Need little assistance

[OmMeE]

Need some assistance in overclocking my Q8200.
My system specs are:

Q8200
Asus P5Q-E
Asus Silent Knight 2
Corsair Twin2X2048-6400C5DHX 800mhz (5-5-5-18) 1.8v

I have a pretty stable system at 3.0ghz (429x7) FSB.
Temperatures are a little high,(63-65 degrees at 100% load) but I've googled alot and have found out that this processor either does get high temps or maybe just shows.
I have ran prime95 and 3dmark06 and all is well till 3.0ghz.

I want to take my clocks to 3.2ghz (457x7) FSB. Till 445x7 system boots and runs for a while unless you dont apply 100% load, but at 457x7 fsb, i either get a BSOD after windows login or the system haults as soon as i put like 10% load on the cpu.

For 3.0ghz (429x7) fsb my voltages are:

CPU voltage... 1.2635v
CPU GTL volt.....auto
CPU PLL VOLT.... 1.5V
FSB termination voltage.... 1.20-1.22
DRAM voltage... 1.8V
NB voltage... 1.10-1.12
SB voltage... 1.0
PCIE voltage... 1.50

Can you guys here help me to reach 3.2ghz stable??
I have seen here people have very low CPU voltages, but their NB are higher...
Does not the NB get hot?

Thanks in advance!

 

[OmMeE]

Can anyone please SWAP the topic title with topic summary. thanks

 

[imported_Redmist]

you could try loosening your memory timings a bit, and/or keeping the ram at 1:1 ratio, or whatever ratio to keep it under or around 800mhz, you may be jacking your ram speed up with the high fsb and that might be what is locking up.

 

[BonzaiDuck]

In addition, I'm pretty sure . . . at those FSB speeds above 400 Mhz, he should bump up his CPU_VTT or "FSB termination" voltage a tad, while keeping it below 1.4V. If it works fine at 1.30V, then don't push it higher.

That's what these Wolfdale cores need -- and in fact, for the 400+ FSB level, I think it applies to Conroe-Kenstfield and Wolfdale-Yorkfield cores. But for the Wolfies and Yorkies, caution is advised per the FSB termination or CPU-VTT voltage.

Also, as far as I can see, he's got "lebensraum" for his VCORE setting. On my (eVGA) mobo, if I set it at 1.3V, the idle reported voltage is around 1.26V and the load value is closer to 1.24V for the vDroop factor. His E8200 voltage spec "safe" range is 0.85 to 1.3625V. He has LOTS of lebensraum.

The trick is to quickly find the minimum voltage setting that is "within spec" to support his target over-clock speed.

 

[OmMeE]

Bro its a Q8200.. not E8200
Im keeping my vFsb and Vnb below 1.3v... still i cant manage... maybe some other voltages need to be tweaked.. for which i need some assistance

 

[BonzaiDuck]

Originally posted by: OmMeE
Bro its a Q8200.. not E8200
Im keeping my vFsb and Vnb below 1.3v... still i cant manage... maybe some other voltages need to be tweaked.. for which i need some assistance

On the matter of VTT voltage, Anandtech tested a C2Q 9x50 at close to 1.5V and burnt it out. The article recommends tweaks that keep the termination voltage below 1.4V. I think the advice was valid and promoted for the Wolfdales as well, but your Q8200 is still in the "Yorkfield" category.

Intel shows the Q8200 with the same boilerplate VID range, or 0.85 to 1.3625V.

Not sure about your Intel chipset and NB spec, but even 1.3 to 1.35V might be "nominal" for the NB setting on some motherboards. I'm surprised you have it set so low, but then, I'm not that familiar with your chipset, so others might speak with more authority. Same (in my opinion) for the vSB you report, but I could be wrong, . . . . too . . . .

