Motherboard: Asus P6T Deluxe V2 CPU: i7 920 D0 (SLEBJ) Cooling: Titan Fenrir Overclocked Speed: 3.6 GHz BCLK: 180 MHz DRAM Speed: 1443 MHz Turbo Mode: Disabled Hyper Threading: Enabled Speedstep: Enabled Loadline Calibration: Disabled CPU Voltage: 1.15 CPU PLL Voltage: 1.80 QPI/DRam Voltage: 1.4 DRam Voltage: 1.66 At the mo I'm gunning for a nice 3.8 24/7 overclock while keep my nice low CPU voltage but this setup idles at ~48, prime95's at ~73, and Linx at ~81.
That's weird. I'm at 1.25v for 3.8 and my temps are around 70c under load. The Fenrir is a great cooler, so that's baffling. Your room hot or something?
You've tried to help me cool down my PC before here: http://forums.bit-tech.net/showthread.php?t=169413 and nothing I do seems to help, since then I've taken off my HSF, cleaned off old TIM, reapplied and reseated. Changed the fan on the fenrir, sped up and slowed down my middle fans to try and fiddle with the airflow.....but I've just learnt to live with it tbh :S My rooms not hot, right now my feet are freezing
TBH, I don't think it's that hot. My DO idles around the 40-45ish mark no matter how much I lower the voltages. Mine even idles at these temps without overclocking. I am tempted though to lap my swiftech GTZ. It was designed for the old socket 775 and is slightly convex. After reading the thread on the Thermaltake True, I think many heatsinks have this problem. I've been looking on other forums and these temps are common, as are 70's at load. I just think that Pookeyhead's got lucky with his.
Feel somewhat better now With a better HSF (although only a little) at a slightly faster speed you were getting slightly higher temps
Motherboard: Asus P6T Deluxe CPU: i7 920 D0 (SLEBJ) Cooling used: TRUE Overclocked speed:3.8GHz Baseclock (bclck):200 DRam speed:1603 Turbo Mode (on/off): off Hyper-threading (on/off): on Speedstep (on/off): On HPET (on/off):? Loadline Calibration/VDroop (on/off): off CPU Voltage:1.20 CPU PLL Voltage:1.8 QPI/DRam Voltage:1.3 DRam Voltage:1.66 Just an update. No faster than it was, but been tweaking things.. now fully stable at 1.2v, so it runs even cooler.... Mmmm... Brrrrr.... Incidentally, recently got another TRUE fan holder and fan, and I've been testing over teh past day or so, and I can categorically say.... it does NOTHING using 2 fans. Temps are exactly the same.
Hey Pookeyhead, could you possibly get a snapshot of your overclocking pages with your phone or something so I can see your IOH and ICH voltage, spread spectrum etc, I'm getting ever closer to 3.8 Also, have you changed *any* of the advanced settings such as C1E function etc? Your're like my guru, same mobo, CPU and very similar cooler so I'm trying to see where yours differs from mine and the extra heat etc. Cheers ...aha I spy 200 x 19 not 190 x 20...explains your 1600 RAM
Okay...I have 3.8 stable and my cores max out at 79 under Prime95, and I'm too scared to run LinX. Is that good, bad or about average?
There you go Nope... all stock. I've found it makes little difference.... just put all back to stock. I've benched this at 4.6 with all those settings on stock. Yeah.. I noticed that 20x multi was less stable than 19x, despite me having a higher BCLK... (shrug).. no idea why. It's not terrible. Remind me again... you got a D0 or a C0? That's about what I was getting with my C0 with my unlapped TRUE. From my experience, you'll get around 83C in LinX. Worth finding out because it's a better test. If you can go 12 hours in LinX I'd consider it fully stable.
+rep; Thanks a lot pal, I have the D0 revision to compound the mystery of my high temps :S To only immediate difference I notice is you have a lower DRAM voltage (but then again different RAM, XMP gives mine 1.4) and you have a higher ICH voltage. Thanks again!
Well I'll be buggered, it's much more stable and slightly cooler than it was. Last night at 1.2V it would survive Prime95 for more than say 10 mins before it started giving me errors, today its been going for about 40 mins quite happily. The only thing different is my PCIE voltages aren't auto anymore and and my southbridge has .1 more volts. Going to LinX it for 2 hours or so now to check it defo is stable like this, then try and bring the DRAM voltage down some.
I see your UCLK is at 3408. As I don't have an Asus board I'm not sure how you go about changing that; on the Gigabyte board it's just a matter of changing a multiplier. If you can increase that, you should find that your benchmarking will improve. Unfortunately, I have found that my DO stepping does not like this to go over 4,000. This is a common issue with the DO stepping, not so much with the CO stepping. Worth trying to see if you can squeeze that little bit extra out of it.
