Real Temp - New temp program for Intel Core processors

[unclewebb]

You can buy a new system with a Core i7-900 series CPU. Those sensors work excellent at idle or at full load.

With a 45nm Core 2 CPU, you're screwed. Your sensors actually work better than most Core 2 CPUs and at least now you know their sticking point. These sensors were not designed to report accurate idle temperatures. For that purpose, they're all junk. Some are worse than others but none of them are worth a damn.

[lowtcell]

Yeah, that's what I thought. This is a build I'm working on for my son. He's about to turn 13, and his current AGP system is a bit out of date. I'm sure he's not going to check the temps once I turn it over to him. I just wanted to get everything checked out before I give it to him.

BTW, is the correct TJMax for the E8600 is 95c?

[SimpleTECH]

BTW, is the correct TJMax for the E8600 is 95c?

I believe it is 100°C.

@unclewebb, I think I got my issue resolved. It turns out to be an issue with my mosfets overheating. I put some high-speed fans on them and I'm not getting any more throttling.

[unclewebb]

The Intel TMax spec is 100C for an E8600 but they also admit that is just a spec and actual TJMax can and does vary. They have never stated how much variation to expect or whether that variation is on the plus side the minus side or a bit of both.

SimpleTECH: Sounds like an interesting fix for throttling. Whatever works is a good thing.

[bing]

The Intel TMax spec is 100C for an E8600 but they also admit that is just a spec and actual TJMax can and does vary. They have never stated how much variation to expect or whether that variation is on the plus side the minus side or a bit of both.

That is because during the binning process, they marked and sort of burned that temp level at the cpu internal tjmax register at which the cpu started to error at that particular temperature level during the toasting process, and each cpu is unique. Of course they also have certain "acceptable" range in order to select the good one from the bad one.

So there is no real technical or business advantage to generalize the same tjmax for all cpu. Other wise they will throwing away a lot of cpu just to make the all cpu have the equal absolute tjmax

Absolute tjmax temp may differs say one cpu at 110 while another one could be at 90 even they're originated from the same wafer.

[timo.rolloos]

Wow thx, works great!

[burebista]

Nice temps. Especially on Core1.

[namurt]

quick question out of curiosity. is the default tjmax reported by realtemp correct for my xeon x3440? i've noticed that realtemp correctly recgonizes my cpu, whereas coretemp thinks it's an i7. however, i'm not sure what my tjmax is supposed to be. not many people are running these xeons, so all i manage to find are numbers for the core2 quads and the i5/i7's.

[The1]

Hi Uncle, thanks for all of your efforts refining and troubleshooting RealTemp. I purchased a QX9650 and shortly thereafter, found your utility.
Have been using and following it's development, which brings me to my post. I am having difficulty getting real temp to read the .5 multiplier in version 3.49.
JohnZ had the same problem in post #3917 and you were going to send him a "beta" too see if that fixed it. Didn't see if that fixed the issue, if so could you point me towards the beta, if not...?what can I do to help??
Vista Ultimatex64
Thanks for all of your efforts.

[burebista]

Here's the latest beta. Give it a try.

[The1]

Hi burebista, thanks for the link. :0)
Fileden "seems" to be down right now??
Will try again later and let you know how the beta works.

[unclewebb]

For the QX9650 try adding this to the INI file:

MSRMulti=1

namurt: The QX is a special processor and the Core i7 method that works on most Core 2 CPUs has issues on that one so usually the older MSRMulti method works OK for that one.

The X3440 should have TJMax written into a register within each core of that CPU and RealTemp should have no problem reading that. The typical value is 99, but I think I've seen 92 or 93 on some of the ES processors and maybe a few that are 100. Each CPU is unique which is why software has to read the info from that register.

There is no guarantee that the info Intel wrote in that register is 100% accurate but that's all monitoring software can do is rely on that register.

[namurt]

thanks for the response, uncle! appreciate it.

realtemp is reporting a tjmax of 97C by default. (coretemp is the same.) i'm content with trusting this number as i never allow my distance to tjmax to get any closer than 25 anyway.

