• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

1055t going down to x4

Well guys I think the thread is getting a bit confused now so for clarity I'm just going to tidy up my ramblings and leave it at that...

1) There are two physical sensors relating to the temperature of the AMD PhenomII CPU: The "CPU Socket" and "Internal CPU" sensors. The former is located on the motherboard underneath the CPU socket. That latter is located inside the CPU case. Before I continue, I shall assume that we're all well aware that PhenomII lacks the physical core temp sensors found in Intel CPUs. So any program that says "Core Temp" on an AMD rig is guilty of mis-labelling the data.

2) Because of the locations of the TWO sensors, you'd expect the "Internal CPU" sensor to react more quickly to changing CPU loads; it is closer to the heat source and it takes time for the heat to travel through the material of the CPU and socket. For similar reasons you'd expect the "CPU Socket" temp to max out at a slightly lower temp than the "Internal CPU" temp.

3) My own tests involving three motherboards and three CPUs show that the figures labelled "Core Temps" in HWMonitor react much faster than TMPIN0/1 to changing CPU loads. Conclusion: The HWMonitor "Core Temps" are reading from the "Internal CPU" sensor and TMPIN0/1 is reading from the "CPU Socket" sensor. Perhaps for AMD systems the writers of HWMonitor should relabel "Core Temp" as "Internal CPU Temp"?

4) AMD's own Overdrive software displays the same figures as HWMonitor "Core Temps". Conclusion: These numbers are important to AMD so they should also be important to us. They may display odd values but a bit of googling will reveal why. It's a scaled offset from some maximum value rather than an absolute reading.

5) The CoreTemp software also displays the same figure as AMD Overdrive "Core Temps" and HWMOnitor. If you go to their website, you can have a read of the FAQs concerning AMD processors. An offset needs to be applied to get a more meaningful reading. Trouble is that the offset varies from CPU to CPU and mobo to mobo. However, this is where TMPIN0/1 become a useful reference. Going back to point 2 above, you'd expect a fully loaded CPU (Prime95 etc) to have an "Internal CPU" reading a few degrees higher than the "CPU Socket". So you just need to apply an offset that gives such a result. For my machine this turns out to be about 20*C. With an offset of 20C, the "Internal CPU" temp runs about 5-6C hotter than the TMPIN0/1 reading. So Gareth, my way of monitoring the CPU temp is actually safer than yours as it turns out.

Now of course this is all utter rubbish because Gareth has had an email from somebody at AMD saying to "read the CPU temp"....

Thanks for listening.
 
Last edited:
pswfps is correct. Phenom cpu's have an internal sensor under the heatsink just forward of centre. This sensor does not read the cores (ATIorNvidia is correct there), but rather they produce a sort of aggregate average temperature based on a formula provided by AMD.
Monitoring programs such as Coretemp attempt to read this sensor and fail dismally, producing spurious figures due to not being yet programmed correctly for AMD 6 core processors.
AMD's vague references to ''overall cpu temperature'' are just that - vague and ambiguous. They are referring to the socket sensor if they say tmpino/1 is the cpu sensor, and that's fine. If you let that get to 62'C then your average core temp will be around 72'C, which is what you would expect and AMD would not set a figure on the cusp of death would they?
If you follow pswfps's method, your cpu would be even safer, 'cause he's not letting his go over 62'C on the internal sensor, around 52/55'C.
 
pswfps said:
Well guys I think the thread is getting a bit confused now so for clarity I'm just going to tidy up my ramblings and leave it at that...

3) My own tests involving three motherboards and three CPUs show that the figures labelled "Core Temps" in HWMonitor react much faster than TMPIN0/1 to changing CPU loads. Conclusion: The HWMonitor "Core Temps" are reading from the "Internal CPU" sensor and TMPIN0/1 is reading from the "CPU Socket" sensor. Perhaps for AMD systems the writers of HWMonitor should relabel "Core Temp" as "Internal CPU Temp"?

4) AMD's own Overdrive software displays the same figures as HWMonitor "Core Temps". Conclusion: These numbers are important to AMD so they should also be important to us. They may display odd values but a bit of googling will reveal why. It's a scaled offset from some maximum value rather than an absolute reading.

5) The CoreTemp software also displays the same figure as AMD Overdrive "Core Temps" and HWMOnitor. If you go to their website, you can have a read of the FAQs concerning AMD processors. An offset needs to be applied to get a more meaningful reading. Trouble is that the offset varies from CPU to CPU and mobo to mobo. However, this is where TMPIN0/1 become a useful reference. Going back to point 2 above, you'd expect a fully loaded CPU (Prime95 etc) to have an "Internal CPU" reading a few degrees higher than the "CPU Socket". So you just need to apply an offset that gives such a result. For my machine this turns out to be about 20*C. With an offset of 20C, the "Internal CPU" temp runs about 5-6C hotter than the TMPIN0/1 reading. So Gareth, my way of monitoring the CPU temp is actually safer than yours as it turns out.

Now of course this is all utter rubbish because Gareth has had an email from somebody at AMD saying to "read the CPU temp"....

