|
|
|
|
| Author |
Message |
-
Guest
|
 Posted: Fri Dec 19, 2003 4:16 am Post subject: How safe is memory when overclocking FSB? |
 |
|
Is memory more likely to give out first when trying to overclock a system?
(I have an ASRock K7VT2 mobo, Athlon 1800+ xp cpu and Hyundai 256 133Mhz
SDRAM chips all running at 137Mhz and with a locked multiplier of 11.5).
Thanks for any advice. |
|
| Back to top |
|
 |
|
|
Wes Newell
Guest
|
| Posted: Fri Dec 19, 2003 5:11 am Post subject: Re: How safe is memory when overclocking FSB? |
|
|
On Thu, 18 Dec 2003 22:16:27 +0000, - wrote:
| Quote: | Is memory more likely to give out first when trying to overclock a system?
(I have an ASRock K7VT2 mobo, Athlon 1800+ xp cpu and Hyundai 256 133Mhz
SDRAM chips all running at 137Mhz and with a locked multiplier of 11.5).
Memory is usually rated for it's top speed so there's less room to |
overclock it. Some won't even reach there rated speed. I always buy faster
ram than required to be safe.
--
Abit KT7-Raid (KT133) Tbred B core CPU @2400MHz (24x100FSB)
http://mysite.verizon.net/res0exft/cpu.html |
|
| Back to top |
|
|
Chip
Guest
|
| Posted: Fri Dec 19, 2003 7:29 pm Post subject: Re: How safe is memory when overclocking FSB? |
|
|
"Wes Newell" <w.newell@SOSverizon.net> wrote in message
news:pan.2003.12.18.23.15.39.695639@SOSverizon.net...
| Quote: | On Thu, 18 Dec 2003 22:16:27 +0000, - wrote:
Is memory more likely to give out first when trying to overclock a
system?
(I have an ASRock K7VT2 mobo, Athlon 1800+ xp cpu and Hyundai 256 133Mhz
SDRAM chips all running at 137Mhz and with a locked multiplier of 11.5).
Memory is usually rated for it's top speed so there's less room to
overclock it. Some won't even reach there rated speed. I always buy faster
ram than required to be safe.
|
Isn't that a bit of an oversimplification, Wes? The manufacturers don't
want 100's of RMA's because every stick is "on the limit", do they. They
build in a certain margin for comfort. And apart from anything else, you
have the issues of what volts you apply and the timings you set. A stick
rated as PC3200 cas 2 at 2.5v might very well do more (a lot more?) at say
2.7v and cas2.5 or 3. Lots of ram is good for more MHz than its rated at
without too much fiddling.
My OCZ 3500 ram is only rated for 217MHz, yet it does 228+ at tight timings,
no problem. And this is typical of most quality brands - Corsair, Mushkin,
Kingston etc. imho.
So my answer to OP would be "not necessarily".
Chip.
|
|
| Back to top |
|
|
-
Guest
|
| Posted: Fri Dec 19, 2003 10:28 pm Post subject: Re: How safe is memory when overclocking FSB? |
|
|
| Quote: | So my answer to OP would be "not necessarily".
|
Thanks for the comments.
Would I be right in thinking that to minimize the risk of blowing memory, it
would be better to o/c from a FSB of 100 (my memory is 133) to see what the
maximum increment is, then switch back to 133 and then perform the same
increment or more sensibly just below it? Or, does the fact that the memory
is overclocked at anything different from the FSB mean that its at risk?
Basically is say, 107 Mhz safe (because its between 100-133) whereas 140 Mhz
is getting risky? |
|
| Back to top |
|
|
Chip
Guest
|
| Posted: Fri Dec 19, 2003 11:45 pm Post subject: Re: How safe is memory when overclocking FSB? |
|
|
"-" <nospam@nospam.com> wrote in message
news:3fe32770$0$384$afc38c87@news.easynet.co.uk...
| Quote: | So my answer to OP would be "not necessarily".
Thanks for the comments.
Would I be right in thinking that to minimize the risk of blowing memory,
it
would be better to o/c from a FSB of 100 (my memory is 133) to see what
the
maximum increment is, then switch back to 133 and then perform the same
increment or more sensibly just below it? Or, does the fact that the
memory
is overclocked at anything different from the FSB mean that its at risk?
Basically is say, 107 Mhz safe (because its between 100-133) whereas 140
Mhz
is getting risky?
|
LONG post....
You won't "blow" the memory. If you overclock it too far, you will just get
memory read/write errors.
