borgqueenx
Well-Known Member
i only really hope the media player will get lots of optimalisations. So we can get about 100 hours of music out of the pandora.
Change In Mhz Ratings..
- Thread starter Exophase
- Start date
greendots said:
for one thing because ram has much less headroom for OCing, you can consider them as already being "overclocked".
mobile cpus are often underclocked for saving on battery power which does not matter much because RAM is already so slow compared to the CPU that in most applications (not emulators) halving the cpu and bus clock will hardly reduce performance.
because the ram is so slow, manufacturers try to push it as much as they can.
one thing that is slow in RAM is recharging the capacitors (little voltage tanks that holds the 0s and 1s) this takes a large amount of time, reading them empties them, and they also leak.
if you change the clock they wont be charged long enough (too fast), or they wont be refilled quickly enough as they leak (too slow), and your 0 will turn into a 1 or vice-versa depending on what is around it.
there's also other issues with DRAM but retaining data is one of the big problems.
Last edited by a moderator:
Squidge
Certified Guru
No, if you just overclock, you'll still get guaranteed > 100Khours.j.pickens said:
However, if you then use OPP5-6 to make the overclocking more stable, then the hours will drop depending on how long the device spends in OPP5-6.
Last edited by a moderator:
Bosbeetle
Terminally lost
Exophase said:
true I was just thinking about having to replace my pandora with pandora 2 within a couple of years and I wouldnt like that Id rather enjoy it a bit longer. Maybe I am to conservative but the technology outruns my wallet all the time.
I just get scared by things as the iphone that has 3 versions already each one quicker than the other. I see all these people being dissatisfied by their iphone 3G and wanting a 3Gs. 3 versions in 3 years is a bit much.
But probably im idle in me wanting a bit of technology that lasts
Last edited by a moderator:
Bosbeetle
Terminally lost
Hopefully if the pandora is a huge succes the team will stay the pandora team and not turn into a big corporate beast like every other computer compagny
Enverex
Active Member
I'm curious to know if the RAM will overclock somewhat (with the SGX with it) even if it's just a little. Does anyone have any way to actually try it at this point? I'm guessing the RAM is set to some sort of 3x multiplier to the speed that the SGX runs at?
fusion_power
Advanced Member
I think we can be happy if we get at least 25 hours of Music. AFAIK, the many (marketing) Power-Saving features of the Pandora's OMAP are not fully working at the moment, MWeston knows more about this.borgqueenx said:
Well, I hope the average Software will be at least so optimized, that I don't need higher clockings than for equal Software onto the GP2X (example: Picodrive is fast enough @ 200MHz onto the GP2X, even with Virtua Racing, I would be confused if the emulator would need 500MHz to run at the same speed onto the Pandora)
Well, and over all I think (power hungry) Overclocking is not the best philosophy for an battery-driven Device.
If 25% overclocking means 25% (ore more) less battery time, I think twice what I overclock and what not. (Quake 3 with All Details, including lightmap is a good candidate for Overclocking if it will run full-speed then )
Last edited by a moderator:
Exophase
Nothing good will ever come of Exophase.
The clocking on OMAP3530 is complicated and I keep having to go over it.. but I'll try to grasp it a little better here.
There are five PLLs, DPLL1 through DPLL5, that generate clocks internally from SYS_CLK. DPLL3 sources SYS_CLK, and generates CORE_CLK, which goes to L3_CLK, the level 3 memory subsystem (which the SDRAM interfaces on). There's a 1 or 2 divider from DPLL3 to L3_CLK. SGX FCLK (the rate it operates at internally) is sourced from either CORE_CLK / { 3, 4, 6 } or from CM_96M_FCLK which comes from DPLL4 and as the name suggests is a fixed 96MHz.
So if I understand correctly, we're looking at a DPLL3 input of 333MHz, generating CORE_CLK at 333MHz, L3_CLK at 333MHz or 166MHz, and SGX FCLK at either 55.5MHz, 83.25MHz, or 111MHz.
Obviously the DDR won't run at 333MHz, so the question becomes one of whether or not we can change DPLL3/CORE_CLK. Although not tied to the clock speeds of the CPU or DSP, this still regulates a ton of things. I'm going to guess that since some of them have MHz ratings in their names, that it really has to be 333MHz. For instance, there are 96MHz and 48MHz outputs, I imagine these go to the SD card and USB amongst other things. I don't think you'd want to try overclocking these things.
