RealTemp General Discussion

[unclewebb]

The RealTemp Power Limits stuff was for Core i CPUs before Sandy Bridge. I never received any feedback about this feature so I forgot about it. I'll try to remember to block this for Sandy Bridge CPUs in the next release since Sandy Bridge CPUs use a different register to control the Power Limit settings.

I have been working more on ThrottleStop recently. If you click on the ThrottleStop - TPL button, it might give you some ability to control these settings but many manufacturers have decided to lock the new Power Limits register so changing this from within Windows might not be possible.

If you want some control over the CPU multiplier then that's a good reason to try ThrottleStop too. With Sandy Bridge CPUs, I haven't discovered a way to increase the multiplier higher than what you boot up at. I am not sure if this is possible anymore.

Don't waste your time trying to calibrate your temperature sensors. I haven't done any in-depth testing of Core i processors so I have quit making calibration recommendations. Just keep in mind that Intel's temperature sensors were never designed for 100% accurate temperature reporting. Intel has used a variety of sensors with a variety of issues over the years. The only thing that is consistent is how vague Intel has always been about the sensors they use as well as what their sensor calibration procedure is.

 

[pipsie]

looks like thy didn’t set me up

the multiplier didn’t work for me (guessing the i7-720QM doesn’t allow this or the bios), but the disable boost did (It went from 1.7 to 1.5GHz then and the temp decreased 10°C).
The funny thing about that i7 is that it can’t go higher than 1.7Ghz (with I find low) but it does the trick gaming, video editing, ….
My adapter is 90W (19V,4.74A). My laptop is from xxodd an unknown (and now broke) custom made company. So you select a base model and you can upgrade it.
Its almost 2 years since I bought it. It works good (for a laptop) the only bad thing is the temperature and the GPU (mobility radeon hd 5650) isn’t very strong so modern games doesn’t always play well.
But basically it’s an clevo because xxodd bought them from them.
But if i see this right there is no big problem. The temperature stays 5°C under Tjmax at 100% load and the cpu isn’t hold back or anything like that.

Just the extreme heat coming out it is the big problem
When I’m gaming (yes its capable of doing that ) the thing is like an hairdryer and I’m not kidding. I don’t have a temperature sensor that goes over 50°C but it looks like the heat coming out is very close to the temperature of the CPU (with isn’t that odd if you think about it).
That explains why i couldn’t put my hand next to the laptop sometimes. (70°C is hot)

I'm going to try to lift it up (put it a bit higher of the desk) and see if that helps
I thank you for the great software and all the help.

 

[n3tmaster]

@uncleweb: thank u for answer me so fast!
Just another question: i should use ThrottleStop instead? or can I use it with RealTemp? can these software run togheter?

 

[unclewebb]

When using ThrottleStop, you have to click on the Turn On button and you also have to enable the Set Multiplier box in order for this program to control your multiplier. When all 4 cores are active, the maximum turbo multiplier is 13. When a single core is active and the other 3 are asleep in C3 or C6, the maximum multiplier is 21. ThrottleStop reports an accurate average of this for each thread using high performance timers within the CPU. Try running the built in TS Bench benchmark and set that to 1 thread and you should see a much higher multiplier than 13.

For the 720QM, both the Turbo Ratio Limits and the Turbo Power Limits are locked by Intel. ThrottleStop can show you what your limits are but Intel has prevented these limits from being adjusted. The Core i7-920XM Extreme CPUs have both of these limits unlocked so with ThrottleStop you can do amazing things with these CPUs as long as you can find a way to keep them cool.

ThrottleStop can make a fully loaded 920XM run over 100% faster than a 720QM. Here's an example of a 940XM.

http://img339.imageshack.us/img339/8104/fullspeed.jpg

Run ThrottleStop with the Log File option checked and then go play a game. Make sure the log file shows that the two types of clock modulation throttling are at 100.0% all the time. I don't know if Clevo uses either of these throttling methods like Dell does.

The default multiplier for a 720QM is 12 so when the CPU is loaded, the log file should not be showing it going below this value.

