OCing Trouble

[Ruger22C]

E8400, stock cooling. I can get 3.5GHZ @ 48C / load.
DDR2 800 RAM.

I can get 390x9 (3.51ghz) with no problems so far. Any higher, and I PrimeTest95 starts producing errors under 2 minutes.

If I increase the voltage from 1.15 to 1.162 or so, I get immediate BSOD when Vista starts to load.

What now?


EDIT: Sorry. 390 is when the trouble starts.
I can get 380 and maybe 385 without problem.

 

[Drsignguy]

Well, I am going to guess that you will need a bit more voltage bump than that as now the chip is starting to require it. Also, you may need to bump up the mch and northbridge a notch.

How much Ram do you have and is your ram set at 1:1 ratio? What motherboard are you using.

Stock cooling will only get you so far as now that you have to start cranking up the voltage as heat will be an issue.


BTW, have you read the sticky at the top of this forum? It will answer many questions that you may have like temps, how to's and others.

 

[Ruger22C]

OK. I'll run 3.4ghz then...
RAM ratio? No idea... It's at 912 or something like that, it was changed automatically? Is that 1:1?
4gB 2.1v (stock)

 

[Drsignguy]

Originally posted by: Ruger22C
OK. I'll run 3.4ghz then...
RAM ratio? No idea... It's at 912 or something like that, it was changed automatically? Is that 1:1?
4gB 2.1v (stock)

Well this depends on the ram you have, 800, 1000, 1066? If 800Mhz, No, its is not set to 1:1 as it is now overclocked. If it is a 1000Mhz, then you have a bit more headroom.
In your Bios, there are settings for this and you have to go in and change them. Auto settings are not really good when you overclock so you are now at the piont that you should start setting these manually.

Edit, I think you are using a Gigabyte board, ( tell us what one ). In bios settings under the MIT area, you will find that the ram settings to be different. Just set the Ram at 2.00. This is the 1:1 ratio setting for Gigabyte motherboards.

 

[BonzaiDuck]

Originally posted by: Drsignguy

Originally posted by: Ruger22C
OK. I'll run 3.4ghz then...
RAM ratio? No idea... It's at 912 or something like that, it was changed automatically? Is that 1:1?
4gB 2.1v (stock)

Well this depends on the ram you have, 800, 1000, 1066? If 800Mhz, No, its is not set to 1:1 as it is now overclocked. If it is a 1000Mhz, then you have a bit more headroom.
In your Bios, there are settings for this and you have to go in and change them. Auto settings are not really good when you overclock so you are now at the piont that you should start setting these manually.

Edit, I think you are using a Gigabyte board, ( tell us what one ). In bios settings under the MIT area, you will find that the ram settings to be different. Just set the Ram at 2.00. This is the 1:1 ratio setting for Gigabyte motherboards.

Ruger (.22 cal.!!) needs to read Graysky's sticky. I'm less familiar with Gigabyte boards, although I have a couple and I set everything manually. Per CPU : RAM ratios, I find it is too much trouble to interpret the board's "canned" settings. So I simply unlink them, do the integer math to dial in QDR-FSB and DDR-RAM speeds, and carefully calculate any ratio that deviates from 1:1. With 1:1, you'd simply set DDR-RAM to half the QDR-FSB or conversely FSB to double the DDR speed.

If I want a 4:5 ratio, I divide the target CPU_FSB ("external" or "reference" frequency) by 4 and multiply the result by 5 to get the resulting RAM frequency (half the DDR speed.) So 333 Mhz (CPU default for a Wolfdale core) gives 416.5 Mhz for the RAM or a DDR= 832+ Mhz.

Beyond that, You want to tweak the RAM voltage and/or loosen the timings if the RAM generates an error at these speeds. And for a slightly quicker-sloppier approach to over-clocking than Graysky's method, it becomes a matter of trial and error on either loosening the timings or bumping up the voltage a notch.

