Overclock.net › Forums › Intel › Intel CPUs › The Truth about Temperatures and Voltages
New Posts  All Forums:Forum Nav:

The Truth about Temperatures and Voltages - Page 52

post #511 of 599
Quote:
Originally Posted by unclewebb View Post
Just to clarify a few things
You clarified more than a few things for me with this post.
Thanks
post #512 of 599
Quote:
Originally Posted by unclewebb View Post
As good as the Core i7-900 sensors are, I still don't believe that TJMax is consistent across all 4 cores. I'm 99.9% sure that it isn't and my best guess is that TJMax is very close to 105C on core 3 compared to 100C on core 0 with the two center cores being somewhere in the middle of this range. There is no documentation from Intel to confirm or deny this. It's just a theory I have based on observations.
From memory, while running the cooldown test in Real Temp my i7 920's Core 0 was notably "hotter" than Core 1 at full load, and the other two cores where slightly "cooler" than Core 1. At idle it was a bit of the reverse. Since Core 1 was roughly in the middle throughout the whole load range and had the least deviation from a linear change in reading (although it was pretty close for all cores), I aligned my other cores' DTS to Core 1's DTS. The readings are pretty consistent across all cores through the whole load range with the following settings:



A little bit unorthodox having TjMax below Tj Target I must admit, but that was needed to bring Core 0 down to the others Of course I could just as easily change all of these settings to make Core 0 the reference, but that would mean all cores are reporting 2-3C below it by default.

Quote:
Originally Posted by leppie View Post
Thanks, it makes sense that it can upset the high turbo multipliers, did not think about that, but then again, I dont use Turbo currently
If you happen to get an LGA1366 i7, it really makes no difference. Turbo will never in any real-world scenario reach the second speed bin. It pretty much requires all but one core to be in a C3/C6 state for >98% of the time. Even background processes will prevent this unless every process has affinity set to a single core (the same core).
Edited by randomizer - 2/9/10 at 5:17pm
    
CPUMotherboardGraphicsRAM
i7 920 D0 MSI X58 Pro-E GTX 560 Ti 448 3x2GB G.Skill DDR3-1333 9-9-9-24 
Hard DriveHard DriveOptical DriveOS
840 Pro Caviar Black LG BD-ROM Windows 8.1 Pro x64 
MonitorMonitorKeyboardPower
Dell U2713HM Dell U2311H Turbo-Trak (Google it :D) Corsair HX-520 
CaseMouseMouse PadAudio
CM690 Mionix Avior 7000 Everglide Titan AKG K 242 HD 
  hide details  
Reply
    
CPUMotherboardGraphicsRAM
i7 920 D0 MSI X58 Pro-E GTX 560 Ti 448 3x2GB G.Skill DDR3-1333 9-9-9-24 
Hard DriveHard DriveOptical DriveOS
840 Pro Caviar Black LG BD-ROM Windows 8.1 Pro x64 
MonitorMonitorKeyboardPower
Dell U2713HM Dell U2311H Turbo-Trak (Google it :D) Corsair HX-520 
CaseMouseMouse PadAudio
CM690 Mionix Avior 7000 Everglide Titan AKG K 242 HD 
  hide details  
Reply
post #513 of 599
Most X58 motherboards do not properly support turbo mode as soon as you adjust the BCLK from its default value of 133 MHz. I think the Intel X58 board is one of the few that can let you access the 2 bins of turbo boost that are built into the i7-920 even when overclocking.

Most motherboard manufacturers decided to read and follow the Intel docs for their P55 boards.

If you want to check for this randomizer then try running i7 Turbo and a single thread of Super PI mod. If the BCLK is at its default setting then you should see a multiplier on the hardest working thread in the 21.0 to 22.0 range.

I won't argue with anyone that can get their reported temperatures to line up from idle to full load on any Quad core.
post #514 of 599
Yep, definitely saw a 21.85x multi bouncing around a bit. I see the problem with my previous testing was forcing 8 prime95 worker threads to run on 1 core, rather than allowing 1 worker thread to run on all 8 logical cores. As long as there's only worker 1 thread, the default Windows thread scheduler does a good job and setting Affinity for SuperPi/P95 to one core actually reduced the average highest multi down to 21.5x even with a single thread running, whether in SuperPi or P95. That explains why I needed to kill almost all processes or set them to run on one core to get the multiplier up higher before.
    