 

[OmMeE]

I have seen while searching.. a Q8200@ 3.2ghz with 1.22vfsb and vnb 1.3.. and his prime95 runs stable for 5 hours :S

 

[BonzaiDuck]

Originally posted by: OmMeE
I have seen while searching.. a Q8200@ 3.2ghz with 1.22vfsb and vnb 1.3.. and his prime95 runs stable for 5 hours :S

"1.22vfsb . . . " I guess the proof of the pudding would be a statistical sample of user-settings for similar chipsets on various forums. Otherwise, with little familiarity on that chipset, I wouldn't know with any certainty. But at least you know what the safe range or limit should be, if you decide to tweak it a tad higher.

A 5-hour PRIME95 run is a preliminary indicator of being "near perfect stability," but the test should be run for at least double that time to confirm it. I've had the program fail after 9-hour runs.

On the matter of RAM or MEMORY: I have a set of Corsair DHX modules rated for [DDR]=800. You don't say whether you have "800's" or "1000's," but the voltage spec should be the same. And I can tell you this, having communicated directly with a Corsair tech about it. You should run them at the "recommended" voltage of 2.1V. Corsair says their warranty allows for a 5% overage, so you're safe up to 2.2V. I've seen them bench-tested with vDIMM setting of 2.4V.

Running the RAM at their full voltage spec and the default spec latencies assures that you eliminate memory as a source of stability problems while over-clocking. Frankly, I think that I tried running the DHX kit initially at the voltage-setting you cite for yourself, and they showed error before I embarked on serious, methodical over-clocking -- based on the Corsair tech's remarks.

 

[OmMeE]

My ram are rated at 800mhz and their voltage is 1.8v as per their website. and wats written on the sticks.

 

[BonzaiDuck]

Originally posted by: OmMeE
My ram are rated at 800mhz and their voltage is 1.8v as per their website. and wats written on the sticks.

My mistake, if you don't have the DHX DDR2 kit rated for the timings 4,4,4,12.

I checked the Egg, and in addition to some of the "Dominator" kits touted as DHX (with either 2.1V or 2.4V specs), there was a DHX kit rated for 5,5,5,15 and 1.85V.

But most of the DHX kits featured at the reseller were spec'd at 2.1V.

That being said, I'm guessing that there's a significant possibility that running those modules at from 2 x 429 to 2 x 457 [your target] would require more than stock voltage, even with the stock timings. I'm also guessing that the 5% warranty-allowance would allow you to take them safely to around 1.9V.

You'll have to confirm this for yourself -- maybe even contact the Corsair tech. It's just my opinion that you're trying to squeeze more out of the memory at those speed settings than the stock voltage will allow.

If you have some extra change, you might want to look at the G.SKILL DDR2-900's after seeing if they are compatible with the P5Q-E board -- or even a set of DDR2-1000's. DDR2 kits are cheap right now. I can tell you that the Black-PI 900-Mhz kits will run at their minimum voltage spec of 2.0V at speeds closer to what you're trying for. At your current 2 x 429 = 858 Mhz speed, you can set the latencies to 4,3,3,8 and they'll run at that minimum voltage -- if not, maybe one notch higher. The upper voltage limit is 2.1V.

 

[OmMeE]

The thing is i have been googling around and have asked some people who really over clock their stuff... One on them suggested me memtest86... Now at 3.2ghz.. rams rated 915mhz i think.. i passed the memtest86 on that people said nothing is wrong with the ram... Im trying to get 1066mhz rams.. lets see if i can manage to

 

[BonzaiDuck]

Originally posted by: OmMeE
The thing is i have been googling around and have asked some people who really over clock their stuff... One on them suggested me memtest86... Now at 3.2ghz.. rams rated 915mhz i think.. i passed the memtest86 on that people said nothing is wrong with the ram... Im trying to get 1066mhz rams.. lets see if i can manage to

I can say from personal experience, and others should support it, that MEMTEST86 or MEMTEST86+ is useful for certain purposes:

* Verifying, after the initial system-building (and even before installation of the operating system) that the RAM-kit at stock settings runs error-free

* Preliminary testing of RAM under over-clock settings

I've found that you can sometimes run MEMTEST86/+ all day and night without error under some over-clock regime, only to have errors halting a run of PRIME95 under "Blend-Test" -- which stresses RAM.

