OK, i am from Germany, i really don't understand what EMC2 has done there Could someone explain this a little bit for me? What do I have to do? :confused: And what gets better than before?
Printable View
OK, i am from Germany, i really don't understand what EMC2 has done there Could someone explain this a little bit for me? What do I have to do? :confused: And what gets better than before?
Any news?
Just looked at this thread and it is VERY interesting. Wish I can buy one.. :(
The memory busses on this board use active termination. The dimms must be able to pull up AND down in order to transfer data. The termination voltage is called Vtt. The device on the motherboard that supplies Vtt can sometimes lose regulation and Vtt then droops as much as 200mV. When this happens, data can be corrupted.Quote:
Originally Posted by Rolf
EMC is working on a circuit that will force Vtt to track Vdimm more accurately and thus allow for more reliable operation and higher overclocks.
What you have to do is wait for EMC to say "it's done". What happens after that is unknown at this time.
Thanks, that sounds very very interesting! :fact:Quote:
Originally Posted by sluggo
Wow EMC2, nice job you did there, how is the design of the circuit going? :)
OK, that's what I found today by chance:
http://www.micrel.com/product-info/SSTL-Term.shtml
This sounds very interesting. :) 5A would be really nice when running 4 RAM modules.
Any chance EMC could give the once over to the new DFI ATI Crossfire motherboard? just to see if Vtt is not an issue. Dont even want to consider buying it unless its clean of issues like this.
This looks very interesting!Quote:
Originally Posted by celemine1Gig
Any updates EMC2?
please emc2 speak to the masses!!!
curious to hear your update
________
Problems from zoloft
i couldnt wait any longer for the board that could.
found this baby:
http://pdfserv.maxim-ic.com/en/ds/MAX8554EVKIT.pdf
this board can be setup to run vtt from what i understand provide upwards of 8A and 550KHz switching freq. they want 60$ shipped in the usa.
i can begin to figure why they have such a wimpy part for vtt on this board.
3A is minimum for ONE DDR channel. it didnt take me long to find this out reading white papers. shame on DFI.
Where can I buy it?
just as an update to my previous post
i have 2x512 BH5 sticks
could not run stable at any speed before
even one stick seemed iffy
after adding parts to adjust the output volts (7K resistor and 10K pot)
running the board off the 12v line
using 200Khz switching freq
(rt part on nf4 apears to run a 1khz freq)
http://www.maxim-ic.com/quick_view2.cfm/qv_pk/4247
now @245 tight timings its rock solid.
I dont have any scope that would be useful here, my dmm shows no more than 20mv voltage swings. the output coil gets very hot, i have a feeling this is due to the length of the hookup wire. (about 6 inches).
This seems very interessting and all, but I'm not good enough at electronics to understand this. Is there any way I could use some of these stuff for my neo2 platinum of msi, my ram just isn't running at what it's supposed to be able of.
If someone could write me a guide... (I know it's lame not figuring it out myself, but I'm really a noob in electronics :()
Gosh, I can't wait for the solution here.
EMC2's last post on these forums was almost 1 year and two months ago. I'm tired of waiting for this mod however. Does anyone know anything about how it worked? To my understanding the base of the circuit was a difference or differential amplifier.
From one of EMC2's posts in another thread. (I was reading this wrong earlier, the Vref, Vref2 and Vtt aren't the labelling of the DIFFAMP but the labelling of the connections from the RT9173A).
http://xtremesystems.org/forums/atta...5&d=1121305069
Instead of having the diffamp connected to pin 4 and pin 5 for the inputs, connecting it to pin 1 and pin 5 would give us the desired results. Otherwise you need a different circuit where Vout = 2 x vref - vtt (and Vref = 1/2 Vmem, so using Vmem seems a bit easier).
However, for this circuit to provide 2.29v to pin 4 (EMC2's device's result in one of the graphs on the first page), Vtt would have had to drop to 0.51v which is unlikely. The graphs suggest that a difference in Vtt ofaround 90mV resulted in a 900mV shift in Vref input (a gain of 10). Meaning that the equation required is Vout = 10(Vref - Vtt) + Vref. (So for every mV that Vtt is out of range, Vref will be adjusted 10mV, hopefully resulting in Vtt tracking Vref more closely).
Edit: Supposedly a 'passive averager' circuit could be used. The output from the diffamp and Vmem can be averaged using two equal resistors. The resulting equation would be: Vout = (20(Vref - Vtt) + Vmem)/2, which happens to equal the target equation (Since Vmem = 2Vref). I can't find any evidence that EMC2's circuit works in this way, but the end result should be the same. This leaves one semiconductor device for the 'augmentation signal'.
Edit2: By the looks of things, most diffamps / opamps cannot provide more than a fraction of an amp. The 'augmentation signal' was pumping out almost 3A peak, so I'm currently looking at different methods of power regulation. It should be possible to use the output from the 'passive averager' as an input for the power regulation to track. If the power regulation method can output 3A is should average out with the Vtt output from the RT9173A, but another passive averager might be needed.