View Full Version : Is there anything that one can do to memory?
felinusz
06-22-2005, 03:14 PM
I mean in the way of physical modifications?
I have a half gig of Corsair BH-5 that peaks with 3.78V. I want more out of it, and have a feeling that it's a combination of noise, and PCB limitations that are holding it back from more gains at higher voltages. I have seen other people
s BH-5 gain at ~3.9-4V. My set of Mushkin BH-5 gained right to 3.9V. This Corsair stuff should have more in it.
I'm wondering if there are any physical component modifications that can help this memory out. By component modifications, I refer to soldering stuff to the memory PCB.
Mainly, I want to decrease noise at high voltages, and if neccessary, alter the PCB to decrease any DDR VDD "signal leakage" (resistances) at high voltages.
I can supply a scanned picture of both sides of the memory stick if neccesary.
Forgive me if this is an ignorant request, but I honestly think that memory modifications should be possible and possibly beneficial :).
dippyskoodlez
06-22-2005, 07:54 PM
I mean in the way of physical modifications?
I have a half gig of Corsair BH-5 that peaks with 3.78V. I want more out of it, and have a feeling that it's a combination of noise, and PCB limitations that are holding it back from more gains at higher voltages. I have seen other people
s BH-5 gain at ~3.9-4V. My set of Mushkin BH-5 gained right to 3.9V. This Corsair stuff should have more in it.
I'm wondering if there are any physical component modifications that can help this memory out. By component modifications, I refer to soldering stuff to the memory PCB.
Mainly, I want to decrease noise at high voltages, and if neccessary, alter the PCB to decrease any DDR VDD "signal leakage" (resistances) at high voltages.
I can supply a scanned picture of both sides of the memory stick if neccesary.
Forgive me if this is an ignorant request, but I honestly think that memory modifications should be possible and possibly beneficial :).
heard someone did a dimm cap mod once. never seen it though.
if your wanting to get around resistance, why not do what everyone else does with CPU and gpu? cooling! :D
its XS. lets see some uber cooling. :D
definatly sink it... hmm.. someone else will probably have more info. :)
ChongL
06-27-2005, 09:56 AM
that is actually a great idea...sink each individual chip and get an 80mm fan blowing over it. I pumped 4.06V through my BH5 to get a suicide screenie before, but I never felt safe going that high...
Let us know your results!!
Why not water cool them? I forget the guys name, but he water cooled his! He made these nice little blocks, very sexy. He is around here somewhere at XS, I think he actually left because he got a lot of crap for those memory watertblocks being useless. I would love to have my memory water cooled, that would be tight. You would have to have them on a seperate loop if you didnt want restriction and extra heat dump into your existing loop.
Or you could build a phase change system, make an evaporator to cover the memory, and put two evaporators on it!
Post a close up pic of it please....the clearer, the better (of course without any HS on it)...
digitalfrost
06-30-2005, 10:43 PM
http://www.kaltmacher.de/viewtopic.php?t=32848&postdays=0&postorder=asc&highlight=ram&start=0
ultron5's ram wakü :D
felinusz
07-01-2005, 11:36 PM
EMC2
Post a close up pic of it please....the clearer, the better (of course without any HS on it)...
My Corsauir BH-5 uses the same PCB as the RAM in these pictures:
http://kubokubo.jp/memory/corsair_cmx512-2700llpt/cmx512-2700llpt.html
http://kubokubo.jp/memory/corsair_pc3500_256mb/03.jpg
The PCB Revision Number is "50-00107 Rev A1" - AFAIK it is a Corsair PCB made for Winbond components.
If you need it, I can scan one of the sticks tommorrow and post up better pictures :).
Watercooling or Peltier Watercooling memory has been done before. The gains are typically signifigant, but not worth the massive expense to me. "Make a memory waterblock" is FAR easier said then done :(. I'm willing to do voltmod work on the sticks if it will help, and heatsink the memory ICs.
FrozenPC4Brain
07-02-2005, 07:30 AM
Ram Water Blocks, in FACT are fairly easy to build.
When I'm ready I'll post a how to.
I've now seen two RAM Water blocks built the hard way, I tried and found an easy SUPER-FLOW system for this. I'll show it of soon. (This month, maybe a couple weeks. Have "Other life" and stuff to finish on Multi Loop First water cooled, 64 bit I'm creating).
Have a GREAT weekend.
alexio
07-02-2005, 07:42 AM
http://www.kaltmacher.de/viewtopic.php?t=32848&postdays=0&postorder=asc&highlight=ram&start=0
ultron5's ram wakü :D
Aircooling is more efficient than this, looks very well manufactured but the design is just crap. If I had the tools I could do much better.
