Discrepancies on Chipset temperatures and possibly CPU temperatures

[BonzaiDuck]

Scenario: Two systems use the same motherboard with a 610i chipset, the same memory, and different processors (E2140 and E2180) which are really "the same."

The E2180 has a multiplier of 10 and is over clocked to 266 with a 4:5 divider to meet the full DDR2-667 memory spec.

The E2140 has a multiplier of 8 and is over clocked to 333 with a 1:1 divider to meet the full DDR2-667 memory spec.

CPU cores on both systems all seem to be within a 3 or 4C degree range. The 2180 seems to run hotter. Cooling uses the stock coolers. Fans are set in BIOS to max out. Both systems are set with VCORE = "Auto" and report idle 1.33V and 1.28V load -- identically.

So if the 2140 is showing 46C on the cores for the very same Large FFTs iteration, the 2180 is showing around 50C for both cores.

The chipset temperatures also vary, despite the same HR-05 coolers, the same diamond thermal-paste, the same consistency of installation. Running Large in-place FFTs on either machine shows the 2180 topping out with an MCP temperature of 30C, while the 2140 tops out at 23C. The chipset coolers both have the same fan installed running at the same rpm.

Would running a 4:5 divider increase chipset temperatures? Is it possible that the 2180 would run cooler by dropping the multiplier and running 8 x 333? This latter question seems contrary to conventional wisdom -- that lower multipliers would cause higher temperatures at the same (2.67 Ghz) speed. Maybe I got it wrong.

Any takers here? I was thinking there could be a calibration problem with the built-in sensors on the same motherboard, but I don't have one of those infrared thermometers . . .

 

[DSF]

Are the systems really identical? What about graphics cards, hard drives, case, power supply, intake & exhaust fans, etc?

 

[BonzaiDuck]

Good point, there, DSF.

On these Gigabyte mATX boards, I'm using the built-in nVidia 7050 VGA for both. The hard drives are identical Hitachi 160 GB SATA150.

BUT -- the E2140 system is in a fulltower case [which, contrary to many opinions around here, I'm convinced offers better cooling than a midtower with more input CFMs and well-directed exhaust]. The E2180 (which is running warmer) is in a plain-vanilla PC Power&Cooling Midtower ATX case. But the CPU temperatures on the E2180 are only a couple degrees -- if that -- above those in the E2140. One more detail for that -- before I forget -- I lapped the E2180; I didn't lap the E2140.

But I lapped both HR-05 coolers and used diamond paste on everything.

The PSUs are different. The E2180 uses an Antec Earthwatts 430 or 480. The E2140 re-deploys my old OCZ PowerStream 520.

Now the OCZ PowerStream used to get warm. In this setup, it feels like it's turned off, except for the slight flow of air from its exhaust fan.

Per BIOS monitor and Everest, the voltages on both power supplies are within +/- 1% of each rail spec. The 12V rail on the E2140 (cooler chipset) is showing about 12.05V. The 12V rail on the Earthwatts E2180-system shows about 11.98V. The 3.3V and 5V rails are also very close to spec on both systems respectively.

The warmer E2180 has a 0.48A 92mm exhaust fan which can put out about 70 CFM, but it's tuned down to 60% top_end rpms; the cooler E2140 has a 120mm Evercool aluminum (2,000 rpm exhaust) with a bout 70 to 80 CFM. Both systems use Panaflo M1 120x38 for intake, pulling somewhere betwen 0.40 to 0.50 Amps.

That's about it.

Whoops. Edit one more -- although I already mentioned it. Both HR-05s fitted with Zalman 80x15mm OP-1 optiun fan, and both spinning at rpms between 2,900 and 3,000.
[You wouldn't believe it, but the reason I got a pile of these OP-1 as spares lies in the fact that at 3,000 rpm -- ya still can't hear 'em -- hardly.]

 

[BonzaiDuck]

Now I'm reviewing my own lengthy saga here, going over the facts, and I realized something.

a) Full-tower v Midtower, the air throughput for the full-tower is greater.
b) I lapped the E2180, which only shows between 2 to 4C excess of load temperatures compared to the E2140. Add back to that difference what I already know was the improvement through lapping, or about 4 to 5C degrees.

c) Reconsidering the difference between load values on the MCP chipset of 23C for E2140 and 29 to 30C on the E2180, We're lookin' at maybe a 6 to 7C difference there.

And considering b), we're lookin' at what may only be a 6 to 7C difference as well.

False alarm.

My typing skills are always improving, though! Almost as good as the Few-ture, when we can walk around with 3.5 mm phone jacks to plug into the back of each other's heads and communicate gigabytes in seconds!!

:laugh: [eeee-eeee-yuck-yuck!]

 

[Mondoman]

Apples and oranges because of the different cases and power supplies. In my experience, the system with the Earthwatts will run hotter, because the PS's internal fan doesn't run as fast/often, and thus provides less airflow through the system.

 

[BonzaiDuck]

Well, the PowerStream used to warm up considerably. Now, both the newer Earthwatts and the OCZ seem to be at room temperature.

It's always been my experience that midtowers present slightly higher case-interior temperatures than full-towers, for various reasons -- provided that both case-types have good intake and exhaust. But in the case (ha) of these two configurations, there is equal intake CFMs, with perhaps better exhaust CFM from the full tower case.

But I measured the improvement in cooling for the lapped processor, and the numbers add up on the temperatures. Anyway, the Earthwatts is used in the midtower, with the OCZ in the full-tower. So, either way, there's a potential difference in these combinations, consistent with my higher temperatures with the E2180 in the midtower and the E2140 in the larger case. It is also consistent with your speculation -- we're just opined differently about why interior air may be warmer.

My biggest concern was some direct electronic cause for the chipset in the E2180 system to be warmer by 7C while the processor-- seemingly -- was warmer by 2 or at most 3C. The difference there would only be the lapping of the E2180.

One thing is especially nice, though. That cheap motherboard measures the Northbridge temperatures. I don't think I found that feature with Everest on my Striker cadillac, but then it's set up with the licensed version of Everest from last year, and the trial version on these budget machines is the most recent software version.