Printable View
Simple, they are trying to assist the user with a high performance board. Your right about it being elaborate even though the things you mention are everyday stuff, but show me another board that has the features or ability to scream like this board does.Quote:
Originally Posted by hovo73
The fact is it doesn't exist.
btw, if you change ram on any mobo without unplugging or switching of the PSU you run the risk of frying your ram.
For those who did not know about the electromigration effect ...
http://www.cadence.com/whitepapers/4...gration_WP.pdf
It's something like that when you apply a very high current on a bulk wire , the wire resistance will make it become very hot , once the temperature of the bulk wire becomes hot , it's resistance increase , and continues this
high current density -> high temp -> even higher current density --> even higher temp ................ LOOP ...
Until the bulk wire is melted ... :banana:
As the pdf file explain , VOLTAGE itself does not cause electromigration effect , the HIGH CURRENT DENSITY and HIGH TEMP did ...
If these UTT chip is killed by electromigration effect , that means even this chip can handle the high voltage , let's say even at 5.0V ... IT can not handle the high current density and high temperature followed by this high voltage for a long period until any conducting wire in it is melted ~~~
I was referring to this thread:Quote:
Originally Posted by hovo73
http://xtremesystems.org/forums/showthread.php?t=49750
Quote:
oskar: so if user directly switch the vdimm source from 5v rail but use actual vdimm value under 3.3v,
then that would cause serious heat on Mosfets. That 's why the factory default vdimm source is set from 3.3v rail.
**(OPB: SO, if you are not going to do vdimm over 3.3v for 24/7, I and Oskar would not suggest you to switch vdimm source to 5v rail to get that heat on your MOS)
My point exactly :rolleyes:Quote:
Originally Posted by hovo73
--- hovo ---
Because there are circuits that are powered via 5VSB, which only goes away by AC removal. And that has been the safe way to do things for many years.
--- wtz54321 ---
Wrong conclusion in your last paragraph bro ;) You still have oxide breakdown possibilities at high voltages, which has nothing to do with electromigration. The thinner the gate oxide thickness, the less V it can handle.
Peace :toast:
didn't read through this last page too thoroughly, but think about this.
back when bh-5 was out everyone and his m8 bought it and turned their noses up at all ch-5 with its higher ras to cas timings. NO one on these forums liked it, EVERYone laughed at it mainly as things like twinmos 256mb bh-5 sticks were all of £22 2 years ago. no one was paying £20 for ch-5 sticks.
so most people overvolting mem back then were using bh-5 and not ch-5. i had mainly bh-5, 2/3 sticks of ch-5 after prices skyrocketed and towards end of bh-5 ch-5 started to run ras/cas 2 at high voltage but i, and most people didn't really run them for long or that hard.
look at p4's, xp's, ath 64's, personally my top voltage limits i've used since t-birds haven't dropped much, but i don't think i'd put 2.2v through my 0.09 ath 64's, though i did on my newcastles. smaller process is by design not going to react to voltage in the same way and WILL ALMOST CERTAINLY not last as long under same voltage as same chip in a larger die size.
the first and, IMO only thing to gather together info wise is , outta what died who had ch-5 and who had bh-5. its completely unsurprising to me that ch-5 would die quicker or simply not run at the same high voltages that bh-5 would.
the way that some/most boards get reduced voltages from a line is simply , for want of a better word, pulling the extra voltage out as heat before sending the new voltage to the memory. each 0.1v you drop will drop a crap load of heat into the mosfet where its done. so dropping from 5v line down to 4v isn't too bad, dropping and extra 0.8v down to 3.2v starts making that heat extremely high and as it doesn't have a cpu type heatsink on that mosfet well, who knows, that mosfet could definately overheat, wether that could result in some possibly extra high voltage going to memory or simply board dying i don't know, more likely that later though i would think. anyway thats why its dangerous to run low voltages off the 5v jumper.
i'm ordering some more bh-5 this weekend, mainly as this ch-5 utt simply won't do ras/cas 2 over 210-215Mhz or so whatever voltage. i dunno why we aren't just all buying bh-5 garenteed stuff, its out there and costs very little extra. £14 more for a gig of twinmos with bh-5 rather than ch-5 in the uk. and ocz's bh-5 is piss cheap(considering some ram prices and the fact that their limited edition stuff before with the last of the bh-5 was like £200 in the uk, the new bh-5 is £130).
I am impressed that Oskar himself is not only looking into this, but letting folks in on his findings as well.
