Intel's Conroe / Core 2 CPU

[Deeko]

Here is some more info for you luck :



The cpu is running with a 1.5vcore to achieve 3.4ghz (a 1Ghz overclock). I just installed the Noctua NH-U12 cooler to ensure cool running, the cooler is up there with the Tuniq Tower and Thermalright Ultra 120. As you can see in the screenshot folding is running and has been all night and the cpu temp is 42C, it idles at 32C. At stock vcore and clocks with this cooler my idle temps are 24C and load temps while running Orthos (dual prime 95) were only 34C.

Dont worry about not being able to produce any in depth results, Im lucky i could produce those as I myself have just started back my second year of college after a lengthy break. There might not be the time for me to indulge myself as much anymore

Its all about using our graphics card having untapped potential if its powerful enough. Its well known that high end SLI/Crossfire rigs need the best possibe cpu powering them to exploit this potential otherwise you will certainly be bottlenecked by the cpu. Single card configurations will also benefit but probably not on the scale as a dual card setup as luck has already stated.

[Beach]

Intel: 80 Cores similar to Cell Processor
http://www.anandtech.com/cpuchipsets...spx?i=2840&p=6

I don't want to make too many threads

[CreamyPoo]

Hey Deeko, isnt 1.5v vcore a little high for everyday use? I mean it'll shorten its lifespan and stuff. And why only 3.4ghz? Some people are able to do 3.6ghz with only 1.45v stable.

I've got mine at 3.0GHz stock voltages running 2 folding@home instances with orthos prime for about 5 hours without a hitch. Maybe you should consider lowering your overclock to that level. I hate to see peoples cpu's burnout after only a year.

Whats your coretemps btw? You dont show it in your pics. Oh yeah, and I noticed you got your ram underclocked. Is it better to have a 1:1 fsb:dram ratio than to just have high ram clockspeed? I know that conroe reacts better to higher ram speeds than to tighter timings.

[Deeko]

1.5v isnt that high at all if you have very good air cooling for the Cpu with a good case that gives excellent airflow like the Antec P180. Normally to achieve a proper stable overclock of over 3.6ghz you will need about 1.55v upwards that moves into watercooling territory with the voltage/clocks unless you have a very good sample that overclocks well on low voltage. I have the Noctua Nu-120 tower cooler so temps arent a problem at my running overclocked spec, 31C idle and 42C underload on the thermal diode with Orthos (dual Prime 95) . Intels TAT(thermal analysis tool) reads the digital coretemps as 67C and 64C with Orthos and its more accurate than coretemp, well under the 85C Tcase maximum when the thermal throttle kicks in. In easy speak I have a headroom of 20C before I hit thermal throttle which wont happen given my current maximum underload temps.

With lots of Core 6600's you will hit a a wall at around 3.2-3.4 and to move above this you will need to raise the voltage considerably, I was running 1.4v at 3.2ghz but at 3.4Ghz the voltage required a considerable boost to 1.5v. I could run mines at 3ghz stock voltage to as can most people with the 6600. Every Cpu is different and they dont overclock the same, I seem to have hit a wall at 3.4ghz with anything higher requiring a lot of voltage raising the temp far too much for air cooling unless I went onto water cooling. Im happy with the performance and the gain from 3.4>3.6 isnt important , as the heat and voltage generated to hit it is way too much. Also running a 1:1 ratio is best for performance by a long shot and stability for a large overclock. The P5W DH Deluxe doesnt like dividers when your running a very high FSB, 1:1 is the best for stability. The Ram is running at 760mhz, very close to 800mhz and at 4-4-4-12 which gives me a much more lower latency. Ive had the system running at 3.2ghz with the Ram at 900mhz at 5-5-5-15 but performance was a lot better at 1:1 with 3.4ghz and ram at 760mhz @ 4-4-4-12.

The cpu wont burn out after a year with the cooling I have in place as its well within tolerance of the heat it generates, secondly I dont normally keep cpus for anywhere near a year before I sell them on ebay and upgrade to the next best thing for minimal outlay









Ive also added another 120mm fan to the rear of the cooler for better performance.

[CreamyPoo]

I tried that intel TAT thingy just to see my temps and it gives me the same readings as in coretemps and everest. But after uninstalling it... ARGH!! it broke speedstep!! What used to be 2000MHz <> 3000MHz is now down to 360MHz <> blah... now I have to disable speedstep and permanently run my cpu at 3.0GHz... no big deal.. but still... Anyone know how to fix it besides reinstalling winxp?

[Deeko]

Sorry I cant help you there, perhaps a proper search will bring something up. I tried everest as well but I found the digital temps were reported approx 4C lower than Intels TAT, so I decided to trust Intel on that one rather than trust the lower temps Everest gave even though it would have been nicer to believe them Everest did report the normal thermal diode temp the same as Asus probe though.

[CreamyPoo]

Well, my speedstepping fixed itself for some reason after playing around with winxp power settings, disable/enable speedstep in bios, and a couple restarts... for awhile i was running the most underclocked conroe ever at 360MHz...

