Thermal curiosities, A64

[DrZDO]

See sig. for specs, but I'm using the stock Coolermaster alum. hs/f for testing here.

I'm using Speedfan to monitor CPU temp., which it reads as "Remote." Idle ~48C, Load~61C. Shutting off Prime95's torture test results in an instant (one second) drop from around 61C to around 52C.

Can I assume:

A, that it's reading the A64's internal thermocouple

B, that 61-63C isn't hot enough to worry about?

I believe the complete range of temps I've seen at this slight overclock is 46-64. I'm leaning toward believing it's reading higher than most people's rigs due to the method of measurement, and not that it's actually hotter. Sound right?

Note: I didn't do temp. testing with the Tt hsf, but both hsfs sit at ~50-52 in BIOS.

 

[nJett]

that is pretty toasty - I'd probably remount the HSF just to be certain everything is fine there and double check that you aren't overvolting the CPU.

 

[zephyrprime]

The stock heatsink isn't so good. Those temps are normal for an overclock with the stock heatsink.

 

[DrZDO]

Originally posted by: nJett
that is pretty toasty - I'd probably remount the HSF just to be certain everything is fine there and double check that you aren't overvolting the CPU.

Nah, like I said, similar BIOS cpu temps with 2 diff. heatsinks. It's installed properly.

CPU Voltage is 1.425, 0.025 more than stock simply because I put ~100MHz more speed on. Reads same in cpu-z.

 

[JohnAn2112]

Set it back to stock voltage. You shouldn't need to ad .025 more just for a measly 100MHz OC.

 

[F1shF4t]

Also that really hot for a 3200+, even a stock heatsink is overkill for non overclocked single core athlons. (3000+, 3200+, 3500+, the higher ones get more toasty)
Also reduce the voltage, at 1.425v u should be able to hit like 2.4ghz.

EDIT: also i suggest remounting the heatsink, and checking that the fan is spinning properly, cause they should not run that hot. Use better thermal paste also the stock stuff is not so good.

 

[DrZDO]

Originally posted by: Dark Cupcake
Also that really hot for a 3200+, even a stock heatsink is overkill for non overclocked single core athlons. (3000+, 3200+, 3500+, the higher ones get more toasty)
Also reduce the voltage, at 1.425v u should be able to hit like 2.4ghz.

EDIT: also i suggest remounting the heatsink, and checking that the fan is spinning properly, cause they should not run that hot. Use better thermal paste also the stock stuff is not so good.

Yes, I know it sounds hot, that's why I OP'd. But I don't think it's actually overly hot (The heatsink is barely warmer than ambient to touch). I think it's reading the cpu internal diode temp, not the cooler socket temp like most motherboards would. I really wanted some feedback to see if that's probably the situation.

I'm quite sure it's installed correctly, with Arctic Alumina, and gee, the fan is on too. It barely runs hotter than the Tt hsf I've used for months. The reason I swapped was to eliminate that hsf as a cause for high temps.

btw, I'm only overclocking 100MHz because I can't be sure SATA ports are locked. This is a rare board model with no support. I have a better one coming, and some Ceramique as well.

Edit: This screenshot shows the rapid temp. fluctuations from relatively idle to P95 back to idle. The duration of the chart is about 1 min. 10 seconds total: http://img139.imageshack.us/img139/245/heat14kn.jpg

 

[imported_h04x]

First off.. All the 8th Gen Athlon series processors use an on-die thermal diode for temperature measurement. What that means is that the device used to measure the temperature for the chip is actually a part of the chip itself. This design is useful because you get the truest reading of actual die temperature with the measurement device as an integrated part of the die.

Secondly, the maximum thermal design specification for all Athlon series chips can be found in the OPN, or ordering part number. For your chip (assuming that you have a socket 939 .09µm part) the OPN is ADA3200DIK4BI which breaks down as follows:

ADA 3200 D I K 4 BI

ADA is the processor brand name (Athlon64 in this case). 3200 is the series/model number. D is the package type (939 pin lidded OµPGA here). I is the operating voltage (1.40v nominal for the 3200+). Now the important one, K which is the maximum operating temperature for the die. In this case it is 65°C, which you have been within one degree of, as the on-die temperature monitor is reporting temps of 63-64°C. 4 is the L2 cache size (512KB) and BI is the Part definition (939 package, CPUID model number F1, Revision D0, 0.09µm).