[Monstru]

The1 - whenever there are problems with fileden, know that we also host th latest RealTemp here.

[M4rk]

Hi, you're going to hate me, but i would like to ask for some assistance with calibrating RealTemp.

I have a Q9550 (E0) running at 3.33GHz at 1.144V. With the Noctua NH-D14 it idles at 39-40C (ambient of 22C), which is quite disappointing. Though, I am aware that 45nm Wolfdale/Yorkfield chips often have stuck sensors.

Idle:



Sensor Test:



I'm not sure what to make of it, but i'd like to use the idle temps of the cooler running cores. Although I think i've seen you state that Core0 is generally more accurate, logically thinking it seems to be the wrong one.

I plotted the data in Excel, and came up with this:



Judging from that ^, it actually seems that Core1 doesn't move linearly, and that Core0 may have a different tjmax.

Any help would be greatly appreciated, thanks

[cjbrown80]

Wondering if anyone can help on this issue. RMA'd my Eclipse+ in mid Decmber and opted to wait for a new one instead of a refurb. It arrived two weeks ago and I just rebuilt the machine last night.
The PCIex bus died on the first one but I had it at 4GHz (20x200) 1.416v CPU, 1.44v QPI, realtemp read usually in the mid 60s C. This new one's been happy at 3.9 (21x187), 1.3v CPU, 1.46 QPI, but RealTemp is now reading 77C! WTF?!? Can that be right?
I installed speedfan and it was reading 12 to 15 C lower on each core, so one of them is lying. I do realize they interpret the sensors differently but that is a HUGE margin. This all made me very paranoid, but P95 has been running with no errors for a couple hours now so it seems pretty stable. It's a Corsair H50 (exhausting) that I attached with AS5, with it's stock fan usually running at appr 1700 rpm under 100% load. The idle temp on the original Eclipse+ was around 32 with Realtemp, but this one idles at around 38 to 40. Is it just because it's a different board? Any ideas?

EDIT: Scratch that! It just crashed, aaaargh!
Also just noticed that it's throttling my multiplier back to 20x from 21x dropping me to 3.75gHz. So the system is noticing the excessive temps I guess.

[SimpleTECH]

unclewebb, I noticed that Real Temp 3.50 RC6 does work with my X3440 and Windows XP.

When I double-click it, it does not open. If I click on RealTempGT.exe it shows 3 cores.

I then downloaded version 3.40 off of TechPowerUp! and it works fine.

Just thought I'd let you know.

[Wishmaker]

unclewebb, I noticed that Real Temp 3.50 RC6 does work with my X3440 and Windows XP.

When I double-click it, it does not open. If I click on RealTempGT.exe it shows 3 cores.

I then downloaded version 3.40 off of TechPowerUp! and it works fine.

Just thought I'd let you know.

I've encountered this issue. You launch RT and it does not open yet the process appears in the task manager. What you need to do is to end the process, reboot, delete the program, extract it again and run it. It will run fine afterwards.

[cincyrob]

yes they do like to have stuck sensores for some reason. i have 2 of them 0core and 2core. both wont go lower than 39c. and 3core wont go under 36c but once under stress they are just fine. core 1 is the only one that isnt stuck for me.hence the nice low of 26c. i love that..lol im getting the new HAF922 in the next day or 2 waiting on it to be shipped. hope that will drop a few degrees..

[unclewebb]

cjbrown80: RealTemp and Core Temp both follow the Intel directions for reading TJMax directly from the Core i7 CPUs. If SpeedFan does not do this correctly then it won't be able to report your temperatures correctly.

SimpleTECH: Thanks for that. That's not the first I've heard of that issue. RealTemp 3.50 always opened up fine for me so I'm not sure if there is a borked version running around in the wild or what's gong on. I'll plug my XP drive in and see if I can figure out any problems. I mostly use Vista or Windows 7 these days and haven't had any problems.

Wishmaker: Your testing might help me find the RealTemp hangs at start up issue.

cincyrob: Your version of RealTemp is so damn bright it's hurting my eyes.

