Improving the thermals on an old B3 Kenstfield Q6600

Page 2 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

BonzaiDuck

Lifer
Jun 30, 2004
16,867
2,184
126
Originally posted by: Thraxen
Originally posted by: aigomorla
fsb termination voltage needs to be up'd to 1.4+ on asus boards to hold 400fsb+ stable. on quads.

the fsb termination votlage is what holds asus boards back. bump it to 1.4V and it should hold 400-450fsb.

Would you recommend setting it manually or leaving it on 'Auto'? Also, is there any point to setting it to 1.4 if running <400 fsb?

On the "Striker Extreme Nightmare" thread, see my latest [recertification, revalidation] over-clock summary for Q6600 @ 3.2 4:5. If I were to go "1:1" the memory would run at [DDR] 712 Mhz. Under either set of circumstances -- with either ratio ( I mean) -- Aigo is right about the 1.40V+ VTT setting.

Previously I'd set VTT, 1.2V_HT and NBCore to 1.45V, but now I discover that I can trim the NB Core and VTT voltages to 1.40V (set). I might even be able to trim the NB Core to 1.35V, but have yet to try. But the V_HT on this board needs to go to 1.45 with a monitored result >= 1.50V. And the VTT (FSB) voltage definitely must be 1.40V for this board.

Ooops -- 1 more here: I'm guessing the 1.2V_HT (this is a 680i board) may be memory-dependent. This is the factor mentioned in the recent Anandtech article on OC'ing the QX9650 -- among the first "problems" to clear up in finding your "stretch" with low multiplier, suggesting tweaks to this voltage.
 

BonzaiDuck

Lifer
Jun 30, 2004
16,867
2,184
126
Originally posted by: jim1976
Originally posted by: BonzaiDuck
Like I said, room-ambient is key to understanding anything about relative cooling performance. What is your room-ambient, jim1976?

I wish I had a G0. I'm in decision-limbo at the moment, for reasons not all related to the technology or its market-values.

I suppose I could flip the B3 and pick up a . . . . . . well, like I said, I'm gonna think about it. . . .


Of course room temp is critical to cooling performance..My avg room temp is around 28C..If it wasn't for W/C then I'd have seen greater load temps for sure. Air wouldn't do the trick with such room temps..
Well I wouldn't jump to a Q6600 if G0 hadn't proved to be good a good revision

And I might wanna jump to higher 24/7 clocks later or bench at higher clocks with exotic cooling sometimes. Given the pros/cons of a Q9450 I went the G0 way..

See my epic-length thread in "STriker Extreme Nightmare" just posted -- referenced above. WC is always better for lower temperatures, but 82F room-ambient would yield just about the same results I already show with my system here -- for B3 Stepping.
 

jim1976

Platinum Member
Aug 7, 2003
2,704
6
81
Originally posted by: BonzaiDuck

See my epic-length thread in "STriker Extreme Nightmare" just posted -- referenced above. WC is always better for lower temperatures, but 82F room-ambient would yield just about the same results I already show with my system here -- for B3 Stepping.

Yes that is why I mentioned that you did a great job with this B3..
Who knows? I still haven't lapped mine though yet.. And my water loop needs some cleaning..
Also I like the fact that I can have my 24/7 3.6GHz @ only 1.35V.. It might help with higher clocks.. Or not.. It depends on the O/C scaling..
 

BonzaiDuck

Lifer
Jun 30, 2004
16,867
2,184
126
Ya know -- I'm getting an itch from change in my pocket (so to speak.)

Since it's the voltage that produces the most thermals, I can't see why I wouldn't be able to get to . . . . . somewhere between 3.4 and 3.6 at that same voltage. You're saying that your core temperatures are hitting 60C on water-cooling at 82F room-ambient?

AigoMorla got that far with his G0 on water, but I THINK it was a G0 and he'd pushed the voltage to 1.45+. Maybe it was a B3-stepping.

And if it's a G0, maybe it's the luck of the draw. 1.35V. Geeesh!!


"Ah be sorely temp-et-ed to ge-at mah-self a G0!!"
 

jim1976

Platinum Member
Aug 7, 2003
2,704
6
81
Originally posted by: BonzaiDuck
Ya know -- I'm getting an itch from change in my pocket (so to speak.)

