Asus P5Q Pro,auto voltages - Are all voltages increased when overclocking? - Some are,& I've now created a partial table

[Assimilator1]

If you vcore is jumping up under load then I believe you must have Load line calibration enabled.

Except for the ROG versions of our mbrd the bios AFAIK has no 'hooks' for the NBv, SBv, CPU PLL & FSB term voltage, which means no software can access the info either, cos it's just not there! . The only way to measure it is to use a digital multi-meter like I did. However I'm not removing my mbrd just to measure the NBv , the review that n7 linked gives a good table for the various voltages, which is the best you'll get to knowing the actual voltage.
Make sure you set CPU PLL to 1.52v & the FSB term to 1.22v to stop the mbrd massively over-volting them (as per my updated 1st post).


Oh & Asus insist they don't know what the FSB to CPU PLLv & NBv settings etc are & that Intel should be able to tell me .

 

[Tullphan]

OK...can you give me any ideas on the following settings:
Load Line Calibration
CPU & PCI-e Spread Sprectrum
CPU & NB Clock Skew
CPU Margin Enhancement
CPU GTL Reference
PCI-e SATA Voltage
Thanks!

 

[Assimilator1]

Hmm, I'm still very much learning myself , however.....

About LLC see n7s last post about it & my replie, however I think it ought to be off, see this article by Anandtech for an in depth explanation.
Next one off (unless you have a radio or TV sat right next to it which gets interference), no idea on the rest of them except eSATA leave on standard.

 

[walk2k]

All mine are set to the minimum level (for 45nm).

FSB is 400mhz (x 9 = 3.60Ghz E8400)

http://i231.photobucket.com/al...1/walk2k/p5qvolts1.jpg
http://i231.photobucket.com/al...1/walk2k/p5qvolts2.jpg

(Except for CPU and DRAM that is.)

DRAM is 2.10v standard for 1066 DDR2.

CPU is 1.2500 in bios, in XP (CPU-Z) 1.232 idle (2.4Ghz), 1.240 load (3.6Ghz).

(Btw I just bought another 2GB of the same OCZ 1066 ram for.... $28.95 .. hahah that's $0.014 (1.4 cents) per megabyte...lol.)

 

[Assimilator1]

Hmm, interesting I don't have 'CPU GTL ref' or 'CPU margin enhancement' is that a 45nm CPU thing? or maybe you have a latter bios than me?

Anyway, if you read my updated 1st post (or some of the latter ones) you'll see that setting the CPU PLL & FSB term to their lowest setting over volts it significantly, its the same as setting auto!
So for your 45nm CPU with CPU PLL set to 1.5v you're actually getting 1.74v & with FSB term set to 1.1v you'll actually be getting 1.24v (which isn't so bad but still probably more than needed).

See this thread for the 45nm figure updates to the review linked by n7.
If you want to lower the voltages knock it up a step! lol, although bear in mind that may lower it too much for your o/c & you ought to stability test it over again.

 

[DMF]

Originally posted by: Tullphan
OK...can you give me any ideas on the following settings:
Load Line Calibration - still under discussion, see below
CPU & PCI-e Spread Sprectrum - only for EMI. I Disable both
CPU & NB Clock Skew - Auto. I'm not that sophisticated yet
CPU Margin Enhancement - Optimized: nothing else will POST
CPU GTL Reference - I use Auto since I'm not looking for max CPU speed
PCI-e SATA Voltage - set to 1.50V and don't change it

 

[DMF]

Originally posted by: n7
I use LLC, ... you get lower idle temps & less voltage thru the CPU most of the time, as with it off, you end up with a lot higher idle voltage than is ideal IMO.

I'm not happy about having to run 1.35V for FSB=416 (yes Virginia, there is degradation). So I tried Enabling LLC and dropping the V. Ran Prime95 Blend for 26.5 hrs at 1.325V (1.30V in CPU-Z). Thanks for the tip.

I'll leave LLC Disabled until I want to qualify a 24/7 OC, though.

 

[DMF]

Originally posted by: walk2k
FSB is 400mhz (x 9 = 3.60Ghz E8400)

CPU is 1.2500 in bios, in XP (CPU-Z) 1.232 idle (2.4Ghz), 1.240 load (3.6Ghz).

