Printable View
my pc overclock stable super pi memtest prime95 at 255x10 but more than that give erros in all programs,test 8 memtest sometimes give error from 258 to 265mhz,test5 only give error more than 265,i think this could be cause by this problem,so making the mod is the only way to find iff overclock more and more stable.
my cpu can do 2800mhz(xp-120) at 1.475v but give errors too(255x11)
This explains the behavior I am seeing with my Redlines. Passes test 5 no problem but test 8 will fail at random intervals during a 100 pass run. Looking forward to the fix for this.
--- blander ---
Short answer - two completely different critters based on what I know of George's Maximizer... their only commonality being they help you to OC your memory higher ;) The only thing the LBTC does is either control, augment, or replace the actual Vtt voltage on the MB and must have its connections soldered in (test clips are fine for testing or benching if careful, but too easily dislodged for "real use" IMO).
--- Portugal ---
To tell for sure, drop your CPU multi down to 8... but if you can run T5 for a long time (hour) at 265 without any errrors and you get errors in T8 above 258 and errors in windows and things like Spi32M, OCCT, etc. with a lower CPU multi, then the answer most likely is yes (take it as a given that you have your timings properly sorted out). Prime95 I can't speak too, as I don't use it usually.
--- Q4S ---
What happens during Spi32M or OCCT? T5 speeds or T8 speeds?
Peace :toast:
BTW... 3.6V BIOS gave 255M Spi32M tonight
so,here is the results from tests at diferent multipliers:
264mhzx8 no errors in test 8(tested 1hour)
264mhzx9 no errors in test 8(tested 1hour)
264mhzx10 9 errors in test 8(less than 30minutes) here is the pic:
http://img300.imageshack.us/img300/6577/test89ex.th.jpg
i see the errors is in diferent adresses and mb in all loops,so what thinking about that?
I just finished testing and concluded that the VTT fluctuations is limiting the memory OC.
For the testing I Used SuperPi32, Sp2004 & Memtest.
I was hitting a wall at 265 with any voltage. Errors in memtest 8 and fails SuperPi 32M
After I did the VTT mod and raised the voltage .02 it now runs clean at 265 :woot: and passes all tests without errors. I am now testing at 267.
I know the vmod is just a band-aid so I am looking forward to the fix.
Ive been having issues with my BH5.. Passes test 5 fine at 260-265 but errors out test 8, occt and 32m superpi in a bad way.. It doesnt seem to affect my TCC5 much though.. This same BH5 runs great and doesnt show this when running in my 250gb mobo..
Looking foward to your fix EMC2.. :)
I don't want something that I plug into anything. I want to be able to solder it to components on the mobo. I hate plugs. I solder wires to everything, the back of my power headers on my mobo and my videocard power header, sometimes directly to capacitors or chips, depending on the application.
I support the solder only installation of it.
Just a few random items to add.
The 9173 datasheet includes some weasel-words about how the customer is "encouraged" to duplex this part to reduce power consumption in dual-channel DDR SDRAM applications (like the NF4). As a former applications manager for a semiconductor company, I can tell you that "encouraged" nearly always means "we're out of margin - do this if you want it to work". The NF4 board design probably would have included two of these puppies if RT's advice had been trusted.
The SOP-8 package version of the 9173 has the same Rds(on) as the TO-263 package, which means RT is using the same die size for the two parts (but current limiting the SOP-8 for thermal reasons). In a package the size of the TO-263 they could easily fit a MOSFET with half the Rds(on) of the SOP-8. Looks like they'd rather sell you two parts than use a bigger die. Cheesy. :rolleyes:
ChiefTech/Fortune is working on a pin-compatible replacement for the RT9173A, but it's not sampling yet (FS8863B). Might be worth following.
sluggo
So do we think just piggybacking a second 9173A would be helpful?
you know i seen that on the datasheet, so this makes me wonder will the fluctuations still occur out of dual channel? personally i dont see much differance between the two for power consumption but it might be worth a look.
that would explain why memory overclocks are higher/more stable in single channel mode then dual channel right there haha.Quote:
Originally Posted by tennvols_69
also ive read many post that say something like i got two sticks of ram that do 250 2-2-2-5 each but together my system run stable even at stock settigns of 200mhz and im thinking this might have something to do with it on the more unlucky boards
________
Honda jazz specifications
btw i forgot to mention emc2 you are my freaking hero man i think i will try to improve my knowlage of electronics. hey you need to be pretty darn good at math to become an electrical engineer? anyways im real gratefull of your work just ask if there is something that i can ever help you with
________
Zx14 vs hayabusa
It's a worthwhile question. I don't know the answer, but it seems to me that since dual-channel operation will increase the duty cycle of this part (which, aparrently, was designed only for single-channel applications), the part willl be running hotter than it would be in single-channel. If it's current limit is appropriate for single-channel use, then it should also be fine for dual-channel, as only one channel is active at a time. If, however, the part was only characterized for single-channel and it's running hotter in dual-channel, then the thermal and current limiting data on the datasheet may not be applicable for the NF4 design.Quote:
Originally Posted by tennvols_69
I'm not sure I'd be willing to try it with my own board. It looks like it might work, but RT recommends having the two parts drive alternate lines rather than using them in parallel. I'm guessing they recommend doing it that way for a reason.Quote:
Originally Posted by uwackme
sluggo
subscribe :clap:
More random info. The early version of the datasheet for the 9173 called for a 100uF cap on the 2.5V line, while later versions call for 470uF. The later spec also adds a 1000uF output cap to the original 10uF (!) recommendation.
