn't that would mean that, if using all 4 cores, it could be something on the order of 40 times faster on a multi-threaded scenario?  

Someone needs to run this (or other comparable) on a 1Ghz Pandora to see how much of this is benchmark BS and how much of it is real.  My Pandora is sick with a bad internal video cable - that I can't order until the external video cable vouchers come out to combine shipping.  Besides, it's only believable if someone else runs the benchmark.

Can anyone find something in this set that could be run directly on a 1Ghz Pandora to get close to lemons & limes?

http://us.hardware.info/reviews/4792/intel-atom-z3770-lbay-trailr-review-strong-competition-for-arm
 
Back
Top