My Dual Harper - Asus Z7S WS living review *56K warning*

[STEvil]

You have a good fan on the ram, right? And airflow for the northbridge too? They both get incredibly hot.

[milkmandan]

You have a good fan on the ram, right? And airflow for the northbridge too? They both get incredibly hot.

Yup I have the OCZ Memory cooler right over the RAM, and I have the thermalright hr-05/ifx for NB, both are nice and cool.

[jcool]

If you are getting errors in prime SmallFFT immediately, your Vcore is too low for that clockspeed.
Seems like the CPUs you got aren't very good, just like my E5440 were a great disappointment.
If you open up CPU-Z, what revision does it show? Should show something like C0 or C1, if it's B something you have your answer, B step Harpers suck (in my experience anyway)

[milkmandan]

If you are getting errors in prime SmallFFT immediately, your Vcore is too low for that clockspeed.
Seems like the CPUs you got aren't very good, just like my E5440 were a great disappointment.
If you open up CPU-Z, what revision does it show? Should show something like C0 or C1, if it's B something you have your answer, B step Harpers suck (in my experience anyway)

My is Revision C0 O_o
The core voltage is 0.080V

So this is a CPU problem not a ram problem?

Well either way this compy is still blazing fast O_O @ 3.0Ghz.
Heck i could even go back to 2.66Ghz :\ cinebench only went up by 4 seconds when i benched at 2.66. really dont think that 4 seconds is going to make my life phenomenally better anyways

creating the build log as i write this.. might take a few days as i want to be detailed

[Jaco]

The core voltage is 0.080V

So this is a CPU problem not a ram problem?

Core voltage is a little on the low side

just kidding , cpuZ reads it all wrong

[jcool]

Hum, if your CPUs are C0 (like mine) they should at least do FSB 400... strange.
You are right of course, these things are fast even at stock speeds. And 4s don't matter, but what does matter is the WCG performance. And there, every Mhz helps - that's why I keep pushing

You might wanna try FSB400 with 6x mult again (remember to disable EIST and C1E under CPU features for this) and run memtest from CD or USB stick. That way you'll see if it's the ram crapping out.

[milkmandan]

Hum, if your CPUs are C0 (like mine) they should at least do FSB 400... strange.
You are right of course, these things are fast even at stock speeds. And 4s don't matter, but what does matter is the WCG performance. And there, every Mhz helps - that's why I keep pushing

You might wanna try FSB400 with 6x mult again (remember to disable EIST and C1E under CPU features for this) and run memtest from CD or USB stick. That way you'll see if it's the ram crapping out.

memtest from CD/USB stick? you mean prime?

and i dont see EIST in the CPU config menu O_o

[Jaco]

memtest from CD/USB stick? you mean prime?

Dan,


Jcool is talking about memtest86. look over here : http://www.memtest86.com/

[milkmandan]

Dan,
Jcool is talking about memtest86. look over here : http://www.memtest86.com/

argh the toils of having to boot from CD
gna need to reconnect the IDE drive, which i really don't wanna do haha..

tried 385 x 6 and with x 8 and it failed during a cinebench run.
tried 380 x 6 and with x 8 and it ran fine.

i'll try memtest when i reconnect my IDE drive.
can't seem to get memtest on my usb stick.

----

it could be i got some DDR2-667 ram? maybe newegg was wrong or Cruicial lied about that DDR2-800 speed

should I be worried that prime95 crashes when i run it at 375 x 8? Cinebench10 is fine :\ On normal use, will i run into problems?

[Jaco]

When your system is not prime stable , you can experience all sorts of instability. Random reboots , lockups , games crashing , WCG errors , etc...
I wouldn't recommend it . Not primestable = failed overclock in my book .
I think jcool will agree with me

[milkmandan]

Any recommendations for fix?
or other BIOS settings i should be aware of?

I still can't find EIST in the BIOS :\

been using the 64bit version of prime95 as well.

[Jaco]

EIST is called 'Intel speedstep technology' in the bios.
just above 'clockgenerator'
EIST stands for 'enhanced intel speedstep technology'

You could try to raise the NB voltage and/or FBdimm voltage.
(jcool is running stock voltages , I use 1.35v NB and 1.86v FBdimm , ymmv)

[milkmandan]

EIST is called 'Intel speedstep technology' in the bios.
just above 'clockgenerator'
EIST stands for 'enhanced intel speedstep technology'

You could try to raise the NB voltage and/or FBdimm voltage.
(jcool is running stock voltages , I use 1.35v NB and 1.86v FBdimm , ymmv)

Ahhh the speedstep, i turned that off a while ago.

what does tweaking these voltages do?
I tried it, and not my prime doesn't CRASH anymore, but one of the workerthreads just doesn't work :\, after some odd 30 seconds of running prime, it says 'error: rounding was 0.5, expected less than 0.4' and it stops, all the other cores, just keep pumping, no problems. :\ OC at 375 right now.

[jcool]

If prime smallFFT fails at FSB 375 on Core 0, then that core can't handle the Mhz (that error you describe is typical for too low Vcore). Back down to 370 or something and try again.
Running a system on WCG that fails prime that fast (even if it's only 1 core) is not advisable, since it can errors of all kinds (like Jaco said).

I'm still not clear as to what fails in your rig. Just follow the testing instructions I gave you and you'll know exactly what's holding you back.

