For you engineers in the group, here is the information you seek..

I could not pull it up in the developers area at Intel but located it here:

http://wiki.imt-systems.com/(S(waeplme04ac5wa55tlqgusvj))/GetFile.aspx?File=/FA0...




DTS Enhancements

• Nehalem has improved the Digital Thermal Sensor circuit
– Expanded temperature range – unlikely to ‘bottom out’
– Calibration accuracy is improved
– Slope error is reduced


Software visible register contains the target TJ

– A new feature in the Intel® Core™ i7 processor is a software readable field in the IA32_TEMPERATURE_TARGET register that contains the minimum temperature at which PROCHOT# will be asserted. The PROCHOT# activation temperature is calibrated on a part-by-part basis and normal factory variation may result in the actual activation temperature being higher than the value listed in the register. PROCHOT# activation temperatures may
change based on processor stepping, frequency or manufacturing efficiencies.

• IA32_TEMPERATURE_TARGET register
– MSR 1A2h Bits [23:16]
– Data format is decimal degrees C








or you can use what we know as safe for max load stress testing since the real world use will remain under that value and we found the average which plays best for just about all systems with i7, which is 80c and can vary from system to system by about 2-3c up or down


Unlike Intel procs of the past there is an advantage with i7 the C2/Q procs did not allow and required a lot more with those to find the correct TCC limit


Should you RUN the system all the time in the mid-high 70's?   NO! not if you can avoid it but it wont hurt the system to run that in peaks and for short periods as long as its not a constant.

You can also run >80c somewhat the same way (shut down the BIOS protection) however you WILL be reducing the rated life of that proc if you are always in that 75-85c range, no question about it which is why intel specs 67.9c for the TS of the i7 900 series



It was also discovered some time ago shutting down CPU TM Function in any clock will provide a more stable base to hit those high notes but none the less, even if you stay within the official specs noted below clocking will reduce the rated life. By how much can not be defined by any single equation because our towers do not sit in a lab under optimal monitor and environment conditions. Even the socket temp itself which is another variable can alter that result.



Thermal Specification: 67.9°C
@
VID Voltage Range: .80V-1.375V



5-7 years on a well considered clock is the established norm. On a poorly considered clock it could be 1-3 years and with a very poorly considered clock, past 1 second and you are on borrowed time.   

Just an FYI

I keep my responses in forums to a level where everyone who reads the information GETS the message. If I were to address information in MSFS forums on the same level as I address one of my JEDEC workshops no one other than real engineers with solid backgrounds would understand what hell I am talking about.

Your suggestion that damage would not occur which did not address the fact that damage will in FACT occur, just not right away, is a not how an engineer with a lick of sense and concern for the readers addresses a topic in a non technical forum.

People need to know the facts so they can make informed decisions. Based on the right information someone may not risk a clock that runs their system @ 75c under normal running load with demanding applications since they dont buy new hardware every other year and instead opt for a lower clock to protect their investment. They may also opt for the better timing memory because they want every advantage possible with a new purchase since so many of them have learned the HARD WAY in the past when they skim on quality they get what they pay for in MSFS.

According to your statement a layman could assume they were completely safe and their proc would continue to run for the rated 10 year life overclocked at the temps you suggested. Your information was false the way it was presented which negated the laws of physics, regardless of how you meant it.

And as a participating test engineer I was at the Intel conference on Bloomfield back in 2008 where that presentation above as well as the advances the proc provides with respect to memory performance based on its timing was also shown. Although it is true because the bandwidth situation has changed with 2nd generation DDR3 and timing is not as critical as it was with DDR2 or 1st Gen DDR3, the result (positive or null) is centered to the application in use and how demanding it is on the proc cache

Since MSFS is old school rendering and somewhat poor design in coding which brings on issues very similar to serious memory disambiguation, the communication path between the memory and the proc cache is cluttered as hell and every cycle counts to the result. Therefore, memory speed in relation to Column Address Strobe and tRD http://www.thetechrepository.com/showthread.php?t=195 are still very valid elements to obtaining better results in MSFS.



I also know exactly what advances the new Intel design has with Windows7 and NONE OF THEM will have any effect with MSFS because the software was not designed under the Vista or W7 driver and memory management model

Nothing in the Windows7 or Vista OS (other than moving from 32 to 64bit, just like with XP) will make FSX run better other than tighting up the loose nut behind the keyboard.



Although I do agree that W7 is a far cry from the Vista disaster the differences people MAY see in MSFS and report as "W7 is the best OS for FSX because my frames went up" are not a result of the OS itself but changes they made due to a reinstall, period!

idahosurge wrote on Jan 19^th, 2010 at 11:36am:



Rod for the clock you are running the settings as you listed above are fine but in order to go higher I am sure you will need to increase QPI and CPU voltage. Some 950s are known to run 4Ghz @ CPU 1.25-1.28v but usually it requires right around 1.30-1.35

The 950 is far more stable in a high clock than a 940 since it is a D0 stepping proc and as Intel points out with the stepping change came changes to the internal thermal monitoring system as well.

