Overclock.net › Forums › Components › Memory › H2 Burn in your RAM
New Posts  All Forums:Forum Nav:

H2 Burn in your RAM

post #1 of 15
Thread Starter 
Before you go any further, download this file! http://files.extremeoverclocking.com/file.php?f=125

Although burn in is usually equated with CPUs, any device that uses ICs will be affected by a good burn in. Nobody knows what burn in actually is, or how it works, but the fact is, it does work. I was quite excited when I did my first burn in. I had found a way to increase my memory bandwidth by tightening timings and upping the FSB, effectively making both my CPU and RAM faster. It's the best of both worlds, and it will make both of them better.

The goals of burning in a CPU are two fold. The first is to be able to achieve the same clock speeds with less voltage. The second is to be able to OC more at the current voltage. With RAM, the concept and process are basically the same as with a CPU. It varies mainly in this one thing; with RAM, you can also burn in to get tighter timings at a given clock and voltage.

Since RAM has three factors (timings, FSB & voltage), that means there are (primarily) three ways to burn it in. The first method involves changing the timings, the second is to alter the FSB and the third (come on now, it's not that hard to guess) is to change the voltage. I will say this before I go into an explanation of what to do, FSB and timings are the more successful route to go, this is because most mobos aren't going to have extensive options for changing your VDIMM in minute enough steps (In fact, none that I know of ATM). But since most boards do have good options for changing timings and FSB, those will be your best bet. But don't forget about changing voltages yet.

Keep in mind that when doing a burn in, make sure your CPU is within specs. You don't want it to crash and you think it's your RAM when the CPU is the real issue. For those of us with unlocked CPUs, do a burn in at the multi you want to use, or as close to that as possible, otherwise you get slighlty different results for different multipliers.


METHOD I -FSB (Best used when your CPU is nowhere near it's max)

The first RAM burn in I did was accomplished by upping the FSB. This is a fairly easy enough process. You just up the FSB while keeping the timings and voltage constant and run memtest between every increase. Go in 10mhz increments until you get errors. After you get errors, decrease by 5mhz and then raise or lower the FSB from there depending on whether or not you get errors. The goal here is to find what your RAM maxes at. From here you have a choice to make. How much do I value my OS? If you would cry at the thought of messing up your installation, do the burn in at the fastest error free FSB. If you don't give a rip about your OS, go to your max error free FSB (hereafter referred to as "MEF FSB") and then up the FSB another two or three mhz. It will speed the burn in up, but you risk data corruption if you do this.

After you have found your MEF FSB and decided how much you love your OS, open up Hot CPU and go to the Burn-in tab. Do the memory burn in and make sure you've selected the appropriate amount of RAM. **WARNING** %DO NOT SELECT TOO MUCH RAM. GO TO THE SYSTEM INFO TAB AND SEE HOW MUCH RAM IS AVAILABLE AND SELECT THE NEXT LOWEST AMOUNT TO TEST% Just do the burn in for a good 6-8 hours while making sure not to do anything else on your compy during that period. After the time is up, reboot, go into your BIOS and up the FSB some more in an attempt to find your new, higher MEF FSB.

After you have found the new MEF FSB, repeat the process until you can't burn in anymore or you achieve the desired result. If you want to be sure that you have completely burned in, run a good 12 hour RAM burn in and if you see no improvement, then you're finished.



METHOD II -timings (Best used if your CPU is already maxed out at the current FSB)

I did a second round of burn ins after I found my new MEF FSB. This time, my goal was tighten the timings. Using Method I, my timings were at 2.5-2-2-11. I really wanted that to be at 2-2-2-11. So I naturally decided to give a burn in with tighter timings a shot. It goes like this; set your timings to whatever you want them to end up at. Find your MEF FSB with those timings and do a burn in at your MEF FSB. Do the burn in for at least 6 hours. After this, keep your timings where they are and find your new MEF FSB. Continue with the process until you get tired of it or can't go any further with it.



METHOD III -voltage (Works best when used in conjunction with the other two methods)

The same thing goes here as with the other methods, find your MEF FSB with a given voltage and burn in the RAM to perfection. I don't recommend this method by itself as it doesn't offer the precision that the others do.



OTHER METHODS

These consist of mixing and matching the other 3 methods. i.e., getting your timings AND voltage where you want it at, and then proceeding to burn in at your MEF FSB. Or getting your FSB where you want it and tweaking your voltage and timings accordingly. There is a lot of freedom with mixing and matching like this and how much you want to work on each parameter in your mix at a certain time.



