I like Micron D9GMH, when they are binned properly (like in Crucial Ram kits) the results can be astonishing. I played a little with my 2 kits with GMH, you can see the results in my sig :)
Printable View
I like Micron D9GMH, when they are binned properly (like in Crucial Ram kits) the results can be astonishing. I played a little with my 2 kits with GMH, you can see the results in my sig :)
That's some incredible speed, what board were you using?
I've had very similar results with my GMH & GKX kits, in fact my CellShock GKX kit is a little better than my TG GMH kit, but I haven't pushed them anywhere near as hard as you.
EDIT - just saw valid - you did that on a P965 commando? Wow, my old P5B would never go over ~680 with my Corsair D9's What vNB did that require?
Awesome man! Seeing these kind of results tempts me to pump some more voltage through my kit. I know it can do 1300 (1302 is the highest I've taken it on 2.1v which wasn't stable) but because I don't have active cooling I'm kinda scared of killing it. There's no way I can afford another kit right now if anything bad happens! As a guess, I think it'd take at least 2.2v to get it stable, but with this kind of voltage, active cooling is a must, right? I'm not sure it's worth the extra noise from a fan. On the other hand, knowing there is significantly more performance which isn't being used, it's getting hard to resist.Quote:
Originally Posted by CryptiK
I could lower the CPU overclock a little and get the RAM to 12xx easily. DDR1200 will finish Pi 32M on only 1.9v set in BIOS. It's probably not worth losing the clock frequency for a little extra MHz on the RAM, though.
What to do...
Damn! :shocked:Quote:
Originally Posted by Monstru
Guys. I have a problem. I can pass Memtest86+ the whole night with no errors. I can pass IntelBurnTest using all available memory the whole night. I can pass Prime95 and OCCT with no errors. BUT! I have errors when using Memtest in Windows !
Help me, guys!
@ Dee - You do need some active cooling, even when running 1.9 - 2.0v I use it. You don't want to risk hurting your ram. I have 2 x 80mm fans running at 70% duty cycle and they are no louder than the 120mm fans I have for cpu (each running ~1300rpm) & case and the 140mm blowhole fan on low (~1000rpm).
I have to re-do the fans for the ram though they are too far away and are supported by cable ties as I've been swapping ram in & out a lot recently.
That's strange behaviour. I wouldn't worry about Windows MemTest too much if you're stable in everything else.Quote:
Originally Posted by WaterFlex
OK, I'ma rig a fan up in a sec. My sticks are running nice and cool, though.Quote:
Originally Posted by CryptiK
EDIT: Just remembered I have a ton of old 80mm's and a few low noise adapters. It'll probably not be very loud after all.
Weird. The other apps stress ram a lot more than Memtest in windows. Have you tried upping the vdimm one or two notches, or bumping up the vNB or vcore to see if it makes any difference?Quote:
Originally Posted by WaterFlex
That sounds like a plan, two 80mm fans at 5v should be fine and keep them cool. Unless of course you go into the 2.3 - 2.5v area for benching, then I'd recommend running the fans at full speed.Quote:
Originally Posted by Dee
I can officially say any D9 ICs you DON'T need active cooling for up to 2.1v as the ICs won't get that warm even under heavy load, you could even skate along with 2.2v and be ok. I would advise that active cooling is used though with voltages beyond 2.1v.
The CAS4 was done with Asus Commando, not bad for a chipset 2,5 years old, no? The voltage was pretty high, in the 1.7-1.8 area.Quote:
Originally Posted by CryptiK
The CAS5 was done on Biostar TPower I45, because the Commando would not go above 690MHz, not even in CAS 5.
every set is differentQuote:
Originally Posted by Ket
most of mine get average temp of 26c with 2 90mm fans
i got a set of d9gkx that is 29c with 2 fans 2.1v
not everyone has a lasergun to measure temps they better off with a fan no matter what the volts
Theres no point in adding something that isn't needed or simply won't increase any threshold. Fans for memory are only beneficial once ICs start to feel quite warm.
