Video ram speed MHz. Should only certain frequencies be used?

[Killrose]

I know that video ram speed MHz is usually due to a manufactures specs. But are there certain frequencies to avoid that may cause "intermod" with other devices within the computers bus/board components?

Why are speeds such as 143, 166, 183MHz chosen? And when overclocking should you try for the next speed step and not some random number?

 

[Superdoopercooper]

Intermodulation distortion??? I think that you may be looking WAY too far into the issue.

Think about the frequencies chosen: 166, 183, 267, etc.... It is easy to tell about the 166 and 267... they are all integer multiples of 33.3333MHz... which is the PCI bus frequency... or 1/2 the AGP 1x bus frequency... and 183Mhz is 5.5x 33.3333 Mhz.

It all has to do with multipliers... it is easiest to just use a 1/4, 1/2, or whole value of a system clock frequency... Of course with programs like PowerStrip, you can change this to whatever single MHz value you want. Doing that... I don't think you would ever see a case where 133, 134 and 135 work and 138-158, but 136, 137 dont... due to Intermod Dist. I think the limits are where the thing craps out at... and manufacturers use the 1/4, 1/2 or whole factor of 33Mhz or 66Mhz for ease of manufacturing... they pick a number that has a high enough speed to satisfy the consumer but low enough that their yeilds remain high.

They could sell the Radeon 234.4MHz and the Radeon 209.8Mhz... but the yeilds on those would be terrible, test time would be high, and the counsumer wouldn't want to pay for that type of spec'd performance.

EDIT: crappy spelling

 

[Killrose]

Alright. I just did'nt know if there were certain frequencies that should be avoided.

 

[AndyHui]

Those RAM speeds that you see are simply a function of the NS rating.

The MHz speed is simply derived by taking the number 1000 and dividing it by the NS rating, hence:

10ns = 100MHz
8.0ns = 125MHz
7.5ns = 133MHz
7.0ns = 143MHz
6.5ns = 153MHz
6.0ns = 166MHz
5.5ns = 183MHz
5.0ns = 200MHz

and so on.

 

[Peter]

Nonetheless, one should avoid using exact harmonics of common system frequencies as this might kick the EMI emission spectrum peaks above regulatory ceilings.

 

[Killrose]

Peter, that is what I was wondering. I own a 2-way Radio shop and we deal with intermod problems with radio frequencies all the time. I'm also wondering if when picking core/mem speeds (frequency) whether or not some of them cause problems between the two (intermod) causing trash (the spotting and flashing of textures we see while overclocking) which actually slows video performance.

I guess I could play around with different settings of core and mememory but then again I'm sure all boards are filtered differently and you probably could not come up with a one size fits all sort of chart.

My heads starting to hurt just thinking about it.

 

[Peter]

Such an optimization would have to be done during system design. However, most systems (outside major brand operations or the
industrial/telecomm/aerospace business I'm in) aren't exactly designed, slapped together would be the right word.

Thorough EMI spectrum analysis and optimization costs time and money (quite a lot of both actually), so don't expect any off-the-shelf
generic PC to be anywhere near good at it - let alone the homemade stuff. The slightest modification might mess everything up completely,
this isn't easy at all.

All the end user can do is flatten out the pin sharp peaks in an x86 box's spectrum by enabling Spread Spectrum Modulation, and if the
mainboard's BIOS offers that, disable the clock lines that go to unoccupied DIMM or PCI slots ("DIMM/PCI clk autodetect").

regards, Peter