READ GRAYSKY'S STICKY.

Also, I speculate that your VCORE is too low, given what we know here about that processor. If you look at the retail box, you'll see a "maximum voltage" which is a conservative estimate of near-total safety. On the E8400, I think it is about 1.25V to 1.26V. Many of us agree that this spec is more significant for load voltages than for idle, and since there is a vDroop on motherboards of between a hundredth and a few-hundredths of a volt, you could kick up your set voltage to 1.30V without doing any harm. In my saying that, there are many here who would say I'm a sissy for being too cautious. Many of us are running the E8600 @ 1.32+V to get 4.0 Ghz stability. On my rig, that means about 1.28V "sensor-reading" and 1.27V per the sensor-reading at load.

OF course you want to minimize that setting as much as possible.

Also, in the higher ranges for some motherboards close to or beyond 1,600 FSB (CPUFSB=400 Mhz), you'll need to adjust the CPU_FSB, HT, SPP and MCP voltages slightly. Usually, the culprit is among the first two, and CPU_FSB may be more important at that speed. But read the Anandtech blog about how they burned out a QX9650 processor by applying voltages above 1.5V, and how both the testers and Intel suggest anything of 1.45V or above is "taking risk."

 

[Ruger22C]

I went up to 1.18v, BSOD went away... but then windows loaded half-way and my PC restarted .. -_- I clocked the RAM down to 780mhz to make sure it wasn't that.

 

[BonzaiDuck]

Originally posted by: Ruger22C
I went up to 1.18v, BSOD went away... but then windows loaded half-way and my PC restarted .. -_- I clocked the RAM down to 780mhz to make sure it wasn't that.

And? . . . .

As for me, I have to revise what I said. I've seen a lot of posts on the "show your results" thread" and used 1.32V as a starting point. It's still rock-solid at a setting of 1.30V for 4 Ghz. Since the voltage monitor values are now within the 1.26V spec, I'm wondering if it isn't possible to knock it down further. I still haven't found a failure point so I know what the minimum good voltage is.

EDIT: Before I let my runaway enthusiasm confuse you, I've got the E8600.

Many people have clocked the E8400 to 4 Ghz. I had more difficulty with it between 3.75 and 3.82 Ghz, and 3.6 Ghz was a great setting for it.

EDIT: I just checked my over-clock log entries for the E8400. To get to 3.6 Ghz for a 9-hour PRIME95 run with deliberate termination and 0 errs, 0 warns, I had to set the VCORE to 1.3125V and the voltage reading of 1.29V. Figure though that this is a "luck-of-the-draw" on the processor you get, depends on motherboard and other things.

In the end, I'm less interested in producing drag-strip speed-records and more inclined to find a nice balance between processor and RAM speeds with decently tight timings. And voltages set so as not to kill any of my parts.

 

[Drsignguy]

Originally posted by: BonzaiDuck

Originally posted by: Ruger22C
I went up to 1.18v, BSOD went away... but then windows loaded half-way and my PC restarted .. -_- I clocked the RAM down to 780mhz to make sure it wasn't that.

And? . . . .

As for me, I have to revise what I said. I've seen a lot of posts on the "show your results" thread" and used 1.32V as a starting point. It's still rock-solid at a setting of 1.30V for 4 Ghz. Since the voltage monitor values are now within the 1.26V spec, I'm wondering if it isn't possible to knock it down further. I still haven't found a failure point so I know what the minimum good voltage is.

EDIT: Before I let my runaway enthusiasm confuse you, I've got the E8600.

Many people have clocked the E8400 to 4 Ghz. I had more difficulty with it between 3.75 and 3.82 Ghz, and 3.6 Ghz was a great setting for it.

EDIT: I just checked my over-clock log entries for the E8400. To get to 3.6 Ghz for a 9-hour PRIME95 run with deliberate termination and 0 errs, 0 warns, I had to set the VCORE to 1.3125V and the voltage reading of 1.29V. Figure though that this is a "luck-of-the-draw" on the processor you get, depends on motherboard and other things.