I played a little bit more with my beloved GMH, you can see the thread here :)
Arrgh! Sick!Quote:
Originally Posted by Monstru
:up: God bless Crucial.
500 FSB, 625 MHz
http://i266.photobucket.com/albums/i...m139vNBPL8.jpg
530 FSB, 635.7 MHz
http://i266.photobucket.com/albums/i...m139vNBPL8.jpg
You lie! Its not 530 FSB its 529.8! :p:
Guys... I can`t solve a problem which i described upper... I am being depressed...
Can you pass Memtest in Windows?
:(
oops lol :p: missed that :DQuote:
Originally Posted by Ket
Don't get depressed, did you try with more voltage on ram, vNB and cpu like I suggested? If so what was the result?Quote:
Originally Posted by WaterFlex
I have not tried to run it TBH I use Memtest in bootable CD form, Orthos, Linpack, and gaming to test system stability, so i cannot say whether or not I pass.
Yes, I tried like you suggested...unfortunately, no improvements...still 2-4 errors after 100 coverage.
If everything works ok, why worry about 4-5 errors only in WIN Memtest? :)
@Ket
I noticed the geil sticks in your sig. I just ordered a kit of the 2x1 pc8500 CL5 sticks for the family rig, any comment as to the ICs they might have or performance i might see?
should i tack on a spare hs?
Same amount or less errors after increasing the voltage? I honestly wouldn't worry about it, I have used that app before and didn't like it. I think you are stable if you pass Orthos 12 hrs+, Linpack, and all gaming/3d marks etc.Quote:
Originally Posted by WaterFlex
If you are determined to try and pass it, try raising the PL to 10 and retest, try dropping your cpu clock, or try dropping the ram clock (each done separately of course). Whichever one makes you stable, you know where to start looking.
Same amount, yeah.Quote:
Originally Posted by CryptiK
I have PL=8 at the current moment. When I set it to 9 i can pass Win Memtest with no errors. I can`t understand what the hell is going on.
I thought so, it's the MCH causing the errors. Set PL back to 8, and bump up just the vNB 2 notches. What happens?
It seems like your memory controller is on the verge of stability. Give it slightly more vNB :up:
The sticks run perfectly cool. You will probably hit ~1150 as top speed @ 5-5-5, 2.2v or so. As for the ICs.. no idea, some in-house geil stuff is all I know.Quote:
Originally Posted by Slovnaft
If so, it meen that Win MemTest is more hard for RAM than IntelBurnTest, OCCT, Prime95 and DOS MemTest86+.Quote:
Originally Posted by Oj101
Will try it when come home. I am at work at this moment.
Keep in touch my friends ;)
I don't know if the first three can be considered memory stability tests. I do not, however, have much faith in MemTest86+, as I've had systems which pass it but won't boot into Windows :up:Quote:
Originally Posted by WaterFlex
Me too. when you run MemTest from a CD or bootable ISO it doesn't take the Windows load into account. My usual method for testing if RAM is "stable enough" is Super Pi 32M test, OCCT "Mix", and Windows MemTest. If it ain't stable, crank the voltage up or loosen some timings. If it passes these tests but crashes in gaming, you're probably clocked just a little too high. Don't be greedy!
For me, the best stability tool is 24/7 usage :up:
thanks! that sounds like just what i need.Quote:
Originally Posted by Ket
although i'll probably try to hit some cl4 numbers.
Everest @ 550 FSB, 660 MHz CL 5. Not bad bandwidth and latency for a P45 :D
http://i266.photobucket.com/albums/i...SB660MHz_1.jpg
665.8 MHz screen (TeamXtreem 8500's at 2.3v real)
http://i266.photobucket.com/albums/i..._23vscreen.jpg
CryptiK
Wow! Nice job!
Sorry for the offtopic, can u give me your bios config for 550 fsb stable? ;)
Good work :)
Thanks :)Quote:
Originally Posted by WaterFlex
I can't remember the vNB & vFSB to more than 2 d.p but that should guide you.