CPUMotherboardGraphicsRAM
i7 920 D0 MSI X58 Pro-E GTX 560 Ti 448 3x2GB G.Skill DDR3-1333 9-9-9-24 
Hard DriveHard DriveOptical DriveOS
840 Pro Caviar Black LG BD-ROM Windows 8.1 Pro x64 
MonitorMonitorKeyboardPower
Dell U2713HM Dell U2311H Turbo-Trak (Google it :D) Corsair HX-520 
CaseMouseMouse PadAudio
CM690 Mionix Avior 7000 Everglide Titan AKG K 242 HD 
  hide details  
Reply
    
CPUMotherboardGraphicsRAM
i7 920 D0 MSI X58 Pro-E GTX 560 Ti 448 3x2GB G.Skill DDR3-1333 9-9-9-24 
Hard DriveHard DriveOptical DriveOS
840 Pro Caviar Black LG BD-ROM Windows 8.1 Pro x64 
MonitorMonitorKeyboardPower
Dell U2713HM Dell U2311H Turbo-Trak (Google it :D) Corsair HX-520 
CaseMouseMouse PadAudio
CM690 Mionix Avior 7000 Everglide Titan AKG K 242 HD 
  hide details  
Reply
post #515 of 599
Quote:
Originally Posted by FSF-Foxhound View Post
it's just to scare us all, been running 1.45 for a few months now, still going strong. With a 920, you'll lose a memory channel after a few months @ 1.45+ most of the time. Only time will tell!
What? Seriously?
My System
(20 items)
 
  
CPUMotherboardGraphicsGraphics
5960x Rampage V Extreme GTX 980 Ti GTX 980 Ti 
GraphicsGraphicsRAMHard Drive
GTX 980 Ti GTX 980 Ti 16GB Intel X25-M G2 160GB Raid 0 
CoolingCoolingCoolingCooling
EK Supremacy EVO 2x MO-RA3 420 18x Noiseblocker PK-1 140mm MCP35X2 
OSMonitorPowerCase
7 Ultimate 3007WFP-HC Corsair AX1500i Antec 1200 
  hide details  
Reply
My System
(20 items)
 
  
CPUMotherboardGraphicsGraphics
5960x Rampage V Extreme GTX 980 Ti GTX 980 Ti 
GraphicsGraphicsRAMHard Drive
GTX 980 Ti GTX 980 Ti 16GB Intel X25-M G2 160GB Raid 0 
CoolingCoolingCoolingCooling
EK Supremacy EVO 2x MO-RA3 420 18x Noiseblocker PK-1 140mm MCP35X2 
OSMonitorPowerCase
7 Ultimate 3007WFP-HC Corsair AX1500i Antec 1200 
  hide details  
Reply
post #516 of 599
Thread Starter 
Quote:
Originally Posted by unclewebb View Post
Just to clarify a few things from the last page, RealTemp only reads the temperature sensor for each core once per second. I believe that Core Temp does the same thing.

leppie: Reading the temperature sensors 50 times per second is a good idea in theory but can cause a problem with the newer Core i7/i5 turbo boost feature. To obtain the maximum multiplier on these CPUs, you need to enable C3/C6 in the bios which lets inactive cores go into a sleep state. The maximum amount of turbo boost is based on the number of active cores at any instant in time so the more cores asleep in the inactive state, the faster your CPU will run.
Thanks for clearing that up Uncleweb. Like I said earlier I thought it might be more than once per second, but I didn't write it so I didn't know and definitly didn't want to speak for you.



Quote:
Originally Posted by unclewebb View Post
The newer Core i CPUs have TJ information written into a register within each core of the CPU so there is less to argue about now but there is only one problem. This data that every one assumes is TJMax really isn't. The register is called:

IA32_TEMPERATURE_TARGET

Intel also stated at their IDF conference that actual TJMax is not the same as TJ Target. The target for many newer Core i CPUs is 99 according to that register but actual TJMax can vary from that number. It can vary from one CPU to the next even if they have the same model number and I'm convinced that it can vary from one core to the next on the same CPU.
Once person actually did show me a screenshot with a TJ Max set as 95, which if it is being read from the register would indicate that on at least some CPUs that they were being changed (note this is the only one I have seen).