Some games are not very CPU demanding so using ThrottleStop to Disable Turbo Boost can help lower overall system temperatures without negatively effecting the frame rates. You can also use ThrottleStop to use two different profiles. The first profile could use turbo boost and then once the core temperature goes over some preset value like 85C, turbo boost could be turned off until the temperature gets back under 85C. The ThrottleStop Alarm feature is used for this purpose.

n3tmaster: ThrottleStop was designed as a solution for the laptop throttling issues that are industry wide. You could use both tools at the same time but I can't think of a reason why you would need to do that. Use whatever tool is best for you.

 

[unclewebb]

The RealTemp Power Limits stuff was for Core i CPUs before Sandy Bridge.

n3tmaster: I screwed up. That statement I posted earlier is not true. I spend most of my time working on ThrottleStop these days so I completely forgot that I added the ThrottleStop Power Limits feature to RealTemp for the Sandy Bridge CPUs. If you are using RealTemp 3.69.1 that I posted above, this feature should work correctly. It's the same code that is in ThrottleStop so this code has been well tested on Sandy Bridge.

There is a register in Sandy Bridge CPUs that controls when your CPU is allowed to use Turbo Boost. There is both a short term power limit as well as a long term power limit. Sandy Bridge CPUs can operate beyond their designed TDP limit for a short period of time (approx. 30 seconds) and then the amount of turbo boost will be reduced if it needs to so that the CPU power consumption decreases.

Most desktop motherboards allow you to set the turbo power limits high enough in the bios so they don't interfere with the amount of turbo boost you get. If these values are set too low, Intel Turbo Boost will be reduced to keep the CPU TDP under the power limit. If you have any specific questions, post a screen shot of this window so I can see how your CPU is setup.

 

[n3tmaster]

Ah ok, I see this setting..
So, for my purposes, I'm going to use RealTemp (I need an alarm system for Core temperature, in ThrottleStop I couldn't see anything about Core temperature alarm). I found TDP settings on RealTemp but I don't known how it works. There ara a lot of settings as "Campling Limit", "Lock Power Limit" and so on.. so Can you explain me how it works?
Thank u so much

 

[unclewebb]

Post a screen shot of the power window so I can explain it to you and help you out.

www.imageshack.us

 

[n3tmaster]

This is the Power Limits Window:

Thank u again for helping me

 

[n3tmaster]

the correct link:

Uploaded with ImageShack.us

 

[unclewebb]

Your bios has set your turbo power limits so high that you will be guaranteed to get maximum turbo boost performance. There is no need to adjust any of these values unless you are interested in getting less than your maximum performance.

If you want less performance the easiest thing to do is use the Disable Turbo function. If you want more control of your CPU, use ThrottleStop. If you have any specific questions about the above screen shot, just ask.

 

[deep-packet]

unclewebb thanks for all your work with RealTemp. I'm a new RealTemp user and I was hoping you could clear up some confusion on my part. I have built a mini-ITX HTPC using the Core i3 2100-T CPU and an ultra-small case (Habey EMC-600B). Currently I'm running the system with the cover off because of my thermal concerns.

At idle, RealTemp (3.69.1) shows a temp of about 40C and distance to TJMax of 45C. In looking at the Intel spec's, I understand that TJMax is an unpublished value (according to your documentation), however, I do see a TCase value of 65C. When I have examined the CPU temperature reported in the BIOS setup screens, I get around 50-53C with the cover off, and over 60 with the cover on. Based upon the RealTemp values, it seems like I have plenty of headroom. (Even the worst case on the RealTemp calibration test reported 25C from TJMax.) However, the BIOS readings in relationship to the TCase value seem to indicate that I might be close to the thermal design limit.

Do the readings from my BIOS indicate a potential problem for the i3 2100-T, or am I misinterpreting this data? Thanks!!

 

[unclewebb]

TJMax was never published for Core 2 processors but it is written directly into each core of the newer Core i CPUs. Most CPUs tend to be 100C but some of the newer Core i3 CPUs are set to 85C by Intel so they don't have as much headroom before they start to throttle and slow down. RealTemp can now read TJMax directly from a register within the CPU so there is no more guessing with Core i CPUs.

I don't know how accurate the bios temperature is. When you are in the bios, the CPU usually isn't able to use any of its power saving sleep states so if it runs hot in the bios, that shouldn't be a surprise.