If you are interested, this is what I would do:
Run the memory and fsb *synchronously*. Don't mess about with memory at one
speed and FSB at another - its a waste of time. Simply gently increase the
FSB bit by bit to see how far it can go. What I like to do is to prepare a
memtest86 floppy disk and boot from floppy first for each FSB increase. If
memtest86 runs OK for a while, you can then try booting to windows. If you
get into windows (which shouldn't be a problem - see below) then run Prime95
for a few minutes. If Prime95 runs OK, reboot, increase the fsb a notch or
two and repeat the process.
The reason I do it this way is because if Prime95 runs OK for a while, then
you can be reasonably sure that Windows will at least *boot* when you
increase the setting further. For example if Prime95 will run for 10
minutes at 140 MHz FSB, then you can be pretty sure Windows will load OK at
142MHz. This way you minimise the risk of windows failing to load and
corrupting system files: you spot errors in Prime95, not when Windows loads
up.
All the time, keep an eye on the CPU temperatures. You don't really want
them higher than 50C for any of this. Remember they will go higher when you
start stressing the PC, which you haven't done yet.
Once you get to a point where Prime95 shows errors, then you know you have
reached a limit. Its then up to you whether you want to stop now or try to
go further. If you've had enough, just back down the FSB a bit and make
sure everything is stable by running Prime95 for an hour or so. And settle
for what you have.
If you still want to go further, you have a bit of investigation to do. You
don't know what's limiting things: is it your ram, or is it your CPU? So to
find out, reboot and run memtest86 from the floppy again at these same
settings for about an hour. This is long enough at this point to show you
where the problem lies. There are then 2 scenarios:
Scenario 1: Memtest86 runs OK for 1 hour or more.
This means your memory is good at this FSB, and therefore it must be your
CPU thats the problem: you've hit the limit for your CPU. You could
consider increasing the CPU voltage (if your board allows it, and if your
CPU temperatures are nice and low). If you do this, try booting Windows are
running Prime95 again to see if its now stable. If it is, then you can
increase the FSB 1 or 2 MHz more and repeat the whole process.
Scenario 2: memtest86 shows errors.
This means your CPU is probably OK at this FSB and its probably your memory
that's hit the limit. See if you can get your memory to go higher by
playing with the bios settings, i.e. timings and voltages. Again, it
depends what options are available in your bios. But for example, if you
are running the memory at 2.6volts, you could try increasing it to 2.7v. I
would not go over 2.8v unless you know your memory can handle it: you *can*
blow a memory stick by putting too much voltage across it. Alternatively,
if the memory is running timings of (say) Cas 2, then perhaps you could try
setting it to Cas 2.5. Whatever changes you make, run memtest86 again and
see if the errors go away. If they do, you can boot Windows again and then
run Prime95 again. If that now runs OK, you can repeat the process all over
again and try increasing the FSB a little bit more.
Using the above approach, you can - bit by bit - increase the speed of the
system, without placing it under any sudden stress or stain, and find out
what components limit its performance and where the limit is. It isn't as
labourious as it sounds, either. The little tests along the way only need
to run for a minute or two to test for basic stability. Its only when you
have finished (or you think you have) that you should run memtest86 and
Prime95 each for a couple of hours or so, to make certain everything is rock
solid at your chosen settings.
A word of caution however, since your motherboard is KT266 board, when you
overclock the FSB, you are also overclocking your PCI bus and therefore your
harddisk(s). This is where the greatest danger lies. Hard disks really
don't like running much out of spec. Most are OKish at perhaps 37.5MHz, but
once you get near to 40MHz, you are running high risk of some major disk
corruption or even a trashed hard drive. So be warned, and be careful!
This means 150MHz FSB is as fast as you can go with reasonable (but not
100%) safety. Beyond that you are running the risk of significant disk
corruption, windows reloads etc and even broken hardware! So be careful.
HTH?
CHip. |
|
| Back to top |
|
|
Fix your Windows Problems - FAST.
FREE Safe Scan Registry Check. Locate & Fix Errors in Minutes!
|
|
-
Guest
|
| Posted: Sat Dec 20, 2003 1:04 am Post subject: Re: How safe is memory when overclocking FSB? |
|
|
Brilliant stuff - very clear and interesting, I look forward to
experimenting in the ways you have suggested when I get the time.
Thanks Chip : ) |
|
| Back to top |
|
|
|
|
| |
Watched Topics
Search
Register
Log in to check your private messages
Profile
Log in
Memberlist
Usergroups