Do load temps raise quicker on an a64 then an xp?

S

samiam696

Limp Gawd
Joined
Sep 29, 2002
Messages
147
When I run the "small FFTs test (max cpu)" in prime95 my temps seem to jump from their low idle temps to their high peek load temps very quickly. Im talking like a 15-20 degree jump in a matter of 10 seconds, im going from like low to mid 30s to low to mid 50s... With my old amd xp 2400 I think it took a while for the cpu temps to climp to their max when the cpu was placed under load. Could this have just been a delayed response because i was reading from a motherboard sensor rather then an chip sensor? Does the a64 die temp sensor that im reading off of or something?

Im running an 939 A64 3000+ 90nm @ 2400 mhz ~1.5v
MSI Neo2 Plat Motherboard with bios 1.37.
CPU is cooled by a dangerden RBX with AS5...

Also, something else i was wondering about, my vcore seems to jump around by about .03 and they dont match the bios vcore settings: Ill set the vcore to 1.5v in the bios, and itll show it as running 1.4v in windows. shouldnt i be able to hit a 1.7v core with the neo plat2? 5.55v +10% = ~1.7? Could this be a PSU issue, motherboard problem or bios version problem?

For power im running a Forton Source 350, Besides powering the motherboard, cpu and ram, im only powering a r9800npm dvd burner, cd burner, 1 raptor and 3 case fans.

As far as stability goes, i can fully boot to windows and run some benchmarks at 2.7 ghz, however, i cannot bench anything over 2.4ghz without failing prime95 in a matter of seconds, no matter what voltage im running or how low i have the memory/htt clocked, is 2.4 ghz basicly the wall for my cpu?
 
I have the same thing. As soon as I load folding folding temps climb 15-20C's in seconds. I have a Neo Plat socket 754 board. I am guessing its the sensor as I have a custom W/cing loop as well. A64 running @2.5GHZ 1.75voltcore.
 
Vengance_01 said:
I have the same thing. As soon as I load folding folding temps climb 15-20C's in seconds. I have a Neo Plat socket 754 board. I am guessing its the sensor as I have a custom W/cing loop as well. A64 running @2.5GHZ 1.75voltcore.
Click to expand...

I origionaly thought it was just a bug in the core temp program or the mtherboard bios, but the other day i put my finger on the bottom of the heat spreader and waterblock, then started the benchmark test. I could feel them go from a little over room temp to almost untuchable very quickly, im starting the think these temps im reading are accurate... It still dosnt seem like the cpu should get so hot that quickly though,

I think it might either be caused by:
A. Bad contact between the cpu and waterblock, maybe cause by too much as5?
B. Although my RBX is made to attach to an A64 socket, it still seems like its designed for an axp, as almost all of the cooling power seems to be directed twords the very center.
C. The CPU is a crappy overclocker and does not deal with higher temps well.
D. Incorrect reading from motherboard bios temp sensor.
E. Maybe thats just how its sposta be?
any of these ideas sound like they make sense? Right now im sitting here at 33 core and 32 case temps... Let me run prime real quick: juse went from 33 to 53 in about 20 seconds... Thats reading from motherboard monitor, my core center temps are 5c hotter. :(
 
you are going from no load to full load. What do you expect the temps to do?!

On teh opposite end, I can tell you that my temps drop 10 degrees in 20 seconds when I turn my fans from "quiet"->"on"
 
GLSauron said:
you are going from no load to full load. What do you expect the temps to do?!

On teh opposite end, I can tell you that my temps drop 10 degrees in 20 seconds when I turn my fans from "quiet"->"on"
Click to expand...

I guess my last motherboard/cpu (epox 83ka+ and amd xp2400) was very inaccurate as far as the speed in which the temp climbed, cuz it took the cpu a while to heat up when under full load.
 
did you consider that maybe the chip itself might just be better at controlling heat when you're not stressing it? not sure, but didn't a64's have cool 'n' quiet or something like that? was it just semprons?
 
Woofer00 said:
did you consider that maybe the chip itself might just be better at controlling heat when you're not stressing it? not sure, but didn't a64's have cool 'n' quiet or something like that? was it just semprons?
Click to expand...