The only temperature number that is important is the Distance to TJMax number that RealTemp shows. When that number counts down to zero, your CPU will start to throttle and slow down so it can cool itself off. 85C is very conservative for an Intel CPU. If your Distance to TJMax tops out at 25 when running Prime95 then you have absolutely nothing to be worried about. Your CPU will live a long and healthy life. Many laptop users are gaming for hours at over 90C so no worries.

 

[deep-packet]

Thanks for the info. BTW i found a few minor bugs in the 3.69.1 beta version that you may be interested in knowing about:

1. The temp does not always display in the notification area. Not sure what the patter is here.
2. When you click on the "Show hidden icons" arrow, RT shows up as a blank icon, i.e., a white square. The fly-over pop-up and the right-click menu work fine.
3. For the times that the temp does display in the notification area, it seems to always disappear on its own after 45 sec. Not sure if this is intentional, if it is, I don't see any settings that might control this behavior.

Thanks again for a very useful utility!!

 

[unclewebb]

I know there is a problem when you allow Windows to hide your Notification Area Icons.

By default, the RealTemp temperature number is in white and shows up nicely on the typical black or dark system tray bar. When Microsoft hides icons, it can change the background color from black to white so the same white text becomes invisible and then you can't see what temperature it is anymore. I don't know any way to correct this issue so I just set Windows to always show all icons.

When set like that, RealTemp's icons show up OK without any disappearing issues.

 

[dr4g0n]

Hey gents,

I just put together a new computer and ran a Real Temp sensor test to make sure the temperatures are running at a good range. The numbers look fine to me, but I don't really understand how to tell if the CPU is getting throttled or not. Maybe there's something here that I'm not seeing, and I just want to make sure.

Thanks!

 

[Mussels]

thermal status "OK" means no throttling has taken place.

 

[clockazilla]

Hi (appoligies in advance for my spelling - not my strong point),

Yes, its yet another person new to the world of overclocking looking for advice. I though I had better try and calibrate Real Temp before I start poking around those Vcore settings in BIOS! I have tried to do as much reading as possible but now feel I have hit a wall (maybe I just need some hand holding).

I think I understand how Real Temp measures temperaturs: I have a Q6700 - it has 4 temperature diodes - one in each core. the primary function (only intended function?) of these diodes is to report on one temp (Tj Max) when this temp is reached throttling takes place. I have no idea how an actual diode does this but I presume it works on some sort of frequency principle - as the temps go up the frequency signal increases until it hits a frequency which indicates Tj Max has been reached?

From what I understand Real Temp knows the frequency at which throttling takes place but estimates what this means in reality (this is beacuse intel could not release this information) - so Real Temp will set Tj Max at a certain degree c (in my case it has set Tj max at 100) againts which it measures the slower "frequencies" to estimate tempretures.

there are esentially two things that can be altered in Real Temp - Tj Max which is the base line Real Temp uses and you can also elongate or compress the "frequency" scale to address idle temp correction.

I hope I have the above correct?

my rather long winded question is how do you know if you have reasonably approximate TjMax values - the only reason I ask this is because core 2 on my Q6700 reports a temperature difference of around 9 degrees C which does not narrow when approaching higher temps (its a contant 9 degrees cooler than core 0 and 1) - this means that Tj Max is set incorrectly in Real Temp - am I correct? (oh and at idle Real Temp reports this core at 3 degrees below ambient)

If so how do I go about assessing the most likley Tjmax for this core:

PS I'm new to forum posting and have no idea how to upload images - it asks for URL address - I presume I have to upload my immages to an external website and then link here to them?)

If anyone does have the time to read over the above - thanks in advance

 

[clockazilla]

figured out how to post images

 

[unclewebb]

I have seen a lot of Core 2 Quad CPUs that are very similar to yours where Core 2 is reported significantly cooler than core 0 and core 1. My theory is that Intel does not set TJMax consistently across all 4 cores and they do this deliberately to better control thermal throttling. If Intel set the first two cores to throttle at 100C then I think on these early quads, they deliberately set the throttling point of the other two cores at approximately 105C. On your CPU, it might be closer to 100, 100, 108, 105. Doing this would be a great way to prevent all 4 cores from throttling at the exact same time so a user might not ever notice if only 2 cores were throttling. These early Quads consisted of two separate Dual Cores under the heatspreader so it is possible to run each Dual Core at a different multiplier and a different speed from each other.