During a burn in, your RAM can get very hot. In fact, heat is often what keeps your RAM OC from working. If your RAM is hot to the touch during a burn in, I would suggest getting an active memory cooler, unless of course you have a heatpipe cooler on your RAM, in which case a case fan would still help out.. One with at least one 60mm fan will work the best.

Well, here you have the methodology behind getting some RAM that will kick the competition. If you want to get your benchmark that much better, just do a good burn in and you'll be on your way to accomplishing that goal. But words without numbers are basically meaningless. So here's some numbers for you.



Crucial Ballistix DDR 400 (Dual Channel, 2.9v), golden-orange heat spreaders. Micron -5B G chips

When I first got my RAM, its original MEF FSB was 212mhz at 2-2-2-11. Keeping those same timings, the MEF FSB is now 220mhz.

At 2.5-2-2-11, the original MEF FSB was 228mhz. That's now a happy 236mhz.


As I burn in more RAM and continue with this burn in I'll update the numbers. I hope this helped out some and that you can get your FSB up a bit with this.



Unless otherwise noted, all testing and burn in was done with the following:
NF7-S v2.0
OCZ XTC active memory cooler- 2x60mm fans
XP-M 2800 running at 11xFSB
Corsair HX 520

MEF FSB is gotten by doing 7 passes of memtest #5, 1 pass memtest #7 and 3 passes memtest #8. I reccommend an all night memtest run to check for 100% stability after you are satisfied with your OC.

v1.0 -hopefully this will be getting updates very soon


The intellectual rights of this belong to me, LC_STiK. If for any reason you want to use this guide for any other publication on the web or elsewhere, please contact me first via PM and we'll take it from there.

Thanks.
My System
(13 items)
 
  
CPUMotherboardGraphicsRAM
4850e M3A78 4850 4x2GB 
Hard DriveMonitorPowerCase
1TB WD Blac Hanns G 28" Corsair HX 520 TT 8003BWS 
  hide details  
Reply
My System
(13 items)
 
  
CPUMotherboardGraphicsRAM
4850e M3A78 4850 4x2GB 
Hard DriveMonitorPowerCase
1TB WD Blac Hanns G 28" Corsair HX 520 TT 8003BWS 
  hide details  
Reply
post #2 of 15
Hey nice guide...I firmly believe in burn in myself...I'm like you I don't know why it works reason would say it wont make a difference...But it seems to work for me...I've always used Memtest or blend on Prime and OCCT I'll have to try this other program...
My System
(13 items)
 
  
CPUMotherboardGraphicsRAM
E8600 4545Mhz*9 w/1.29V DFI DK P45 XFX 8800 GS D9GKX & D9GMH 
Hard DriveOptical DriveOSPower
2x Seagate 160 Perp. RD-1 2x Asus Vista Premium Corsair HX 620 
Case
Open Bench 
  hide details  
Reply
My System
(13 items)
 
  
CPUMotherboardGraphicsRAM
E8600 4545Mhz*9 w/1.29V DFI DK P45 XFX 8800 GS D9GKX & D9GMH 
Hard DriveOptical DriveOSPower
2x Seagate 160 Perp. RD-1 2x Asus Vista Premium Corsair HX 620 
Case
Open Bench 
  hide details  
Reply
post #3 of 15
Thread Starter 
For lack of an explanation from Hot CPU, I'm figuring it's like running large FFTs through Prime95. Whatever it does, it works.
My System
(13 items)
 
  
CPUMotherboardGraphicsRAM
4850e M3A78 4850 4x2GB 
Hard DriveMonitorPowerCase
1TB WD Blac Hanns G 28" Corsair HX 520 TT 8003BWS 
  hide details  
Reply
My System
(13 items)
 
  
CPUMotherboardGraphicsRAM
4850e M3A78 4850 4x2GB 
Hard DriveMonitorPowerCase
1TB WD Blac Hanns G 28" Corsair HX 520 TT 8003BWS 
  hide details  
Reply
post #4 of 15
this has been proven again and again. Burn it is NOT a good idea. Even the MaximumPC and PCGAMER magazines agree with this. There is absolutlely NO BENEFIT to doig this and all it will do is up your hydro bill.
post #5 of 15
Thread Starter 
No benefit you say? After a 6 hour burn in yesterday (2-2-2-11 @ 223mhz), I noticed a significant reduction in errors.

Before the burn in, I got 7132 errors in 7 passes of memtest #5. After the burn in? Well, it was 113 errors for the same amount of passes.

I know a measly decrease from 7132 to a whopping 113 isn't much, but it works for me.