[Jaco]

I would try memtest first , I suspect something's wrong with the ram.
Or maybe you need more NB voltage , but I guess you tried that already.

[milkmandan]

Running memtest now....30% done, no problems so far. I am hoping RAM will pass 100%
if thats the case, I might just be unlucky with some CPUs that aren't that overclockable...

NB voltage? guess i should play around with the voltages more?
What about the dual adjustment one?

tweaking SB voltage has no effect right?

[D_A]

Don't forget to let it do a few runs through. Just one pass won't pick up deep heat related issues.

[milkmandan]

Don't forget to let it do a few runs through. Just one pass won't pick up deep heat related issues.

a few runs? this test as already ran for 2.4 hours O_o and its still going...

the total completion is now at 80% still no errors

[D_A]

You'll see a %age completion. Once that hits 100% it starts again from the beginning but it indexes a total number of "passes".

I had a RAM problem on one machine a while ago. It would boot and SEEMED to run just fine, but it would frapp out WCG jobs randomly. I had to let it Memtest run through 8 complete passes before I could pinpoint the problem. One module had "issues" but only once it warmed up enough.
Of course the modules warmed up much faster running CPUburn, but then I couldn't tell if it was a CPU, northbridge or RAM issue.

If you want to know this thing is 24/7 stable, you should at least consider just leaving it running memtest overnight or longer.

[milkmandan]

You'll see a %age completion. Once that hits 100% it starts again from the beginning but it indexes a total number of "passes".

I had a RAM problem on one machine a while ago. It would boot and SEEMED to run just fine, but it would frapp out WCG jobs randomly. I had to let it Memtest run through 8 complete passes before I could pinpoint the problem. One module had "issues" but only once it warmed up enough.
Of course the modules warmed up much faster running CPUburn, but then I couldn't tell if it was a CPU, northbridge or RAM issue.

If you want to know this thing is 24/7 stable, you should at least consider just leaving it running memtest overnight or longer.

will memtest tell me exactly which module is problematic? or what?
each pass is taking like 4 hours O_O is that how long urs was?

[D_A]

How long each pass takes will vary depending on how fast you're running your RAM and of course how much you have. My machine was somewhat smaller than yours is, so it only took about an hour for a pass.

As for reporting, it will tell you what test the error happened on, what memory address is happened at (in hex) what it was supposed to read and what it DID read. Since I don't speak hex that well I just pulled modules till I found the culprit, but if you're running in single channel mode, higher addresses tend to be on the second (or later) sticks. In dual channel I don't know. In quad channel, I hate you for having hardware that fancy and I REALLY have no idea.

[milkmandan]

well #1 pass went through successfully with no issues (took 3 hours), i'll leave it running and see if anything else comes up
i am running this at 400FSB with 6 multiplier. :\

no issues so far.

leaning on the fact that maybe it is not a RAM issue, any suggestions I can do at this point? More voltage to the NB? over the Dual Adjustment? vice versa?
---

I read up on some other forums, with my problem with one core crapping out and reporting in prime95 "Rounding was 0.5, expected less than 0.4" they say not enough voltage to the cpu :\ besides a volt mod, will adding voltage to NB, Dual Adjustment, help at all?

[STEvil]

Put your fingers on the ram heatspreaders.. if you cant hold them there then they're running too hot. I suspect this is part of the issue because that OCZ cooler is rather small and pitiful iirc...

[milkmandan]

Put your fingers on the ram heatspreaders.. if you cant hold them there then they're running too hot. I suspect this is part of the issue because that OCZ cooler is rather small and pitiful iirc...

I can place and hold my fingers on the heat-spreaders just fine and hold them as long as i need. Not even as hot as a car steering wheel on a baking day. I got some good airflow down there as I have a 120mm fan next to the RAM as well. and I have 2 more 120mm fans blowing at the OCZ cooler as well. Though aren't FB-DImms suppose to run hot? :\

OCZ works fine for me :\ It's better than the dominator in my opinion.
If ram heat is the issue, i prolly should purchase some heatsinks like jcool. but not until i find out what the problem is.

2nd pass for memtest, 70% done and no problems so far :\

-------
edit:
aww now this is messed up...

I am underclocking, 2.25GHz, at 375FSBx 6. And prime is giving me that stupid rounding error again.
I've even fed the NB 1.40V, which is the highest setting O_O...

I am unstable even at underclocking. wow....

memtest ran perfectly at 400x6 :\ 3 complete passes, 6 hours of testing, no errors *sigh*

I am getting a strong itch that maybe although my RAM is 'okay', the speed clock is lying, and that its not really 800 but 667.

the BIOS - system information - memory information shows my RAM at 667Mhz, but i duno if thats because of the E5430 :\
what does ur say jcool? I know you got E5420's, so theoretically, ur BIOS should also report your RAM as 667. :\

----
any more suggestions?

at stock voltages....going to slowly step up from 333 and see when prime starts to crap out on me.

[jcool]

375 x 6 gives the same errors as 375 with max mult? Ok then, it's probably VTT. Even though it shouldn't be at that kind of FSB with C0 chips...
Maybe (just maybe) your prime version is ed up? You said you are using 64bit prime.. don't. Use the normal 32bit V25.5 or 25.6 and retest.
If the problem persists, you will need to run FSB 370 at max multiplier to remain stable