Infact I'm getting the tools this vacation so you'll see some cool waterblocks from me. The RAM waterblock is very high on my list of to do things.
felinusz
07-02-2005, 11:06 AM
If someone makes well designed, installable with good mounting pressure, pure copper, 3/8" ID barbed (because 1/2" is seriously pointless overkill for memory watercooling, and the larger tubing would make routing up there a real hassle), not ridiculously overpriced memory waterblocks, I will buy two of them and write a review for you :). A double-sized 80mm radiator and a tiny little pump would be sufficient.
However, I don't think that all of the above criteria will ever be met, mostly just the price factor - I have faith in your guy's ability to make an amazing waterblock :).
But, I don't think too many people are willing to spend ~$30+ USD on each waterblock when you need two of them, and will only see ~10 MHz gains from their use. And if they are aluminum, or not mounteable with pressure, they're semi-useless :(. Just my two cents on the memory watercooling issue, I would love to see inexpensive developments.
In the meantime, if anyone knows who did that cap mod to their BH-5, or how I could lower the noise or increase the voltage tolerance through a voltmod with my sticks, I would love to try and replicate it and see if I get any gains :).
L33T M33P
Sticking extra capacitors on increases the charge storage capacity of the whole converter for a given voltage. Translated that means more charge can be drained from the capacitors for the same level of voltage drop, or conversely less voltage drop for a given transfer of charge...
At an extremely high VDIMM voltage I should think that a Cap mod would have some signifigant impact on the memory - with a hefty frequency overclock and enormous (3.75V+) active voltage overvolt we're looking at some serious strain on a memory PCB specc'ed for 2.6V.
felinusz
07-02-2005, 07:07 PM
EMC2,
http://i23.photobucket.com/albums/b394/felinusz/CorsairXMSSide2.jpg
http://i23.photobucket.com/albums/b394/felinusz/CorsairXMSSide1.jpg
Here are some high-detail scanned pics of my memory sticks :). YGPM.
Once you click on the link, you need to click on the picture that loads to get the full-size picture. It loads a thumbnail.
Both sticks are the same, except that my other stick is half 310WE BH-5, and half 322WF BH-5.
jumanji969
07-02-2005, 10:01 PM
I think Hipro5 did some mods to his bh-6 ram.
stealth17
07-02-2005, 10:05 PM
I think Hipro5 did some mods to his bh-6 ram.
and you are correct. ive seen the pic. he said he gained like a couple mhz, not worth the trouble in his opinon
felinusz
07-03-2005, 03:43 PM
"Worth the trouble" is relative, isn't it? :)
$25 worth of SMD ceramic capacitors for ~5 MHz in gains... not cost effective at all. But I don't care, I'm doing it anyways :). Voltmods are fun - but most importantly, charting the gains (if any) from this will be extremely interesting to me.
To the non-overclocker my entire machine is "Not worth the trouble", and "Not cost effective" - when taken to an extreme this 'sport' is not a practical one by any means, we all know that.
A related discussion: http://www.xtremesystems.org/forums/showthread.php?t=51210
dippyskoodlez
07-03-2005, 04:06 PM
"Worth the trouble" is relative, isn't it? :)
$25 worth of SMD ceramic capacitors for ~5 MHz in gains... not cost effective at all. But I don't care, I'm doing it anyways :). Voltmods are fun - but most importantly, charting the gains (if any) from this will be extremely interesting to me.
To the non-overclocker my entire machine is "Not worth the trouble", and "Not cost effective" - when taken to an extreme this 'sport' is not a practical one by any means, we all know that.
A related discussion: http://www.xtremesystems.org/forums/showthread.php?t=51210
exactly... this is a place where even 2 mhz is worth $1 in caps. letalone 5mhz! :p:
[XC] moddolicous
07-04-2005, 07:20 AM
PM hipro about what he did on his ram. He added some types of caps and gained a few MHZ. Also, if u havent already, burn those sticks in completely. Find the highest stable overclock, let it run memtest for like 6 or more hours, then increase the clocks till u get alot of errors.
stealth17
07-04-2005, 07:26 AM
PM hipro about what he did on his ram. He added some types of caps and gained a few MHZ. Also, if u havent already, burn those sticks in completely. Find the highest stable overclock, let it run memtest for like 6 or more hours, then increase the clocks till u get alot of errors.
i dont liike that type of burn in. what i do is get highest stable clockes and reduce voltage untill you can hardly boot or you get a ton of errors then run memtest #5 for many hours. i turn the reporting mode to count only so it doesnt freeze too
felinusz
07-04-2005, 08:14 AM
I've burned these sticks quite thoroughly, but I'll try your method stealth.