Since it's the voltage that produces the most thermals, I can't see why I wouldn't be able to get to . . . . . somewhere between 3.4 and 3.6 at that same voltage. You're saying that your core temperatures are hitting 60C on water-cooling at 82F room-ambient?

AigoMorla got that far with his G0 on water, but I THINK it was a G0 and he'd pushed the voltage to 1.45+. Maybe it was a B3-stepping.

And if it's a G0, maybe it's the luck of the draw. 1.35V. Geeesh!!


"Ah be sorely temp-et-ed to ge-at mah-self a G0!!"

No my load temps are hitting 60C with stress tests..
As far as concerns the voltage..
It has to do with the vid (Voltage identification). If you look at Coretemp there's an indication that says a number. Mine's 1.2250 which is good. Not the best but really good and hard to find. Usually vid is higher in G0s, hence the higher need for vcore. If you are very lucky you may even find G0s with lower vid than mine. But it is very difficult and 1.225 is very good as I said..You wanna know the fun part? I can post and run PI and benchmarks at 1.26v@3.6GHz!! :laugh:

Some info that you might find interesting.. ;)

From the C2D datasheet.
2.3 Voltage Identification
The Voltage Identification (VID) specification for the processor is defined by the Voltage
Regulator-Down (VRD) 11.0 Processor Power Delivery Design Guidelines For Desktop
LGA775 Socket. The voltage set by the VID signals is the reference VR output voltage
to be delivered to the processor VCC pins (see Chapter 2.6.3 for VCC overshoot
specifications). Refer to Table 14 for the DC specifications for these signals. Voltages
for each processor frequency is provided in Table 5.
Individual processor VID values may be calibrated during manufacturing such that two
devices at the same core speed may have different default VID settings. This is
reflected by the VID Range values provided in Table 5. Refer to the Intel® Core?2 Duo
Desktop Processor E6000 and E4000 Sequence and Intel® Core?2 Extreme Processor
X6800 Specification Update for further details on specific valid core frequency and VID
values of the processor. Note this differs from the VID employed by the processor
during a power management event (Thermal Monitor 2, Enhanced Intel SpeedStep®
Technology, or Extended HALT State).
...
The processor provides the ability to operate while transitioning to an adjacent VID and
its associated processor core voltage (VCC). This will represent a DC shift in the load
line. It should be noted that a low-to-high or high-to-low voltage state change may
result in as many VID transitions as necessary to reach the target core voltage.
Transitions above the specified VID are not permitted. Table 5 includes VID step sizes
and DC shift ranges. Minimum and maximum voltages must be maintained as shown in
Table 6 and Figure 1 as measured across the VCC_SENSE and VSS_SENSE lands.
The VRM or VRD used must be capable of regulating its output to the value defined by
the new VID. DC specifications for dynamic VID transitions are included in Table 5 and
Table 6. Refer to the Voltage Regulator-Down (VRD) 11.0 Processor Power Delivery
Design Guidelines For Desktop LGA775 Socket for further details.


Vid chages due to energy savings from Intel VRD 11.0 spec (cpu voltage regulator).:

Dynamic-Voltage Identification Functionality
VRD11 architecture includes the Dynamic Voltage Identification (D-VID) feature set,
which enables the processor to reduce power consumption and processor
temperature. Reference VID codes are dynamically updated by the processor to the
VRD controller via the VID bus when a low power state is initiated. VID codes are
updated sequentially in 12.5 mV steps and are transmitted every 5 microseconds until
the final voltage code is encountered. Processors are capable of transitioning from
standard operational VID levels to either the VR11 or extended VR10 VID table
minimum values. They are also capable of returning to a higher VID code in a similar
manner. The low voltage code will be held for a minimum of 50 microseconds prior to
sequentially transitioning through the VID table to a new voltage reference which can
be any higher VID code, but is generally the original reference VID.
 

aigomorla

CPU, Cases&Cooling Mod PC Gaming Mod Elite Member
Super Moderator
Sep 28, 2005
21,128
3,653
126
Originally posted by: BonzaiDuck
Ya know -- I'm getting an itch from change in my pocket (so to speak.)

Since it's the voltage that produces the most thermals, I can't see why I wouldn't be able to get to . . . . . somewhere between 3.4 and 3.6 at that same voltage. You're saying that your core temperatures are hitting 60C on water-cooling at 82F room-ambient?

