Overclocking the UD5 with 12gb of ram.

Decided to try lowering CPU PLL from 1.8 to 1.7 and QPI from 1.35 to 1.3 in the hope that doing so would let me hit 4.0GHz stable, as I'd gathered that decreasing them can help with stability.

Did so. Booted into Windows, all seemed to be going well and then bam, irrecoverable hardware error. I've put them back up to 1.8 and 1.35 respectively now, BUT: I was running a virtual machine and Spotify (amongst other things) when it happened and now, VirtualBox won't even start and Spotify decided that it was going to revoke my Offline playlist synchronisation ability. I then enabled it again, and in doing so it decided to delete entirely, without trace, *TWENTY FIVE JIGGABYTES* of synchronised offline music :(. I used up half of this months broadband download allowance to sync that!

So essentially a warning to anyone who doesn't realise - and you probably all do, but I thought I'd share my story - overclocking kills! :p

I'm now back on the same settings that I posted earlier and that have been stable for the last week, but have pushed up the base clock to 190 for 3.8GHz (from 3.6) and am going to see how this does - it's not 4.0 but it's closer.

Currently trying to repair my virtual machine, and then sleep will be in order while this thing runs IBT as a first test.
 
Last edited:
Ouch! That's properly bad luck :(

A reminder for the rest of us to only run stress testing apps on a potentially very unstable system, having it nuke spotify sucks. Can't help with the recovery I'm afraid, I know close to nothing about windows. Hope it works out for you,

One thought is 25gb of music doesn't just vanish, this is probably recoverable. Please do report back, I'll be plagued with guilt if my advice inadvertently cost you a fortnights bandwidth.

Decided to have a play with ram. Starting from the stable 4ghz settings, except with bsck dropped to 150 to help ensure the cpu isn't failing on me. Unclock "auto" sets it at 3ghz for x6, x8 or x10 ram, seems reasonable. Were this left at x14 it's likely to go a bit wrong at higher ram multipliers. Appended test results to the OP, basically the ram is working well enough to pass memtest.

Trying 4.4 still. Saw 78 degrees on one of the cores, suspect my cooling to be inadequate. Not the end of the world as I have bigger fans and better thermal paste waiting for me to find a milling machine and the courage to write off the warranty on my i7, but still a bit of a surprise.
 
Last edited:
JonJ678 said:
One thought is 25gb of music doesn't just vanish, this is probably recoverable. Please do report back, I'll be plagued with guilt if my advice inadvertently cost you a fortnights bandwidth.
Click to expand...
It wasn't when it locked up that the music vanished. I think the lock up made maybe one or two cache related files corrupt and as a consequence, when I put the machine back to known stable settings and booted back up again, I launched Spotify and a message telling me that my offline synchronisation capability had been disabled came up when I logged in. I noticed that it had just reset my cache location to the default, so I tried putting it back to the location I use (on a separate partition) and on doing so it deleted everything in it. So the lock up didn't directly cause the deletion, it was me trying to fix the situation. I'm going back to uni either tomorrow or later in week and surprisingly the bandwidth is actually cheaper there than at home (get 50GB included at home and then it's £1/GB!), so I'll just resynchronise at uni. I wouldn't worry that my Spotify cache went AWOL Jon.

I've got the virtual machines working again. Turns out my VirtualBox.xml was completely corrupt (was just showing up as null bytes), so I deleted all my VirtualBox configuration and rebuilt my machines from scratch. The virtual disks seem fine, in that both my VirtualBox machines boot fine.

Interestingly I noticed this morning that one of my Steam cache files was downloading at ~25% - too low a completion level to be just an update, so I think that must have been corrupted too. Just getting a little worried now about what else may have been open at the time and is corrupt as a result, so once I'm sure this configuration is stable I'll copy all my essential documents back from a backup just to be sure they are consistent - though I didn't have any open when it locked up so they should be fine anyway.

Moving back onto topic, I let it do 200 runs of IBT overnight at this new OC of 3.8 and it had passed when I woke up, which is a good sign but not all that surprising given that it was 24 hour Prime stable and LinX stable at 4.0 - as you seem to have found Jon it seems that IBT/LinX/Prime are no substitute for days of running, which is what I'm just waiting on now.
 
