i7 920 - Hitting a brick wall at 4.2ghz.

W

WalkedAirplane

Limp Gawd
Joined
Apr 19, 2005
Messages
378
Relevant details:
- i7 920 D0 revision
- Watercooling loop, 2x120radiator, only for the CPU.
- Gigabyte UD3R motherboard
- OCZ Gold 6gb DDR3 1600mhz kit

So there is the setup, and here is where I start running into trouble:
- 3.8ghz, I can run this at stock voltage with load temps topping no higher than 55C.
- 4.0ghz, this took a minor voltage bump, again, temps were sub-60C
- 4.2ghz (200x21) took 1.325vcore or so (just got in the office, so only have scribbled notes) and otherwise virtually no effort. Load temps stayed right at about 60C.

Now trying to break this barrier and its just not working. Been trying to get to 4.4ghz, and nothing I try works. I HAVE managed to boot windows / run OCCT for a minute or two, but even that is requiring me to bump the QPI voltage, and the CPU voltage to 1.425v or further. Temps are getting up to 75C or so in this situation.

I've seen reference to needing additional tweaks / changes to get past this. I've made sure my RAM voltage and timings are good, I've ensured that all the multipliers are manually set, etc.

Any ideas? Any tweaks or modifications that are commonly needed to break past one of these walls? All the overclocking guides and information seem to be "hit 4.90ghz, sit there happy, never touch again". While I'm happy with 4.2ghz, I'd really like to see if I cant squeeze a bit more out of it :)

Thanks for any insight.
 
I'm having the same issue with mine. I think its more related to my board though. I was able to hit 4.5ghz on my old x58, but my current board wont play nice.
 
Hmph. I hope its not the UD3R, I picked it over the UD4P for pricing considerations and a few others having good luck with it. I would have shelled out for better otherwise.

I'll play some more tonight - anyone got any ideas?
 
Your temperatures are climbing too fast. Not sure it will improve your overclock - though pretty sure it will improve your stability - if you solve whatever's messing up your cooling.

2x 120 should be enough air through your radiator. What is your radiator and waterblock?
 
saw that your multiplier was set at 21. Maybe you should turn of turbo mode. Since turbo does its own tweaking of the voltage and such, it might make the setup unstable.

I could be wrong im still a rookie but its a thought
 
I've tried with and without turbo.

Temps seem about on par with what I have seen via googling around, but I'll see what I can find / tweak to lower it. 4.2ghz at 60C load seems about par for the course.

I'll check things out to see what could be causing it - I know the waterblock is seated nicely, but it is a new loop, so I'll do another round of bleeding to ensure there isnt any air in there.
 
Load temp at 4.2 with 1.325vcore looks pretty good, was that with prime95?
Also, do you still have HT enabled?
What is your memory set at for speed when you start going over 4.2?

I have been hitting similar results with my build, but 4.2 seems to be where I am running into issues.
 
Good luck breaking 200MHz BCLK on that board. Most X58 boards I've tested, including that one usually have a hard time going past 200MHz BCLK. Usually if they do it isn't by much. I'm not saying yours won't, but don't be surprised if you can't.
 
Thats OCCT, Prime95 gives pretty similar results.

HT is enabled.

Memory is set at 6.0 (so ends up being around 1300mhz as I get up there).

Probably going to work on bleeding my loop a bit more, make sure everything is clamped tight, make one more push to get above 4.2 - and if I fail, I'll just sit with 4.2ghz.

Any other suggestions still welcome, wont be able to work on this until later this evening anyways :)
 
I'm having trouble getting past 4.2GHz on my EVGA X58 3X SLI Classified. My CPU seems to want a great deal of voltage in order to hit those speeds. This causes extremely high temperatures past 4.2GHz.
 
I have heard people get good results, lower temp/voltage, and higher clock speeds with out HT.

Though, honestly for my every day OC, I would like HT enabled.


From my understanding as long as we are taking ram to a lower speed and keeping that mostly out of the equation, we still have to worry about blck, core speed, and qpi speed. Due to the 920 having locked multipliers on corespeed and qpi, its hard to find the real limitation of the cpu. So, we could be running into road blocks with the qpi speed but the rest of the core is ready to go higher.