If you would care to provide links to places that say burn in doesn't work, feel free to. Maybe it doesn't work in their universe, but in my little corner of the world, burn in is a real thing that works consistently.
My System
(13 items)
 
  
CPUMotherboardGraphicsRAM
4850e M3A78 4850 4x2GB 
Hard DriveMonitorPowerCase
1TB WD Blac Hanns G 28" Corsair HX 520 TT 8003BWS 
  hide details  
Reply
My System
(13 items)
 
  
CPUMotherboardGraphicsRAM
4850e M3A78 4850 4x2GB 
Hard DriveMonitorPowerCase
1TB WD Blac Hanns G 28" Corsair HX 520 TT 8003BWS 
  hide details  
Reply
post #6 of 15
What exactly do you mean by 'burn in'? Run the component at 100% load for many hours?
post #7 of 15
Quote:
Originally Posted by thealmightyone View Post
What exactly do you mean by 'burn in'? Run the component at 100% load for many hours?
Basically yes...

You can do about the same thing using Memtest...just run it up till it errors and back it down till it stops producing errors and run it 8 hours or so...

You must NOT let memtest run in loop mode while the ram is producing errors...it will most of the time make less errors over time but it will still burn errors into the chips..

I just clock them up as high as I can get them with FSB and voltage and let it run...

Say if I get it up to 330fbb and it produces even 1 error on a pass I will back it down to 299fsb and let it run 8 hours or more...alot of times when you run it back up it will make it to 335 or so before starting to error again...

thats how I got my ddr400 to run ddr670...
My System
(13 items)
 
  
CPUMotherboardGraphicsRAM
i7-860 ECS P55 Black Edition 2-4850X2 (Quadfire) 2x2GB Geil Black Dragon 
Hard DriveOptical DriveOSMonitor
2x640 Raid 0 DVD-RW XPpro, Vista, Win in x32 and x64 flavors Acer 22in Wide Screen 
KeyboardPowerCaseMouse
MS Ergonomic Corsair TX-850 Black ATX MS Optical 
Mouse Pad
yea right... 
  hide details  
Reply
My System
(13 items)
 
  
CPUMotherboardGraphicsRAM
i7-860 ECS P55 Black Edition 2-4850X2 (Quadfire) 2x2GB Geil Black Dragon 
Hard DriveOptical DriveOSMonitor
2x640 Raid 0 DVD-RW XPpro, Vista, Win in x32 and x64 flavors Acer 22in Wide Screen 
KeyboardPowerCaseMouse
MS Ergonomic Corsair TX-850 Black ATX MS Optical 
Mouse Pad
yea right... 
  hide details  
Reply
post #8 of 15
Burn errors in the chip? Do you know what you're on about? Errors cannot be 'burned in', they are simply generated; a 1 being read as a 0 most of the time.

Once I understand how IC's work once I start my uni course, I will be able to see if burn in is a load of c***, or if it does actually work.
post #9 of 15
Quote:
Originally Posted by thealmightyone View Post
Burn errors in the chip? Do you know what you're on about? Errors cannot be 'burned in', they are simply generated; a 1 being read as a 0 most of the time.

Once I understand how IC's work once I start my uni course, I will be able to see if burn in is a load of c***, or if it does actually work.
Yes I do know....

I have seen many, many sticks of ram rendered useless from the owners overclocking ram and running memtest when it's erroring during a pass and letting it run...

It's definately NOT good to stress ram and let it test while erroring...It does no good... to effectively burn it in it must NOT be producing errors while doing so...
My System
(13 items)
 
  
CPUMotherboardGraphicsRAM
i7-860 ECS P55 Black Edition 2-4850X2 (Quadfire) 2x2GB Geil Black Dragon 
Hard DriveOptical DriveOSMonitor
2x640 Raid 0 DVD-RW XPpro, Vista, Win in x32 and x64 flavors Acer 22in Wide Screen 
KeyboardPowerCaseMouse
MS Ergonomic Corsair TX-850 Black ATX MS Optical 
Mouse Pad
yea right... 
  hide details  
Reply
My System
(13 items)
 
  
CPUMotherboardGraphicsRAM
i7-860 ECS P55 Black Edition 2-4850X2 (Quadfire) 2x2GB Geil Black Dragon 
Hard DriveOptical DriveOSMonitor
2x640 Raid 0 DVD-RW XPpro, Vista, Win in x32 and x64 flavors Acer 22in Wide Screen 
KeyboardPowerCaseMouse
MS Ergonomic Corsair TX-850 Black ATX MS Optical 
Mouse Pad
yea right... 
  hide details  
Reply
post #10 of 15
Do you have a link that explains this in absolute detail? I really want to read up on this. I am most interested.
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Memory
Overclock.net › Forums › Components › Memory › H2 Burn in your RAM