Q9650 Cores 0,1 6C higher than 2,3

P

prasvt

Gawd
Joined
Dec 8, 2005
Messages
557
So I just installed a Q9650 a few min ago and as the title says, cores 0 and 1 are idling around 34C and cores 2 and 3 are idling at 26C according to Coretemp. I cleaned the water block, spread an even line of thermal paste (TX-2) across the CPU like i've done a million times and then put the waterblock back on. I ran a quick small fft run on prime95 and under load cores 0 and 1 showed 45C while 2 and 3 showed 39C. It's kind of bizarre...should I tweak the 2 screws on the core 0,1 side of the waterblock or what? Though I'm pretty sure I tightened all the screws about the same amount.

I just wanted to check with you guys before I stress out about it.
 
Same thing on my Q6600. I don't think it's a big issue....I think it's fairly normal for Quads. Cores 0, 1 will run about 38C, and 2,3 will run 29~32 on mine.
 
Okay that's good to hear. My Q6700 cores were more close together so this threw me off a bit.
 
My q6600 often shows some variance as well, but not quite that much. Most I ever see is a difference of 4-5ish. In the few remounts i've done i found the line method (and not trying to spread it) has worked best.

Another thing i've heard around is that some of the quads have a slightly concave / convex shape to them & some wb's are made to try to match that, hence the suggestions to do both cpu and wb when lapping. Other than lapping them, maybe you can try remounting your block 90deg. to where you have it now (i guess ability to do that may depend on what block your using), or even as you suggested yourself adjust the tightening level at each corner/side.

Other than that stuff i really don't know, but hope it helps some.
 
Well if it was random variance among the cores I'd probably redo the mounting..but it's literally the first 2 higher than the second 2 on idle / under load. I put more than enough tim this time around (2 lines actually, and 1 dot in the center in case it was concave). I mean it's fine..I'm on water and overall cpu temp is still in the low 20s. I'll have to see how it looks once I start pushing to 4.0ghz.
 
Ok running intel burn test @ 467 x 9 = 4.2ghz @ 1.3625. Cores 0 and 1 are at 66C while Cores 2 and 3 are at 58C. At least it's consistent...and no crash thus far.
 
Q8200 has the same issue.

I messed with it for weeks then gave up. Lapping it did lower the temps 1 or 2 degree on all cores so I still had a 5 degree difference.
 
I hear ya...well Cores 0 and 1 are idling at 42C while 2 and 3 are at 33C @ 4.2ghz. Really weird. Ok well I guess I'll take out the cpu and check it over once and then adjust the TIM application accordingly and see if it makes a difference. Otherwise I guess it's just one of those things.

Update: reapplied the TIM and remounted the waterblock...almost identical results @ 4.2. 42C idle for Cores 0 and 1 and 32C idle for cores 2 and 3. Oh well...as long as it's stable I guess.
 
Its the same with my q6600, it idles at 32-30-28-23, at full load they go to 68-68-62-62. This is after lapping both the cpu and base of my TRUE. Ive seen very few quads that have even temps across the cores. Btw nice oc on the new cpu:)
 
Alright, so I tried this a third time. This time, I tried supaspoon's suggestion and turned the waterblock 90 degrees. Initially it was sitting sideways (the swiftech logo was sideways), and now it's seated so that the logo is vertical. I'm running prime95 right now and Coretemp is showing: 64, 64, 61, 60 for 0 - 3.

So I guess that helped the load temps. There's still a 10C difference on idle though..so kind of funky. The first few seconds of prime, cores 0 and 1 shot up to 70..and then 10 secs later dropped down. So bizarre.
 
setter said:
Btw nice oc on the new cpu:)
Click to expand...

Yeah no kidding. Did those just get hit w/ a price drop? I thought they were over 500ish, but i'll be damned if the egg isn't showing it at $330. I'm feeling a case of 'i-want-i-tis' coming on.
 
Yeah, they just price dropped ...ordered yesterday and it came this morning. I started a q9650 oc thread so it's a work in progress. I dropped the voltage down a bit and it failed prime, so need to tweak some stuff.

But yes, these things out of the box will do 4ghz it seems haha. Mine is a 1.2125 VID which is decent as some people have 1.25 VID chips. But it looks like most of these OC well despite varying VIDs.
 
Considering changing to a q9550, they can be had for about £220.00 gbp, ive taken my q6600 as far as i can and now fancy a new (less power hungry) cpu to play with, 4ghz has a nice ring to it:D
 
That's what I was thinking ..4ghz has a nice ring to it. And it's the same...pushed my Q6700 as far as it would go under 1.5V vcore - looks like you had better luck with your Q6600.
 
Hey, glad to see I actually helped someone lol. I keep trying to get rid of my n00bstink but the axe just hasn't been working.

