XMS3 timing help needed

HeX · 14 Oct 2011 at 12:58

Glad you've got it sorted.

ChrisGTL · 14 Oct 2011 at 13:01

Plec - I see you have a similar rig to mine.

Have you done any testing to see the difference between 1600mhz with CAS 9 and 1333Mhz with tighter latencies?

khare · 14 Oct 2011 at 13:01

It's just one problem I've got now and while you two magicians are here I may aswell point you in the direction of it

http://forums.overclockers.co.uk/showthread.php?t=18327186

It's the boot up issue that sometimes it won't boot

ChrisGTL · 14 Oct 2011 at 13:05

HeX said:
Also your NB frequency is underclocked, you might want to knock the multi up one.

I knew I'd seen some other problems somewhere. This is why I was saying try run RAM as per spec to rule anything out.

However u've already said you can't run 1600Mhz RAM as the FSB ratio is out, have you corrected the NB freq?

khare · 14 Oct 2011 at 13:07

Yes NB frequency is now 2234mhz with multiplier 9

Plec · 14 Oct 2011 at 13:08

ChrisGTL said:
Plec - I see you have a similar rig to mine.

Have you done any testing to see the difference between 1600mhz with CAS 9 and 1333Mhz with tighter latencies?

Sorry i haven't, mainly because reviews have shown that 1600MHz memory is the sweet spot for sandybridge setups. They indicate, thus far, that there's very little gain in higher frequencies or lower frequencies with slightly tighter timings.

Plus, i'm on my second batch of memory (2x 4Gb - my sig is out of date) as my first set were corrupt so i've been monitoring it at standard spec but at 1.5V.

Plec · 14 Oct 2011 at 13:10

khare said:
It's just one problem I've got now and while you two magicians are here I may aswell point you in the direction of it

http://forums.overclockers.co.uk/showthread.php?t=18327186

It's the boot up issue that sometimes it won't boot

Having a read through as i type...

ChrisGTL · 14 Oct 2011 at 13:14

Plec said:
Sorry i haven't, mainly because reviews have shown that 1600MHz memory is the sweet spot for sandybridge setups. They indicate, thus far, that there's very little gain in higher frequencies or lower frequencies with slightly tighter timings.

Plus, i'm on my second batch of memory (2x 4Gb - my sig is out of date) as my first set were corrupt so i've been monitoring it at standard spec but at 1.5V.

I already had 2x2GB or XMS3 so I bought another 2x2GB - I would much prefer 2x4GB sticks.

I didn't know if I had read about SB not having any gains over 1600Mhz memory, I must have done if you confirm that is correct.

Plec · 14 Oct 2011 at 13:19

ChrisGTL said:
I already had 2x2GB or XMS3 so I bought another 2x2GB - I would much prefer 2x4GB sticks.

Having all the DIMM slots isn;t the problem it used to be with when installed in SB setups - few clocks have been effected by the addition of extra memory.

ChrisGTL said:
I didn't know if I had read about SB not having any gains over 1600Mhz memory, I must have done if you confirm that is correct.

I'll search out the link for you...

Plec · 14 Oct 2011 at 13:22

ChrisGTL said:
I didn't know if I had read about SB not having any gains over 1600Mhz memory, I must have done if you confirm that is correct.

Here you go - link

Plec · 14 Oct 2011 at 13:27

khare said:
It's just one problem I've got now and while you two magicians are here I may aswell point you in the direction of it

http://forums.overclockers.co.uk/showthread.php?t=18327186

It's the boot up issue that sometimes it won't boot

Do you ever lose your clock settings after it's failed to boot and do you have any power saving features enebaled in windows - if so what are they?

(I will add i'm unfamiliar with AMD setups - but your problem seems clock related.)

EDIT - how old is the PSU?

khare · 14 Oct 2011 at 13:31

Clock settings stay as they were last before. Asus has this clock protection thing that stops VRM's from overheating and therefore these mobos are apparently not the best to OC due to this feature. No power saving features enabled in windows.

Plec · 14 Oct 2011 at 13:35

The fact that the clock settings aren't getting reset would 'seem' to indicate that the BIOS settings aren't overstretching the components.

There's a possibilty that it's PSU related (it can never be ruled out with problems like these) as your PSU isn't the best make in the world - do you own a multi/volt meter?

khare · 14 Oct 2011 at 13:39

Yes I do think it may be PSU related. I always think that the PSU buckles under pressure and fails to deliver. I do have a voltmeter yes. Do the voltages in HWmonitor mean anything to you? I could post a screenshot of them if you wish.

Plec · 14 Oct 2011 at 13:43

khare said:
I do have a voltmeter yes.

The paperclip test using a voltmeter is far more accurate than HWmonitor.

If you want to help rule out the PSU and if you haven't got a spare PSU to hand you could do a basic PSU test - below are links to Huddy's site which has a step by step guide on how to test your PSU using a volt/multimeter:

'Use a multimeter to check your PSU' and one to 'How to check if your PSU is faulty'

The process is easier than it sounds and although it's not a definitive test it can rule out fundemental issues.

khare · 14 Oct 2011 at 13:45

Thanks Plec. Will have a read of the guides and report back.

HeX · 14 Oct 2011 at 13:46

How are you testing your overclock? just Prime? If so that's probably not enough to confirm stability.

Grab LinX:

http://www.youwatched.com/datajay/linx(0.64).zip

Extract it, run it, set memory to use to 2GB, loops to 20 and set it off.

(don't use your rig while it's doing this just let it run through)

It'll take about 40mins or so to complete.

You want to check that all the Gflops are roughly the same, and the time taken should again be roughly the same.

If there is a major error it'll just auto-quit.

If you can pass 20 loops of that your rig is stable.

I've had rigs that'd pass 8hrs+ of Prime but fall over in 5mins on LinX. Whereas I find that if my rig will pass 20 loops of LinX it'll pass anything.

Plec · 14 Oct 2011 at 13:56

HeX said:
I've had rigs that'd pass 8hrs+ of Prime but fall over in 5mins on LinX. Whereas I find that if my rig will pass 20 loops of LinX it'll pass anything.

Heh, i know what you mean. I've had systems where i've thrown every stabilty test going at it but will still BSOD with light apps use, browsing etc...

Stress testing is great for trying to draw out a weaknesess in a clock early on - but until it's proved itself in day to day usage over a few weeks i wont trust a new rig's clock until it's been 'general use stable' for at least 4 weeks.

Makes the process more interesting i guess - especially now that SB has made overclocking so simple.

khare · 14 Oct 2011 at 14:18

Here are the results for the PSU.

12v Rail reads 12.15v. Initially it goes to 12.3v but then stabilizes at 12.15v.
5v Rail reads 5.1v
3.3v Rail reads 3.3v.

EDIT: The PSU calculator in the link you posted recommends a 435w PSU, this one is 700w so I presume it's more than enough even though it's not the greatest of PSU's?

Plec · 14 Oct 2011 at 14:26

Well they're well within tolerance - but as mentioned earlier it's not a definitive test as it doesn't test the CPU under load - or indeed initial load (when you switch on) which is when the PSU is put under the most stress.

But at least it's one thing you can cross out (the test) - can you get your hands on a different PSU to test in your rig?