For me at this point its more about curiosity to see what speeds I can get to, while using relatively "safe" voltages. I am more then happy with the speed I was able to get, as you also said you are fairly happy with your speeds. :)
 
Yeah, right now I'm leaning towards using 4.2ghz with HT enables for day-to-day use. I'm going to just do one more round of troubleshooting / tweaking tonight - if that doesnt net me more at comfortable temps /voltage, I'll just be happy with it. It'll still be plenty fast enough and I can get away with turning my fan speeds down I suppose :)
 
I'm having the same issue. I could hit 4.4ghz with 1.4v on my x58 SLI. I bought a UD3R to hold me over while I stepped up my x58 SLI to a Classified. I could not get the same exact chip to do 4.4ghz less than an hour after it was pulled from the x58 SLI and put into the UD3R. I thought it was the board, I'm still not sure it isn't, but I got my Classified yesterday and I'm still having trouble. I upped my CPU PLL to 1.3 and I ran it without vdroop and I was able to get it up to 4.25ghz, but I haven't had any luck with anything higher. 4.2ghz even was the most I was able to get out of this chip on the UD3R.

I think I'm just going to give up. I have another D0 that should be here tomorrow. I think I'm just going to pull this chip and sell it. I don't really need two D0's any way and I want to give a w3520 a try.

I used HT for all of these clocks.
 
Dekar12 said:
I have heard people get good results, lower temp/voltage, and higher clock speeds with out HT.

Though, honestly for my every day OC, I would like HT enabled.


From my understanding as long as we are taking ram to a lower speed and keeping that mostly out of the equation, we still have to worry about blck, core speed, and qpi speed. Due to the 920 having locked multipliers on corespeed and qpi, its hard to find the real limitation of the cpu. So, we could be running into road blocks with the qpi speed but the rest of the core is ready to go higher.

For me at this point its more about curiosity to see what speeds I can get to, while using relatively "safe" voltages. I am more then happy with the speed I was able to get, as you also said you are fairly happy with your speeds. :)
Click to expand...


About the QPI, QPI multiplier is locked but you CAN bring it up to 6.4GT/s. That was supposed to be locked originally but the production 920s have that unlocked. I've actually changed it to 6.4 on my EX58-DS4 on stock speeds, havent tested yet on my 3.5ghz overclock though so i'll check tonight to see if it gives me better performance scores.
 
1.5V with HT off just worked at 210x21 - 4400mhz.

Ran OCCT for a few minutes (definitely havent tested long-term stable or anything) but temps topped at about 65C.

Impressive the difference that HT makes. Very much so. Will keep playing once I'm done with the dentist!
 
No doubt about it. Turning HT off will certainly get you a better overclock. Though as far as I am concerned that really defeats the purpose of the Core i7 in the first place. At least, you lose out on one of its most compelling features aside from the IMC.
 
Benchmarks for HT are so varied, not sure how much of what I use actually is programmed with it in mind.

I am probably going to see if I cant get it HT stable, but we'll see where this adventure takes me.
 
Well HT doesn't always help and can hurt your performance in some applications, but I doubt that it hurts most of them to the point where you would notice any reduction in performance. The increased memory bandwidth and improved processor performance as a whole compared to Core 2 probably makes up for most of that. In other words; when comparing performance degredation in applications that do not work well with HT against those applications that do, you will probably find that the difference is largely academic. I doubt you'd notice it most of the time in real world usage. In benchmark tests, you may see noticable gains or drops in performance but I doubt its something a user would notice.
 
Stable at 4.4ghz with HT enabled. 1.5vcore in bios, 1.46 reported. Going to disable HT and see how far I can go.

I'll make a decison once I know what I can pull off with and without HT :)
 
4.6 looking stable at 1.5 and change. 1.504 in cpuz.

Looking like I'd need to push past 1.55v to go further. Anyone have experience at these volts? My temps are still low 70s load..
 
Currently:

4578w.jpg


Stable, decent temps off this. 50/50 on running that regularly, probably wont :)
 
try turning turbo mode off to see if it will increase stability at higher clocks, since turbo likes to play arround with voltage. which might throw stability out the window
 
You are referring to the Intel Turbo (21x multiplier) correct? Not the Gigabyte turbo option. If the latter, already disabled. Will try 20x230 at 1.5XX here in a few, see what happens..

edit: no luck, but trying 220 x 20 to see how it does for my temps / stability at 4.4ghz.

edit: 220 x 20 brought my temps down by 2-3C and running at lower voltage. Thanks for the tip. See if I cant push 4.5 out of it at reasonable numbers :)
 
Last edited:
im referring to the intel turbo mode, which i can see its turned on because of the 21st multiplier.
 
Cool glad it worked out for the best :p

Im a noob but im aware that if a feature plays arround a lot with voltage it might impact stability and temperature. But i was not expecting you to be able to run lower voltage too, nice nice!
 
Looks like 4.4 is where I'm sticking it out. Ended up going back to turbo - 220x20 worked well enough, but under Linpack stress testing it failed after a bit of time. 210 x 21 is stable as can be though. Slightly more volts, and a few degrees more, but still within acceptable range.