No one knows for sure and Intel has never documented how accurately they set TJMax so it is anyone's guess what they really did and why. I gave up on doing calibrations because there is too much information that is not known so it becomes more like a guessing game than science.

Enjoy your CPU. As long as it isn't throttling and as long as it is running reliably then there isn't any reason to be concerned about what core temperature it is running at. If you are overclocking then running a CPU cooler is a good thing but other than that, it really doesn't matter too much. My best guess is that your first two cores are probably reasonably accurate.

 

[clockazilla]

Hi Unclewebb,

thanks for your reply (and Real temp) - I have read through some of your previous posts - you must get a bit peaved off with people asking the same questions again and again - so sorry about that - you seem to have quite a lot of patience! thanks for your reply though - it is helpful - all I need is a rough idea of whats going on with temps _ I don't intend to go to extreme levels - not intentionally anyway - just give my middle aged chip a bit of boost!

 

[rjflory]

RT Not recognizing all present cores on Xeon W3680

RT up to version 3.69.1 seems to only recognize 3 of the 6 present cores on a 6-core hyper-threading intel chip.

Info attached, hopefully helpful. Notice the APIC cpu/thread IDs are non-consecutive on these chips- maybe a consideration.

Thanks-

ron

 

[95Viper]

RT up to version 3.69.1 seems to only recognize 3 of the 6 present cores on a 6-core hyper-threading intel chip.

Try using Realtemp GT 3.64

 

[rjflory]

Try using Realtemp GT 3.64

http://img27.imageshack.us/img27/7897/realtempgt.jpg

While the image shows version 3.60, I have since updated to 3.69.1, which I imagine is even newer than 3.64 .

Also, your chip is a "Gulftown" series:
http://ark.intel.com/products/58664...cessor-(12M-Cache-3_33-GHz-4_8-GTs-Intel-QPI)

which is a difference family than my "Westmere" chip:
http://ark.intel.com/products/47917...W3680-(12M-Cache-3_33-GHz-6_40-GTs-Intel-QPI)

I suspect there may be some differences in how the cores are mapped and presented to the OS. If you look at my original post, you may notice the internal cores are grouped in two separate sets of three (0..2) then (8..10). This *may* or may not be be different than your chip. If it is, it could very well explain why RT only displays info for 3 cores.

ron

 

[unclewebb]

I think the latest GT version I released is RealTemp GT 3.64 as posted above. I wish I had more time to get everything updated but with no time and no new hardware, updates are hard to release. Intel's 6 core CPUs are organized internally like 2 separate 3 core CPUs.

RealTemp GT is designed to correctly read all 6 cores and organizes the data in the correct physical order. When you run the regular version of RealTemp on one of these CPUs, it will only read data from the first 3 cores because it considers the other 3 cores to be from a separate and second CPU.

If RealTemp GT doesn't read all 6 of your cores then that's a problem. The APIC ID in these 6 core CPUs is usually

0 1 2 3 4 5

for the first set of cores and then

10 11 12 13 14 15 (Edit: these numbers are in hexadecimal, 0x10 to 0x15)

for the second set of cores.

If this is not the case then RealTemp GT might have a bug and might not be able to handle reading these correctly. What CPU-Z is showing might just be its way of displaying APIC ID information. Post a screen shot of RealTemp GT 3.64 so I can have a look.

Edit: Now I understand the CPUID data. The numbers I posted above are in hexadecimal. So 0x10 is equivalent to 16 and what I call 0x15 is equivalent to 21 decimal so RealTemp GT should work correctly on your CPU.

Edit #2: The regular version of RealTemp used to have an INI option called CPU=1 or CPU=2 which would let you read the first or second CPU in a dual CPU system. This might let you read both halfs of your 6 core CPU with RealTemp 3.69.1 but there is not much point in doing that when RealTemp GT 3.64 is available.

 

[rjflory]

Attached is a screen-shot of RT 3.60, 3.64, 3.67, and 3.691, they all display 3 cores.

Also note that 3.64 displays version 3.67 instead of 3.64. Also, I am confused that 3.67 and 3.691 could be older that 3.64...

I included display directory listings of the downloaded zipfiles so you may check them against your release files, perhaps the public webpage has a bogus link that mixed up 3.64 with 3.67, or may you just forgot to update the version string before recompilation

ron