Question about my q6600 GO overclock

[EvilSponge]

Hi all,

I just put in my go rev Q6600 and I am in the middle of Prime95 testing as I write this
So far here are the results with about 20min of prime so far:
Q6600 GO VID 1.2875v @ 3.6ghz
Vcore: 1.45 though cpuz ver 144.1 reports vcore of 1.384?
Coretemp reports: core 0: 65c core 1: 65c core 3: 57c core 4: 56c
core 0 and core 1 now at 63c
How are my temps/voltage?
Mem: 5-5-5-15 1:1 800mhz

Thanks new to ocing the quad

-sponge

 

[flexy]

whats your OC? 3420? Sounds good. Mine goes up to 3537, VID 1.42/1.45. I try to stay at/under 1.45Vcore and coretemp under 71/core.
You can try how far it goes with lower Vcore (mine, Batch L741) seems to be "bad" since i cant hit 3600, but i am OCCT stable at 3537. Your OC and temps look all fine to me.

 

[10acjed]

Sounds good to me.
The Vcore issue is called vdrop, when ur cpu loads up to 100% the vcore will back off, its an intel thing.
Mine is 1.375 in bios, 1.39 (cpu-z) and goes down to 1.31 when loaded.

the tJunction temps are what you are refering to, and they sound very good, tCase (Processor Temp - cpu wiz) is the one I pay attention to, as long as its under 65, its ok for me
core 0-core-1 etc.. temps will get higher

 

[dasracht]

Looks good here too. I'm curious what the speed is at with those temps though. 65c is the absolute maximum that I'll let my cpu get to under load, but others will let it go higher and it seems to work out fine.

 

[flexy]

you're talking about CORE temps as seen in current coretemp? 65C maximum???? Definitly hitting 68-69 easily on the cores with 1.42+ Vcore...this with very good air-cooling already.


"""
coretemp reports: core 0: 65c core 1: 65c core 3: 57c core 4: 56c
"""

10 deg difference....uhm...is not normal, assuming you stress all 4 cores at the same time. 3-4, maybe 5 degrees diff between cores is common...but 9-10? Sure you applied thermal paste right?

 

[dasracht]

Originally posted by: flexy
you're talking about CORE temps as seen in current coretemp? 65C maximum???? Definitly hitting 68-69 easily on the cores with 1.42+ Vcore...this with very good air-cooling already.

That's true, but I'm very conservative since I can't afford to replace stuff if I kill it. That's why I won't let it go over 65, even if that's well within limits

Originally posted by: flexy
10 deg difference....uhm...is not normal, assuming you stress all 4 cores at the same time. 3-4, maybe 5 degrees diff between cores is common...but 9-10? Sure you applied thermal paste right?

This is completely normal for quad cores. Mine usually run at full load: 62, 61, 52, 51. Others on the forum report the same thing.

 

[Idontcare]

Originally posted by: 10acjed
The Vcore issue is called vdrop, when ur cpu loads up to 100% the vcore will back off, its an intel thing.
Mine is 1.375 in bios, 1.39 (cpu-z) and goes down to 1.31 when loaded.

I think you mean Vdroop. Vdrop is a term some folks use interchangeably with the term Voffset, which is not the same as Vdroop.

See Anandtech article: http://www.anandtech.com/cpuch...howdoc.aspx?i=3184&p=5

Specifically the Voffset and Vdroop terms as labeled in this Anandtech graph: http://images.anandtech.com/re...2/transient_vdroop.jpg

 

[10acjed]

great article, thanks.

this is kinda the reason i dont like setting my vcore near the recommende threshold.

if intel says 1.55 i say 1.45, therefore if the article is accurate my peak voltage should stay below the 1.55 mark, right now im at 1.39 (bios 1.375)

because while running ur normal day to day activity you are more likely to hit these high peaks, extreme stress testing only verifies that its stable fully loaded, nothing will match that in day to day use.
Watch your temps during gameing and other things you do that load ur cpu usage, vdroop (vdrop-potayto-potahto) only occurs under "heavy load"

dont wana burn up my cpu, ill leave that the the sub contracted testers, let them see the limit. but then again they only push the limits and benchmark, i like the forums where common day to day use is most relevent.

 

[EvilSponge]

Ok guys here is what I have experimented with so far:

Test#1
Vcore set in bios: 1.48v vdrop per cpuz ver 144.1 1.45v q6600 GO @ 3.6ghz

Temps recorded during prime95 torture test small ftts: Core 0: 66c, Core 1:66c, Core 3: 58c, Core 4: 57c per coretemp the latest
2 hour run perfectly stable

q6600 max temp 71c so then 5 deg margin

Test#2
Vcore set in bios: 1.44v vdrop per cpuz ver 144.1 1.38v q6600 GO @ 3.51ghz

Temps recorded during prime95 torture test small ftts: Core 0: 61c, Core 1:61c, Core 3: 55c, Core 4: 54c per coretemp the latest
2 hour run perfectly stable
Margin: max spec 71c margin 10 deg

My question for those who are more experienced with the q6600 is it worth it to run this guy at 3.6ghz or just leave it at 3.51ghz with more margin?
I ask as I'm not sure how robust these GO rev q6600 chips are.