M4rk: Intel's sensors on the 45nm Core 2 Quad chips don't start to become consistent until beyond 70C which is where Intel calibrates at least one of them. TJMax can vary significantly from one core to the next on these CPUs but you can't determine or take a wild guess at that until the CPU is at a much higher temperature than what you are running at.

There is so much error in the Core 2 45nm sensors caused by a variety of issues that accurate temperatures is mostly a guessing game. I tend to trust core 0 the most. Run your CPU as cool as possible and you'll be able to overclock it reliably as high as possible. Don't put too much effort into trying to get an exact temperature out of these crappy sensors. That's not what Intel designed them for and it shows.

[Wishmaker]

cjbrown80: RealTemp and Core Temp both follow the Intel directions for reading TJMax directly from the Core i7 CPUs. If SpeedFan does not do this correctly then it won't be able to report your temperatures correctly.

SimpleTECH: Thanks for that. That's not the first I've heard of that issue. RealTemp 3.50 always opened up fine for me so I'm not sure if there is a borked version running around in the wild or what's gong on. I'll plug my XP drive in and see if I can figure out any problems. I mostly use Vista or Windows 7 these days and haven't had any problems.

Wishmaker: Your testing might help me find the RealTemp hangs at start up issue.

cincyrob: Your version of RealTemp is so damn bright it's hurting my eyes.

M4rk: Intel's sensors on the 45nm Core 2 Quad chips don't start to become consistent until beyond 70C which is where Intel calibrates at least one of them. TJMax can vary significantly from one core to the next on these CPUs but you can't determine or take a wild guess at that until the CPU is at a much higher temperature than what you are running at.

There is so much error in the Core 2 45nm sensors caused by a variety of issues that accurate temperatures is mostly a guessing game. I tend to trust core 0 the most. Run your CPU as cool as possible and you'll be able to overclock it reliably as high as possible. Don't put too much effort into trying to get an exact temperature out of these crappy sensors. That's not what Intel designed them for and it shows.

No worries. Let me know what I have to do. Sorry for my absence, I've had a month of exams and I used my main desktop like a netbook : editing documents , slideshows and printing projects. Talk about having and overkill system for stuff like that .

I have the RT problem again after fixing it two days ago . It seems that it randomly stops working. The process is shown in the taskmanager, 2nd process top to bottom but the program does not load. I have to end task, reboot, etc.

[unclewebb]

After you reboot does RealTemp appear normally?

This might be caused by the ATI code I added. You can try adding this to your RealTemp.ini file next time you have a problem.

NoATIGPU=1

I plan to make the ATI code optional in the near future. It samples my 5770 with no problems but for other combinations it's hit and miss. It might work OK when not in CrossFire but might choke when you are in CrossFire. I was hoping ATI would release a driver by now that would magically fix a few things. Did you update to 10.1 recently or are you using an older driver?

[stasio]

Uncle,I saw Real Temp 3.50 and i5-660,didn't match CPU speed and multi.
1.http://www.xtremesystems.org/forums/...&postcount=101
2.http://www.xtremesystems.org/forums/...2&postcount=21
Any particular reason?Thanks.

[humeyboy]

After you reboot does RealTemp appear normally?

This might be caused by the ATI code I added. You can try adding this to your RealTemp.ini file next time you have a problem.

NoATIGPU=1

Thanks, that fixed this PIA issue I had.

I would end up with multiple RealTemp.exe's in the Task Manager but not see it running (I clicked it multiple times).

I am on Vista 64 SP2 and have a Nvidia GPU and RT 3.50.

EDIT : Just to let you know as it was totally random, I did not need reboot to run it after adding that line to the " .ini file ", I simply edited then saved and ran it and it opened 1st time.

[M4rk]

Thanks uncle. Given that the 45nm C2Q sensors are crap, I decided to just calibrate my idle temps, regardless if it's the right thing to do.

Also, I just noticed this:


My FSB is at 390MHz, as reported my CPU-z. New bug?