The 64s do, yeah.
 
Woofer00 said:
did you consider that maybe the chip itself might just be better at controlling heat when you're not stressing it? not sure, but didn't a64's have cool 'n' quiet or something like that? was it just semprons?
Click to expand...

good call... I didnt think of that.
 
I actually have a sempron 3100+ oc'ed to 2.34 GHz and at least for this chip, the temps will spike up within 3 seconds of starting a prime test from 40 to 56 on air. Once it gets going, it settles in at 53 or so. When I kill the test, it'll drop to 44, 45 within two seconds, and within two minutes I'm back down to 40 idle. Cool 'n' Quiet is NOT enabled on my mobo, it's just the way the chip gives on heat - ideally it'll drop the multiplier to drive down clock speed. Semprons run 10-20 Watts cooler than real A64 in general, so I don't know how this will affect your chip.
 
What is making you temps jump so fast is this: On a Socket A system, the CPU heat travels from the core, to the TIM (Thermal Interface Material - Arctic Silver, thermal grease etc) to the heatsink and gets dissipated there. With an A64 system, the heat generated by the CPU has to travel from the core, through the TIM between the core and the heatspreader, then through the heatspreader, then through the TIM between the heatsink and the heatspreader, and finally to the heatsink where it is then dissipated. That is why an A64 heats up so fast, because of all the thermal barriars between it and the heatsink. I have a hot running .18 micron Palomino core 2000+ that jumps 2-4 degrees very quickly when I run CPU Burn-in utility. it then drops 2-4 degrees fairly quickly when I go from full load to idle.
 
Agent_N said:
What is making you temps jump so fast is this: On a Socket A system, the CPU heat travels from the core, to the TIM (Thermal Interface Material - Arctic Silver, thermal grease etc) to the heatsink and gets dissipated there. With an A64 system, the heat generated by the CPU has to travel from the core, through the TIM between the core and the heatspreader, then through the heatspreader, then through the TIM between the heatsink and the heatspreader, and finally to the heatsink where it is then dissipated. That is why an A64 heats up so fast, because of all the thermal barriars between it and the heatsink. I have a hot running .18 micron Palomino core 2000+ that jumps 2-4 degrees very quickly when I run CPU Burn-in utility. it then drops 2-4 degrees fairly quickly when I go from full load to idle.
Click to expand...

??? According to that theory, the A64's should just overheat and meltdown - the extra thermal barriers should make the processor retain heat, and therefore unable to cool down. However, in these cases, the temperature falls just as quickly as it rose. I have that same core, and the temperature jump is only a couple of degrees, whereas we're talking about a temp spike of 15-20 degrees. If anything, I'd say the heatspreader is a FAR more effective way of dissipating heat, and that the A64 simply puts out that much more heat when under load than when idle
 
The heatspreader doesn't spread heat as well as we would like to thing it does. The AS5 you put between the heatsink and the heatspreader is far more effective at conducting heat than the thermal interface material used between the core, and the heatspreader. The heatspreader moniker is more misleading than anything. The heatspreader is more of a protective barriar to keep the core from being damaged, unless you try to take the heatspreader off. I've looked at the thermal power dissipation of various AMD cpu's and an Athlon 64 dissipates about as many watts as my hot running .18 micron palomino core 2000+. I think the cpu I have dissipates 72+ watts of heat. About the same as the A64's. The new .09 micron a 64's put out even less heat.

I have seen some thermal images of a cpu core to show where it's hot spots are. It shows that the heatspreader doesn't spread heat as well as one would think. As I stated before, the heat has to travel through two thermal compounds, and a protective shield, before it reaches the heatsink. It would be the equivelant of you standing on a beach, only to make a running start towards a body of water only to be slowed down by the water as soon as reach it. You satanding on the beach is the CPU sitting idle. When you make a beeline for the water that represents the heat from the CPU rising, and than you hit the water which DISSIPATES your ENERGY. Make sense now? The heatspreader transfers heat fairly well, but it acts as a buffer between the core, and the heatsink where it's energy is dissipated. So the heat races through the heatspreader, and then slams into the heatsink. Just look at how hot the A 64's run, and tell me they aren't boiling under their protective shield. My brother A64 (3000+, .13 micron, 764) at full load reaches 61C. That's in the 150 degree range.
 