Heres to a quick resolution
m
does this mean that old school BH-5/6 is skill King? :)
Sure , the high voltage will cause breakdown issue , but if it's this case , the cpu will be dead as well because k8 cpu should have thinner gate oxide thickness compare to .17um BH5 or .15um CH5 ...Quote:
Originally Posted by EMC2
This theory would be acceptable, only the consensus I've seen is UTT-CH scales as high if not higher than UTT-BH when the higher voltages are used.Quote:
Originally Posted by drunkenmaster
Plus, your comparing apples to oranges with your CPU analogy. The delta between the BH (0.15nm) and CH (0.13nm) process is small. Where the P4/A-XP (0.130nm) to A64's (0.09nm) process, not to mention the type processes and transistors used is very large.
And, the same kind of anecdotal analysis can be used to refute your numbers of damaged modules. If damaged UTT-CH modules do out number damaged UTT-BH, than couldn't that simply be the result of more people owning the former rather than the later with this specific mobo?
I've been running off the 5V rail since day one, and haven't had a single problem. Even if the mobo fries my RAM - I'll finally have no excuse for not upgrading to a gig :lol:
Cherries & Peanuts comparison ;)Quote:
Originally Posted by wtz54321
There are different types of scaling and different types of oxides...
UUT DRAM & A64s use quite different processes and materials...
Also keep in mind that the core and the I/O are physically different in the A64...
/me places bets on TDDB effects for memory issues based on BigToe's posted info to date
Folks,
I don't know if this has come up before, but have any of you fried the DIMM slot (not in the literal sense, I mean have any of the DIMM slots been rendered unusable) ?
This happened to me when I used the 5V jumper for just about 3 days. Both my TCCD and BH-5 started giving memtest errors at stock. After a MoBo RMA, everything seems to be fine.
And I have a good PSU, the Bluestorm AX500-A, so I can rule out the PSU. The failure occured after an overnight memtest run at 3.3V.
3.3v is a bit low for using the 5v rail....did you cool the vdimm mosfets?Quote:
Originally Posted by Super Nade
Absolutely! I had 1x92mm fan full speed on the 4V FET and 1x80mm fan blowing over my RAM. Both were cool to the touch. Even if 3.3V is a bit low, why should I encounter problem after I use the 4V option?Quote:
Originally Posted by stealth17
Till this matter is cleared up, I am sticking with my OCZ booster, with both TCCD or BH-5.
dad 5v rail? mosfets dont produce as clean of power off 5v on your board...Quote:
Originally Posted by Super Nade
who knows?
Yes, I did - one dead orange, one yellow - both closest to the CPU socket. Managed to do this by changing the 5v jumper back to 3.3v according to ALL instructions, except that I also cleared cmos simultaneously (and it was nowhere said I shouldn't do it). Afterwards, the slots were dead, but the remaining ones clocked as high as before. Complete mystics to me.Quote:
Originally Posted by Super Nade
Man, that sucks. I could not find any information on your particular case. Help anybody?Quote:
Originally Posted by mrlobber
---mrlobber---
Could you define "dead" in this case please?
Have you tried TCCD mem since then in those slots?
If not have you tried booting with 1 stick, setting timings slack and set FSB to low freq (200 or less) and then retrying DC?
Have you tried another CPU?
Peace :toast:
Please allow me to answer for him :)Quote:
Originally Posted by EMC2
By dead I/he mean(s) defective. Both TCCD and BH-5 show memtest errors on stock settings. TCCD was run with the 4V jumper in the disabled position. Reflashed multiple times (exact procedure as given by RGone/Angry) to no avail. Both TCCD and BH-5 were 24 hrs memtest stable on another board (ASUS A8V). Tried with a fresh 512 Mb stick of Corsair v1.1 BH-5. Same result.
I have a new board after I RMA'd that one. I have not tried using the 4V option. I'll be getting a booster for that. Untill I see a definitive answer, I won't use the 4V option.
Bumping this 1 back up please post to my comments on page 1
What voltage were you running your nb at ?
Default Voltage on the NB.....
Its a Problem with the 5 volt jumper......DFI have admitted this
Ditch this board while you can.......
Also there is some juicy information about to come out about the DFI boards....i cant say anything just yet until i get the authors permission....but OH MY GOD are DFI users going to cry and jump ship.
Everyone deserves a second chance. Lets see them come clean and fix the issue. I wonder how Abit and Sapphire (soon) is able to provide high vdimms without jumpers and switches and stuff like that.
Quote:
Originally Posted by topboy
please do keep us posted. kinda glad I bought the Abit now
Same here i have moved to the ABIT And it is a much much better product.