And sometimes, you can get that result, and after verifying that the stop-error wasn't caused by another core voltage, discover that the RAM still needs another notch's increase in voltage.

A lot of us experienced the troubles with much-touted Crucial (Tracer and Ballistix) RAM kits. You could run the voltage up to a comfortable level below the maximum-recommended, and after some period of time, they would still burn out. Some of my colleagues may agree that Crucial/Micron had inadequately tested their product and set warranty specs that were too high, maybe in eagerness to satisfy the enthusiast market.

I don't think this is the case with Corsair or G.SKILL. I DO think it is worthwhile to buy memory rated for levels above your over-clock settings, provided the modules are "downwardly elastic" and allow for tighter latency timings at lower speeds. That way, you don't need to push the voltage to riskier limits.

 

[OmMeE]

Aiite bro... Will be getting some 1066mhz sticks soon then..

 

[n7]

Whoa whoa whoa...

RAM isn't the issue here.

The issue here is that OmMeE wants to get a quad core stable at 457 FSB on a P5Q-E.

As soon as you hit around 440-450 on most P5Q boards, it starts getting to be a major challenge getting things stable, & it's not a CPU or RAM issue, it's a motherboard one.

Generally you need to work on GTLs to maintain stability at a high FSB like that with a quad.

You've got NB voltage set really low; you can do 1.3v without worrying about it being too high, so you have lots of room to increase that.
vFSB (vTT) also might need a bump or three, though it's more likely NB voltage in your case.

Read up on GTLs: http://www.xtremesystems.org/f...howthread.php?t=202292

And then if you can't get as far you want, sell your P5Q-E & get a UD3R/P.

Getting 1066 RAM isn't going to help you here.

BTW, you can easily ensure the RAM isn't the issue.
Clock down to a lower FSB, set a higher than 1:1 ratio like 5:6 or 4:5, & you'll be able to test the RAM at higher speeds so you know it's not the issue.
Example:

Instead of running your current 7x429 1:1 (DDR2-858), try
7x400 5:6 (DDR2-960), or something a bit lower then work up.
If that's stable, you know your RAM is more than adequate for the DDR2-914 you want.

 

[BonzaiDuck]

Originally posted by: n7
Whoa whoa whoa...

RAM isn't the issue here.

The issue here is that OmMeE wants to get a quad core stable at 457 FSB on a P5Q-E.

As soon as you hit around 440-450 on most P5Q boards, it starts getting to be a major challenge getting things stable, & it's not a CPU or RAM issue, it's a motherboard one.

Generally you need to work on GTLs to maintain stability at a high FSB like that with a quad.

You've got NB voltage set really low; you can do 1.3v without worrying about it being too high, so you have lots of room to increase that.
vFSB (vTT) also might need a bump or three, though it's more likely NB voltage in your case.

Read up on GTLs: http://www.xtremesystems.org/f...howthread.php?t=202292

And then if you can't get as far you want, sell your P5Q-E & get a UD3R/P.

Getting 1066 RAM isn't going to help you here.

BTW, you can easily ensure the RAM isn't the issue.
Clock down to a lower FSB, set a higher than 1:1 ratio like 5:6 or 4:5, & you'll be able to test the RAM at higher speeds so you know it's not the issue.
Example:

Instead of running your current 7x429 1:1 (DDR2-858), try
7x400 5:6 (DDR2-960), or something a bit lower then work up.
If that's stable, you know your RAM is more than adequate for the DDR2-914 you want.

I stand corrected. I forgot that he's trying to OC a quad-core. Even so, at 1:1, I thought his target objective was a bit ambitious. And -- yeah -- I thought that NB voltage was low. On some of the mobos and chipsets I've used, the NB "auto" default is closer to 1.4+V.

On the other hand, I'm not sure the memory-kit he's using would go as high as 960 Mhz without pushing its voltage higher, anyway. But the only way to find out would be . . . as you say. Personally, I still think he should find his overclock settings by testing at 1:1 first. Running them below spec or at spec would enable him to assure that they aren't the source of stress-test problems.