That's quite widespread corruption but it makes sense. Especially the virtual machines, I've had the xml files corrupt on me too. If you've got a few of them running similar hardware it can be a lot quicker to make one, copy and paste, rename, change the uuid. I've got six or so running operating systems in varying states of distress, but all based on the same two core - 2gb template.

It's always nice to pass stability tests even if you don't quite trust them :)
Recently found this thread on some dfi forums, it has a fairly convincing argument for using x19 + turbo instead of x20 + turbo though I don't really follow it. Can't work out what I'm meant to be setting in the bios. Oh, qpi pll doesn't seem to help me, increasing it a notch ruins stability.
 
Last edited:
JonJ678 said:
Recently found this thread on some dfi forums, it has a fairly convincing argument for using x19 + turbo instead of x20 + turbo though I don't really follow it. Can't work out what I'm meant to be setting in the bios. Oh, qpi pll doesn't seem to help me, increasing it a notch ruins stability.
Click to expand...
When I can next justify the downtime, I may try my 3.6 known stable voltage settings with a base clock of 200, multiplier of 21 and disable any power management stuff (the latest UD5 BIOS lets you set a multiplier of 21 - perhaps for the reason that odd multipliers are meant to be more stable). Perhaps I'll have better luck at 4.2 than at 4.0!
 
Walked back to the PC about half an hour ago after leaving it for a bit to find both screens black and the whole thing locked up. Interestingly the room was pretty hot - it turns out the thermostat on my radiator is hopeless. Perhaps the lock ups that have been occurring with these IBT/LinX/Prime stable OCs I've done are due to the room getting a bit hot. That said, I don't want an OC that fails in high room temperatures - obviously an automatic shutdown in the event of overheating is fine, in fact desirable, but lock ups are not.

Given that I was going to have to change something then, I took the opportunity to try out a base clock of 200 and multiplier of 21. It failed within a few minutes of IBT, so I'm now running IBT with a 200 base clock and multiplier of 18 - I want to identify for sure where the bottleneck is. If this succeeds and runs stable over the next few days, at least I will then know that I can run a base clock of 200 OK, and that my memory settings are good - then it will presumably be a case of more Vcore, if I can do so without taking temperatures beyond reasonable levels.
 
Any luck? It sounds possible that it's stable at low temps but freezes at higher ones, stabilty on mine is pretty much absent over 70 degrees in hwmonitor.

Mine is folding peacefully, want to see it complete a unit before I try anything new with it.
 
200 base clock at 3.6 was good for a while but then locked up when the room got hot.

I then realised, the only reason I decided to overclock this thing was so that I can turn around bigadv WUs in reasonable time and to be able to meet the deadlines as well as actually use the power of this machine. At 3.6GHz with a 180 base clock and 20x multiplier, I can do those things, and I've never had it fail due to high room temperatures, so I'm going to settle for 20*180=3600. Also I came back to uni today which means workload will increase rapidly soon, and having the distraction of 'Oh if I could just reach 4GHz...' is something I'll be better off without.

Good luck and have fun to all the rest of you i7 overclockers with 12GB RAM!
 
Sounds good to me. My idea that cad work will go faster at 4ghz justifying spending days overclocking is appealing but clearly a false economy.

When term ends, hopefully I'll have something more concrete for you to go on. Good luck and work hard :)
 
Is 201 x 20 more stable than 200 x 20 with 12GB ?

It's so damn hard getting 12GB 1600MHz linx stable with my Core i7 920 on an Asus P6T. Been trying for months and months :(

So I just dropped down to 1200MHz with tight timings.
 
Last edited:
Very nice it's nice to see the i7 920 is still a great OC, l'm quite happy at 4.2ghZ even thuo custom H2o cooled. Who know's might get the urge to see how far my i7 920 will go!!!!

My Vcore Bios is 1.3125v at 4.2GHz, QPI/VTT Volts - 1.315.
 
You must log in or register to reply here.
Back
Top Bottom