Temperature Differences Between Similar Cpus

Hi,

I am in the process of 'replicating' my Compaq Presario S5160UK ('upgraded')

To this end I have procured a 'identical' m/b, a PSU (XFX TS450), 512MB PC2700 RAM, a Ge7800GS graphics card, k/b mouse and monitor, and CPUs (replica).

This replica works ok. The matter I am concerned with in this post is the temperature of the CPU. The replica's 1.7GHz CPU runs 5C hotter than the 'original' S5160UK's 2.7GHz under test. My test is to run Prime95 (small FFT) and speed fan, in order to set the PU fan at 100% (about 5000 rpm effective/indicated) and to monitor the temperature. I have a thermometer to measure ambient temperature.

I am comparing my S5160UK with my replica with 2 different CPUs, a 2.7GHz Northwood Celeron in the S5160UK and a test processor of the lowest spec I have in the replica, a 1.7GH Celeron given by CPU world as a 180nm Williamette but by Intel Ark as 130nm Northwood (both sources give the same S-numbers). The heatsinks, of the cuboid aluminium design for socket 478, are identical with very similar fans attached
There some are differences in the thermal and electrical specs, such as the
1.7GHz 2.7GHz
core voltage V 1.75 1.5
TDP W 63.5 66.8
Ma Tcase C 76 74
Max Dissipation W 80 77

Another difference in detail is the layout of externals, principally the PSU's extraction fan is very near the CPU heatsink in the S5160UK (in case with side panel removed, so CPU fan is mainly pulling uncirculated air down from space, which is why I think it similar to the replica) and the replica is free standing m/b with the PSU fan not affecting the CPU)

So there are (small ?) differences. But are they enough to account for a 5C difference in the right direction in (amb 20C - 35 & 40C) under test of the two CPUs ? IDK. Can one even tell ? Should I be looking to improve the heatsinking of the replica's CPU before trying a 3.06GHz Pentium (highest rated supported by the m/b) ? I am using AS Ceramique 2, have tried a few different application methods (credit card pre-spread made no effective contact with the heatsink, so failed)

I apologise for the complexity of this post, basically it is 'have I got the heatsinking [TIM] right ?'

Dumb_Question
7.March.2016


Compaq Presario S5160UK DT261A under XP/SP3
Processor - Celeron 2.7 GHz
Motherboard - MSI MS-6577 v2.1
RAM - 1GB +1GB DDR PC2700
PSU - OCZ 500W StealthXStream {upgraded from Octigen 300W model 10270PSOTG ('upgraded' from original Bestec 250W PSU [in 2011?])}
GeForce 7800GS graphics card in AGP slot

compared to

Processor - Celeron 1.7 GHz
Motherboard - MSI MS-6577 v2.1
RAM - 512MB DDR PC2700
PSU - XFX TS450
GeForce 7800GS graphics card in AGP slot

 

CPU's with different clock speeds run at different temperatures.
Also, the 1.7 GHz CPU has only 512 MB RAM to 'work' with, and you'll also notice the page/swap file of the second PC is bigger than that of the first PC.

 

Case/fan/PSU layout can make a significant difference to expelling hot air and temperature readings from the different sensors, as can differences in fan design specifics (blade design, speed,...) between otherwise similar fans.

Different CPU architecture and manufacturing methods can make a difference to temperature even with the same clock speeds.

Don't bother spreading TIM, just put a small rice grain sized blob in the middle of the CPU and ensure that the heatsink/fan combo are torqued down evenly.

 

I have been looking at the two fans in question.
The one in the S5160UK is an AVC L33637 (1056, smeared out in the picture)
The one in the replica is an AVC L3503T (1046)
[apologies for poor picture quality. I do not have steady hands]
I cannot further more detailed information about these fans, though they look identical and speedfan gives the same (top) speed
Pictures are attached, dust is on the cooler CPU !

As there is no indication of a clear reason for this discrepancy (possibly it's because the replica is not yet in a case), do people believe it sensible to go ahead with the upgrade to a 3.06GHz Northwood Pentium CPU in the replica ?

Dumb_Question
12.March.2016

 

No clear reason but plenty of variables to consider, including the possibility that the XFX PSU's fan simply isn't exhausting as much air from the case as it would be under much lower % load than the stock PSU. Do you have a PWM connection on the PSU and a PWM fan header on the 'board?

There are still aftermarket Socket 478 coolers around from less than a tenner and some pretty good ones further up the scale. Amazon UK have 2x different Akasa models at sub-£11:
http://www.amazon.co.uk/Akasa-AK-67...&qid=1457814206&sr=1-2&keywords=Akasa+AK+675B
http://www.amazon.co.uk/AK-675-S-So...0038O2J1I/ref=cm_cr_arp_d_pl_foot_top?ie=UTF8

 

I stopped using the 'stock' PSU (the 250W Bestec) a long time ago !
The replica is using the XFX 450 PSU and that fan is not exhausting air from anywhere, it's nowhere near the CPU/cooler - all components are laid out on the workbench. It's the S5160UK that's in the case with the OCZ PSU exhausting air pretty much directly from the CPU cooler

I don't have a PWM (4 wire) connector on the CPU cooler fan, nor on the PSUs, nor on the m/b

The Akasa coolers (though different in dimension and probably fan) are different to the ones I am using, being expanded to ~80mm x 80mm, whereas mine are straight sided and have a 70mm square fan. They're all aluminum. I also have a copper-cored intel stock cooler (untried).

Dumb_Question
12.March.2016

 

OK.

You mentioned the temp differences but not the actual peak temps when testing.

For comparison, the 3.06Ghx Pentium has a TDP of 81.8 W and a t-case of 69°C, it would probably start throttling from somewhere above 75°C.

Ark comparison of the 3x CPUs: http://ark.intel.com/compare/27172,27181,27499

 

the actual peak temperatures are about 37C for the in-case S5160UK (Celeron 2.7GHz, TDP 68W Tcase 74C) and about 42 for the free replica (Celeron 1.7GHz, TDP 63W Tcase 76C) under stress.
These temperatures are not the junction or core temperatures as in later CPUs; in fact I don't where they are measured or what method is used by Speedfan. Also these are the numbers from the control panel of speedfan, but an estimate from the graphical display of the relevant temperature at some level between peaks and the baseline (the number shown the control panel). The peak temperatures are 2-3 C higher, and because they are sampled the true peaks maybe even higher
I am unsure whether Northwood Pentiums are able to throttle performance to reduce temperature, or whether that feature was introduced in the Prescotts

Thank you for working out the comparison document. I'm not sure what the thermal information is telling me though ie what it's implications are, but I would think that with the same heatsinking arrangements (ie cooling power) the higher the TDP, the higher the CPU case temperature would be, in proportion: ie if I replaced the Celeron with the Pentium in the replica, the measured temperature would be expected to be
(81.8 x 42 / 63.5) = 54C
which although within the bounds of the operating range is higher than I would like, but ok for a quick experiment.

Dumb_Question
12.March.2016