Quote:
Originally Posted by unclewebb View Post
On some CPUs like the 45nm Core 2 Quad I think you can have 4 different TJMax values over a range of 10 degrees or maybe even a little more. Intel has never publicly stated how much variation is possible.

The graphs they presented seemed to indicate that TJMax will always be not less than the TJ Target. If TJ target is 100C then actual TJMax might be any number between 100C and 110C.
I would bet that pairs of cores would be very similar in TJ Max values, but I definitly agree that each pair may have a slightly different TJ Max from the next pair. However, just for caution here I would state that I would always expect Core 0 to run slightly warmer than the Core 1 as most proccesses will default to Core 0 and that even in idle, Core 0 might still be slightly warmer but without a different TJ Max.

Like you said, Intel hasn't really stated this, but it seems like it is the truth.

Quote:
Originally Posted by unclewebb View Post
Once you understand how poorly documented the sensors that Intel uses are and how much slop and slope error variation is in these sensors, you soon learn that 100% accurate core temperatures from idle to TJMax is simply not possible.

Only the Core i7-900 series have excellent sensors. I think TJMax on core 0 is very close to 100C, there isn't much slope error at all and these sensors don't get stuck until the core temperature goes below -10C which is far better than some Core 2 45nm sensors that can get stuck when the core temperature goes below 50C.

As good as the Core i7-900 sensors are, I still don't believe that TJMax is consistent across all 4 cores. I'm 99.9% sure that it isn't and my best guess is that TJMax is very close to 105C on core 3 compared to 100C on core 0 with the two center cores being somewhere in the middle of this range. There is no documentation from Intel to confirm or deny this. It's just a theory I have based on observations.
Nothing to confirm or deny, but generally I would venture a guess that mine is 100 105 105 105. The load temps line up pretty nicely on load, but it's iffy to me. I'm still no where near throttling or shutdown so it's a moot point, but the TJ Max in the register may only be for Core 0...nothing to confirm or deny that.

Quote:
Originally Posted by unclewebb View Post
All of these sensors are only designed and calibrated for two purposes; to control thermal throttling at approximately 100C and to control thermal shut down at approximately 125C. All monitoring software that uses these sensors to report accurate core temperatures are doing a lot of guessing and making a lot of assumptions that may or may not be true. I'd like to believe that RealTemp is closer than most programs but without 10,001 CPUs to test, that's impossible to prove.
I've always wondered if they are in some way amplifying the the voltage made with the Seebeck Effect. If not for a J-type thermalcouple at 100C, the voltage is still only around .39mV. If you've ever wanted a better explaination on why the low temperatures are so horrible, that's it. 0C on a J-type is roughly .0039mV, which is nearly impossible to measure especially on such a small scale.


Quote:
Originally Posted by unclewebb View Post
After putting far too much time into project RealTemp, I finally concluded that core temperatures really aren't that important. The cooler you run your CPU, the more you'll be able to reliably overclock it but as long as your CPU is stable and not thermal throttling then it really doesn't matter what temperature it is running at.
Nicely put. Pretty much the same conclusion I came too.
Gaming PC / HTPC
(17 items)
 
Bedroom PC
(13 items)
 
 
CPUMotherboardGraphicsRAM
i7 7700k Asus Prime Z270-A GB Auros-X 1080 Ti G-Skill (2x8gb) 3000mhz  
Hard DriveHard DriveHard DriveOptical Drive
Samsung 850 Western Digital 5900rpm Seagate 7200rpm LG 22x DVD Burner 
CoolingOSMonitorKeyboard
Noctua NH-D14 Win 10 x64 Pro LG 55" 4k B6 OLED Microsoft Desktop 400 
PowerCaseMouseMouse Pad
Corsair TX750 V2 NZXT Source 210 Microsoft Desktop 400 Black 
Audio
Custom Surround 5.1 
CPUMotherboardGraphicsRAM
Intel 2120T Biostar H61MH Intel HD 2000 G-Skill F3-12800CL9D-8GBXL  
Hard DriveOptical DriveCoolingOS
Samsung Evo 128gb Asus 24x DVD Burner Stock Win 10 x64 
MonitorKeyboardPowerCase
LG 40" 1080p HP Wireless Keyboard/Mouse Silverstone 400W ST40F-ES Silverstone ML03B 
Other
Linksys WMP600N Wireless Card 
  hide details  
Reply
Gaming PC / HTPC
(17 items)
 
