Finally - accurate TDP for your individual AMD processor

[pcoltrane]

It's always annoyed me when people use the rating they find on some 3rd party website (or even on AMDs) to compare A64 TDPs with other chips. The TDP listed is generally the absolute maximum. The reality is often half as much since power leakage varies from batch to batch.

Check out this program to find out the TDP for your individual processor:
http://www.thecoolest.zerobrains.com/forums/viewtopic.php?t=83

People rave about Dothans, Turions etc, but consider this. The real TDP for the 3000+ venice I run in my HTPC is just 32.4W. Since it happily runs at 1.1v at stock speed that's just 20W! I haven't tested yet but there is a good chance it will run at 1.0V ... just 16.5W!
No wonder i can so easily cool it passively!

Incidentally AMDs TDP figures are absolute worst case scenario and will typically never reach the TDP. Other manufacturers including Intel give typical instead of absolute maximum ratings. This is misleading and makes chips like the dothan etc appear better than they are in terms of power consumption when compared to the turion etc.

Perhaps everyone could list their chip below to see what kind of spread of chips we have at ocuk.

I'll start off with mine:

3000+, 939, DH8-E3, 53C, 32.4W

Marc

 

[pcoltrane]

According to AMDs specs Opteron 165/170/175/180s can vary massively. Anything from 35W to 110W :O

I think the majority come in at around the 35W mark! It would be interesting to see how the X2s and 165s people here have vary.

Marc

 

[pcoltrane]

It reads the information stored in the chip hence why older chips (the winchesters etc) don't give TDP. Also it doesn't go by your current clock speed - it just goes by the data encoded on the chip. The variation is purely down to the batch that your chip came from.

Marc

 

[pcoltrane]

Wrong!! Unless 2.8GHz is stock for my 4200

It may list your current CPU speed, but my point is that the TDP it lists does not change with your clock, it just reads the data stored on the chip and works out the stock TDP from that.

 

[pcoltrane]

I'm not convinced. For example my "32.4W" 3000+ venice A64 certainly runs extremely cool as the TDP suggests, but it's not the best chip I've seen at undervolting. Some people can run stock speed at 1.1v or less with their chips but at 1.1v with mine the best i can do is about 1.5ghz.

I suspect this TDP is dependant on power leakage / internal resistance of the chip and therefore current draw. From what i can gather these numbers - the TDP and tcasemax represent the thermal resistance (C/W) of the chip. Ie - AMD measure the chip under controlled conditions (stock voltage, ambient temperature of 42deg etc). The tcasemax represents the temperature that the test rig stabilises at under full load (one would assume the temp measured say exhausting from their test rig enclosure). Once you know the temperature increase you can calculate the thermal resistance and TDP of the processor. Of course this could just as eaily be done the other way around with AMD measuring the current draw of the chip and calculating tcasemax and TDP etc from there.

The question seems to be - is there a link between this TDP and overclockability? Processor design, manufacture and semiconuctor behaviour is pretty complex stuff. i won't pretend to understand it. I am not yet convinced though that there is a direct the link between TDP an overclockability. There may or may not be one, but there are certainly other factors involved too.

I'm going to have a hunt through the AMD white papers next week and see if i can find some more info on how it is calculated. for now though there is an interesting thread on SPCR theorising on the significance of this TDP rating.

http://forums.silentpcreview.com/viewtopic.php?t=30774

Marc

 

[pcoltrane]

Someone got out the wrong side of bed today

Not everyone reads XS, and as can be seen by the number of posts in this thread it seems to be a topic of discussion quite a few people are interested in talking about here. Or perhaps we should forget about these forums and just move to XS?

I for one am finding it interesting to see the range of TDPs we have among the ocuk members. They all seem to be nice chaps so I'm quite happy to chat with them about it even if it is rehashing ground covered elsewhere

Marc

 

[pcoltrane]

Interesting. I wonder which underclock further - the high tcasemax or the low tcasemax chips. I suspect the ones that overclock best may also underclock best (but will still pobably give out more heat than a low TDp chip)