Thanks to those who have already given their input and those that can help with this question

-sponge

PS> Interesting side note runs stable in both tests with speed step enabled

 

[dasracht]

If it's running 3.6 with load temp of 61, I say it's good to use all the time. Mine runs load temp of almost 70 at 3.6 with the same hs and a scythe fan, so I'm a bit jealous =)

 

[EvilSponge]

Nope it's running 3.6ghz fully loaded 66c on cores 0,1 and 57-56c on cores 3,4

fully loaded 61c on cores 0,1 and 55-54 on cores 3,4 running 3.51ghz

-sponge

 

[dasracht]

Oh, sorry didnt notice you changed the speed on test two. Have you tried running it at 3.6 using 1.44 vcore? They usually recommend to try lower numbers, then when it fails in prime, up it a little and try again.

Maybe your vcore is much higher than it needs to be... mine runs 3.6 at around 1.42. Try it at lower and maybe you'll get those temps down. If you can get it closer to 60c then keep it at that speed 24/7 imo.

Some would likely say that it's fine as it is, 24/7 at 3.6 with 66c full load, but thats a little too hot for my taste.

 

[EvilSponge]

Yeah I tried a lower voltage just not stable on prime95. There is a decent vdrop on my board so higher is really lower and lower just may be too low. I guess I'm trying to decide would I even notice a performance difference between a q6600 running @ 3.6ghz and one running @ 3.51ghz

-sponge

 

[dasracht]

It's better to run your fsb and ram at the same speed, so I would go with 400mhz fsb (1600) which is the 3.6.

Honestly I dont know the technical reason for why this 1:1 ratio is better, but I see it in almost every faq. Maybe someone else here can give the proper reason.

 

[core2slow]

since its a bad batch, I wouldnt push it too much. You're not going to get much benefit when only two of the cores are pushing it while the other two just sit back and relax under load. When im idling, my temps are in the mid-high 30C (3-4C difference), and under load theyre ALL in the low 60C (1-2C difference!).

Try and limit your bios core to 1.4v and OC within that boundary.

 

[EvilSponge]

Yeah I was wondering why two core are working harder than the other two so bad batch is the reason?

there seems to be a 7-8 degree difference between cores 0,1 66c and cores 3,4 59c

-sponge

 

[EvilSponge]

It's looking like 3.51 ghz is my limit it matches the same temps as yours core2slow fully loaded on prime95 and mid 30's idle 1.40 vcore

Thanks

-sponge

 

[EvilSponge]

Update reduced my multiplier to 8x450 for 3.6 ghz now prime stable max load 63c - 65c @ 1.38v vcore per cpuz

-sponge

The weird thing is the difference in core temps has changed as well they are all within 3 deg of each other? It was a difference of 7-8 before

 

[flexy]

Originally posted by: dasracht
It's better to run your fsb and ram at the same speed, so I would go with 400mhz fsb (1600) which is the 3.6.

Honestly I dont know the technical reason for why this 1:1 ratio is better, but I see it in almost every faq. Maybe someone else here can give the proper reason.

we had this yesterday, and then i also did some tests.

Running ram higher has a benefit since faster mem-speed translates into better latency. So...running mem at 530 5:6 was a bit better than running mem 1:1 at 442. Definitly NOT the case that 1:1 and lower mem was "faster".

Add:

I still say 10deg difference is unusual, 5-6 is usually seen.

And IMHO all temps til 70C under extreme load (OCCT, Prime etc.) is ok. So 65 would be an excellent core temp for under full load using OCCT or Prime.

 

[demiurge3141]

Originally posted by: core2slow
since its a bad batch, I wouldnt push it too much. You're not going to get much benefit when only two of the cores are pushing it while the other two just sit back and relax under load. When im idling, my temps are in the mid-high 30C (3-4C difference), and under load theyre ALL in the low 60C (1-2C difference!).

Try and limit your bios core to 1.4v and OC within that boundary.

How can you tell it's a bad batch? Just from VID?

I don't think the temp difference means anything about cores working harder.

 

[flexy]

no..but if you have 10+ temp diff it *could* be that TIM is applied wrongly....when i applied AS5 the first time i didnt know that for quad core you're supposed to apply a *line* across the chip...i did the old-school method of just putting a small dab in the middle. This then could contribute to some less efficient tempsa cross cores. Just my $0.02 sicne thats what happened to me.

10 deg diff seems a LITTLE unusual...but if temps are fine otherwise (under 70 full load) i wouldnt worry about it. Was a pain in the *** here since i had to reapply TIMm and with ultra extreme and my case and board this meant disassembling literally everything. I gained 1-2 degrees less across cores tho..for what its worth

 

[EvilSponge]

Yeah it did prime 4 hours with a max temp of 63c - 65c I did the line of AS5 but it was AS5 from an old tube I ordered a new tube and am going to reapply even if it has no effect I thought I read somewhere AS5 has a shelf life

-sponge

 

[EvilSponge]

Screen of over clock

http://bluekarmann.com/web%20pages/espcoc.html

-sponge