Anyway, those coretemps at over 60 degrees is way too hot. According to this doc http://download.intel.com/design/pro...s/31327802.pdf, the max tcase for conroe is 61.4c. From my own experiments, thermal throttling occurs over 60+c and at 85c, the cpu shuts itself down. I was curious to see if I seated my zalman right so I fiddled with it while in windows monitoring temps. Then I made a slight blunder of removing the heatsink... I watched the temps rise slowly and when it hit about 80c, winxp crashed.

"In the event of a catastrophic cooling failure, the processor will automatically shut down when the silicon has reached a temperature approximately 20 °C above the maximum TC."

[atrophocy]

you're a brave guy Creamy. I'm pretty relaxed about my hardware and dont mind leaving cpu's and gfx cards lying around the room and so on, but there's no way i'd ever risk running my chip without a heatsink. i guess nobody can argue with you about those thresholds now though

[Deeko]

Originally Posted by CreamyPoo

Well, my speedstepping fixed itself for some reason after playing around with winxp power settings, disable/enable speedstep in bios, and a couple restarts... for awhile i was running the most underclocked conroe ever at 360MHz...

Anyway, those coretemps at over 60 degrees is way too hot. According to this doc http://download.intel.com/design/pro...s/31327802.pdf, the max tcase for conroe is 61.4c. From my own experiments, thermal throttling occurs over 60+c and at 85c, the cpu shuts itself down. I was curious to see if I seated my zalman right so I fiddled with it while in windows monitoring temps. Then I made a slight blunder of removing the heatsink... I watched the temps rise slowly and when it hit about 80c, winxp crashed.

"In the event of a catastrophic cooling failure, the processor will automatically shut down when the silicon has reached a temperature approximately 20 °C above the maximum TC."

I checked out the pdf you linked to and the Tc (61c) is the processors max temp from the thermal diode(analogue) that is on the die and takes the maximum temperature of the IHS taken at the center, as shown in the diagram below , its NOT referring to the digital sensors temp that are in each core ,they are independent and at a different level. This thermal diode reading is the very same reading that programs like Asus Probe, Speedfan and everest report for Cpu, although Everest has now included the DTS readings.

My thermal diode reads 42C underload ,nowhere near 61C, its not the digital core temp its giving thermal throttling specs for in the tables in the PDF. You can clearly see in the diagrams that they are measuring the temp close to the processors heatshield that is the thermal diode, by measuring the temp from the thermal diode in the geometric 'center' of the processors case or IHS (Tcase). The digital sensors are embeded in each of the cores , therefore Intel will use some kind of formula to calculate the digital Tcasemax using other sensors as it seems the digital temperatures are also employed for Fan control execution, the info below also clearly shows that the chip does employ two seperate sensors systems.

Tcase for thermal diode reading:



Digital Thermal Sensor:



Coretemp or TAT takes a digital thermal sensor reading (DTS) from deeper in the cores themselves, the temps we see on TAT or coretemp will be converted from digital to give us an analogue temp representation on TAT/coretemps GUI for the temps in the cores. The table and temps you are referring to are for the thermal diode. The only way I have seen thermal throttling activate is when the digital temps reached 80C-85C , my digital temps are in the mid-60's approx underload. Ive never seen my thermal Diode go over 47C and that was with the stock fan. In no way whatsoever does thermal throttling occur with the digital core temps in the 60's or even the 70's for that matter. Coretemp clearly shows the Tcase max for digital temps is 85C, mines are 20C below this.

You must understand that the Digital Thermal Sensors (DTS) and the thermal diode are taking different readings from different places with entirely different sensors and are not to be compared with each other. From what I can see and have read, Intel do not have specific data for thermal throttling for its digital temps in their specs in the PDF, as its already dealt with by the thermal diode, it seems the DTS sensors are indeed used for more accurate readings with regard to fan control.

Digital temps system for Core Duo:



Sadly Intel didnt provide us with any information in the PDF regarding the digital core temps with regards to throttling etc.

Originally Posted by atrophocy

i guess nobody can argue with you about those thresholds now though

Im not arguing with his point as he is correct, but its only correct insofar as its the thermal diodes Tc temp he is referring to and not the DTS temps, Intels thermal table in the linked PDF is based on this thermal diode. My thermal diode Tc temps are only at 42C, 19C under the stated temp when thermal throttling occurs.

Creamypoo also stated correctly in his above post "thermal throttling occurs over 60+c and at 85c", to quote it more accurately it should have been "61.4C for Tcase thermal diode (IHS) reading and the seperate DTS tcasemax digital reading of 85C.".

Given the information we now have at hand we can now say that if the thermal diode (analogue) Tcase temp reaches 61C then we will experience thermal throttling, and with the entirely seperate digital thermal reading on each of the cores we will experience thermal throttling when it reaches 80-85C as is shown in TAT/Coretemp. My system temps read at around 20C below for both of these different systems thresholds and are completely safe.

[bigtabs]

All that time and effort Deeko, and you could have used the built in overclocking tool that Asus provides and had yourself an small easy OC without having to learn all of this stuff. All you have to do is raise a percentile figure in the BIOS and it does it all for you.