I looked on the board to see what DFI did about RichTek's changes to the reference design. They went 'em one better - DFI added two 470's and two 1000's. I guess they had some concerns even back then.
sluggo
Wow huge hand for EMC2!!! :clap: :clap: :clap: :clap: Thanks much bro! Why arent you an XIP yet?!?!?
Now, what do I have to do to build one of these gadgets? Maybe a scematic? Also, were can we get those PCBs you have or can we get them from you?
THANK YOU! WOOOOOT FOR EMC2!!!! :woot: :woot:
(I wount tell you I love you as last time you said "But if you ever try to kiss me again, I'm gonna use a LEAD PIPE on your head!") :rofl:
ok i'm gunna play devils advocate.
what rev of dfi board is that ?
how many different DFI boards did you test ?
no board is perfect they all got nasty drops in vdimm and vcore some worse than others.
but you said you only got like 210mhz on stock DFI board with UTT, that man is just the worst mobo i have ever seen. no one i have seen was limited by that much on a dfi nf4..
now that being said....
omfg!!!! what a brain bender of a thread!!! thats some seriously nice work youve done!
ok back to play devil's advocate.
not all the boards are like that i really believe you got a massively crap dfi baord. that hated that utt and took some pretty massive vdroops.
but then again WOW!! i just cant say that enough WOW...
My board performs similarly. I have a much lower end scope and I can watch where it drops as it switches from off to on. It's not just his board.
WOW, just wow.. exellent engineering work there emc2 :clap:
I will do some testing, the following week and see if this could explain my problems with S-Pi, it wont run with high mem speed (bh-utt) whereas it is totally memtest5 stable (i have never bothered with test8)..
Will it be possible to buy the solution in europe/norway when your testings are complete?
Id also like to ask one (probably) stupid question :p: what does java mean?
Java = Coffee bean
--- Portugal --- You have a several week old PM ;)
--- Craig --- If I do, it will be however one wants to do it... flexibility is always a given ;)
--- Uwackme --- As I said before, it might work... question would be the differences in the Vos between the chips.
--- tennvols --- Go look in the original thread I referenced... a single 1x512MB doesn't create the problem. The loading is half ;)
--- brandinb --- Yes, lots of math is a requisite... geometry, trig, calculus, differential equations, partial differential equations, etc. are required for any engineering degree. But, you can learn a lot about electronics without it... just limits you in some areas without it.
--- looks at Stealth and brandinb and taps keyboard with pipe --- :p:
--- sluggo --- No, in DC, both "channels" are active at a time... it's a 128-bit memory bus configuration rather than a 64-bit configuration ;) Regarding the output capacitance... it helps with *short term* fluctuations, supplies current until the regulator catches up... but does nothing for long term fluctuations like the nF4 has.
--- Lestat --- the limitation depends on CPU/memory/timings and board. Regarding if it's only one board... go back and read the references that I gave to Michal in post #24 of this thread and my own comments playing advocate ;)
--- un0kai --- if it goes to sale, yes... BTW, there is no such thing as a stupid Q when it is an honest quest for knowledge ;) In fact, such questions are smart ones :)
Oh... and regarding Java:
Quote:
"It is by will alone I set my mind in motion. It is by the juice of Java that thoughts acquire speed, the lips acquire stains, the stains become a warning. It is by will alone I set my mind in motion."
Fantastic stuff EMC2, takes me back to my uni days! Wish i had a scope here to do some playing myself. Keep up the great work!
and wath about trying to ad output mosfets to get more current?
do anyone know if dfi have fixed this vtt problem on the new revison (dr-extra) board? My guess is it is still there?
and btw. i love your avatar craig588
and emc2: regarding the quest for knowledge, I could not have said it better myself...
First up a massive hand to EMC2, brilliant stuff, well done! :toast:
This is the kind of overclocking stuff I like to see, people putting to use electronic engineering knowledge to improve overclockablility (not like the "ADD MORE CAPS!" approach :stick: ).
Second, are you planning to make and sell this device or just release the a circuit digram or both? Sign me up for a copy of the diagram if you release it. :cool:
Edit: I had a little thought, I wouldnt mind paying $5-10 for the diagram and a little guide, just to justify your time and effort spent on it. Doesnt sound like much, but given this forums prominence in the world of overclocking, you could be sell a few thousand copies of it easily.