Real World Core 2 Temperatures

unclewebb

[H]ard|Gawd
Joined
Jun 21, 2006
Messages
1,807
It's great if your computer is Prime or Orthos stable for a week and I realize that some users need 100% stability but many users will rarely stress their system anywhere near what Orthos, TAT or OCCT do.

Here's a good example for gamers. Most games are still single threaded which means they can only use one core at a time. What XP does is it takes one task running 100% on one core and shares it equally so that both cores are running at about each 50%.

Running both cores at half throttle obviously creates a lot less heat than Orthos does. This is why users can be perfectly game stable at a higher overclock than they are Orthos stable.

For testing I played some Need For Speed - Carbon for a while. The temperature stabilized on both cores at about 57C. At the end of a race I did an ALT+TAB and went back to the Desktop and immediately started Orthos which increased the core temperatures to 75C.

realworldhe2.png


If you're doing some Folding or encoding while gaming then by all means make sure you're system is 100% Orthos stable. For most users though that surf, chat, check their e-mail and play a game or three, being Orthos stable for an extended period of time isn't nearly as important as it was during the single core days.

If I stop running Orthos then I probably really don't need a shiny Thermalright Ultra 120.
http://www.anandtech.com/casecooling/showdoc.aspx?i=2941
It would drop my temps like crazy for benching and I'd really like one but I really don't need it! :D
 
For most people including myself, there really is no need to overclock these C2D cos there plenty fast .... Money spent on cooling, fast rams and what not is really just to bench and see how high an overclock you get but after that it all becomes unnecessary..... Thats why I was never a fan of stability benches like orthos cos even if you can run Orthos say at maximum 3600mhz for say 10hr, all youve proven is you can run that speed..... It will be stupid to make that your 24/7 speed cos you are just putting unnecessay stress on your machine for nothing..... 3600mhz or 3300mhz will not make a differnce for what we use out computer for, so to test stability is fact useless......

In a closed case I bet not too many if any can tell the difference between using a machine running 3700mhz 4-4-4-12 and one running 3300mhz 5-5-5-15.... you can probably even put a Opty 170 at 2900mhz and it wont be noticed..... Thats what I think anyways...
 
There are a few behaviors that are improved with faster system components.

Games are certianly smoother and you can play them with much more detail. ( I can notice big differences at 500mhz +/-)

I think 800mhz and cas4 would deliver an obvious performance boost. You are talking the difference between an e6300 and e6700 in proc speed alone.
 
In my case, I'm folding 24/7 so I value 24+ hours Orthos stability very much and keeping heat in check is also very important. That's why I bought a Thermalright SI-128 so I can have a peace of mind.

However, I agree with luihed and trying by all means to run the fastest 24/7 is a bit pointless since this will provide extra stress for nothing. It's very important to find the right speed who can keep the temperature at around or under 65C, not using too much voltage. In my case, I was able to get 3510 MHz Orthos stable for over 6 hours but I know it's too much since there are random crashes occuring 1 or 2 times a day and temps is hovering at around 70-73C, too high. Currently, i'm settled to 3380 MHz at 1.475v, which give me around 65C under Orthos and below 60C while folding.

But, it's nice to be able to run at this speed since the bottleneck will disappear and everything is very fast.
 
It has been years since I folded and yeah it does make a big difference there, folding heats up your rig pretty good.... Who do you fold for? ... When we started folding, we actually used that as a stability test cos it is pretty stressfull... Wow that has to been late 02- early 03, time flies....... lol

Saw your sig, S754 rocks:D .... Most fun I had overclocking especially when the Shuttle AN5Or came out, first ( and probably only) board capable of giving unlimited vdimm... well not really unlimited, mine topped out at 4.06v:eek: .... Its unreal how that cpu matured in that I needed to freeze my cpu to hit 2600mhz and they do that now probably on stock cooling lol.......
 
It has been years since I folded and yeah it does make a big difference there, folding heats up your rig pretty good.... Who do you fold for? ... When we started folding, we actually used that as a stability test cos it is pretty stressfull... Wow that has to been late 02- early 03, time flies....... lol

Saw your sig, S754 rocks:D .... Most fun I had overclocking especially when the Shuttle AN5Or came out, first ( and probably only) board capable of giving unlimited vdimm... well not really unlimited, mine topped out at 4.06v:eek: .... Its unreal how that cpu matured in that I needed to freeze my cpu to hit 2600mhz and they do that now probably on stock cooling lol.......