I'm defo on the home stretch of my O/C'ing but I have one last hurdle to overcome...LinX I can run any of the torture tests in Prime95 indefinatley, for example the "Blend" torture test is coming up for the 2 hour mark now and its not put a foot wrong...LinX on the other hand crashes after 5 iterations.. I'm gonna leave Prime going a bit more, but if my PC is perfectly stable in prime, and not in LinX, is it still okay to call it finished and use it normally? I ask as when I used to O/C my C2Q I'd never heard of LinX and I used to think that 2-3 hours of Prime meant a stable system
Linpack (LinX, IntelBurnTest) is supposed to be a better indicator of stability than Prime95, but it varies - some people are 24 hours Prime stable but fail out of Linpack on their first loops, and some people can run 10 loops of Linpack with max memory but can't run Prime for more than a few hours. In general though, Linpack seems to be a better indicator of stability because more people are Prime stable/Linpack unstable than vice versa. Linpack's also faster, and doesn't suck up a full day's worth of CPU power just to test stability. I would personally run Prime for 24 hours and Linpack 15 loops on max memory to be absolutely sure of stability, but if you don't have the time or are too lazy just Linpack should be fine. On a slightly unrelated note, IntelBurnTest failed out on Very High on the second loop at stock clocks on my i7 920 but did 10 loops on Very High after I bumped the bclk to 160 .
20 loop of linpack and my 3.8 system is stable. My settings are identical to Pokeyhead except mine needed 1.215 VCore instead of 1.2, and my RAM requires 1.4. This could be lower though, now I've sussed it was the VCore too low crashing it I can try to bring the RAM down again. My system idles at an average across all 4 of 53.5, primes at 79-82, and LinX at 85-88, now is this too toasty and should I revert back to my happy 3.6 or am I safe in the knowledge my PC will never see the sort of load that LinX puts on it. And you were happy to run your system 24/7 with this O/C? I will shut up with all the nooby questions soon; just I poured so much into this PC I want the most out of it, and I'm desperate not to cook it
It will probably be fine in every day use, and unlikely to give you a problem, but you'll always have that nagging thought in the back of your mind that it couldn't run linpack. If that bothered you.. give it slightly more volts until it gets stable. Yeah I know all about UCLK (uncore). I used to run 3.6, and yes, it gives a marked increase. Now I have mine utterly bombproof I'll start to increase uncore in steps. Nah... that's fine. I'd consider anything under 85 OK for 24/7. Remember, you can take your prime95 temps as a more realistic idea of hard core gaming temps. NOTHING punishes your rig like LinX on a daily basis.
Motherboard: Asus P6T Deluxe CPU: i7 920 D0 (SLEBJ) Cooling used: TRUE Overclocked speed:3.8GHz Baseclock (bclck):200 DRam speed:1603 Turbo Mode (on/off): off Hyper-threading (on/off): on Speedstep (on/off): On HPET (on/off):? Loadline Calibration/VDroop (on/off): off CPU Voltage:1.20 CPU PLL Voltage:1.8 QPI/DRam Voltage:1.35 DRam Voltage:1.66 OK.. been playing with Uncore (UCLK). Interesting. It has a noticeable effect on memory bandwidth, but not MASSIVE... just noticable. Everest memory marks: uncore 3.4: read - 17817 write - 14827 copy - 21342 uncore 3.6: read - 18411 write - 15656 copy - 22106 uncore 3.8: read - 18287 write - 16306 copy - 22732 At 3.8 and higher the read speeds start to drop off, but everything else still shows gains. Higher than 3.8 and everything starts to drop off.. at 4GHz Uncore there's a slight decrease in memory bandwidth, and past 4GHz I get stability issues. I've settled on 3.8 for now. I've had to increase QPI/DRAM voltage from 1.3 to 1.35 which has had increased my temps slightly (from highest core temp 65ish to around 68ish (but this was with LinX so it doesn't really compare to my last screenshot). It would appear the sweet spot for Uncore, is 2xRAM speed + BCLKx3. So if your ram is at 1600, a 3800 uncore is the sweet spot. I'll stress it overnight and if it's stable, I'm done. Any hint of trouble, and it goes back to 3.6, or even 3.4... as the gains aren't really worth the hassle. Next stop water cooling I think, then I'll head for a 4.3GHz 24/7... but on air.. I'm happy now. I've also tried for a DDR3 2000 ram clock... but this Dominator stuff aint all that... flatly refuses, even at 9-9-9-24-2T with 1.8v. I wish I'd got 2000MHz RAM now. There may be some RAM for sale soon LOL There you go..... It's all part of the service I provide