In the end, I'm less interested in producing drag-strip speed-records and more inclined to find a nice balance between processor and RAM speeds with decently tight timings. And voltages set so as not to kill any of my parts.

Bonzai, you have better and more thorough explanations than I and it shows with patience. Nice work!


And as I posted earlier, Ruger should read the sticky...

Originally posted by: Drsignguy
Well, I am going to guess that you will need a bit more voltage bump than that as now the chip is starting to require it. Also, you may need to bump up the mch and northbridge a notch.

How much Ram do you have and is your ram set at 1:1 ratio? What motherboard are you using.

Stock cooling will only get you so far as now that you have to start cranking up the voltage as heat will be an issue.


BTW, have you read the sticky at the top of this forum? It will answer many questions that you may have like temps, how to's and others.

 

[BonzaiDuck]

Originally posted by: Drsignguy




Bonzai, you have better and more thorough explanations than I and it shows with patience. Nice work!


And as I posted earlier, Ruger should read the sticky...

Originally posted by: Drsignguy
Well, I am going to guess that you will need a bit more voltage bump than that as now the chip is starting to require it. Also, you may need to bump up the mch and northbridge a notch.

How much Ram do you have and is your ram set at 1:1 ratio? What motherboard are you using.

Stock cooling will only get you so far as now that you have to start cranking up the voltage as heat will be an issue.


BTW, have you read the sticky at the top of this forum? It will answer many questions that you may have like temps, how to's and others.

It's not as though I'm starved for things to do -- this is it. The room is littered with dust and cyber-junk. And with these fingers on the keyboard -- I can spew a lot of hot air while waiting for OCCT or PRIME95.

 

[Drsignguy]

Lol, I feel the same way sometimes... I babysit 3 rigs right now (F@H) and during that time, I'm here too....

 

[Ruger22C]

You are right, I apologize. I should read his thread before I post any more questions.

 

[Drsignguy]

Originally posted by: Ruger22C
You are right, I apologize. I should read his thread before I post any more questions.

No worries Ruger. Read on and return. There will be more questions to be answered. This is why we all are in this... to learn, to help and to have fun!

 

[Ruger22C]

Hey!

I can't see any setting in BIOS to force PCI to stay at 33mhz.
I only have pci-e, which I set to 100mhz......
??

---

Also, I didn't understand.
What's "safe" temps for tjunction (not tcore?)

 

[myocardia]

Originally posted by: Ruger22C
Hey!

I can't see any setting in BIOS to force PCI to stay at 33mhz.
I only have pci-e, which I set to 100mhz......
??

It's automatically locked @ 33.333 Mhz. You don't have to manually lock it.

 

[BonzaiDuck]

You'd best attend to the PCI-E settings. The default is "Auto" for about three of these. Two of them should be fixed at 100 and the third at 200 Mhz.

I'm foggy about which is which and what the labels are, but at stock settings and "auto" you should see them at default 100,100,200.

On the topic of how low-volted these Wolfdale cores can be, and especially my new E8600:

First, apparently the E8400 has appeared in the E0-stepping flavor.

Second, when I dialed in my initial voltage settings according to what people showed on the "show your results for E0" C2D and quad 9xx0 processors, it was safe but still too high. I'm surprised it took me so many 1 and 2 hour interations of PRIME to get it to fail.

The system -- THIS SYSTEM I'M ACCESSING AS WE SPEAK -- is now set tentatively for 4.1 Ghz at "set" VCORE of 1.29+V. I got it to fail in 17 minutes with 1.28375V, and bumped it up one notch on this eVGA 780i mobo. I just shut down PRIME95 after 3 hours, 45 min 0 errs/0 warns. The monitor readings show 1.25V idle, 1.24V load.

