help with 9550 overclock

Ha, I love your "I can't let it win" comment.

Anyway, believe it or not, the Small FFTs doesn't test the ram at all.

Large FFTs, according to OCCT, "will stress the CPU/Chipset/Memory chain with huge data transfers."

And the "Blend" test is a mixture of Small and Large FFTs (I'm personally assuming it's a 1:1 mixture).

Anyway, based on your current overclock and the voltage that you're using, it looks like both my E8400 and your Q9550 are similar overclockers. What I mean by this is that if you want a 1.170GHz overclock, I very much believe that at least the core voltage may have to come close to 1.3625V, or even exceed it a little bit. But, not to worry: you have the Xiggy protecting your CPU (I am hoping it has a 120mm fan on it).

However, having said all that, I am really quite curious: have you disabled the BIOS settings below?

• Spread Spectrum
• Intel Speedstep
• C1E Enhanced Halt State/C1 Enhanced/C1E
• CPU Thermal Control (it has options of "TM1", "TM2", and "TM1 & TM2")

If not, then be sure to do this as it will help tremendously. However, I admit that it does look like these are already disabled (or at least 1 or two are).

That's pretty much my limit. I hope this helps, though.

Quote:
I have disabled all 4 of those. I have some new info. I had been working with a NB of 1.38-1.42... neither affected stability. I decide to drop the FSB back to 400 and lower the NB to 1.10 and VTT to 1.20 with the RAM still at 1:1. Last night I tested it for 2 hours on large FFT.. pass. Then went to 410, pass all the way up to 475 all passing with only increasing the NB to 1.14. This is way less voltage than I had ever even tried on the NB. I usually just started it at 1.20. I pumped up the RAM all the way (about 1800) and tested until I passed 2 hrs with the NB only at 1.26. I am currently running P95 on blend and it has gone over 3 hours. If I can get the P95 to go error free overnight I will call it good as it is at 4.075 mhz and the voltages seem pretty good.

In my old OC I had the VTT at 1.34 (my last stable had been 1.32) and NB at 1.42... perhaps I had somehow managed to overvolt, especially the VTT i guess... I dunno.

Quote:
What happens when you run Prime95's Small FFTs test for 12 hours? And furthermore, what happens if you try running the Linpack test (like the Intel Burn Test, or the latest version of OCCT)? If you can run like an hour or two of Linpack without any problems, then you're certainly able to call it stable. However, if you run Intel Burn Test, then please know that you will see a column of numbers called "Residual<norm>" If there's any instability, then at least one of those numbers will be different from the rest. They must all be perfectly identical to indicate stability. [Source@XtremeSystems]

However, with OCCT, I don't know how to know whether or not it's perfectly stable other than just seeing it go for a long time without my system locking up or anything like that. But, you can tell when a new test starts when the CPU usage drops significantly. Mine was starting a new test approximately once every minute.

Anyway, as for the rest of the stuff you mentioned about voltages, I'm not sure, so I'll let somebody else take care of that who's actually qualified. But I do want to see how well this overclock does with a CPU stress test vs. a blend or a Large FFTs test.

Quote:
I'd love to try and help, but we need to focus on campb292 right now. So, make a new thread. And when you do, provide more information if you can. I mean, be detailed. You can even post CPU-Z screenshots (just use the Manage Attachments button beneath the Message text box. Scroll down; you'll see it). Not only that, but you should fill in your System Specs so it goes in your signature just like the rest of us (it'll greatly help others help you). We affectionately call them or "sig rigs".

Here's how to fill in your sig rig:

1. Click on "User CP" on the black bar at the top (alternatively, click on the Quick Links menu off to the left, and choose "User Control Panel" which is located on the 3rd dark bar down from the top)
2. Once you're in the User CP, click on "Edit System" located on the Left.
3. Fill it in as thoroughly as you can. Try not to leave anything out if you can.
4. If you have any additional systems, then use "Add System".
5. Once you're done, you can save it by either pressing Enter in any of those fields, or just click on "Save Changes" at the bottom. You'll know it's saved when you're automatically scrolled back to the top (there's no confirmation).

By creating a new thread, you'll significantly increase your visibility thereby simultaneously increasing your chances of getting the help you need. Good luck!

From the beginning...

Quote:
ding! Yes it does. I tried a couple other coolers, I luv that thing.

Quote:
The basic ones are like this:

Multiplier 8.5, FSB 475, DRAM timing 8-8-8-24, CPUV 1.3275, GTL .667, PLL 1.52, FSB Term 1.20, DRAMV 1.90 (that is the stock of mine), NB 1.26.

Quote:
One core junked out at 11hrs 13min. The other 3 were still ticking. I ran a couple runs of memtest which were fine, but I am going to bump the NB up to 1.28 and run large test with OCCT while I am out to breakfast and on a hike with my lovely wife. I never use the IntelBT, I will download that later and take a peak at it.

Quote:
Yeah, running a Linpack test is a great addition to running Small FFTs (not at the same time. Never multitask while stress testing. It's less accurate). In fact, I think the best Linpack tester I have ever seen is LinX as found on XtremeSystems. You gotta check this thing out. Intel Burn Test is fine, but this program makes it even more obvious when it's not stable.