MP = 7
FSB = 550
Ram Speed = 1320
CPU Voltage : 1.3425v
CPU PLL Voltage : 1.50v
FSB Termination Voltage : ~1.25v
DRAM Voltage : 2.3v
North Bridge Voltage : ~1.39v
South Bridge Voltage 1.5 : 1.5v
South Bridge Voltage 1.1 : 1.1v
These are not optimized at all and may be overkill, especially the vNB, I just guessed and benched.
Thanks :)Quote:
Originally Posted by Monstru
CryptiK
Thank you very much! Will try it soon! But, first of all:
I`ve just decided my problem about passing Win MemTest!!!
As I`ve already said before, I can pass IntelBurnTest, Prime95, OCCT and DOS MemTest86+ with no errors @2,06VDimm (real).
But! I need 2,10(real) to pass Win MemTest with no errors!
Why??? How? Win MemTest is the most killing Ram test?
Yes, seems to be true, I wouldn't have thought it. I am running 2 instances of it now with 890MB set in each one (I have found this to be the best way to run it if you have 2GB of ram, as it produces a higher NB temp meaning it's causing more stress).Quote:
Originally Posted by WaterFlex
NB temp is 50*C, and I had to bump up vdimm from 1.97 to 1.99v and I'm now able to run it for hours with no errors. It seems like a nice little program, I seriously underestimated it.
Good to hear you got it stable.
thanks, my friend!
By the way, I have 50*C NB temp under load too ;)
Dho.. :slap: I think I might start getting somewhere with OCing my memory now. Being so used to the 975X and various voltages being "tied", I completely forgot about needing to raise NB voltage. I'm going to hit the NB up with 1.7v and monitor how styable things are. My case has great airflow so I shouldn't need to slap a fan on the NB.
:rofl::ROTF:Quote:
Originally Posted by Ket
That's hilarious, oversight of the century :p: Get to it then, I wanna see some benches :D
Oh, 1.7v is a bit xtreme, 1.5 should suffice. I haven't used more than 1.45v on my board yet.
Working on 1143MHz now :p: I don't think my CPU will be handling 1200+ stable. Kinda impressed with 1143 tbh out of a cheapy E4400, I've heard many stories of E4400s crapping out with memory much beyond 900MHz. I think the CPU gods blessed me when buying this E4400.
Just been sprinkling some ket magical pixie dust on an old BIOS rev for the P5Q Pro that doesnt have that weird FSB enhancement option, and low and behold, it seems I can now do 1200MHz+ stable! Can't post any screenshots right now as I'm letting that system run OCCT v2a, over 20mins in stable so far. At least part of the key for 1200+ has been to set trfc to 42 and a performance level of 8. Will post screens when OCCT is finished.
Nice one! I haven't seen that FSB Enhancement option before, seems like all it enhances is user frustration :p: Awaiting screens :D
Hmmm guys. I have problems with passing Win Memtest again. Again 3-4 errors after 100% coverage.
It seems, Asus MIIF is a bad mobo, it can`t keep 1200mhz stable.
Yo i just got my geil sticks on Biostar I45, and they're unable to do stock speeds at 2.4vdimm and 1.5vnb. I'm hoping this is a BIOS support issue, as i've heard this board has some RAM support holes. Anyway you could sprinkle some of that magic pixie dust on the 710 BIOS from this board:rolleyes:? it's really starting to limit my qx :(Quote:
Originally Posted by Ket
Ket
It would be nice to know if you can pass Win Memtest.
What vdimm and vNB are you using when it gives you errors?Quote:
Originally Posted by WaterFlex
Bump up the VDIMM.