See, 0.03V set or -0.02V load voltage reading above or below the retail-box max VCORE spec is -- for all intents and purposes -- "safe." I don't think there would be electromigration at these settings. So I'm stunned at this E8600 E0-stepping. And this is only a 23% over-clock. I think it could be reasonable to push the volts back up to between 1.32 and 1.34V to squeeze another 200 Mhz out of this sucker. And since I'm on the low-end of the range for these G.SKILLs, I can probably maintain the 4,4,4,12 timings into the high-800's [DDR]. Or I can maybe tighten them further at 800 Mhz and drop it back to 4.0Ghz.

Interesting that I was so set on simply building an "afterthought extra-computer" from last year's spare parts, knowing that the Striker mobo would only take a Wolfdale, that when Striker failed and I ordered this 780i, I ordered the E8600 instead of a Q9xx0. The 780i is good for a quad.

Striker RMA has been requested. We'll see what they do. I miss some features -- upset about the bad BIOS implementation of the fan-control features for the 780i. But the STriker was more trouble to OC to this 4Ghz-plus level.

 

[Ruger22C]

What's the Safe temp on tjunction?

 

[BonzaiDuck]

Originally posted by: Ruger22C
What's the Safe temp on tjunction?

It's hard to remember all the details. TJunction on these Wolfdales is 100C, I think.

The maximum temperature spec in the Intel spec-sheet is around 75C -- for TCase. I haven't yet found the TCase-to-core spread for these -- it had been discussed in an Anandtech article for the 65nm Conroes. That is, you would expect the cores on average to be 10C above TCase. I noticed for the Kentsfields (specifically my old B3-stepping) it was somewhere between 15 and 18C.

I cannot remember for sure whether the system shuts down at TJunction or at the equivalent for the TCase spec. But figure at 75C for TCase, the cores would have to be between 85 and 90C.

Except for my deceased (but RMA may resurrect it) Striker board with Aug 08 BIOS revision, I see people all over the place with boards by non-Intel makers and either nVidia or Intel chipsets complaining about the temperatures. But from my Striker experience, it is the BIOS code that interprets the new Penryn sensors. They're really not running as hot as the software tells you. BIOS programmers should be very busy in coming months with these last-of-the-LGA775 boards.

I'm Oc'd to 4.1 Ghz with this E8600, and small-FFTs is not currently pushing TCase above 53C @ 74F room-ambient at this moment. I believe this 780i eVGA mobo is biassing TCase upward by maybe 5C degrees or so, and the cores are just not reported accurately at all -- appearing to be stuck at 56C. I can't just believe their "stuck." I think it's the BIOS. But go figure. The processor has a TDP of 65W @ stock speed. The quadratic voltage component of the overage due to OC'ing is practically zilch, because my VCore is right at the spec-maximum of 1.26V. If software and BIOS say "37C (idle)" -- it should really be below 30C -- probably lower than that for these room ambients. All that I can tell for sure from these temperatures is that TCASE rises as it would be expected to at higher-than-TDP thermal power with this TR Ultima 90 cooler -- about a 15C idle-to-load spread.

That may help YOU to answer your question. That is, the threshold between "safe" and "not safe" is somewhere between 85C and 95C for the core temperatures. As I said, the Intel maximum TCASE spec is about 75C.

 

[cmdrdredd]

Also don't forget that many motherboards have vdroop which basically means that when you use the CPU in windows and under load the voltage you set in BIOS actually drops below what it's set for. So 1.3v may become 1.27 or so. It varies from board to board.

 

[BonzaiDuck]

That's an interesting point about vDroop.

On my Striker 680i board, it was between 0.02 and 0.03V.

This eVGA 780i has some built-in compensation feature, so I never see a droop of more than 0.01V.

It makes me hesitate at the thought of making a pencil-mod.

I've seen other boards show a more drastic vDroop, and the paradox for me is this. Once the Intel board touted to sport "Native SLI" capability appears, nVidia's chipsets could go the way of Rambus. Not sure what to anticipate on that count . . . .