297 MHz?

Surac

I see many code samples and even working code using 297 MHz.
Why not 300. I fear I am missing the catch of this 297 MHz setting

Can someone clarify?

Cluso99

297MHz and 148.5MHz are VGA 1920x1080 sweet spots.

rogloh

Yes HDTV (1080i/p and 720p) use both 148.5MHz and 74.25MHz frequencies and we typically need the P2 running at 2x pixel clocks to do much video work like rendering any text (though you can still stream raw scan line data from hub memory at the 1x pixel clock). That's one reason why you'll often see 297MHz used.

Rayman

300 MHz works just fine too for 1080p on everything I've tested with....

Monitors are pretty forgiving these days...

Surac

Thank you all for explaining this to me

Best regards

evanh

Rayman wrote: »

Monitors are pretty forgiving these days...

They always have been.

Mickster

I see that the recommended frequency is still 180MHz but at what frequency does the P2 start to become flaky?

evanh

Depends on die temperature. If it's kept cool it'll run sweet at 360 MHz. 380 MHz will need refrigeration.

250 MHz will nominally be below 1 Watt on Vcore. I'd expect that to be easily kept in a runnable condition. That said, 250 MHz can exceed 2.5 W if certain resources are hammered at max bandwidth.

That's something I had intended to roughly map out but had forgotten about it ... off to bed now though and new week in the morning ...

Surac

So even the lowest HDMI pixel clock means using the P2 out out spec as we need min 251.75 MHz?

wmosscrop

evanh wrote: »

Rayman wrote: »

Monitors are pretty forgiving these days...

They always have been.

Early VGA monitors (1980-ish CRTs) were not forgiving. It was possible to damage them by using an unsupported refresh frequency.
These were analog devices and did not have any protection circuitry.

Rayman

The polarity of the sync signals used to be very important too...

evanh

Fixed frequency VGA monitors don't count. Those died a very long time ago. Most weren't even VGA either. You'd be classing the last 30 years as "these days"!

evanh

evanh wrote: »

Depends on die temperature. If it's kept cool it'll run sweet at 360 MHz. 380 MHz will need refrigeration.

250 MHz will nominally be below 1 Watt on Vcore. I'd expect that to be easily kept in a runnable condition. That said, 250 MHz can exceed 2.5 W if certain resources are hammered at max bandwidth.

I can achieve 400 MHz at 20 degC. Smartpin generating a simulated clock at sysclock/1000 to prove actual clock rate. It was at lowest idle power of 694 mW with 1.83 V. All cogs running this code:

program1
.loop
		waitx	#500
		jmp	#.loop

Hard to say how stable a more complex program would be. I think I've seen the I/O pins exhibits erratic relationships with code execution. Maybe once temperature is moved well below the PLL frequency limit, reliable operation can be expected even at 400 MHz.

I was going to map out a graph of the PLL frequency limit against temperature but my test setup for measuring and controlling temperature isn't really that accurate at this stage.

Cluso99

IIRC I found hard code failure at ~385MHz. Not a heat or power problem. Info on the SD Driver thread.

evanh

Did you refrigerate it?

Cluso99

No air temp around 20C

evanh

I did do some temperature recording of this a while back, and wrote about it. Only thing I remember though was I spent most effort with the revA chip instead of revB even though lots of us had a revB glob-top at that stage.

I'm guessing 10 degC would make 400 MHz reliable. But that gets tricky when the power is ramped up because the thermal conductivity is not great.

Surac

I have mounted a 10x10 mm dram cooler Heat sink on top. It is made from copper with the typical comb like features at the top. My P2 keeps cool to around 360 MHz all cogs pumping io pins

evanh