Bedroom PC
(13 items)
 
 
CPUMotherboardGraphicsRAM
i7 7700k Asus Prime Z270-A GB Auros-X 1080 Ti G-Skill (2x8gb) 3000mhz  
Hard DriveHard DriveHard DriveOptical Drive
Samsung 850 Western Digital 5900rpm Seagate 7200rpm LG 22x DVD Burner 
CoolingOSMonitorKeyboard
Noctua NH-D14 Win 10 x64 Pro LG 55" 4k B6 OLED Microsoft Desktop 400 
PowerCaseMouseMouse Pad
Corsair TX750 V2 NZXT Source 210 Microsoft Desktop 400 Black 
Audio
Custom Surround 5.1 
CPUMotherboardGraphicsRAM
Intel 2120T Biostar H61MH Intel HD 2000 G-Skill F3-12800CL9D-8GBXL  
Hard DriveOptical DriveCoolingOS
Samsung Evo 128gb Asus 24x DVD Burner Stock Win 10 x64 
MonitorKeyboardPowerCase
LG 40" 1080p HP Wireless Keyboard/Mouse Silverstone 400W ST40F-ES Silverstone ML03B 
Other
Linksys WMP600N Wireless Card 
  hide details  
Reply
post #517 of 599
Just for the record, I've seen a couple of ES processors that had 92C and 93C written into the TEMPERATURE_TARGET register. These seemed like low ball numbers and actual TJMax seemed similar to most other CPUs somewhere around 100C. It might have been a game by Intel so early testing would show some nice low temperature numbers.

Crappy sensors and poor to non-existent documentation makes accurate temperatures mostly a guessing game.
post #518 of 599
Interesting stuff for sure. That would explain why the tjmax on my 3520 was read as 100c while my 3540 reads as 98c.
xeon
(13 items)
 
  
CPUMotherboardGraphicsRAM
W3520 x58 ud5 9800gtx+ tridents 
Hard DriveOSMonitorPower
wd black x2 W7 64 acer 23 ocz 850hx 
Case
haf932 
  hide details  
Reply
xeon
(13 items)
 
  
CPUMotherboardGraphicsRAM
W3520 x58 ud5 9800gtx+ tridents 
Hard DriveOSMonitorPower
wd black x2 W7 64 acer 23 ocz 850hx 
Case
haf932 
  hide details  
Reply
post #519 of 599
*sigh*
Erm... +rep and good job explaining.
I am learning a lot, lol. But even nooby me has noticed that manufacturers keep secrets from us buyers.
Intel has 2-3 public pdf's to show around per processor and that is not even close to good info.
Asus guys are close to secrecy when i call and ask them: "whats my NB+SB temperature and voltage specifications?"
Gigabyte and OCZ are in the same situation.

I think they just stopped caring what they are doing as long as it sells, at some point. And yes, they don't provide us with info because they don't pay someone to and because they don't really give a rat's arse about what we think. Our new god is publicity and they force us to worship it; ...don't know if this is an understatement.

