AMD XP vs TBird heat production

[Freshbrain]

The Athlon XP is supposed to have a 20% less power consumption compared to its old brother the Tbird. That would also make the CPU produce less heat, but how much? a 20% less? Would a Tbird 1.4ghz run at 50C while a XP 1600+ (1.4ghz) run at 40C? Thanks a lot.

 

[Mikewarrior2]

if it were 20% (its more like 10-15% cooler clock for clock), it would mean 20% cooler in terms of CPU temp rise over ambient...

So a 50C temp, with 25C temp on a T-bird = 25C rise over ambient

IN theory, with a 20% temp drop, and accurate temp measurement, the XP chip would show 20C rise over ambient.



Mike

 

[Freshbrain]

Yeah, that's what's stated in AMD datasheet, 20% less power consumption IN THEORY, but... how does it reflect in real life? Anyone tried both of them in the same board with same HSF? I had my tbird on a different board with higher readings, so I can't compare my tb 1200 with my 1700+

 

[Mikewarrior2]

In terms of real life, 10-15% cooler clock for clock....

Unfortunately, due to the different PCB, direct comparisons, even in the same motherboard/heatsink/paste, etc are not possible....



Mike

 

[grunjee]

Here's my personal experience... I was curious about this very issue the other day, so I plugged my old 1.33 TBird into my 8kha+.... here's the results compared to my 1700+...

1700+
--------
Idles at 30-32 C, 39 C at full load (running f@h) -- this is at .1 V over default, running 150x11 (shows 1662 MHz or 2000+).


1.33 Tbird
-------------
Idles at 42 C. That was all I cared to see. I didn't do stress testing, and as dumb as this sounds, I'm not sure whether voltage was at default or +.1.

I use watercooling without a pelt (just a standard inline system with a radiator).

So at any rate, after seeing the difference in temps, I'm sticking with the XP line

Edit: According to MBM5, my case/mobo temps sit around 21-23 C.

 

[Mikewarrior2]

Grunjee,

Thanks for hte post... it re-affirms that XP chip temps cannot, under any conditions, be compared with T-bird chip temps...

Your 2 chips vary in max CPU wattage by: 5.8 (T-bird is ~69.8 and XP is ~64).

Even with the worst air-cooling possible, the temp difference (witha 25C ambient case temp) would be ~4C. Since you're using a watercooling setup, it would be much less in terms of DIE temp if the PCB and other things were equal (but they are not).

Is this a case of where AXP pCB really throws off the socket-thermistor? Or is simply just hte socket-thermistor being stupid? Or both .



Mike

 

[grunjee]

Bump for an informative thread, plus I edited my above post.

 

[Freshbrain]

Well, from looking at your test results we could say that the XPs are really cooler than Tbirds... Even considering an error margin it is still clear that there is a very important difference in temps between both of the CPUs.
Besides, how can I calculate a CPU's max wattage? Where do I find specs on the Wattage of my CPU? Thanks a lot

 

[grunjee]

Something else to note here... don't know if this will help or not... but this past week I built both my mom and my sister a system each. For my sister, a 1700+ and my mom, a (Morgan) Duron 1 GHz.

On both of them, I bought retail cpu's and am using AMD's included HSF, and the HSF's both appear to be identical.

We all know that stock HSF's suck, but I never imagined they would suck this bad.

On both systems, temps are almost identical. At idle, mobo temps are about 29 C, and processor temps are about 40-42 C. At load, processor temps are 49-52 C.

Both of these cases have poor airflow inside, so that is probably a factor. Still though, I would never use a standard HSF, even on the 'cooler' Durons and XP's. There is no way in hell I would let my proc run at 52 C!

I don't know how these stock HSF's for the newer procs would do on older TBirds, like my 1.33, but I'm not about to do any testing.

 

[redoak77]

I just changed my Athlon 1.4 to an XP1800+ last night. Neither was overclocked. I used the same heatsink/fan combo (Thermaltake Volcano 5) and Artic Silver II for both. Realistically, I saw an across the board reduction of about 5C with the XP1800+. It sits at about 39C idle and after several hours of gaming it was 46C. The XP1800+ gives a reduction in heat and power usage and runs faster than the "old" Tbird.

 

[Mikewarrior2]

Freshbrain and Redoak77,

Remember that the way AMD chip temps are measured is highly inaccurate, and the differences in temp between T-bird and XP chip are exaggerated by the socket-thermistor (the Ceramic PCB was not as efficient as the new Organic PCB in forcing heat through the primary heat pathway -the CPU to Heatsink pathway).

So, under normal circumstances, you are going to see a fairly sizeable difference in socket-thermistor temps between a t-bird and XP chip due to PCB and other factors. IN terms of actual "die temp" difference, it is much closer.



Mike

 

[Rand]

<<
Besides, how can I calculate a CPU's max wattage? Where do I find specs on the Wattage of my CPU? Thanks a lot >>



AMD Technical Documentation
Intel and AMD both have all the whitepapers and tech docs available on their site, you can find the typical and maximum wattage therein.
Though in the case of Intel they are markedly reluctant to give out exact specifications on their later processors maximum power consumption.

 

[Freshbrain]

I understand what you say Mikewarrior2, that would mean that we should be looking for a better way to measure temps to acurately compare both CPUs. Maybe placing a probe on the side of the core?

Now, would that also mean that the already inacurate reading of temps in the Tbirds has even become worse on tha XPs? I know that temps are usually read lower than they really are, but how low? 6Cº 10Cº??? Thanks

 

[Mikewarrior2]

Freshbrain,

The problem with a side-mounted thermistor is: They still don't register full DIE temp... they often measure a bit of die, some ambient air around the die, and other temps all mixed into its reading. At best, 70-80% of a core -temp rise over ambient is shown with these probes, and at worse, maybe 50% is shown.

Yeah, we should be looking for a reason why motherboard manufacturers have been slow or unable to provide athlonxp/mp diode support. I'm personally beginning to think that current motherboard IC monitoritng chips are incapable of reading the diode. Looking at various people's setups where htey've modded the on-motherboard winbond chip to read the diode, the results have been horrible. C/W figures calculated from the diode should be linear, not vary with different CPU wattages.

Until motherboards consistently provide solutions to reading the diode, temp readings will continue to be inaccurate. The inaccuracy that existed with t-birds is significantly worse with XP chips, so you can probably, probably to the point where XP Die temp is 2-5C lower than T-birds, but socket-thermistors show 5-10C lower. I'd like to say that you could add 10-15C to your temp reading to make the reading closer to true temps, but that's not possible... every motherboard can be so different in readings.


Mike