On the other hand, maybe you could alter DPLL4 to change the SGX clock speed..
There are five PLLs, DPLL1 through DPLL5, that generate clocks internally from SYS_CLK. DPLL3 sources SYS_CLK, and generates CORE_CLK, which goes to L3_CLK, the level 3 memory subsystem (which the SDRAM interfaces on). There's a 1 or 2 divider from DPLL3 to L3_CLK. SGX FCLK (the rate it operates at internally) is sourced from either CORE_CLK / { 3, 4, 6 } or from CM_96M_FCLK which comes from DPLL4 and as the name suggests is a fixed 96MHz.
So if I understand correctly, we're looking at a DPLL3 input of 333MHz, generating CORE_CLK at 333MHz, L3_CLK at 333MHz or 166MHz, and SGX FCLK at either 55.5MHz, 83.25MHz, or 111MHz.
Obviously the DDR won't run at 333MHz, so the question becomes one of whether or not we can change DPLL3/CORE_CLK. Although not tied to the clock speeds of the CPU or DSP, this still regulates a ton of things. I'm going to guess that since some of them have MHz ratings in their names, that it really has to be 333MHz. For instance, there are 96MHz and 48MHz outputs, I imagine these go to the SD card and USB amongst other things. I don't think you'd want to try overclocking these things.
On the other hand, maybe you could alter DPLL4 to change the SGX clock speed..
MWeston
Internal Development
I've seen the SGX core clocked at 166MHz based on CORE_CLK/3: 500/3
I think the core wasn't doing much at the time but you can physically set it to 166 or even 200MHz. I'm not saying it won't go nuts but it will be tested by someone sometime soon enough I'm sure.
Oh, and upon further inspection, the 720MHz specified chips have to be ordered as a separate part number. They have the suffix 72 at the end of the part and they cost a lot more! I'm asking my favorite TI hardware engineer what this will mean for us Pandora folks in the future since we've had no trouble with these speeds already.
I think the core wasn't doing much at the time but you can physically set it to 166 or even 200MHz. I'm not saying it won't go nuts but it will be tested by someone sometime soon enough I'm sure.
Oh, and upon further inspection, the 720MHz specified chips have to be ordered as a separate part number. They have the suffix 72 at the end of the part and they cost a lot more! I'm asking my favorite TI hardware engineer what this will mean for us Pandora folks in the future since we've had no trouble with these speeds already.
r00tw00t
Member
So if I understand correctly it isn't 720MHz but 600MHz. The newer produced OMAP3530 has a slightly higher MHz? I'm confused
Exophase
Nothing good will ever come of Exophase.
Sigh, unfortunately MWeston is right - they buried this information down quite deep and didn't bring it up in the clocking section in the datasheet either.
I'm concerned that this will mean that the 600MHz ones are going to be the bin losers that won't even clock at 720MHz at all, putting us in a worse situation than where we started. I wonder if later Pandora shipments are going to be using the 72 core? The difference in price is $8.35 at 1K quantity. It's still about $10 cheaper than the chips were when first listed over a year ago.
I'm concerned that this will mean that the 600MHz ones are going to be the bin losers that won't even clock at 720MHz at all, putting us in a worse situation than where we started. I wonder if later Pandora shipments are going to be using the 72 core? The difference in price is $8.35 at 1K quantity. It's still about $10 cheaper than the chips were when first listed over a year ago.
r00tw00t
Member
I hope I'm misunderstanding you- but are you saying: we can't overclock the Pandora to over 600MHz?
senorgomez
Member
- Joined
- May 3, 2009
- Messages
- 150
Has the Pandora team already placed their 4000 chip order? Is it possible we might get a last minute upgrade? I ordered later on, it would be awesome to have this new 72 core.
I see no reason why they wouldn't use this slightly different chip if it is the same model.
What is the relationship between pandora and TI anyways? I thought they were giving them some slack in order to get some of these chips out there... on the other hand the pandora has been delayed pretty long and the Cortex8 ship sailed with the Iphone GS. Any 'first adopter karma' would have been used up by now I think. Hope they still smile upon the our project.
I see no reason why they wouldn't use this slightly different chip if it is the same model.
