How much faster is the 1Ghz unit - what we know


I wonder if there will be any differences in max throughput rates on the SD slots.


Quick - someone send me a unit so I can find out. :)
 
Yeah, but the high end is pushed up a lot by the 32X one which is really peculiar, since it achieves superlinear scaling on the OMAP3530 too.
 
Looks more like 69% and 60% for the two games tested..
Sorry, I meant to say on the tests at stock speeds.
There are for the OMAP3530@600MHz vs DM3730@1GHz test, yeah. Is my table hard to understand?


53% figure is from memory bandwidth test, so it's not affected by CPU clock rate much.


32X one probably has a measurement error.

I wonder if there will be any differences in max throughput rates on the SD slots.


Quick - someone send me a unit so I can find out. :)
Preliminary tests show they are the same (maxing out at 17-18MB/s), it's exactly the same host controller after all.
 
Last edited by a moderator:
So at the same clocking like the old OMAP from the Pandora 1 (600 MHz) there is no speed improvement in the new chip? I mean the "per MHz efficiency" or how this is called.
 
So at the same clocking like the old OMAP from the Pandora 1 (600 MHz) there is no speed improvement in the new chip? I mean the "per MHz efficiency" or how this is called.

It's a little faster, by the looks of things maybe a few percent on average?
 
Well, I usually run the DM3730 Pandora at 1,2 GHz (which is smoother for webbrowsing) and I also had the feeling battery life was longer compared to the OMAP with 800 Mhz.
 
Well, I usually run the DM3730 Pandora at 1,2 GHz (which is smoother for webbrowsing) and I also had the feeling battery life was longer compared to the OMAP with 800 Mhz.

You use less power while the unit is doing nothing than while it's busy, so while not using 100% CPU time power consumption goes down. You use less power at lower clock speeds even if you use more CPU, in order to do the same amount of work I mean, but the combination of more idle time and a more efficient chip could be putting you ahead.


But I wouldn't expect 1.2GHz DM3730 at full CPU utilization to use less power than 800MHz OMAP3530 at full CPU utilization.
 
Yes, that's pretty much what I am thinking as well.


With 100% CPU usage this probably won't be the case.


But with a higher clockrate, the CPU is idling more, so it uses less power :)


I never made any measurements, it's just what I was experiencing based on the time left on the battery meter when doing the same things as before.
 
There should be a small reduction due to process shrink (65nm to 45nm), but not much, as they're already low consumption circuits.


The power management circuitry has been tweaked, not sure how but this table should give you an idea of the effect. Look at the core domain reduction, other reductions could be from the process shrink.


And, you can refer to this table for voltage supplied at different OPP (ya you know me) levels


I'm thinking a 600MHz clocked 3730 will last longer, a lot longer.
 
Additionally, if you're using a kernel that can put the CPU into a low power state (where a lot of the chip is clock gated) the smaller process should have lower leakage as well.
 
Any changes to Wifi faster, better, stronger? So I can stop using this mini dongle with a new 1GHz unit?!
 
Any changes to Wifi faster, better, stronger? So I can stop using this mini dongle with a new 1GHz unit?!

They're just changing the SoC (and if you have 256MB, the RAM. And if your unit is particularly old, the nubs and LCD cable). Wifi is the same. But since the wifi quality seems varies a lot from user to user, and therefore potentially unit to unit, you may get luckier..
 
@notaz


can you check with BBQ ?


its got an fps counter in-game.


the game is GPU and/or memory bound once the world is generated and loaded, CPU's only at 40-50% when the window is maximized @720Mhz.


I'm curious how much of a difference just the GPU and memory does.


just click local play and let the game sit for half a minute without inputs, should give a stable reading.
 
Isn't there a timed demo run mode with an FPS report in one of the Quake series?
 
Okay, I think I'm going to crack and upgrade my March 2009 preorder to one o' these babies. I intend to use it for development, so the extra horsepower will be quite useful.
 
Okay, I think I'm going to crack and upgrade my March 2009 preorder to one o' these babies. I intend to use it for development, so the extra horsepower will be quite useful.
Sure will.Ive done the same upgraded my second old preorder.Im no developer though.I will test out lots of software though.I didnt expect the 1ghz units to be available so soon.Sweeet
 
Wrath of Khan said:
I didnt expect the 1ghz units to be available so soon.Sweeet
Production is expected to happen in about 4 - 6 weeks. I think thats pretty fast. With only 150 boards and no big problems the run could be over in two weeks, making it a total of 6 - 8 weeks. Note it's only an estimation but hey cool ; D


soooo excited ; )
 
Last edited by a moderator:
Back
Top