New Build Haswell Machine has a Fault, please help me find it!

b.cummins

b.cummins

b.cummins

Associate
Joined
23 Mar 2006
Posts
906
Location
Guildford, Surrey
Morning all,

So I have moved up a chipset to coincide with downsizing to m-atx and in doing so bought the following components to go into my SG10B:

Gigabyte GA-Z97MX-Gaming 5 m-atx motherboard
Haswell i5 4670K 3.4Ghz
G.Skill 8GB (2x4GB) DDR3 2400MHz TridentX Memory
Samsung EVO 840 250GB (Basic) Sata 3 SSD
WD Caviar Blue 1TB Sata 3 HDD

I also plonked in a couple of my existing components:
Pallit Geforce 570 GTX
Samsung Evo 830 60GB SSD
Coolermaster Silent Pro 600W Modular PSU

I also purchased a copy of Windows 8.1 OEM 64-bit to run it.

I'll mention everything that I've done as it might have some significance or part, however small!
So installation when smoothly, set the SATA controller to AHCI, no overclocks, windows installed without any issues or hiccups. Subsequently I've installed a few staple bits of software (steam, origin, ccleaner), chipset and hardware drivers and also had a go with the gigabyte utility software.
I then updated the motherboard BIOS from the F2 to F4 revision. After that I had a play with the settings using the BIOS (I enabled fast boot) and gigabytes' own OC windows utility. Picked their "moderate" overclock which was 4.2Ghz, rebooted, ran fine into windows. Picked the 2nd X.M.P profile for my RAM, ran fine, then tightened the command rate to 1T, again ran fine, ran a game for a couple of hours no issues.

I noticed the problem after leaving windows idle for a few hours. I'd come back and the mouse would respond, but no ability to interact with the icons in the taskbar, extremely slow, couldn't even draw selection squares on the desktop without heavy lag, took an age to CTRL+ALT+DEL, wouldn't run the start menu. In the end I had to do a hard reset.

So this is where I get concerned and decide to run Prime95. Immediately, the computer restarts when I initiated the test. So first thing I do is remove the overclock. Back into windows, the non-responsive behaviour persists on reboot this time - very odd. So I lax the memory to 2T. System seems fine. I started Prime95 last night and it ran overnight for 3/4 cores. 1 of the cores failed after 1hr29mins as it "rounded to 0.5 when expecting 0.4 or less". Same unresponsiveness again in the morning as well.

I am now at a point where I am not quite sure what is wrong. Any advice would be much appreciated!

My thoughts are running Prime95 again to see if the 3rd core fails at the same point and also maybe running memtest. Any ideas/suggestions welcome!
 
Good call on the memtest. Test each stick individually - ideally for 8 passes.

Also worth checking the hard disk and ssd's with crystal disk info and with something like seagate seatools.
 
Does it fail with everything at stock ?

Otherwise it seems to be a software issue. Most of that Gidabyte stuff is not helpful. I'd uninstall all that and try again.
 
Thanks - would you say it's worth doing them individually from the start or testing them together first then individually?

the SSD testing hadn't occurred to me so thanks for the tip there!

I should clarify the prime95 failure on 1 of the 4 cores was at stock speed. Not sure I can fully verify stock settings as I manually reverted the overclock in the BIOS.
 
Sounds like corrupted system files/disk due to overclocking both CPU and memory. Did you set a restore point before overclocking? If so try reverting to that, running disk and memory checks at stock settings etc. Try upping VCCSA to 1.25v as this helps immensely with Haswell and fast (>2133) RAM. If everything is OK, keep your RAM at stock and o/c the CPU. When you've determined that your overclock is stable increase your VCCSA and enable XMP for your RAM.
 
I have to say on one of my hard resets a program I run flagged up it had lost it's configuration file so that's probably not beyond plausible!

I have to wonder if it's worth a completely fresh install at this early stage with a CMOS reset and all..? What is the typical VCCSA setting?
 
Also make sure your Sata controller is in AHCI setting rather than NTFS. This can cause issues with some SSD's. Sounds to me like a memory issue but check the hard disc aswell as I know I have had plenty of issues before with SSD's.

My issues seemed to always be fixed though after formatting and reinstalling.
 
si_mats said:
~1.1v I believe. Did you try system restore with everything at stock?
Click to expand...

Thanks - I think as it is so fresh, and with nothing really lost besides a few downloads and a bit of time, I'm going to clear the CMOS, reformat and reinstall windows. If I still have problems after tests then, I will certainly look at things like upping that voltage before I completely give up on it.

HalfBakedJake said:
Also make sure your Sata controller is in AHCI setting rather than NTFS. This can cause issues with some SSD's. Sounds to me like a memory issue but check the hard disc aswell as I know I have had plenty of issues before with SSD's.

My issues seemed to always be fixed though after formatting and reinstalling.
Click to expand...

Mine is currently set to AHCI in the BIOS - there was no NTFS setting - unless this is something you set in windows?
 
I have:

Reset the CMOS
Reset the Sata Controller to RAID (for SRT)
Reinstalled windows 8.1 and set up SSD caching on a non-boot drive
I also have a 2nd kit of RAM to test, the Corsair Vengeance 8GB dual channel kit (1866MHZ)

So far at stock, albeit in "turbo" - that is running at 3.8Ghz, the system has been Prime95 stable for 16 hours.

However, having started memtest this morning before work, by the time I had left there were already 103 errors.

This was on the dual channel corsair kit running to the XMP profile settings.

My thoughts are:

Re-test at the default SPD in dual channel
If errors still return
Re-test 1 stick at a time

Either way does this mean the kit fails to run at the advertised XMP settings?

I also noticed memtest references the CPU caches in the first few tests.. does this mean the CPU cache memory has failed?!
 
b.cummins said:
I have:

Reset the CMOS
Reset the Sata Controller to RAID (for SRT)
Reinstalled windows 8.1 and set up SSD caching on a non-boot drive
I also have a 2nd kit of RAM to test, the Corsair Vengeance 8GB dual channel kit (1866MHZ)

So far at stock, albeit in "turbo" - that is running at 3.8Ghz, the system has been Prime95 stable for 16 hours.

However, having started memtest this morning before work, by the time I had left there were already 103 errors.

This was on the dual channel corsair kit running to the XMP profile settings.

My thoughts are:

Re-test at the default SPD in dual channel
If errors still return
Re-test 1 stick at a time

Either way does this mean the kit fails to run at the advertised XMP settings?

I also noticed memtest references the CPU caches in the first few tests.. does this mean the CPU cache memory has failed?!
Click to expand...


Set everything to defaults and test again. If you still get memory errors try each stick individually. If both sticks are ok individually then try them in the other memory slots as there is always a chance of a faulty memory slot.

There is also a slim chance that the issue could be cpu related as the memory controller is built into the i5 cpu's. It happens but is rare in my experience.
 
So the outcome of the memory testing was quite interesting.

I tested the corsair kit which happened to be my from my previous build.
In dual channel the 1st error occurred within 20 seconds. By the time the test was finished it was over 1150. This was in either dual slot configuration.
Then testing the sticks individually in every slot - no errors.

Finally I got round to testing my new G.Skill Ram.
Test completed at default speed - no errors.
Test completed @ XMP settings (2400MHz) and command rate set to 1T - no errors.

So this tells me all my new hardware is good - clearly the corsair RAM cannot run in dual channel mode and should be RMA'd?
 
Try loading the defaults in the BIOS and run prime then.

If you get errors with default settings then there's bound to be a hardware issue somewhere
 
You must log in or register to reply here.
Back
Top Bottom