Thanks for listening.
Click to expand...
3) nope , not right. i've had 3 motherboards with a 965 and 1090t, the tmpin1 always reacts so more quicker then the "core reading"
4/5) as i keep saying amd overdrive wasn't only made for phenom chip. yes it displays the core temp reading same as hwmonitor/core temp program but it doesn't mean it's the right one to monitor. i've even seen a lot of people that don't eeven have the "core temp reading" they only have tmpin0/1 or cpu temp.

i've even contacted amd by phone and asked to speak to someone who knows about these chips. i explained to the guy about the 2 readings i have (tmpin1 in hwmonitor and cpu in Everest , core temp in both) he said it's the tmpin1/cpu reading that should monitored but he did say it'll be afew degrees out... but that i can understand, it can't be 100% perfect. i've even told him if the cpu dies due to overheating i'll be blaming amd for they wrong information, and be wanting a replacement, and he agreed to that but he sure's me it won't overheat if monitored right. so i've kept the emails for proof
 
Last edited:
'cause he's not letting his go over 62'C on the internal sensor, around 52/55'C.
Click to expand...
Well I'd probably let it go to 71C on the internal sensor which equates to around 65C at the socket TMPIN0 on my Asus board. According to AMD, this 95W chip is good to 71C and like you say, they wouldn't set a figure on the cusp of death. But yeah, you get my drift.

i've had 3 motherboards with a 965 and 1090t, the tmpin1 always reacts so more quicker then the "core reading"
Click to expand...
Well that is odd. Has anybody else experienced this?

he said it's the tmpin1/cpu reading that should monitored but he did say it'll be afew degrees out
Click to expand...
Maybe he said that because he knows that TMPIN0/1 is easier to read BUT knowing that it is the socket reading had better make provision by saying it will be a few degrees out?
 
i really can't see the problem just by monitoring the temp they said.. if the cpu dies due to overheating it'll be amd fault for giving wrong information..

theres no point of all this.
 
Last edited:
pswfps said:
With an offset of 20C, the "Internal CPU" temp runs about 5-6C hotter than the TMPIN0/1 reading. So Gareth, my way of monitoring the CPU temp is actually safer than yours as it turns out.
Click to expand...

Flawed logic, all you've done really is guess a random number that you *think* is correct and applied it as an offset. Anybody can apply a high offset and say their CPU temps are safer.

pswfps said:
These numbers are important to AMD so they should also be important to us.
Click to expand...

Not in the slightest
 
Well no. The idea that the "Internal CPU" sensor is going to experience higher temps than the "CPU Socket" sensor is based on very sound thermodynamic principles. No?

That said, you are right in that my 5-6C margin is something of an educated guess. The problem gets worse with overclocking too as the delta from socket to internal temps will increase quite significantly. I really hope AMD sort this nonsense out with the next gen CPUs and include proper core temps.

But until then, if you're going to measure the CPU temp using TMPIN0/1 and compare it to the quoted AMD max temps, you are almost certainly going to exceed normal operating temps inside the chip.
 
pswfps said:
Well no. The idea that the "Internal CPU" sensor is going to experience higher temps than the "CPU Socket" sensor is based on very sound thermodynamic principles. No?

That said, you are right in that my 5-6C margin is something of an educated guess. The problem gets worse with overclocking too as the delta from socket to internal temps will increase quite significantly. I really hope AMD sort this nonsense out with the next gen CPUs and include proper core temps.

But until then, if you're going to measure the CPU temp using TMPIN0/1 and compare it to the quoted AMD max temps, you are almost certainly going to exceed normal operating temps inside the chip.
Click to expand...

The max temp generally for Phenom II's are 55-62c, that could very well just mean 55-62c CPU temps, and not core temps.
 
rory12 said:
Hi everyone, my brother recently got a 1055t 95w edition and when we were overclocking it and running linx and prime the other day every time it went above 48c it would drop the multiplier down to x4 [x7.5]

his motherboard is a ASUS M4A78LT-M LE
Click to expand...
Hey Rory,

It looks like your abusing that poor little motherboard! . . . if you don't stop it I shall have to report you! :p

m4a78ltmle.jpg


As has been said already in this thread, the boards power circuitry is "straining" with an overclocked Hex Core running flat out full load under LinX and in an effort to keep everything running stable its reducing the CPU multiplier . . . it's a 4+1 board thats rated to run 95watt TDP chips . . . once you clock that 1055T your exceeding 95watts TDP . . .

Adding some cooling to the burning hot VRM's will help a little but that will only get you so far . . . try and keep the vCore & CPU-NB voltages as low as possible and along with the extra cooling you should get a nice little day-to-day overclock going! :cool:
 
Well it could be with ref to the Socket Temp but I doubt it. Their own monitoring software targets the Internal CPU sensor so that's what they think is important, I'd say.
 
I remember reading it somewhere on the AMD site, that was ages ago so I wouldn't be able to find it, either way, whether the CPU/TMPIN1 temp is socket temp or internet CPU temp, it CERTAINLY isn't the core temps, and core sensors don't exist on Phenom II's.
 
Yes, we all know that. But when AMD say "CPU Temp" they could mean either the CPU Socket temp or more likely the CPU Internal Temp. Forget core temps, nobody is talking about core temps.
 
You must log in or register to reply here.
Back
Top Bottom