In my opinion you are getting mightily close to the thermal design limit of that CPU. While it may run fine at those temps, you are most likely severely shortening the life of that CPU by running it that hot for extended periods of time. I have a 3500+ in the box that I am writing this post on, it idles at 38°C and hits 50°C full load with a decent Zalman HFS and some Arctic Silver 5. I think you might want to back the vCore back to stock, as it shouldn't take a volt bump to push 100 MHz, and possibly examine the cooling situation in your case. I don't care if you have the best HSF on the planet, it does no good if it?s bathed in hot ambient air. Barring that, you might want to look into a better aftermarket HSF, such as some of the models from Zalman or Thermaltake.


-edited for spelling-

 

[F1shF4t]

Originally posted by: h04x
First off.. All the 8th Gen Athlon series processors use an on-die thermal diode for temperature measurement. What that means is that the device used to measure the temperature for the chip is actually a part of the chip itself. This design is useful because you get the truest reading of actual die temperature with the measurement device as an integrated part of the die.

Secondly, the maximum thermal design specification for all Athlon series chips can be found in the OPN, or ordering part number. For your chip (assuming that you have a socket 939 .09µm part) the OPN is ADA3200DIK4BI which breaks down as follows:

ADA 3200 D I K 4 BI

ADA is the processor brand name (Athlon64 in this case). 3200 is the series/model number. D is the package type (939 pin lidded OµPGA here). I is the operating voltage (1.40v nominal for the 3200+). Now the important one, K which is the maximum operating temperature for the die. In this case it is 65°C, which you have been within one degree of, as the on-die temperature monitor is reporting temps of 63-64°C. 4 is the L2 cache size (512KB) and BI is the Part definition (939 package, CPUID model number F1, Revision D0, 0.09µm).

In my opinion you are getting mightily close to the thermal design limit of that CPU. While it may run fine at those temps, you are most likely severely shortening the life of that CPU by running it that hot for extended periods of time. I have a 3500+ in the box that I am writing this post on, it idles at 38°C and hits 50°C full load with a decent Zalman HFS and some Arctic Silver 5. I think you might want to back the vCore back to stock, as it shouldn't take a volt bump to push 100 MHz, and possibly examine the cooling situation in your case. I don't care if you have the best HSF on the planet, it does no good if it?s bathed in hot ambient air. Barring that, you might want to look into a better aftermarket HSF, such as some of the models from Zalman or Thermaltake.


-edited for spelling-

Yea the athlons 64 have the diode, what about anything else on chip which will save the chip from burning up if the mainboard does not have a max shutdown temp.

Like the p4's will throtle and i think will just shut off without any intervention from the mainboard (could be wrong about the shuting off) So effectively a p4 can be turned on without a heatsink and will not die.

 

[nJett]

This sounds kind of weird - have you tried measuring the temp. by any other methods?

 

[DrZDO]

Originally posted by: h04x
First off.. All the 8th Gen Athlon series processors use an on-die thermal diode for temperature measurement. What that means is that the device used to measure the temperature for the chip is actually a part of the chip itself. This design is useful because you get the truest reading of actual die temperature with the measurement device as an integrated part of the die.

Secondly, the maximum thermal design specification for all Athlon series chips can be found in the OPN, or ordering part number. For your chip (assuming that you have a socket 939 .09µm part) the OPN is ADA3200DIK4BI which breaks down as follows:

ADA 3200 D I K 4 BI

ADA is the processor brand name (Athlon64 in this case). 3200 is the series/model number. D is the package type (939 pin lidded OµPGA here). I is the operating voltage (1.40v nominal for the 3200+). Now the important one, K which is the maximum operating temperature for the die. In this case it is 65°C, which you have been within one degree of, as the on-die temperature monitor is reporting temps of 63-64°C. 4 is the L2 cache size (512KB) and BI is the Part definition (939 package, CPUID model number F1, Revision D0, 0.09µm).

In my opinion you are getting mightily close to the thermal design limit of that CPU. While it may run fine at those temps, you are most likely severely shortening the life of that CPU by running it that hot for extended periods of time. I have a 3500+ in the box that I am writing this post on, it idles at 38°C and hits 50°C full load with a decent Zalman HFS and some Arctic Silver 5. I think you might want to back the vCore back to stock, as it shouldn't take a volt bump to push 100 MHz, and possibly examine the cooling situation in your case. I don't care if you have the best HSF on the planet, it does no good if it?s bathed in hot ambient air. Barring that, you might want to look into a better aftermarket HSF, such as some of the models from Zalman or Thermaltake.