That seems very low for 400. Have you qualified it by running Prime95 overnight? Mine "ran" at much lower voltages than it was stable at.

C0 or E0 stepping? How much time on the chip at OC?

 

[DMF]

LLC does something else I didn't like. I was seeing occasional 5°C temperature spikes with LLC on that I had never seen before...

Well, despite Prime95 Blend running for 26.5 hrs, I did a quick pass for max temp today and it failed quickly! So I've thrown up my hands and backed off to FSB 401. At 401 I can run 1.25V (haven't fully qualified it yet) and as an added bonus I can drop tRD to 9! All with LLC off.

 

[walk2k]

Originally posted by: DMF

Originally posted by: walk2k
FSB is 400mhz (x 9 = 3.60Ghz E8400)

CPU is 1.2500 in bios, in XP (CPU-Z) 1.232 idle (2.4Ghz), 1.240 load (3.6Ghz).

That seems very low for 400. Have you qualified it by running Prime95 overnight? Mine "ran" at much lower voltages than it was stable at.

C0 or E0 stepping? How much time on the chip at OC?

E0. It "runs" at a couple notches lower (ie 1.22 or 1.23v) but I got some crashes in games so I bumped it up a couple notches (to 1.24). Been running like this for about 2 months, it's very stable. I haven't tested it overnight but it runs Orthos small or blend for 30-40 minutes, heat plateaus after 4-5 minutes minutes anyway.. unless you want to test room temperature cycles (hotter in the day, cooler at night) there's no point in testing further... it's a digital computer, once heat plateaus the results will be the same... it doesn't just decide to fail after 24 hours because it's bored..

It's stable running Orhtos small, 3dmark06, and Itunes playing some music all at the same time.

 

[walk2k]

Originally posted by: Assimilator1
If you want to lower the voltages knock it up a step! lol, although bear in mind that may lower it too much for your o/c & you ought to stability test it over again.

Ok I did that. 1.1 -> 1.12 and 1.5 -> 1.52. I don't have any way to test the voltage, but so far everything seems exactly the same (or maybe even running a little cooler, but then again room temps are down a couple degrees... even here in Nor Cali

 

[evolucion8]

I was unable to overclock my QX6850 beyond 3.52GHz (Multiplier 10x, FSB 350MHz) without failing the Prime95 tests. I put everything on auto except CPU Vcore which was set to 1.3875, anything higher or lower would case errors on Prime95. The funny thing is that I have a Zalman 9500 and the Asus board reports temperatures up to 82C on the CPU, Core 1 90, Core 2 90, Core 3 85 Core 4 85 during the Prime95 session. I found that most websites couldn't overclock the CPU beyond 3.5GHz, but heck, I expected more. Can someone point me what I'm doing wrong? I have an Asus P5Q PRO too and G.Skill PC8500 4GB kit

 

[Assimilator1]

Bloody hell if you're CPU temps really are that high then you are risking serious damage!:Q
You maybe at its o/cing limit (a common max o/c for the G0 quads is around 3.6GHz), however with those temps that will certainly be 'limiting' your speed!.

What temps does it show if you use coretemp?

As for using auto voltages, don't, some of them will be execessive!, read my 1st post about that .

Originally posted by: walk2k

Originally posted by: DMF

Originally posted by: walk2k
FSB is 400mhz (x 9 = 3.60Ghz E8400)

CPU is 1.2500 in bios, in XP (CPU-Z) 1.232 idle (2.4Ghz), 1.240 load (3.6Ghz).

That seems very low for 400. Have you qualified it by running Prime95 overnight? Mine "ran" at much lower voltages than it was stable at.

C0 or E0 stepping? How much time on the chip at OC?

E0. It "runs" at a couple notches lower (ie 1.22 or 1.23v) but I got some crashes in games so I bumped it up a couple notches (to 1.24). Been running like this for about 2 months, it's very stable. I haven't tested it overnight but it runs Orthos small or blend for 30-40 minutes, heat plateaus after 4-5 minutes minutes anyway.. unless you want to test room temperature cycles (hotter in the day, cooler at night) there's no point in testing further... it's a digital computer, once heat plateaus the results will be the same... it doesn't just decide to fail after 24 hours because it's bored.. .