:rotf::rotf::rotf: Very true. That FSB enhacement option is supposed to provide better FSB clocking for lower end C2Ds. All it done for me was drastically cripple my memory and lower my max FSB.. I tried to work with it though to no avail. In a nutshell asus should either tell intel to go swivel on it and take that damn option out, or intel should release some kind of notes on how to make that option work for the benefit of the user. I literally tried everything without any good results.Quote:
Originally Posted by CryptiK
Useless:( I used to bumped it up to 2,2. The same situation.Quote:
Originally Posted by Dee
I think my motherboard :down:
Then maybe it's related to sub-timings?Quote:
Originally Posted by WaterFlex
that option I think is called something else in the P5Q-Deluxe BIOS versions >803. It's causing havoc with Q9450's and guys with them have to run 803 to be stable even at stock clocks. It's a ridiculous option and should be removed.Quote:
Originally Posted by Ket
Try higher vNB mate see what happens. I need 1.27 vNB for 445 FSB PL 8 stable, you may need 1.32 - 1.34v for 500 FSB PL 8.Quote:
Originally Posted by WaterFlex
CryptiKNope. I`m stable at 500FSB PL8 with 1,30 vNB. I pass Linpack, Prime95, OCCT.Quote:
Try higher vNB mate see what happens. I need 1.27 vNB for 445 FSB PL 8 stable, you may need 1.32 - 1.34v for 500 FSB PL 8.
I did 1,40 vNB and 2,1 vDimm and had the same result, several errors with Win MemTest.
When I go to 1000MHz 4-4-4-4 at 2,06V it`s all ok, I pass Win MemTest.
Dee
May be. But I don`t know what timings to set :shrug:Quote:
Then maybe it's related to sub-timings?
Lol! I`ve done 2,2 right now and passed 300 coverage with MemTest3.7 with no errors!
How can that be, guys? Stable with DOS MemTest86+, Linpack, OCCT, Prime95 with 2.06vDimm but need 2,2 to be stable with WinMemTest.
Unbelievable...
Will try now 2.15 vDimm on stableness
Done with 2.1v vDIMM, BIOS 0703 which I've modified. Going to see how much further these sticks will go :cool:
Good work, Ket. ;)
Max these sticks will boot is 1280 @ 2.1 5-5-5-15, haven't tested if their stable at that but pretty promising with just 2.1v.
Heres a screen with stuff boosted up a bit.
Nice clock guys!
Yep, nice clockage!
Had mine running 1230 or so earlier on (also 2.1v). Didn't stress test or anything but they seemed to work well. I was just testing out lower multi's and higher FSB with the P5Q-E. I get quite a nice boost in mem bandwidth using the lower multi's, only thing is I didn't realize until now that I could still run Speedstep on a multi other than the default 10x. :) That's why I always ran 10x.
I love my P5Q Pro in the gayest way now :p: Just gotta tweak things a bit for max performance.
24/7 setup 1.4Vnb, 2.1V ram Transcend PC6400 gmh
Not bad.. Feed that CPU some more juice though and crack 4GHz :up:
Nothing too major, just some tests with my Ballistix 8500 on my DFI LT P35 T2R. VDIMM was set to 2.47 in BIOS.
LOL.Quote:
Originally Posted by Ket
I feel the same way about CellShock. I'm such a fanboy now, it's disgusting...
the problem is that i give 1.3625V in bios for the cpu and i am getting 1.24 under load for ~3.8Ghz.Quote:
Originally Posted by Ket
So for 4Ghz i'll have to give ~1.4V for the cpu and i dont want that for 24/7 settings.
Damn you P5KC.:mad:
I have 1.4v for 24/7 on an E8600. Might knock it down a little at some point but no further than 4.25GHz @ 1.376 (heavy load).
Anyways. Just messing with different OC's and posting this for the sake of breaking 10k in Everest Read, Write & Copy. Not bad for 1.92 vDIMM (in BIOS, so possibly slightly under real vDIMM). DDR1200+ tests to follow at some point.
http://img521.imageshack.us/img521/9256/482yq0.gif
1.4v is perfectly fine still ;) when I do eventually get a wolfie I'll be running 1.42v 24\7, but I'll have to set around 1.45v in BIOS so its 1.42v under load. Wolfies aren't AS fragile as most people would think. In fact, with decent cooling theres no reason to think an OCd CPU wouldnt have the same lifespan if not longer than a CPU run stock with its stock HSF.
