Need help adjusting memory timing

[genetix]

from ''Edited". No, I don't think so

The board automates the NB/HT. That might be too high. RAM you could get stable with simply voltage.

however, here's one comment of fellow from extremeclockers with same board as yours:

I have a wierd problem. For some reason MY Gigabyte GA-MA790X-UD4P motherboard with a OCZ Low Voltage 1066 4GB kit installed will give 2.05v to ram when "normal" is selected for ddr2 1.8 voltage (as listed in bios) If I increase the voltage by +.1 volts, the motherboard instead of giving true voltage of 2.15, applies less voltage of 1.95. From there the scaling is correct, as each +.1 increment is applied, voltage increases appropriately. However this motherboard should allow 1.8 volts (or 1.85 if slightly overvolted), but the least applied to ram is 1.95. I ended up Returning this ram after trying various bios without successfully being able to apply 1.8volts or close.

As you can see it is very low, but your 2.3v on it should in that scale just and just do the 2.15v on RAM, but I still suspect this to be too high voltages on NB/HT and drop of 5 V rail on PSU. Drop the NB to near 1.40v and HT to 1.26v. vCore is only stable up to 3,5Ghz at 1.40v depending. According to this source the NB/HT needs incredible low voltages. Here's couple screenshots of stable 3,679Ghz and 4,0Ghz:

http://forums.extremeoverclocking.co...1&d=1246930628
http://forums.extremeoverclocking.co...2&d=1246930628
http://forums.extremeoverclocking.co...3&d=1246930628
http://forums.extremeoverclocking.co...4&d=1246930628
http://forums.extremeoverclocking.co...5&d=1246930628
http://forums.extremeoverclocking.co...6&d=1246930948

(Although, I know G.Skill memory takes like insane volts at least what I am using on 1066Mhz yet I doubt it needs more than actual 2.24v.)

[retsamradassaT]

Should I leave my memory voltage at 2.2 or 2.3?

Where do I adjust HT voltage? I don't see that option anywhere in my BIOS. If it's the HT link, then mine is at 2000MHz

Basically, my BIOS looks like this:
http://gskill.us/forum/showthread.php?t=1129 (taken from the sticky)


Edited: looks like I'm still getting errors in my prime95 test. My memory voltage was at 2.2V, CPU voltage @ 1.47, NB Voltage 1.4

More edit: tested my memory @ 2.3V and it still failed, should I go back to using 400MHz? looks like 533MHz is unstable or should I just test this thing with 4GB?


Benchmark result looks pretty good with these voltage:

but it sucks how it's unstable when it comes to testing =(

[genetix]

Keep the CPU at 1.4v, verify that NB is at least 1.40-1.45v by your OS. BIOS lies these values check sensors pages on new Everest or the one you're using or AMD Overdrive tools there's got to be some tool to print out valid voltages.

That copy speed is so damn high we would need somehow to balance them together. Write speed again is way too low where most of the values in your BIOS affects. Something isn't correct still there. I assume this has something to do with board voltages still. Drop the tRC to 24T and I think this tWR still bugging here. tWR would need on these speeds 12.6T minimum, but then again your Write speeds are like near nothing which speaks completely against it. I'm still not too sure on the drive strength's they are a lot above normal all should be 1.00x. Assuming that means there should be a lot of room for tweaking.

[retsamradassaT]

CPU is at 1.47V (verified by CPU-Z). No idea where to check NB, since I have trouble finding it on Everest, and AMD doesn't tell me either, the CPU Voltage on AMD Overdrive is way off too, so I'm guessing AMD overdrive is unreliable at this point.

Anyways, here's the new benchmark with your new recommended settings:



Also, if you noticed, this is my BIOS (img taken from sticky, so it's not my actual settings):
http://i40.tinypic.com/4hfgcz.jpg

and noticed how there's NB Volt Control and CPU NB VID Control, which NB should I adjust when I am asked to adjust my NB?


Edited: I just memtested my memories, and it appears that there were a lot of errors.
Don't know what happened between then and now. It seems that it was fine a little while ago.