I can live with that.
 
spi.jpg


cpuzc.jpg


Where I ended up. Staying put here.

edit: Managed to hit 8.72 SuperPi on 4.6ghz, but stability / temps were troublesome in the end. Not really worth working on - 4.4ghz is pretty damn quick. ConvertXtoDVD runs in 5minutes for a full length movie, on my E8400 at 3.6ghz it would take in excess of 15min average.
 
Last edited:
Really nice overclock there, little high volts for my comfort, but if your temps are low you should be fine.

What do you have your memory running at?

And the more I was looking at HT stuff, it really does not seem to help a whole lot of programs, so I think I may go without as well.

I am still having stability issues with anything over 4ghz, games seem to be fine, but stress testing is not having it. Think I am either running into memory or qpi clocking issues.
 
I clocked my RAM down to 6x210 - 1260mhz. Bumping it to 1680 results in stability issues. This OCZ does not want to go above the rated speed whatsoever.

My benchmarks showed no performance difference though.

And yeah, temps are alright - its at home running an all day stability test now. If that passes, and the temps stay low - I'm set. Still going to play with lowering voltage though, if I can get it down, great - otherwise, I will live with it.
 
WalkedAirplane said:
I clocked my RAM down to 6x210 - 1260mhz. Bumping it to 1680 results in stability issues. This OCZ does not want to go above the rated speed whatsoever.

My benchmarks showed no performance difference though.

And yeah, temps are alright - its at home running an all day stability test now. If that passes, and the temps stay low - I'm set. Still going to play with lowering voltage though, if I can get it down, great - otherwise, I will live with it.
Click to expand...

I've noticed that with the Core i7 systems it seems pretty common that the memory won't run above its rated speed for some reason. I usually never get more than 5MHz over. This may be an issue with memory manufacturers only being able to validate these modules at their rated speed with no real room to go above that.
 
Dan_D said:
I've noticed that with the Core i7 systems it seems pretty common that the memory won't run above its rated speed for some reason. I usually never get more than 5MHz over. This may be an issue with memory manufacturers only being able to validate these modules at their rated speed with no real room to go above that.
Click to expand...
Statistically speaking on the manufacturing side, I wonder how much more difficult it is to get "good" (in the sense that they will go well above rated speed) sets of 3 modules as opposed to good sets of two.
 
enlightenedby42 said:
Statistically speaking on the manufacturing side, I wonder how much more difficult it is to get "good" (in the sense that they will go well above rated speed) sets of 3 modules as opposed to good sets of two.
Click to expand...

That's a good question. The Dominator GT's for example are their best binned parts today. I have to wonder how many modules are good enough to be Dominators and Dominator GT's compared to the regular RAM that goes out the door.
 
The crucial value ram 3x2gb on newegg is only like 70$ and they have Micron D9JNL chips on them. Same as the dominator gt's I think,
 
Alright, I dont think anyone actually cares about my clocks, but I ran a whole bunch of testing at 4.4ghz, got my voltage down stable.

Tonight I'm going to make a push for stable 4.6ghz. I'll be keeping this thread sparsely updated with my results. Going to see how far I can push the cheapest X58 board available. :)
 
4.6 is giving me a bitch of a time getting stable.

Any suggestions for settings to alter to try and get it to take? Superpi will run, and cpuz will validate, but occt crashes about a minute in.
 
Aright, 4.6 just wont do it. I threw 1.6vcore at it and 1.5vtt and it wouldnt stay stable. Those are my limits.

Currently working on getting 4.5ghz as low temp as I can and whatnot. May toss another 1-2bclk on it and see if I cant get the superpi in the 8's.
 
cpuzd.png


4557spi.jpg


MAXED. This is where I'll be running day to day, but I'm going to try and milk the voltages downward as much as I can. And also try to bump the ram timings a bit too. We will see where this goes.

Pretty happy with this on a $200 motherboard.
 
Looking really good, excellent work, get those volts down as much as possible to keep it cool. :)



I decided to settle on 4.2, since anything higher was just giving me bad memory multipliers, since I can't push my ram over 1600 by much.
 
Dekar12 said:
Looking really good, excellent work, get those volts down as much as possible to keep it cool. :)



I decided to settle on 4.2, since anything higher was just giving me bad memory multipliers, since I can't push my ram over 1600 by much.
Click to expand...


Volts ended up being a little higher than I'd like, but I am concious of the decision and watching temps extremely closely. Any degredation I risk, I do knowingly and accept with a smile on my face :)

4.5ghz left me running the ram at a 6.0 multiplier, but my benchmark testing showed minimal gains by running higher.
 
You must log in or register to reply here.
Back
Top