What I posted is valid for a 920 or 950. You need to work the clock out on your own as no one can give you "plug-n-go" settings that will work perfectly. You must work the problem out using the test software available to you for that purpose and make the call on temps and stability yourself.

The name of the game is highest CPU and memory speed (with lowest memory timing) stable <80c in full load tests and of course it will be lower with normal or FSX use. You can use REALTEMP 3.0 or above to do some monitoring while you are running FSX and see what the peak temps look like too at any time. You do not need to run RealTemp all the time.

As I mentioned, it is possible to be stable in OCCT stress tests and still bomb in FSX due to the PCIe buss coming into play. You may need to increase the CPU core, CPU PLL or IOH voltage a touch to stabilize if that happens. I gave you all the safe limits to work with

I dont think you will need to go any higher than 1.35v on the QPI/DRAM unless you try to go over 1600 on the memory speed. 1.375 is the Intel max limit although my discussions with the Intel boys reveal they have no problem with 1.42-1.44 as long as the tower has good airflow. Of course that is not a 'official' stance 

As always thanks for the response Nick and the settings!

Rod

I love reading these threads, you guys are so knowledgable. 

http://download.intel.com/design/processor/datashts/320834.pdf

Let me spell it out for you

With your statement you are personally assuming you know the critical point at which yours, mine or anyone else’s processor on this forum will fail or be damaged as long as it is within a the spec of 100c (according to you its not 100c)

“And going over 100c is not the end of the line” <-------- your words

Here is some interesting information for you and it is based on the physics Intel applies to all their processors, including i7.

There is a relationship between the TS of 67.9c and the i7 tj of 100c….   do you know what that relationship is?

Dont know?  Clue: it has to do with the TDP system and all the materials the slug is made from, including the silicon changes used in i7, (those darn physics), in combination with where the tC reading is taken on the slug.

or should I say before I gave you the data, you didn’t know or have a clue or you would not have asked for data.

If you did not know the answer to that question when you posted your statement, then you have no idea what you are talking about sir when you say…

“And going over 100c is not the end of the line. In most instances the test will be stopped. And going over 80c in a stress test with a few higher spikes is not going to harm or lesson the life of the computer. A stress test is only one or two days in the overall life of the computer.”



Hmmmmmm....


The fact of the matter is Intel engineering places the variance of the stated tJ in a NON-CLOCKED processor <--- (remember those words) @ 5c + or – and they also state that any processor run at or over their Thermal Specification spec in relation to tJ according to the calibration math based on real time ambient, socket, DTS and other factors (read boy, read) means system A and system B are NOT alike unless they are under equivalent monitored and controlled laboratory conditions, and that the possible outcomes are;

a. Instantaneous catastrophic failure (bye-bye birdie)

b. Internal damage that may allow processor to function but with critical repeating errors (catastrophic but still boots the computer)

c. Rated life reduced by 1-50% or more (depends on all variables at the critical moment) which means the user does not know HOW much damage they have done.

                                       or

d. No damage, usually because of the stated variance (+5c) which is driven by the final quality of the slug in the actual manufacturing process as well as environmental variables that fell in line with TDP and tJ.    In other words YOU GOT LUCKY and the internal Intel system worked and saved your processor.

And that is based on a non-clocked processor WITH the TM feature ENABLED in the BIOS! Any one of those can precede the processor shutting down correctly via the BIOS TM setting


Then comes the question on my mind after reading what you said....

If you know when/where a chip run by you, me, Rod or anyone else on this forum ran near or over 90-100+c during overclocking will NOT fail during load test spikes and you know for a fact that it WONT fail within 1-2-5sec @ 5c +/-  tJ, or, that its life will be not shortened by 1 year, 2 years, 5 years by running it over 80c… then please, Riddle me this.....




One must wonder how you can presume anything without any specifications which I provided after the fact, and were designed by Intel based on those

gosh darn pesky laws of physics

  in suggesting temps are SAFE to run at 85-95-100c or in your words,.. "over 100c is not the end of the line" for even 1 second when overclocking.


Why?

Intel revokes the warranty for overclocking because

when you overclock you are not following (here it comes) the Processor Thermal Profile which in turn defines Intels specification @ tJ  and you are in FACT exceeding the 0-130w profile of the i7 slug even if you are between 85-100c and therefore you can catastrophically damage or significantly shorten the life of a i7 processor well under (or slightly over) 100c instantly if the right conditions are met while clocked. The Intel specification for thermal dissipation was not made for a processor that was overclocked because under that state you can be exceeding the watt rating @ temp for which the TS-tJ specification was designed.