Nothing wrong with being a fanboy of a product thats genuinely good :toast: What makes me sick is the blind fanboys, particularly nvidia fanboys. Even when all the evidence is there to show theres an issue and its nothing to do with the user these blind fanboys come out in their masses and try slamming people. :shun: Anyway I think I'll be getting a XTC memory cooler in a couple days, thought I would finger my Cellshocks at the end of the 1hr OCCT stress test, and no :banana::banana::banana::banana: those heatspreaders were as hot as the NB! So they must of easily been around 50c :eek: Incredible to think they could be that hot, I'm only using 2.2v. (upped it a tad so I could run PL8 :D ) I'm certain with those heatspreaders off the sticks would run much cooler. If only I could get the farking things off..Quote:
Originally Posted by Dee
Heres a screenshot of my current settings, I think we can agree for just 3.2GHz thats some pretty insane memory bandwidth, almost sickening :p: I know my CPU core temps look kinda hot, its why I want to get that bloody IHS off too.. I tried everything it just won't budge. The next step would be some kind of solvent... but I'm not sure if there is something out there that would just corrode the rubber seal stuff and not the die\sillicon.
I told you they get hot ya n00b :p: I said people would need active cooling with them for sure, as CellShock heatspreaders actually work I think :DQuote:
Originally Posted by Ket
@ Dee - nice mate, see if you can crack 11K across the board :D
:banana::banana::banana::banana: yes, the temps are a bit crazy. Those are really idle temps?!!
I think it's possible with either 542 x 8 or maybe 578 x 7.5. Both land my CPU at 4.33GHz, so should be do-able. I can't find my fans to add some active cooling, so 1200+ is out of the question for now because the RAM takes more juice than I want to feed it. D9 scares me! :p:Quote:
Originally Posted by CryptiK
I never said The Cellshocks didn't get hot :p: I just said any D9 IC can handle up to 2.2v without active cooling, but active cooling is advised beyond 2.1v :cool:Quote:
Originally Posted by CryptiK
Yeah, load the cores get to 84c :down: Anyone know of any solvents I can pick up that will corrode the rubber sealent stuff and leave the CPU itself undamaged?Quote:
Originally Posted by Dee
84C? Wow, you're thrashing that thing good and proper.
Well done. :D
Its why I wanna get the damn IHS off :p: don't gemme wrong I don't care about how much heat the CPU is dealing with, I just want to significantly reduce temps so I can make the chip go faster... thus thrashing it more :D
Yeah well... there's overclocking, and then there's rape!
ROFLCOPTER
I prefer to look at it as unleasing the "true power" of the chip :p:
you need a new chip manQuote:
Originally Posted by Ket
time to get more power so u can see what d9gkx can do
Agreed. But I'm sure Ket would buy a new chip as soon as possible if lack of immediate funds wasn't a major factor.
Indeed I would order a new chip right now if I didnt have to buy\pay for so much stuff atm. Over the space of about 5hrs alone on the weekend I spent about £50 while I was out with mates :yawn:
I also want to unhood the E4400 simply just so I can find its very top foot-to-the-floor speed. Nows a good time for somebody to grab all 3 RAM kits I don't need :D
I was just kidding around :p: :cool: Active cooling is recommended, they get so hot when running Windows memtest or even the bootable cd memtest, hotter than just Orthos/OCCT etc. Same with the NB, that Windows memtest app makes my NB hit 50*C.Quote:
Originally Posted by Ket
I would try to avoid industrial solvent, acetone may be too harsh, and isopropyl alcohol (that I use to clean pretty much everything) may not be strong enough.
Cut it through with a razor blade, then gently heat the IHS with a lighter (not too hot) and remove it. Should come off I think the 4400's are just paste aren't they not the indium solder?