I'm folding for Stanford to find a cure for cancer, and this for Team 33 :D

About the 754, My C2D setup remplaced a 754 Venice 3000+ @ 2.5 and the Venice is a great chip indeed, unfortunately outdated by today's standard. The 2nd PC is my media server, using parts which is originally in my main PC 2 1/2 years ago. It's still running strong, albeit at stock voltage to reduce heat and power bill because it's also folding 24/7 while serving media files.
 
I first folded for OCAddiction then later on for Xtremesystems....Cant remember the name but I had a 3200+ with 1mb cache, ran it 266x10 , 2-2-2-5 on my HyperX 3200 ..... I reach my goal of making the first page in 3dmark2001 and after that I packed it in.... That was my last hardcore rig, now its all air for me....:(
 
What XP does is it takes one task running 100% on one core and shares it equally so that both cores are running at about each 50%.

If i understand what you are saying correctly, I don't think that is accurate. XP can't split a thread between two cores, one thread can only be assigned to one processor core at a time. I could be misinterpreting what you are trying to say however.

 
What he means is when you run say Prime95 or superpi, instead of 1 core running 100%, youll have 2 cores both running at 50% according to task manager.... Run a single threaded app and look at the task manager, youll see what he means..
 
Mine shows one core running 100%, but 50% overall cpu usage since the second core is unused.
 
Mine shows one core running 100%, but 50% overall cpu usage since the second core is unused.

yep, sorrry to say OP's analysis has flaws, however his point about temps is probally valid. sorta, regardless of load, it never hurts to run electronics as cool as possible within reason.
 
I don't know about XP Home or Vista but with XP Pro it takes a single threaded process and constantly switches it back and forth to both cores. Instead of one core running at 100% and the other core sitting idle running at or near 0%, this allows both cores to run at 50% which balances the load and heat between the two cores.

It's possible to override this behavior by going into the Task Manager and right clicking on the process and setting Affinity... If you set this process to use only one of your cores then it will but if you don't specifically set Affinity then both cores will take turns working on the process until it's done.

Does this happen in XP Home as well or just XP Pro?
 
I don't know about XP Home or Vista but with XP Pro it takes a single threaded process and constantly switches it back and forth to both cores. Instead of one core running at 100% and the other core sitting idle running at or near 0%, this allows both cores to run at 50% which balances the load and heat between the two cores.

It's possible to override this behavior by going into the Task Manager and right clicking on the process and setting Affinity... If you set this process to use only one of your cores then it will but if you don't specifically set Affinity then both cores will take turns working on the process until it's done.

Does this happen in XP Home as well or just XP Pro?

I run XP Pro and I've never seen it do this, can you point me to where you got this information?
 
Unclewebb is correct. I have read this/his info a few times in articles pertaining to the foundations and definitions of a hyper threading cpu.
 
To test on your system press CTRL+ALT+DEL and start up Task Manager and click on the Performance tab so you can have a look at the 2 graphs.

Start up a single threaded application or game and use it for a minute or two and then have a look at the graphs. When gaming with NFS Carbon, the two graphs are squiggly lines varying between say 40% and 60% as the 2 cores are shared.

If your graphs show one core at 100% and the other at 0% then it is not sharing.
 
balanceds8.png


I originally thought XP shared processes 50 / 50 but after seeing this graph it's pretty obvious it shares the work based solely on keeping the core temperatures of the two cores equal.

If you run CoreTemp or SpeedFan v.4.32 you can watch as the two cores take turns. If Core 0 is running hotter than Core 1 then more of the work load is given to Core 1.

When Core 1 becomes hotter than Core 0, it transfers some of the load back to Core 0 to keep the temperatures of both cores in sync.

In this case instead of 50% / 50% the two cores are loaded 25% / 75% which keeps both cores happily running at 53C. It seems to be the Intel C2D doing the balancing based on balancing core temperatures so it should work similar no matter what OS you're running.
 
Back
Top