Acive voltage signal "starving" for extensive time periods in order to make the transistors more responsive to overvoltage..... it sometimes works with A64 processors too. I would guess that it somehow improves transistor signal tolerance.
[XC] moddolicous
07-04-2005, 10:00 AM
i dont liike that type of burn in. what i do is get highest stable clockes and reduce voltage untill you can hardly boot or you get a ton of errors then run memtest #5 for many hours. i turn the reporting mode to count only so it doesnt freeze too
Yea, u can do that too. Thats more for like a stability thing though isnt it. Thats a good way to burn them in. I think running all the test is better than just test #5, but I guess thats an opinion.
Edit: http://www.thelab.gr/reviews/ic7mods_en19.php Picture of Hipro's modded ram.
felinusz
07-04-2005, 07:15 PM
Well, better stability at any given speed/voltage should also get your sticks that much higher for peak screenshot/benchmark stable speeds. My sticks are stable at 273 Mhz, bench stable at 275 MHz, and they're good for peaked screenshots into the 280s.
Before a 'uwackme' style burn-in at 3.78V, all three of these numebrs were 2 MHz lower.
stealth17
07-04-2005, 08:49 PM
Well, better stability at any given speed/voltage should also get your sticks that much higher for peak screenshot/benchmark stable speeds. My sticks are stable at 273 Mhz, bench stable at 275 MHz, and they're good for peaked screenshots into the 280s.
Before a 'uwackme' style burn-in at 3.78V, all three of these numebrs were 2 MHz lower.
whats a uwackme style burn-in?
FrozenPC4Brain
07-04-2005, 09:36 PM
OK, Everybody seems pretty sure about Ram being water cooled as a wast of time and effort. LMAO.
Just the "Bling" factor was enough for me!
My design uses "Brass", So b4 I post a how to, am I going to get slammed for that?
And the next version will have better built in turbulence being that the flow is to un-restricted.
As I recall brass was pretty good for heat transfer and I couldn't find a source of copper anywhere on-line or off, And I looked very hard, Emailed a bunch of metal dist places and none wanted any sales that small.
You certainly wouldnt get slammed by me, as I think it is super cool! Like you said, the same goes with me, that the bling is more than enough for me! I'm going to have a second loop on my water cooling system, built with my old Thermaltake Aquarius II, and that will be cooling the NB and the RAM hopefully. That way, people cannot say I am hurting my system at ALL. I'm not even putting an extra heat dump into the loop on my CPU.
felinusz
07-05-2005, 07:19 AM
whats a uwackme style burn-in?
'uwackme' of the DFI stree forums developed it in the first place :)
Max out voltage, tighten timings, run at memtest86 test #5 looped at an unstable speed for a specific time increment (~6 hours).
After 6 hours, write down the number of errors that occured.
Repeat the 6 hour increment at the same settings, record the number of errors after the time increment.
After three or four time increments, the number of errors per time increment should decrease signifigantly when compared to the first time increment run.
Jayw28
07-05-2005, 11:58 PM
do burn ins really serve a purpose? before i would just run my stuff o/ced as high as i can right out of the box.
[XC] moddolicous
07-06-2005, 07:24 PM
do burn ins really serve a purpose? before i would just run my stuff o/ced as high as i can right out of the box.
U burn in so that u can reach higher stable OC's, and u can also make some items run with lower volts. IE: Before your CPU may need 1.6v to run 2.5ghz, and now after burn in u can make it un with 1.5v or less. Samething applies to ram.
So what would be the ideal way (even an overkill way) to burn in all of your components as best as possible?
stealth17
07-07-2005, 01:00 PM
So what would be the ideal way (even an overkill way) to burn in all of your components as best as possible?
yes....just about anything will burn in thats made from silicone. ive heard of ppl buring in their gpu too
dippyskoodlez
07-07-2005, 01:41 PM
yes....just about anything will burn in thats made from silicone. ive heard of ppl buring in their gpu too
silicone is rubbery.. silicon is what computers use ;)
I wasnt asking if it was possible, i was askign the best way to do it. Sorry should have been a little more descrptive. Got a new build, not wuite done yet, am going to OC it with a very healthy OC, but I know i need to do a nice burn in before I can attempt that. What would be the best way to burn it in? Run Prime95 for days on end or something?
stealth17
07-08-2005, 06:15 AM
I wasnt asking if it was possible, i was askign the best way to do it. Sorry should have been a little more descrptive. Got a new build, not wuite done yet, am going to OC it with a very healthy OC, but I know i need to do a nice burn in before I can attempt that. What would be the best way to burn it in? Run Prime95 for days on end or something?
http://forums.extremeoverclocking.com/showthread.php?t=171646&highlight=burn+san+diegos
Sweet, there we go, thanks. I'll go super overkill with mine and just burn-in the hell out of it! :conniving:
EDIT: It's a new build, but I'm using a processor that I had already set up in my other system for a little over a year (non overclocked) as well as the ram (temporarily until I can afford my LA's) is it too late for a burn in?