Woofer00 said:
??? According to that theory, the A64's should just overheat and meltdown - the extra thermal barriers should make the processor retain heat, and therefore unable to cool down. However, in these cases, the temperature falls just as quickly as it rose. I have that same core, and the temperature jump is only a couple of degrees, whereas we're talking about a temp spike of 15-20 degrees. If anything, I'd say the heatspreader is a FAR more effective way of dissipating heat, and that the A64 simply puts out that much more heat when under load than when idle
Click to expand...

You misunderstand what I said. The thermal barriers don't allow heat to be dissipated until it reaches the heatsink to be dissipated. The TIM's used don't conduct heat perfectly, as in they don't transfer 100% of the heat. They may only transfer 90% of the heat, so the other 10% is retained and results in higher CPU temps. I never implied that the thermal barriers trapped heat, they don't transfer all of it, and even less efficiently as a core to heatsink contact will.
 
Looking for and at signatures, I don't see that any of you have an X800 or 6800 videocard. This same thing happens the moment you start a 3d app with a 6800GT. My temps will jump 15c within the first 20 seconds of starting. It's weird to watch because it also shows an external temperature which moves much slower and not to the same amounts that the core does.
The thing is that these chips (A64, X800, 6800) have internal temp sensors so now we see how dynamic the two states are. We see what's actually happening in the core rather than outside of it.
 
the AMD 64 runs cooler at idle, because of C n' Q and other things. It is normal for the temp to shoot up and down very quickly when loaded and unloaded. This means everything is doing its job.

Even with the stock HSF mine will drop from its uber high (130 F) to an idle of 85 F or so in about 10 seconds. My CPU idles @ only ambient + 15 F, cant ask for much better than that on air.

instant low voltage and clock speed courtesy of C n' Q doing its job
 
Hmmm checking sources...

Hmmm.....


A-XP 130 nm process and less transistors
A-64 90nm process and a LOT more transistors + often higher frequencies

Now according to my High School physics (and some geekiness) the thermal density of the a64 should be alot higher. This means that when it heats up it will heat up faster and also lose heat faster since the temp-delta will be higher.

comments?
 
Frallan said:
Hmmm checking sources...

Hmmm.....


A-XP 130 nm process and less transistors
A-64 90nm process and a LOT more transistors + often higher frequencies

Now according to my High School physics (and some geekiness) the thermal density of the a64 should be alot higher. This means that when it heats up it will heat up faster and also lose heat faster since the temp-delta will be higher.

comments?
Click to expand...

The .09 micron a 64 shouldn't have any more transistors than the .13 micron part. All AMD did was a die shrink, unlike Intel which did add transistors to their .09 micron part that we know as Prescott. The Rev E A64 coming out next year may have more transistors on the .09 micron process. Yes, the Cool & Quiet does play a part in A 64 temps, but CPU-Z doesn't show any core frequency change. Maybe it's just CPU-Z in that case.
 
samiam696 said:
now i wanna rip off the heat spreader :eek:
Click to expand...

Your heatspreader transfers most of the heat just fine, it's the TIM between the heatspreader and the core that probably has the greatest impact on heat transfer. I doubt they use AS Ceramique, or another high quality TIM between the core and the heatspreader.
 
Well the compaarison is between the Athlon XP and the A-64 so no its not between the winchester and the Newcastle :D

And as for ripping heat spreaders of Ull run the risk of destroying the core and having uneven surfaces. And if U use good cooling U also have higher risk of destroying the core when mounting the cooling system (especislly if U use a Vapo).
 
you still haven't explained just why the temps drop so fast once the chips return to idle - i stand by my statement that i think the "heatspreader" does a decent job of dissipating heat, as opposed to my AXP which takes much longer to drop temps, a drop which is far smaller in magnitude.
 
You must log in or register to reply here.
Back
Top