-edited for spelling-

Then why is it running so bloody hot? I reduced vcore to 1.4, and Prime95 hit 61 within a minute. My case has an air tunnel direct to the CPU, and the temps are the same OR HIGHER with the side panel off. Again, like I said, I also have a Thermaltake Venus 12, and it was running hot too (50-52 in BIOS at stock clock, while this Coolermaster runs 51-54).

As far as I'm concerned, unless the readings went wrong somewhere (Occam's razor), something's wrong that has nothing to do with me, like they forgot thermal paste under the spreader, or put a toaster in there instead of a core. Right now, at stock everything, I'm sitting at 47C, 0-2% CPU Usage. Room temp's in the mid 70sF, and I've got that cold air inlet.

I have an A8N32 on the way, and I'll post whether or not it reads differently. If not, I'll just clock out with my c*** out until it cooks, 'cause 2 GHz isn't fast enough.

Something which I specifically find confusing is that (as noted above, see pic link) the temp drops so fast when cpu usage drops. If the heatsink isn't working well, then where are 9-10 degrees C going in 1-2 seconds?

Edit: This is a 90nanometer Rev. E chip (different from D0?). And here is everything scribed on the heatspreader:

right:
amd athlontm 64
ada3200daa4bw
lbbwe 0529dpgw
q382026g52356

lower:
amd
r c 2001 amd

left:
assembled in malaysia

 

[DrZDO]

Originally posted by: nJett
This sounds kind of weird - have you tried measuring the temp. by any other methods?

lol, tell me about it. No, I haven't. I don't think I have anything else I could use for that, but thanks.

 

[imported_h04x]

Originally posted by: Dark Cupcake

Yea the athlons 64 have the diode, what about anything else on chip which will save the chip from burning up if the mainboard does not have a max shutdown temp.

Like the p4's will throtle and i think will just shut off without any intervention from the mainboard (could be wrong about the shuting off) So effectively a p4 can be turned on without a heatsink and will not die.

The Athlon64 provides a hardware-enforced thermal protection mechanism. When the processor?s die temperature exceeds a specified temperature, the processor is designed to stop its internal clocks and assert the THERMTRIP_L output. The THERMTRIP_L assertion indicates the processor die temperature has exceeded normal operating parameters and the chip will come to a screeching halt before the the proc blows it's top.

Edit: This is a 90nanometer Rev. E chip (different from D0?). And here is everything scribed on the heatspreader:

right:
amd athlontm 64
ada3200daa4bw
lbbwe 0529dpgw
q382026g52356

ADA3200DAA4BW is the revision E6 variant at .09µm. Same D package but the A for temp and voltage means "variable" which for that revision processor is 45-65°C for operating temp and 1.35 - 1.40 vCore. Given the post in reply to mine before, I'm left scratching my head as to why the temps are that high given the setup that you have. I am starting to agree with your original assesment that the temps may somehow be artificially inflated by something past the diode adding a few degrees to that reading. I am interested to see what readings you get from the A8N32 in comparison. Just keep an eye on her for now and hope she doesn't blow. Always remember the golden rule:
All electronics are powered by smoke. Once you let the smoke out, they no longer function.

 

[DrZDO]

Originally posted by: h04x

The Athlon64 provides a hardware-enforced thermal protection mechanism. When the processor?s die temperature exceeds a specified temperature, the processor is designed to stop its internal clocks and assert the THERMTRIP_L output. The THERMTRIP_L assertion indicates the processor die temperature has exceeded normal operating parameters and the chip will come to a screeching halt before the the proc blows it's top.

ADA3200DAA4BW is the revision E6 variant at .09µm. Same D package but the A for temp and voltage means "variable" which for that revision processor is 45-65°C for operating temp and 1.35 - 1.40 vCore. Given the post in reply to mine before, I'm left scratching my head as to why the temps are that high given the setup that you have. I am starting to agree with your original assesment that the temps may somehow be artificially inflated by something past the diode adding a few degrees to that reading. I am interested to see what readings you get from the A8N32 in comparison. Just keep an eye on her for now and hope she doesn't blow. Always remember the golden rule:
All electronics are powered by smoke. Once you let the smoke out, they no longer function.

Thanks. While searching, I found some reports that older Asus A7... boards added several degrees to the socket temp. reading to be more "accurate." Maybe this board carries on that tradition. Since I'm the only one on the planet with one, who knows.

 

[DrZDO]

I just set up the A8N32, and it reads about 34 idle. So the old motherboard was either reading the temp. wrong, or overheating the chip.