 

[n7]

1.1v is default NB voltage for P45 boards AFAIK, or at least all P5Qs anyway.

1.4v would be higher than ideal or needed.

Somewhere in between is where he'll need to be.

He already has tested his RAM @ 457 (DDR2-914):

Originally posted by: OmMeE
...Now at 3.2ghz.. rams rated 915mhz i think.. i passed the memtest86 on that people said nothing is wrong with the ram...

I was saying to try even higher just to be certain it's not RAM.

When you are OCing toward the limits of various components, you want to narrow down the possible issues as much as you can.

1:1 is what he'll end up using, yes.

But if we do not know how far the RAM can go, it's good to make sure it's not going to be an issue, which is why i said to try testing the RAM at an even higher speed just to be positive it's not going to be an issue.

I guess not everyone understands how to test their individual components for their maxes, but you were basically suggesting he gets new RAM when he hasn't even found out how far his RAM can go.


It's like with my RAM.
It's DDR2-800 4-4-4-12 2.0-2.1v rated.

Yet it can easily run DDR2-800 5-4-4-12 1.8v, or it can do DDR2-1080 ~ 2.0v.

Too many people assume things with RAM rather than simply figuring out their own RAM.

I don't personally believe in worrying about RAM's rated specs; i worry about what it's capable of under my own testing.
I'd rather have generic PSC DDR2-533 1.8v than some branded kit with DDR2-800 5-5-5 D9HNL 1.8v, because i know that the PSC is likely going to hit 1000+ with low voltage & is actually better than the "higher rated" 800 stuff, just as a random example.

Anyway, i got way off track...just trying to say that RAM too can be OCed, & in this case, it's already obvious RAM isn't the issue.
Plus, the OP can loosen to 6-6-6 & get farther if 5-5-5 was really holding him back (which it's not).

 

[OmMeE]

^Thanks for the comments you guys, a lil occupied atm will try my ram out soon..

 

[BonzaiDuck]

Originally posted by: OmMeE
^Thanks for the comments you guys, a lil occupied atm will try my ram out soon..

I think the point we're both trying to make is that there is a "trade-off" between voltage, RAM speed and latency settings. You can take the RAM to an above-spec speed by either loosening the latency timings, increasing the voltage, or both. You can tighten the latency timings by decreasing the speed, but the tighter settings may require the same or higher voltage.

n7's other point is that you need to eliminate the RAM as a source of error -- which also means eliminating anything associated with the MCH or memory-controller hub. My view is that you'd do this by setting the CPU : RAM ratio to 1:1, for otherwise, OCing the front-side bus might push the RAM beyond its limit, so you'd want to keep the RAM closer to its rated spec, increase the RAM voltage so that it is at spec or slightly higher (within warranty), and let the timings default to the spec.

On the other hand, I think he's saying that you want to first assure the RAM isn't defective, so you would run it AT SPEC in other than a 1:1 ratio -- first testing with MEMTEST86. If you can't OC the system to run 1:1, your next best option is probably a 4:5 ratio. Of course, if you OC the CPU for any multiplier value at or below stock to 400 Mhz, the RAM will have to run at 500 or DDR=1000 Mhz -- which may be beyond the ability of DDR2-800-rated RAM at voltages that are safe for it.

These days, and with the Penryn's, we'd always say that it's best to buy DDR2-800 or DDR2-1000, since over-clocking would nominally have a target around 400 Mhz or higher.

But again, it's important to limit the source of trouble so that the RAM settings don't figure in as a source of error. Otherwise, you're playing with too many variables at once.

As for the quad versus C2D -- you can look around here to find what maximum over-clocks are achievable for the entire gamut of Yorkfield cores. Nominally, I think cautious over-clockers would settle in around 400 Mhz. Pushing them higher means more voltage -- you take your chances. And beyond that, extreme tweaking might involve tuning the GTLREF voltages. I'd say -- don't play with that feature unless you've read some useful guide on the subject, and read it enough times to fully understand it. Especially, the GTLREF tuning is linked to choices for the VTT voltage.