No, it doesn't work like that, heat isn't the only cause of program crashes. Randoom calculation errors can cause very sporadic problems. So their certainly is a point in testing beyond ~1/2hr which is nowhere near long enough to know if its even 'moderatly' stable. I've had Prime 95 & OCCT fail between 16-18hrs. Now it maybe that you don't have any 'mission critical' apps & are not bothered by just trial & erroring it over a few months, fair enough as long as you're expecting that.
If you do want to test it to see if it's at least 'moderately' stable you ought to run P95 for at least 8hrs, personally I run OCCT for 24hrs because I don't want any bad results sent in my any distributing computing projects I run or for it to crash in a long mission in Guild Wars .

 

[n7]

Originally posted by: evolucion8
I was unable to overclock my QX6850 beyond 3.52GHz (Multiplier 10x, FSB 350MHz) without failing the Prime95 tests. I put everything on auto except CPU Vcore which was set to 1.3875, anything higher or lower would case errors on Prime95. The funny thing is that I have a Zalman 9500 and the Asus board reports temperatures up to 82C on the CPU, Core 1 90, Core 2 90, Core 3 85 Core 4 85 during the Prime95 session. I found that most websites couldn't overclock the CPU beyond 3.5GHz, but heck, I expected more. Can someone point me what I'm doing wrong? I have an Asus P5Q PRO too and G.Skill PC8500 4GB kit

All the QX6850s seem to run very hot; mine did as well.

You have a cooler that does poorly with quads, which explains your high temps.

If you want a better OC, you need a better cooler.

My QX6850 did just over 3.7 GHz 1.4v Prime 95 small/blend FFTs stable.

You will not achieve that with your current cooler though...

 

[evolucion8]

In the Zalman's website it states that the Zalman 9500A which comes with a Blue Led is certified for Quad Cores and Core 2 Extreme CPU's including mine. Right now is idling at 40C, but it has been stable all night with Prime 95 which temperatures can reach up to 90C in the CPU main temperature while the first two cores would reach 91C and the last two cores 87C. I know that the T. Junction of this CPU is at 100C but, don't like its high temperatures and this CPU warms up my room terribly. My room temperatures are around the mid 85F, so that may explain why it runs a bit hot. I also used the IC7 Diamond Dust thermal paste and the AS7 and didn't do a difference, but still stable with Prime95 and probably if it were overheating, it would started to crash or shutdown.

 

[Assimilator1]

probably if it were overheating, it would started to crash or shutdown.

No, not necessarily, also if you only just within the limit you could be shortening the life of your CPU alot!
Also tjmax is NOT the max safe working temperature it is the max temp before the cpu will throttle down!:Q Read this article for more info
The max safe case temp for your CPU as shown by Intel is 64.5C!, even allowing for the temp difference between Tcase & Tjunction it won't bump the limit up much.

As for you HSF, I'm not that familiar with it, but when they say its compatible with xxx CPU they often mean at default speeds & vcores.

I'm fairly familiar with ICD7 (I was one of many testers), to get good results from that you need quite a 'decent' clamping force & couple of days to let it 'bed in'. If your HSF just uses push pins it probably won't get enough pressure to spread the paste properly.

Anyway I really thinking you are risking your CPU running at those temps, but hey it's upto you!

 

[evolucion8]

http://zalman.co.kr/ENG/produc...oduct_Read.asp?idx=277 <<That's my cooler. Does mean that the Intel's Max Safe Case temp is about the temperature inside of the case? Not the CPU itself? I'm currently using Artic Silver 5 since I went out of ICD7 ( I plan to buy it again soon). If I knew that I wouldn't be able to overclock that CPU enough and that the cooling would be an issue, I wouldn't buy it and I should instead picked the Q9550 (Sorry for my weird english). But meh, I can't believe that a mere 500Mhz overclock would raise the temperature and heat dissipation by that much. At 3.33GHz it never exceeded the 60C threshold and I was using the Zalman 7700CU which sucks compared to the one I own currently, any advise besides lowering the overclocking or using ICD7? The CPU is running at 3.52GHz with 1.375 Vcore voltage, in a Nine Hundred Antec case.