In fact, I'm going to be running a LinX test in a few minutes. I may be able to post how it went.

I have gotten it up and stable just over 4ghz with 475 fsb and 1800 DDR3 at 7-7-7-20. It passed 4 hours of OCCT on large, 10 hours of P95 on blend, memtest86+ for 4 passes. I could not run the IBT on it, as I started it and canceled it as temps were inching their way up in the mid-70s. I booted up an old OC just to check it out (450fsb, like 3.8ghz) and passed a 5 5 and 10 runs. Dude... that program is outragous. On OCCT large I never get over 60-62 on any core over an hour. IBT on that 3.8 overclock hits 68 on the first run and the other 4 all slam into the mid-70s... almost 15c higher. That program is whacked.

QUESTION OF THE DAY:

Let's say machine A passes memtest86+ over 6 passes. Machine A passes 12 hours of P95 small FFT. Machine A fails OCCT large after 1 hour and 20 minutes. Does the fact that is passes memtest mean the memory and the mch are volted correctly? Does memtest test the memory and memory hub (northbridge)?

Quote:
These temperatures are away too high (+70C under a full load). Anyway, that's impressive overclock, but not very profitable with high temperatures, so, I would suggest to take a look at your heatsink.

Quote:
I am curious what temps you feel are too high. I was under the impression, both from this forum and intel specs, that 60-62C under heavy stress testing load (OCCT or P95) are perfectly fine temps for a 4ghz overclock. The only temps I have that are mid-70 are produced by Intel Burn Test, but that really seems irrelevant as it typically isn't even used to determine stability on an overclock that is so high.

Aside from the OCCT and P95 60-62C max I get, my common use/idle temps are in the low 30's and gaming temps are in the 40's. Intel's thermal spec is listed as 71 degrees for the 9550.

Quote:
I don't know the answer to the last question, however, based on what I've been taught, proper testing of memory should be done by doing several hours of the Blend test and then approximately 10 passes in Memtest (or in the opposite order). One cannot call the memory stable and error-free until it passes both several hours of the Blend test for stability checking and then Memtest for error checking. The Small FFTs test is only for CPU stress-testing in order to check for stability, so it is not a test that is used when an overclocker is focusing on their memory.

Regarding the Linpack test (see my sig), it is most definitely not "whacked". It's probably the best stress-testing tool available right now. It makes the CPU do much more complex calculations than the Small FFTs stress test. This is why the temps are 10-15°C higher. However, don't worry about it. It's not like you're making it go that high 24/7 every single day for 6 months straight. It's just a couple of hours, so it's harmless.

Furthermore, the Large FFTs test is not a CPU-only stress test, so it's actually inappropriate to compare the Large FFTs to a Linpack test.

Quote:
The Thermal Specification isn't talking about the core temp, it's actually talking about the temperature on the surface of the chip (where dust would land on the integrated heat spreader of the CPU if it were just sitting on a desk). Here's what you see if you click on Thermal Specification:

Quote:
And here's the explanation of TDP:

Quote:
So, as you increase the core voltage, you simultaneously increase the TDP which also simultaneously decreases the Thermal Specification. But if the temperature is at 71°C at the "geometric center on the topside of the processor integrated heat spreader", then how hot are the cores?

But this uncertainty may be why people say that, as a general rule, try to keep load temps under 70 for an overclock. But this may also have been taken from the average Thermal Specifcation listed for all of Intel's processors. Besides, a CPU will shut itself off if it gets too hot in order to protect itself.

So, I say do this: Run a Small FFTs stress test and get it to be stable at 12 hours with your temps averaging no more than like 70. If it's stable after about 12 hours, then run LinX for like 30-50 times (and use the largest "Problem size" that it will allow you to use, or rather, click on "All" for "Memory to use". This will be superior to the Linpack test in OCCT (and probably IBT too). 30 times may take up to an hour, to an hour and a half. But, it's said that 20-30 passes of Linpack at 100% stable is almost an equivalent of 12 hours of Small FFTs, except that the Small FFTs is a different test, so Linpack cannot replace it.

So, to summarize, use the Small FFTs (or for OCCT, the Small Data Set) and the Linpack for CPU stress testing, and then Blend (or Medium Data Set) and Memtest for memory stress testing/error checking (respectively). I don't know what good a Large FFTs stress test does, however, OCCT says that the Large Data Set stresses the CPU/Chipset/Memory chain with huge data transfers. But I have never seen anyone on Overclock.net recommend it (and I've actually been around since late January/early February of 2008).

And for temps, I say go with the temps you get from a Small FFTs stress test since this is the most common right now. I mean, when people say what their load temps are, almost everyone is going to assume that they mean the Small FFTs test.

And of course, your temps will be fairly high with a big overclock like that. If I were to overclock my CPU by that much, I'm sure my Linpack temps would go into the 70's too.

Anyway, that's pretty much all I have. And by the way, I apologize for my reply. I feel a little "off" today. hehe