Anyway, sweet deal on those price drops man. I'm definately gonna have to grab one of these. I'd get one right now but I'm already feeling fiscally irresponsible enough for the week after blowing my wad on wc stuff. Kind of pisses me off now that i think about it. :mad:
 
Its just the way some are. Mine has ~ 6C difference (56,53,50,50) even after lapping the CPU & HS
 
Yeah I felt fiscally irresponsible when I bought WC stuff ... plus my dad kept telling me find another hobby, like stamp collecting. That's the downside to living in Cali (expensive), being an IT consultant (cisco hiring freeze), and just starting out...you're living at home lol. Anyway both parents were in FLA for a conference this week so I could be a little impulsive without the backlash when the UPS guy got here :)

Yeah 10C diff on idle temps kinda bugs me..but whatever. I'll say this much..@ 4.2ghz running IBT or OCCT, these things get hot 45nm or not. So pushing back to 4.1 right now and tweaking. May try an older trustier BIOS version first and see if that changes anything.
 
prasvt said:
Yeah I felt fiscally irresponsible when I bought WC stuff ... plus my dad kept telling me find another hobby, like stamp collecting. That's the downside to living in Cali (expensive), being an IT consultant (cisco hiring freeze), and just starting out...you're living at home lol. Anyway both parents were in FLA for a conference this week so I could be a little impulsive without the backlash when the UPS guy got here :)

Yeah 10C diff on idle temps kinda bugs me..but whatever. I'll say this much..@ 4.2ghz running IBT or OCCT, these things get hot 45nm or not. So pushing back to 4.1 right now and tweaking. May try an older trustier BIOS version first and see if that changes anything.
Click to expand...

I get 4.2ghz at 1.25v. Temps stay below 60 unless its very warm in the room. Cant say thats real hot. This is on pretty good water though.
 
I have the same occurance. On coretemp cores 0 and 1 are 34c while cores 2 and 3 are at 31c. I have redone the artic ceramque or artic silver 5 several times, the last two cores are allways lower.
 
Well after adjusting GTL and voltages I'm not getting prime blend errors anymore. 4 hours and counting, but I think I need more than 1.25V. More like 1.3625 .. temps are 68 68 59 59 ...but it's working. I'll save the bios settings and try a lower voltage next.

Anime: I think 10C difference is more worrisome than 3C so you're in better shape than me. At least all the cores seem to OC about the same.
 
Anime: I think 10C difference is more worrisome than 3C so you're in better shape than me. At least all the cores seem to OC about the same.
Click to expand...

Point made but at least your not running with 70c or more! :D
 
@ prasvt
Wanna share you tweak settings? I heard of a few people (Asus forum) getting this same set up
(well RE board) to 4.5 on water and plenty of guys doing 4.0 and 4.1 on air. Once I hit 4.2 she
post but flunks IBT and the OS gets shaky. Core temps get close to 80C on IBT. Not good!
Maybe I reached my limit at 460FSB. I have CPU, NB, and VGU on same loop with a quad
Fessor rad. I have the Apogee block. Maybe I should pull them and lap them and use better
tim.

Heres my volt settings I used to further OC. I just feel like anything higher is asking for trouble.
original.jpg

original.jpg


The only thing different in the photo is the LLC is enabled to get my vCore
lower and make it stable. I do have the 1.25VID. I have kinda given up and
decided that maybe my 4.0GHz is best for a 24/7.

Thanks and congrats on the 4.2!

Keith
 
I also noticed cores 0,1 on my q6700 are usually about 3 - 5 c higher
does the os perhaps preferentially load cores 0 and 1?
 
I did some more searching today and it seems that the "2 cores are 10C higher" issue has crept up on a number of Q9650 and Qx9650 chips that people have. No one really has an explanation for it, but it doesn't seem to affect stability.

If the temps for Cores 0 and 1 were 10C lower I'd venture past 4.2 but even so, the temperatures past at and above 4.2ghz tend to ramp up pretty high so I think 4.2 is more than enough for 24/7 use. Plus for gaming, the CPU isn't going to be a bottleneck. I'm going to rip a blu-ray tonight and see how much the encoding speed for x264 @ 720p improves over the Q6700 (As that's the real reason I upgraded).

I wouldn't worry too much about your VID b/c overclocking results I've seen don't really seem to relate to VIDs with these chips. Mine is a 1.2125 VID.

Keith, you need to drop your FSB Vtt setting below 1.4 ASAP!! Anandtech fried a QX9650 because they had that set above 1.4V.

Also apparently Vcores above 1.4 will cause degradation on these quads so I set a hard limit for myself of 1.3625V or high temps..whichever came first. I'm at work right now so I'll post all my bios settings as soon as I get home (couple of hours).

Aside from the standard voltages, I only tweaked the CPU GTLs and NB GTL. Rough estimates..

466 x 9
RAM @ 935

Vcore ~ 1.356V
FSB Vtt ~ 1.36
Vnb ~ 1.46
CPU pll ~ 1.55
SB 1.1 ~ 1.15
SB 1.5 ~ 1.55

load line calibration on


I think finally setting the GTL values correctly helped alot. Also I've got 8GB of Mushkin 1066mhz RAM that right now underclocked to 935mhz or something. The next highest speed is 1135 and I'm not sure all 4 sticks can oc that much. I'm just happy I can get this high with 8GB installed.
 