[genetix]

Yeah, errors attends to appear as you know when you tighten too much of something in too low voltage or perhaps voltage has little affect, if still around 2.3V. In what test exactly the errors started to appear?

CPU NB VID Control is North bridge stabilize volt for CPU in this case very close to L3 cache stability I assume. I have no experience on this. NB volt is the volt controls the rest as usual. Always start from what is the lowest volt you have on BIOS would probably be best advice or that's how I would start and see where they affect.

* NB voltage control = +0.1V
* SB/HT volatge control = +0.1V
* NB/PCIe/PLL voltage control = +0.1V
* DDR VTT voltage control = Normal
* CPU NB VID control = +0.300V
* CPU VID control = +0.250V

This setup you see above is not same deal, but it's same processor earlier model working at 3,9Ghz with 1.55V CPU. Although the CPU is in actual voltage of 1.405V not on 1.55V. So you should keep it just 1.4V as said above messages. Setup is from 'Gigabyte GA-MA790FXT-UD5P' where: CPU: 3.9Ghz @ 1.55V, Multiplier = 19.5, Bus Speed = 200, Memory Speed and Timings = 2 x 1GB, 1066Mhz 5-6-6-17-2T. Increased NB/HT Frequencies way beyond where we are.

Your CPU should be around 1.405V at CPU-Z.



-edit #01-

That Read/Write/Copy speed difference/balance still bothers me and probably is the problem. I understand that we use huge multiplier on memory so that only increases the memory copy speeds and this is just bad as there is absolute no gain in this. any values on BIOS only affects the entire set total speeds as there's no latencies we can which would specify jump on read only, however, on write and tWR on write would be ideal increase however it's not possible.

The thing is the memory should be working exactly style of speed as your L3 cache is, but it's nowhere near even functioning there which is probable temporary cause of errors because it will become unstable either by read or write. Internal copy speeds might be too fast and generate errors on memory tests or system BSODs, but this doesn't help the cause at anyway because it's the read/write speeds which needs to stability at any cost.

It's a pitty we didn't have everest on start to get Write/Copy speeds. Would of seen the errors coming mile away back on page 2 of this topic. There's the turn point on page 2 where we started to tweak copy speed instead of working on actual realistic speed to use the memory.


-edit #02-

btw, as long the MemTest is not stable you can be very sure system won't be stable. Take the tRRD back 1T. try increasing tWTR/tRTW 1T, if you can or if there is we should isolate any Read/write latencies separate try to balance them and drop the generic speed (if it's possible under on so high FSB : DRAM ratio). 10.5-11GB/s copying speeds only hurts the system performance as well as brings stability issues.

[retsamradassaT]

Okay, I've done everything you've asked. My BIOS does not have tRTW, so I cannot adjust that. tRRD, the lowest I can go to is 4T. Now, back to more testing.

Edited: i memtested it for nearly 3 hours, it passed 1 test with no errors. Will try to leave it on overnight today and post results tomorrow.

[genetix]

Well, you only need like tests 4-7 from memtest passing ones. Don't think over testing has any beneficial. Rest are againt the voltages of north bridge (can the controller be flexible to transform data to memory understandable form from cpu) and vDIMM to maintain the data integrity while at memory.

[retsamradassaT]

Alright, I have been testing it with prime with different NB voltage each time and it seems that when the NB voltage is set to 1.3V (+0.3V) it gets the best result when prime testing. It fails after 5 min (only 1 core fails), the rest just fails within the first 2-3 min and it's more than 1 core failing.

Btw, my memory voltage is set to 2.3V
and here's a pic of the current result with the current settings:




Edited: it's definitely passing memtest.. it's just not passing the prime test. Think it's a voltage problem with my CPU or something?

[retsamradassaT]

Gskill tech?

[GSKILL TECH]

Increase your DDR VTT to 1.4V. Keep NB Voltage under 1.3V if you're only running two sticks. See if that will keep it stable for Prime95.


GSKILL TECH

[genetix]

DDR VTT to 1.4v? lol, don't think so, but indeed NB volts might have affect if the PSU rail would be affected by too high voltage. However, you cannot post this system below 1.39-1.44v in anywhere near stable trust me on that one.