Anyway, you're the spearhead of the guys that think and care for their computers. Don't stop

Me is gonna read some moar, have fun!

p.s. I R addicted to realtemp. starts with windows, ends with windows and is most times always on top
Edited by nemaca - 2/12/10 at 3:23am
Silent PC
(18 items)
 
  
CPUMotherboardGraphicsRAM
3570k Asrock Z77 Fatal1ty Performance inno3d gtx 970 iChill Herculez X4 Air Boss Ultra Corsair 1600 8Gb 
Hard DriveOptical DriveCoolingOS
plates LG VenomousX - air Windows 10 Pro x64 
MonitorMonitorKeyboardPower
Phillips 1920x1080 144Hz v-sync LG 55" 668 4k hdr 4:4:4 60Hz Logitech Illuminated Corsair AX1200 1200W 
CaseMouseMouse PadAudio
Corsair Obsidian 800D Logitech G9 + Logitech proteus core 502 X7Gaming Asus Xonar Essence STX + pro opamps 
AudioOther
Denon dac + stereo setup Asus RoG STRIX GL502V  
  hide details  
Reply
Silent PC
(18 items)
 
  
CPUMotherboardGraphicsRAM
3570k Asrock Z77 Fatal1ty Performance inno3d gtx 970 iChill Herculez X4 Air Boss Ultra Corsair 1600 8Gb 
Hard DriveOptical DriveCoolingOS
plates LG VenomousX - air Windows 10 Pro x64 
MonitorMonitorKeyboardPower
Phillips 1920x1080 144Hz v-sync LG 55" 668 4k hdr 4:4:4 60Hz Logitech Illuminated Corsair AX1200 1200W 
CaseMouseMouse PadAudio
Corsair Obsidian 800D Logitech G9 + Logitech proteus core 502 X7Gaming Asus Xonar Essence STX + pro opamps 
AudioOther
Denon dac + stereo setup Asus RoG STRIX GL502V  
  hide details  
Reply
post #520 of 599
Thread Starter 
bump
Gaming PC / HTPC
(17 items)
 
Bedroom PC
(13 items)
 
 
CPUMotherboardGraphicsRAM
i7 7700k Asus Prime Z270-A GB Auros-X 1080 Ti G-Skill (2x8gb) 3000mhz  
Hard DriveHard DriveHard DriveOptical Drive
Samsung 850 Western Digital 5900rpm Seagate 7200rpm LG 22x DVD Burner 
CoolingOSMonitorKeyboard
Noctua NH-D14 Win 10 x64 Pro LG 55" 4k B6 OLED Microsoft Desktop 400 
PowerCaseMouseMouse Pad
Corsair TX750 V2 NZXT Source 210 Microsoft Desktop 400 Black 
Audio
Custom Surround 5.1 
CPUMotherboardGraphicsRAM
Intel 2120T Biostar H61MH Intel HD 2000 G-Skill F3-12800CL9D-8GBXL  
Hard DriveOptical DriveCoolingOS
Samsung Evo 128gb Asus 24x DVD Burner Stock Win 10 x64 
MonitorKeyboardPowerCase
LG 40" 1080p HP Wireless Keyboard/Mouse Silverstone 400W ST40F-ES Silverstone ML03B 
Other
Linksys WMP600N Wireless Card 
  hide details  
Reply
Gaming PC / HTPC
(17 items)
 
Bedroom PC
(13 items)
 
 
CPUMotherboardGraphicsRAM
i7 7700k Asus Prime Z270-A GB Auros-X 1080 Ti G-Skill (2x8gb) 3000mhz  
Hard DriveHard DriveHard DriveOptical Drive
Samsung 850 Western Digital 5900rpm Seagate 7200rpm LG 22x DVD Burner 
CoolingOSMonitorKeyboard
Noctua NH-D14 Win 10 x64 Pro LG 55" 4k B6 OLED Microsoft Desktop 400 
PowerCaseMouseMouse Pad
Corsair TX750 V2 NZXT Source 210 Microsoft Desktop 400 Black 
Audio
Custom Surround 5.1 
CPUMotherboardGraphicsRAM
Intel 2120T Biostar H61MH Intel HD 2000 G-Skill F3-12800CL9D-8GBXL  
Hard DriveOptical DriveCoolingOS
Samsung Evo 128gb Asus 24x DVD Burner Stock Win 10 x64 
MonitorKeyboardPowerCase
LG 40" 1080p HP Wireless Keyboard/Mouse Silverstone 400W ST40F-ES Silverstone ML03B 
Other
Linksys WMP600N Wireless Card 
  hide details  
Reply
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Intel CPUs
Overclock.net › Forums › Intel › Intel CPUs › The Truth about Temperatures and Voltages