Thanks prasvt. I read that article over at Anandtech but it just hasn't
really soaked in until you brought it to my attention. I have started with
backing off the vTT to 1.39xxx and will do a ten round Intel Burn test
to check stability. Then I will work on lowering the other voltage one at
a time checking the stability along the way. But I know from experience
that once I get to those voltages you have posted above that this thing
will not boot or maybe even post. @466FSB. How tight are your
timings at that FSB?My ram is at a 3:5 ratio (450x9=1350MHz)
with a 7-7-7-21 timing @v1.90. My Ram is Kingston
PC3-10700/1600Mgz.

Do you feel safe calling your FSB466x9 set-up a 24/7?

Thanks so much! -keith
 
You know what...once I adjusted the GTL voltages, it let me back down a bit vcore wise...just gimme a minute, let me go into the bios and get the settings I tweaked. Back in a few min.
 
CPU Clock Skew - Auto
NB Clock Skew - Auto
DRAM Clock Skews - All Auto
DRAM Timing control - Auto
DRAM Static Read Control - Disabled
DRAM Training - Disabled
Mem OC Charger - Disabled
AI Clock Twister - Auto
Transaction Booster - Auto
PCI-E - 100
Vcore - 1.35625
PLL - 1.553
FSB Vtt - 1.365
DRAM - 2.078
NB - 1.471 <--- may need to see if I can lower this
SB 1.5 - 1.553
SB 1.1 - 1.153
CPU GTL:

0 = +50mv
1 = -40mv
2 = +50mv
3 = -40mv

(using the equation given here)
 
I noticed you have quite a few auto settings on the first half of your settings list above.
Maybe thats how you are obtaining stable 466fsb. I just also realized you have a
Maximus II mobo. That board OC's better than the RE boards. I read the ASUS
team was disappointed in the RE at the annual OC'ing tournament last year because
of its NB bottlenecking the RAM. There's quite a few distraught customers over
this quirky chipset if you search the Asus sight. It seems the Maximus, Striker, and
the P series did the best OC'ing. Its not Asus fault completely. Intel done it again
with the new 1366 x58 boards. I read in CPU Today that these new x58 chipsets are
holding back the new i7 cpu's the same way the the x48 are holding back the faster
quads like the q9650 and above. Oh well, I'm still in the 4 gig club. I just though I
could get a little more with the H2O I'm pumping through its veins.

Those gtl's are interesting too with the +-+- configuration. I've seen this done
before. I will study this more.

Thanks for you tweak readings.

Keith
 
[QUOTE
I think finally setting the GTL values correctly helped alot. Also I've got 8GB of Mushkin 1066mhz RAM that right now underclocked to 935mhz or something. The next highest speed is 1135 and I'm not sure all 4 sticks can oc that much. I'm just happy I can get this high with 8GB installed.[/QUOTE]

No doubt your ram will OC more, I believe its more of an issue of getting that freq to
pass through the NB on these modern chipsets. There's always a bottleneck
somewhere. I would say your 935 is very decent with 8 gigs.
 
Yeah I left all the RAM settings on Auto since it was underclocked. So, upping your northbridge isn't helping?

Yeah once I tweaked the GTL voltages, I could pass prime95 blend test. I think the BIOS gives you the equation for each core ...it's something like .635*FSBvtt + item = .667*FSBvtt for cores 0 and 2 and then .667*FSBvtt + item = .635*FSBvtt for cores 1 and 3.

As for the motherboard, yeah initially I was trying to choose between an X48 board and a P45 board and it seemed the P45s OC'd pretty well. I lose SLI capabilities, but I'm doing an eVGA step-up to a 295 so that should suffice for a while.

I would say try and set the GTLs based on your FSBvtt (hell, set that to 1.39) and up your NB a little and give it another shot. I take it the Rampage Extreme uses DDR3?
 
prasvt said:
Yeah I left all the RAM settings on Auto since it was underclocked. So, upping your northbridge isn't helping?

I'm actually trying to get that number lower.
Its to high I believe at v.1.577
compared to others.

Yeah once I tweaked the GTL voltages, I could pass prime95 blend test. I think the BIOS gives you the equation for each core ...it's something like .635*FSBvtt + item = .667*FSBvtt for cores 0 and 2 and then .667*FSBvtt + item = .635*FSBvtt for cores 1 and 3.

As for the motherboard, yeah initially I was trying to choose between an X48 board and a P45 board and it seemed the P45s OC'd pretty well. I lose SLI capabilities, but I'm doing an eVGA step-up to a 295 so that should suffice for a while.

I would say try and set the GTLs based on your FSBvtt (hell, set that to 1.39) and up your NB a little and give it another shot. I take it the Rampage Extreme uses DDR3?
Click to expand...

Yeap, DDR3.
 
You must log in or register to reply here.
Back
Top