@ him ummm I thought the DDR booster was not suppose to work with the IC7 boards???
stealth17
07-11-2005, 08:59 AM
Sweet, there we go, thanks. I'll go super overkill with mine and just burn-in the hell out of it! :conniving:
EDIT: It's a new build, but I'm using a processor that I had already set up in my other system for a little over a year (non overclocked) as well as the ram (temporarily until I can afford my LA's) is it too late for a burn in?
never too late for bun-in
and you cant go overkill imo...i burn-in as when ever im not using the processes for something else :p:
@ him ummm I thought the DDR booster was not suppose to work with the IC7 boards???
No, it can. It came with a compatibility chart, cooincidently I have it hanging right next to me. It's was max tested at 3.6, which is a damn lot of voltage.
stealth17
07-11-2005, 12:52 PM
3.6, which is a damn lot of voltage.
bah :D
lalPOOO
07-11-2005, 08:46 PM
The ddr booster works fine with the IC7 boards, but it only boosts the vdimm. It doesn't boost vtt, which is supposed to be 1/2 the vdimm. Most other boards track vtt to vdimm, but the IC7 doesn't.
The ddr booster works fine with the IC7 boards, but it only boosts the vdimm. It doesn't boost vtt, which is supposed to be 1/2 the vdimm. Most other boards track vtt to vdimm, but the IC7 doesn't.
Ok, so what should I do about that? CAn i track the VTT in the BIOS? And do i have to tweak that every time, or at least should I, I adjust the voltage on the DDR Booster?
lalPOOO
07-12-2005, 08:16 AM
In order to get vtt to track you will have to vmod the board and adjust it manually, because 1/2 3.6v = 1.8v and the board only goes to 1.4v normally. Check hipro's guide in the motherboard mods section of the forum, he has a guide posted on how to mod vtt on the IC7 boards. And yes, you would have to adjust this mod everytime you adjusted the booster.
Well, I didnt realize how difficult it would be to use the DDR Booster. I can use it without tracking the VTT, right? Prolly just wouldnt be as effiecient? Is there a way I could put a pot or variable resistor somewhere on the board as to adjust the VTT whenever I adjust the DDR Booster? Apparently there is alot more to it than previously though.
Gogar
07-26-2005, 02:44 PM
Would it help to ground heatspreaders?
speed bump
08-02-2005, 08:26 AM
semi off topic.
felinusz have you got your phase change back yet? since your testing on an A64 platform you will be held back by the memory controller, so the cold your CPU is the better the mem controller will work.
stealth17
08-02-2005, 08:48 AM
semi off topic.
felinusz have you got your phase change back yet? since your testing on an A64 platform you will be held back by the memory controller, so the cold your CPU is the better the mem controller will work.
thats not always true. sometimes it has a reverse effect :(
enzoR
08-04-2005, 11:15 AM
Would it help to ground heatspreaders?
yes, and also:
find the polarity to solder the caps to with a multimeter
:toast:
Nosfer@tu
08-29-2005, 05:08 AM
Find the polarity to solder the caps WITH a multimeter????
I REALLY dont understand that :D
FireDragon
09-22-2005, 07:20 AM
I have been toying with this idea FOR A LONG TIME but my purpose is to use them with chilled water on htem or with TECS i am making a chiller so that is my purpose right now not the tec part.
Dragon
novagamer
09-23-2005, 08:44 PM
I have been toying with this idea FOR A LONG TIME but my purpose is to use them with chilled water on htem or with TECS i am making a chiller so that is my purpose right now not the tec part.
Dragon
Be sure to post pics of that :p:
ZMarre
09-25-2005, 12:02 AM
Ontopic: some time ago, there was some discussion about lapping your RAM. And I mean the plastic casing of your RAM-chips. Theory behind it is that you decrease the resistance for the heat expelled by your chips.
Lots of ppl who did it, gained some extra MHz. :toast:
Only drawback: the risk of lapping too much of your casing and destroying your RAM :mad:
Jochenp
09-25-2005, 07:48 AM
yes, and also:
find the polarity to solder the caps to with a multimeter
:toast:
where did you solder those caps to?