Therefore, an engineer or even a smart PC tech that GETS IT, would reconfigure the limit based on the increase in watts that the slug will be required to dissipate in average during a clock and come up with a value that is in line with Intels specifications but would be LOWER, ergo we DONT EXCEED 80c when load testing a clock to protect our investment!


"My comments were directed to the fact that even when exceeding the voltage/temp limits when fine tuning an over-clock, the effect on the PCs life cycle (5-7 years) would be minimal. And just because one reaches or goes over the MAX, it is not going to break or blow up. "




When did you become not only an Intel engineer, but an Intel engineer with a crystal ball... not blow up?   Show me the FACTS that overrides what Intel posted about how i7 works.

Based on the data I provided from Intel if Rod lives in Egypt and its summer and he followed your suggestion the ambient in that environment has the high potential to force the socket temp sky high which mean that 85-100c is FAR different with respect to where damage occurs than it is with someone in Canada during the winter!

It is actually possible to see LOWER performance under the right conditions running a high clock even if the temp is between 70-72c! I won’t get into that one here because it starts an entire different subject but it is based the same scientific fact and physics related to Intel TDP and slug design and the HSF in use.





I will say it again with a revision:

Any engineer with a lick of sense would never make such an asinine statement in a technical or non-technical forum because what he just told the readers is not only FALSE, he has no idea how or if that statement is TRUE based on the specifications and formulas he did not know or understand when the statement was made.





The last one is questionable as to its accuracy.

The only person in this thread that has stated anything that is questionable in accuracy when it comes to real facts, engineering and science,

is YOU sir

and you wonder why I said what I did!??

then you follow that statement up with this crap...


And the following rhetoric which you posted is really in poor taste, poorly written, and only ends up reflecting the character of the writer. 

You have the nerve to say something like that to me?

If you are going to insult my character do it somewhere else because I have no patience for an engineer that shoots off his/her mouth without facts while trying to justify a statement made in complete ignorance of engineering fact and pass it off as anything NEAR accurate. It is an INSULT to me, my intellegence, this forum and to the profession and reflects the character of the person who attemped to pass off such manure and follow it up with passive aggressive snide undertones in posts that you know different about temps and memory performance suggesting my information is in "question" clearly summed up by that last statement.

macca22au wrote on Jan 21^st, 2010 at 12:50am:






LOL!!     Bad old puddy'tat I am


GEX uses texture and landclass therefore any photoscenery or addons like airports, etc you use will overlay GEX and work just fine with it. The only issue you would see is the boundary between the photoscenery product and the textures.

Photoscenery and textures are 2 different things. Photoscenery is going to give you a sq-km layout of the actual area using sat images where textures and landclass will never produce a direct image like photoscenery does.

Landclass such as UTX, SceneryTech, etc is a data file which specifies the land type in use based on the texture library classifications in FSX. In other words we make textures that match the library catalog description be it a urban, rural, housing, swamp, rock, grass, ice, farm, etc, etc and the landclass file specifies where those tiles play to produce a pseudo environment. The advantage of that system is you are simply overwriting the default FSX textures (which are backed up) and not taking up any extra drive space. At the same time it works just like default which means there are no issues with adding any freeware or payware scenery you want. It will overlay just like it does with default. One of the disadvantages to texture/landclass is that the textures are used in repeat which part of what we do is work out the geometry so the eye does not pick up on the repeats very easy like it does with default FSX textures.

Another issue is called 'tiling' where from high altitude you see a hard pattern from the textures. That is another area we reduce by a massive margin in our custom texture design, so airline pilots get a great show as well as low-n-slow flyers.



When it comes to your clock, the methodology is the same as what I posted for Rod. You would simply work lower BCLK/RATIO and voltages.

As for a step by step, I could not locate it here since the system has been reworked and I would rather have current information about your rig before posting such a list

I know you live in Australia where the temps in recent years have been pretty high during the summer and the drought conditions there don’t help. I would want to know specifically what your motherboard model is because the voltage regulation design can define settings as well as differences in BIOS lists that are present from board to board.

I would also want to know what heatsink you are using and the memory brand/specifications for speed/timing so I can look all those up and give you a starter list.

Getting into the BIOS is usually done by hitting the DEL key during initial boot. That key can differ from board to board which is another reason why I would want to know what motherboard is in use.

What's REALLY scary is the fact that I actually understood *some* of this.    Just don't ask me to repeat it.

With all due respect, Flight Ace - let me put it this way - if Nick says to avoid goig over 80c in a clock, I'd take his word for it. Even if he's wrong (which would amaze me) he always is on the cautious side. As opposed to others in these fora who spout "trusims" that *could* destroy systems.

You mention the KISS principle - I posit that you did follow that BUT you kept it simple for an ENGINEER not the average joe.  There is a difference if you think about it.

Vic