What is the relationship between pandora and TI anyways? I thought they were giving them some slack in order to get some of these chips out there... on the other hand the pandora has been delayed pretty long and the Cortex8 ship sailed with the Iphone GS. Any 'first adopter karma' would have been used up by now I think. Hope they still smile upon the our project.
Blue Ion
Member
Fixed for you.r00tw00t said:
Careful with what you say. No one has said nothing solid.
Once the OP team receives the first 100, they will tell us what we have. Heck, if the 100 that suffered from the burn test can overclock, then it won't be an issue, otherwise...
As an additional test and final test, it could be interesting to do an overclocking test: test how far can the first 100 pandoras go with overclock (or every pandora for that matter) and finally put in the web something like, "Every pandora should overclock above X and their maximum is around Y"
Phawx said:
They were overclocking with dev boards that were done months ago. Perhaps with this change, TI will start cherry picking the chips and send the weaker ones to the ones that ordered the 600 mhz ones.
Last edited by a moderator:
Squidge
Certified Guru
Like lots of other companies, it's probably the result of just another test. The "72" chips will simply be guaranteed to run at that frequency. The 600Mhz ones used in the Pandora have never been guaranteed to run any higher, but in dev boards they have run upto 900Mhz with no problems.
I'm going to guess that they'll be exactly the same die, the only difference being the additional test at the higher frequency.
Obviously, no one can say "All Pandoras will overclock to 900Mhz", we can only say what is guaranteed from the manufacturer, so we only say "The manufacturer specification is 600Mhz, but you should be able to overclock. No guarantees though."
I see a lot of this at work with speeds, temperature ranges and even built in flash/ram sizes. We bought 2,500 SoCs recently with embedded 256KB/16KB guaranteed flash/ram, but they all actually had 128KB RAM which usually cost $2 more (so $5,000 over all). We also saved $10,000 on a purchase of 10,000 other such smaller micros as a newer part number now only guarantees operation upto 85 deg C rather than 125 deg C.
I'm going to guess that they'll be exactly the same die, the only difference being the additional test at the higher frequency.
Obviously, no one can say "All Pandoras will overclock to 900Mhz", we can only say what is guaranteed from the manufacturer, so we only say "The manufacturer specification is 600Mhz, but you should be able to overclock. No guarantees though."
I see a lot of this at work with speeds, temperature ranges and even built in flash/ram sizes. We bought 2,500 SoCs recently with embedded 256KB/16KB guaranteed flash/ram, but they all actually had 128KB RAM which usually cost $2 more (so $5,000 over all). We also saved $10,000 on a purchase of 10,000 other such smaller micros as a newer part number now only guarantees operation upto 85 deg C rather than 125 deg C.
greendots
Its finally here!
- Joined
- Mar 11, 2008
- Messages
- 610
Is that (72) SOC the same circuit size as the one we have or smaller? Do we have access to the clock speed of all of these things on ours? Maybe this is just like buying an overclocked video card: they test cards that preform well at a certain clock above normal and sell higher as overclocked, the ones that fail are sold as is.
AFIAK craig said all the parts have been ordered so not much chance for an exchange.
AFIAK craig said all the parts have been ordered so not much chance for an exchange.
MWeston said:I've seen the SGX core clocked at 166MHz based on CORE_CLK/3: 500/3
I think the core wasn't doing much at the time but you can physically set it to 166 or even 200MHz. I'm not saying it won't go nuts but it will be tested by someone sometime soon enough I'm sure.
Oh, and upon further inspection, the 720MHz specified chips have to be ordered as a separate part number. They have the suffix 72 at the end of the part and they cost a lot more! I'm asking my favorite TI hardware engineer what this will mean for us Pandora folks in the future since we've had no trouble with these speeds already.
what will be interesting is will it make a significant difference to clock it at 166 or even 200Mhz.
how big and efficient is the texture cache?
I worked on a PSP game where down clocking the cpu, gpu, and bus down to 66Mhz cpu / 33Mhz bus from 333/111Mhz would drop the frame rate by only 10 to 20%, the memory being the major bottleneck.
it gives a big difference in battery life tho, cpus don't power down much while waiting for RAM.
Last edited by a moderator:
Vorporeal
Yes, no, I, this is.
Even if the non-72 versions are the ones that failed to run stably at 720MHz, it's unlikely that first-batchers will have any problems with that. All the chips should have already been ordered and delivered, and its unlikely that TI would have sold us failed ones as a "rule" before they officially split the SKUs.