 

[n7]

Certified for quad cores means jack sh!t.

Quad cores can run on much less than your cooler, like Intel's box one, but that doesn't mean it runs cool on that, or that it can OC well on that.

As i mentioned, the QX6850s seem to run quite warm.

Every one i've seen people post about have had warmer than normal 65nm quad temps, including mine.

IOW, get a damn good cooler (& with one they will generally OC farther than other 65nm quads), or be happy with a less than extreme overclock

 

[Assimilator1]

As i mentioned, the QX6850s seem to run quite warm.
Every one i've seen people post about have had warmer than normal 65nm quad temps, including mine.

What temps do they tend to run at?

evolucion8
Does mean that the Intel's Max Safe Case temp is about the temperature inside of the case?
No, if you read the link you'll see it tells you that case temp is the temp of the CPU in the ctr of the IHS (metal plate on top of your CPU).

If I knew that I wouldn't be able to overclock that CPU enough and that the cooling would be an issue, I wouldn't buy it and I should instead picked the Q9550

Of course a Q9550 hitting 3.5GHz would be a larger o/c in terms of % but really that doesn't count for much, it's main advantage is its price of course .
The only other thing is that the Q9550 is an E0 stepping & the QX9650 is a C0 or C1 stepping, maybe the newer E0s run a little cooler?? ,I don't know TBH, maybe n7 or someone else does?

But meh, I can't believe that a mere 500Mhz overclock would raise the temperature and heat dissipation by that much.
The trouble is your getting close(ish) to the general limit of that CPU family, & you have had to bump the voltage up a lot, increasing vcore has significant impact on increasing temps, check out the overclocking guide I linked you earlier.


Anyway we're going way OT on my own thread! lol ,I suggest you 1st start with that overclocking guide I mentioned for HSF recommendations & then if necassary start a new thread.

Back to P5Q voltages

 

[evolucion8]

Yeah that's true, the QX6850 are only made of the G0 stepping, which is an excellent overclocker, specially for the Q6600 based on them, runs a bit cooler and has a lower thermal envelope.

 

[Assimilator1]

45nm QX6850 is C0 not G0 ([edit] my bad, meant QX9650), but yes the Q6600 is G0 (which is 65nm).
I think the E0 stepping is a newer 45nm core & may run cooler.

 

[walk2k]

Not sure bout the Quads but for Duals (Penryn 45nm), I think keep it below 65C (coretemp). Make sure you are using a coretemp reading program too, and NOT the Asus/Winbond chip which WILL report 5-6c lower temps on the cpu core!

Mine will push just a bit over that, 67-68c in Orthos small, on a hot day... Otherwise it's well below 65c in other stress tests, and WELL below that in most games. In games it generally runs in the low 50s (53-55c).

 

[Idontcare]

Originally posted by: Assimilator1
45nm QX6850 is C0 not G0 , but yes the Q6600 is G0 (which is 65nm).
I think the E0 stepping is a newer 45nm core & may run cooler.

Know thy processors.

Intel® Core?2 Extreme Processor List

Intel® Core?2 Extreme Processor QX6850

Manufacturing Technology: 65 nm

Core Stepping: G0

 

[Assimilator1]

Yea roger that (I posted that link further up), but as I said I was refering to the 45nm QX6850, although I see I wasn't clear about that in an earlier post. I believe evolucion8 said he had a 45nm version.
I didn't know though that the 65nm Q6850 was a G0 stepping :thumbsup:

 

[Idontcare]

Originally posted by: Assimilator1
I was refering to the 45nm QX6850

Do you mean QX9650? There is no such thing as a 45nm QX6850. That's a 65nm SKU.

The point evolucion8 is trying to make is that both Q6600 and QX6850 are 65nm G0's and yet in his experience/perception/opinion the QX6850 SKU tends to deliver higher operating temps at equivalent GHz when compared to an equally overclocked Q6600.

Whether that is relevant to your thread is not for me to decide, I am just butting into the conversation here because I wanted to clear up an obvious miscommunication between the two of you.