AigoMorla got that far with his G0 on water, but I THINK it was a G0 and he'd pushed the voltage to 1.45+. Maybe it was a B3-stepping.

And if it's a G0, maybe it's the luck of the draw. 1.35V. Geeesh!!


"Ah be sorely temp-et-ed to ge-at mah-self a G0!!"

G0 @ 415 x 9 - Ummm dont take my loaded temps as normal.
http://i125.photobucket.com/al.../aigomorla/Temps-2.jpg

B3 @ 425FSB x 8 - Again dont take these loaded temps as normal.
Errrrrr..... Cant find the 7 hour. Heres a 3min prime. The Voltage would flux, so add .02 to the voltage you see. Why i FREAKEN HATE ABIT :\
http://i125.photobucket.com/al...73/aigomorla/CM690.jpg

B3 @ 3.6ghz. VERY VERY DIFFICULT!! should not be done on AIR. [board couldnt take the stress for very long :T ]
http://i125.photobucket.com/al...73/aigomorla/Q6600.jpg
 

BonzaiDuck

Lifer
Jun 30, 2004
16,867
2,184
126
Actually, what I meant to say was that you'd pushed it farther. I definitely remember you'd gone beyond 3.6 or 3.7.

So -- nothing new to prove with your screenies, but definitely interesting information.

[ Time to hit the ol' lemon tree again . . . . . I'll be back soon . . :D ]
 

aigomorla

CPU, Cases&Cooling Mod PC Gaming Mod Elite Member
Super Moderator
Sep 28, 2005
21,128
3,653
126
Originally posted by: BonzaiDuck
Actually, what I meant to say was that you'd pushed it farther. I definitely remember you'd gone beyond 3.6 or 3.7.

So -- nothing new to prove with your screenies, but definitely interesting information.

[ Time to hit the ol' lemon tree again . . . . . I'll be back soon . . :D ]

oh

in fsb? ive gone up to 475x8 but thats on a ES chip. Took about 1.5Vcore for that.

Heres 3.825

http://i125.photobucket.com/al...orla/3.825GHZQ6600.jpg
 

VirtualLarry

No Lifer
Aug 25, 2001
56,587
10,227
126
Originally posted by: jim1976
VID codes are
updated sequentially in 12.5 mV steps and are transmitted every 5 microseconds until
the final voltage code is encountered. Processors are capable of transitioning from
standard operational VID levels to either the VR11 or extended VR10 VID table
minimum values. They are also capable of returning to a higher VID code in a similar
manner. The low voltage code will be held for a minimum of 50 microseconds prior to
sequentially transitioning through the VID table to a new voltage reference
which can
be any higher VID code, but is generally the original reference VID.

Ah, so there IS lag with speedstep! If the CPU is transitioned to a lower VID state, then it is held there for a minumum of 50 microseconds, before it can start ramping up again.

I knew that there was a good reason why I always disable speedstep/C1E!

Edit: Does anyone know how many clock cycles there are in 50 microseconds, at say 3Ghz?
 

BonzaiDuck

Lifer
Jun 30, 2004
16,867
2,184
126
Originally posted by: aigomorla
Originally posted by: BonzaiDuck
Actually, what I meant to say was that you'd pushed it farther. I definitely remember you'd gone beyond 3.6 or 3.7.

So -- nothing new to prove with your screenies, but definitely interesting information.

[ Time to hit the ol' lemon tree again . . . . . I'll be back soon . . :D ]

oh

in fsb? ive gone up to 475x8 but thats on a ES chip. Took about 1.5Vcore for that.

Heres 3.825

http://i125.photobucket.com/al...orla/3.825GHZQ6600.jpg

Ok, Aigo, Ok. 3.8+ Ghz, then.

I've seen web-pages on "annual over-clock competitions" scheduled somewhere in the world. I'd say that would be the Bonneville Salt-Flats of Over-Clockers.

Let me know -- I'll chip in a few bucks for your round-trip-fare, dude.

But I'm still trying to decide whether the virtual Bonneville Salt-Flats pilgrimage is worth it on air. Probably a pun and a rhyme . . . . there . . . . too.

[I have far too much time on my hands . . . ]