Yep supposed to be just TIM with the 4 series, so far I've tried a scalpel, stanley blade and a razor blade, none of them can get under the IHS, the black sealent really is on very thin.
My additional 2x1GB CellShocks "Red Devil" will be in my hands this week. It`s interesting, how much vDimm i need to be stable with 4x1GB :shrug:
A razor does actually work. Cover the blade about 2mm from the tip with tape so it doesn't slip in and use a fair amount of pressure. Start at the corners :up:
You won't need more vdimm to make 4 x 1GB stable, you'll need more vNB ;) The sticks just need what they need normally for stability, it's the load on the MCH that increases thereby needing more vNB.Quote:
Originally Posted by WaterFlex
Thank you for your reply ;) I`m glad to hear my Shock will not ask for more volts.Quote:
Originally Posted by CryptiK
Fast test with 4x1GB (2xCrucial pc5300 and 2xCellshock pc6400) 2.20volt 1225mhz
http://i317.photobucket.com/albums/m...n027-lille.jpg
Both kit should be with D9GMH and do around 1300mhz cl5 2.3volt, think 1333 is max i have seen.
Crucial pc5300 1284 5-5-4-8 2.3volt
http://i317.photobucket.com/albums/m...z-videotil.jpg
Turn that vcore down unless you want to kill that wolfie within a few months. 45nm chips can NOT handle voltages above 1.45v and live a long life. At 1.7v your wolfie will die within a year.
Yeah, that's pretty silly.
That Crucial PC5300 is great, btw. I'm sure you know this already!
Tried it. I attempted to use a disected Gillette Mach 3 blade, it just won't get under the IHS.Quote:
Originally Posted by Oj101
noxon
Like your Crucials! You reached high clock btw. Can you pass Windows MemTest with that voltages?
A year? A year from now the E8400 will be a fond memory :D
It's very strange that the blade won't go in. I had to use quite a lot of pressure (as in far more than felt comfortable) to get the blade in but once it was in it was easy going :shrug:
In a year a E8400 clocked @ 4GHz+ will still be a powerful monster of a chip :p: I have no interestin in Nehalem until it matures a bit, so once its out I may jump on about 8-12 months down the line. I might try one more time to remove the IHS, but its so damn difficult its nuts. When I tried using the disected razor I was using quite a bit of force but the blade just kept bumping the side of the IHS, and when I tried just a very slight angle approach to try squeezing it by it just ended up shaving sillicon.
Nice clocks and congrats on breaking 10sec SPI1M. That sure is an aggressive vcore.Quote:
Originally Posted by noxon
Maybe you need to use a Gillette fusion power blade :p: Sounds like you've had a serious crack at it, perhaps it's just not going to happen. Does the IHS just have such a tiny thickness of silicon between it and the substrate that a blade cannot fit between them?Quote:
Originally Posted by Ket
Yep thats pretty much it. Which is why I'm trying to find a solvent that will corrode it but leave the chip itself undamaged. I know that TIM needs replacing.
Give some pure acetone a go, but first test it on the substrate to make sure it doesnt eat it away. If it seems ok, use a cotton wool ball or cotton tip (ie: cue - tip, not sure what you guys call them) and soak an edge of it and see if it softens.
Acetone? What is that? White Spirit?
No not white spirit, we call that Metholated spirits here, Acetone is what is used in nail polish remover and gives it that characteristic strong smell.
Check it out:
http://en.wikipedia.org/wiki/Acetone
I use it to clean off Asus thermal paste, that really sticky stuff, and other things. It's pretty strong stuff so be careful. It can dissolve some things but Ive cleaned NB chips, CPU IHS's, heatsink bases etc with it. I also have cleaned circuit boards with it so it depends on what the substance is, some things it can dissolve others its ok with.
EDIT: NVM. CryptiK beat me to it.
Perhaps you should first try it on an old stick of RAM or graphics card first, something expendable.