New Video From Mweston


levi said:
Stephane Hockenhull said:
its not guaranteed to work at 800Mhz, its LIKELY to work above 600Mhz.

TI says: cpu works fine up to 600Mhz
On the datasheet TI say the CPU works at 720MHz (and the DSP at 520MHz). Surely they wouldn't say this if NEON is liable to lock up the system at such speeds?

That is a different serial number. As of yet no one know if its cherry picking or a revision of the chip afaik.
 
Last edited by a moderator:
Etinin said:
It's exactly the same processor.
They just raised the specs, since all chips can run at that speed.

Not all. They test the chips and bin the ones that run reliably at that speed at a higher price.
 
Last edited by a moderator:
No one has confirmed this yet... MW said he was going to talk to his friend at TI. He either has not or he cannot reveal the information.
 
greendots said:
No one has confirmed this yet... MW said he was going to talk to his friend at TI. He either has not or he cannot reveal the information.

+1

There are many ways to interpret what data we have. Some suggest future batches will get "worse" processors (though still up to spec, just not AS over spec as the first batch) while some suggest we will get better ones (because TI's manufacturing process has improved) or that there will be no change (if 720-cherry-picked ones leave no "failures" to bin, and there are no other changes). We simply don't know, at this stage.

MWeston is in a tough spot almost whatever happens. Let's say he asked his "guy" at TI, and his TI told him the truth. What does MW do next? Let's have a think. Depending on the answer, what happens if MW reports back to us (apart from potentially getting his mate in trouble)?

MW reports that future batches will have "failed 720" CPUs. Everyone panics that there is no future for the device and cancels their orders. The Pandora project fails.

MW reports that future batches will have 720 graded CPUs because OpenPandora will pay to ensure this. Everyone cancels their first batch orders so they can get a faster-rated 2nd batch machine instead - and the Pandora project fails.

MW reports that there will be no difference between the 720 rated version and the 600 rated version - the 600 rated one will overclock JUST as high as a 720 one (taking into account CPU to CPU differences). The word spreads out and TI cannot charge anyone extra for the 720 rated devices and get really pissed at the OpenPandora team. Plus 74% of us call him a liar and cancel our orders, the Pandora project fails.

Well, OK - I'm taking the mickey a bit but... I see potential dilemas for MWeston. I'll be glad if he CAN report back to us, but even if he can it will be chinese whispers - we know from a bloke who knows a bloke who works in a department at TI that is close to... not something to base a LOT of money on.

At the end of the day, the original spec, at 600MHz, isn't bad at all for a handheld - and software devs should have been developing with that speed (or ideally lower to save battery power) and NOT overclocking, as O/C is not guaranteed to work.
 
Last edited by a moderator:
I think the "binning" process will be exactly the same as Intel/AMD/Nvidia/ATI have been doing forever.

Realize that this can work in the reverse too. Sometimes to stock certain versions, even though the chip *could* be rated at 720, they would drop it down to 600 to have enough stock available for people wanting to pay that price.

There have been many cases where certain revisions of cards are software locked to a *lower* frequency or reduced number of pipelines to fill a pricing segment. Just like this:

http://reviews.digitaltrends.com/guide/13/radeon-9500-to-9700-mod-faq-a-follow-up

People were going crazy trying to find 9500's because they were a good deal cheaper then the 9700 and it was an easy mod to "open" them to their full potential.

So it works both ways. Just because Pandora may be buying 600 MHz binned chips in the future doesn't mean they were given the OK for 720, but had to be dropped to 600 to fill stock.

Sorry for repeating myself, just wanted to get my point across.
 
Phawx said:
Realize that this can work in the reverse too. Sometimes to stock certain versions, even though the chip *could* be rated at 720, they would drop it down to 600 to have enough stock available for people wanting to pay that price.

...

Sorry for repeating myself, just wanted to get my point across.

I'm not sure how that is the "reverse" of several possibilities I outlined, especially this one:

MW reports that there will be no difference between the 720 rated version and the 600 rated version - the 600 rated one will overclock JUST as high as a 720 one (taking into account CPU to CPU differences). The word spreads out and TI cannot charge anyone extra for the 720 rated devices and get really pissed at the OpenPandora team. Plus 74% of us call him a liar and cancel our orders, the Pandora project fails.

Which seems, to me, to cover very similar ground?
 
Last edited by a moderator:
I really dont care about cpu overclocking, the author said it will run 2x faster or higher with NEON. In crazy taxi for 1 second in-game it took 5 seconds in real life, so it will be half speed when pandora arrives wich is very much than expected...

Pd: Jet set radio and Sonic adventure in Pandora? Really??!!
 
guizm: In the video, the pandora is overclocked, so if it can't be overclocked with NEON, part of the speed-improvement may be eaten by that. Anyway, as drkIIIRaziel pointed out here, it's not like fpu->neon is the only optimization they are looking for, we will see this emu getting better and a lot faster. If it will be fullspeed? Probably noone knows, yet.
 
I'm curious to know if frame skipping could be used due to the way the game is being rendered.

If you consider a 2x speed increase + FS1 that's going to be quite close to PAL50 speeds.
 
craigix said:
I'm curious to know if frame skipping could be used due to the way the game is being rendered.

If you consider a 2x speed increase + FS1 that's going to be quite close to PAL50 speeds.
When I asked the same question many months ago, the response was that the graphics weren't the bottleneck. In otherwords, no, a frameskip would not bring the speed up a noticeable amount. Of course, a lot can change in a few months coding time.
 
Last edited by a moderator:
Would rendering at something lower than the native resolution of the Pandora help much, just out of curiousity? :p (Personally, I wouldn't mind running at half that, if the speed boost would be a good one.)
 
Prometheus said:
Would rendering at something lower than the native resolution of the Pandora help much, just out of curiousity? :p (Personally, I wouldn't mind running at half that, if the speed boost would be a good one.)
I don't know that much about emulators but...

I think the actual rendering of the graphics in 3D system emulators isn't that much of a problem, but rather the process of actually producing OpenGL instructions from the game, which is the same work regardless of screen resolution.
Although there might be some things in the graphics (complex shaders emulating something?) that slow it down.

Basically it depends whether the rending is actually much of a bottleneck.
If it's not, even reducing the screen res to 1x1 wouldn't help.
 
Last edited by a moderator:
I'm in with Bosbeetle. Did the final nubs get to Mweston yet?
 
WizardStan said:
craigix said:
I'm curious to know if frame skipping could be used due to the way the game is being rendered.

If you consider a 2x speed increase + FS1 that's going to be quite close to PAL50 speeds.
When I asked the same question many months ago, the response was that the graphics weren't the bottleneck. In otherwords, no, a frameskip would not bring the speed up a noticeable amount. Of course, a lot can change in a few months coding time.

Last I read (see above) graphics were not a bottleneck that could be optimised to improve speed, as graphics isn't where the speed issue/s lie.
 
Last edited by a moderator:
Back
Top