Nick - question re temps across cores

Q6600 was pretty much the same from core 0 to 3 however the q9650 is strange at idle it's
37
33
49
49
and under load it's

56
52
67
67

This is after lapping the CPU and using a small amount of Ceramique. The difference was the same before lapping etc.

Any reason for the 12C difference between core 0 amd 2 & 3?

Normally I'd say I screwed up appliying the paste but I'm pretty careful in tat arena - blew up an AMD chip years ago. Also, I've reapplied AS5 twice and switched to Ceramique with te same results.


Using Coretemp 0.99.3

Should I consider the higher temps valid?

Puzzled.

Thanx,

Vic

Before Clocked Max




After clocked Max





Because of the gear shift on the northbridge at 450MHz you dont need PL6

As you can see your latency remained almost the same but you now have a significat jump in memory write.. much more in line with the read so your now ready to blaze the sky


any questions?

Think of the data as load a truck is hauling.. think of the road as the communication path to and from the CPU to the memory... think of the northbridge as a weigh-station on a highway... the longer the truck is sitting in that station waiting to be cleared to proceed, the longer it takes for the CPU to get the data and write to the memory as well.

When you buy the right motherboard which has the higher quality parts on it which reduces 'noise' and allows voltage increases, and, you buy the right memory quality that will work with the higher frequencies and a CPU that is designed to run at the FSB/multiplier you are after... you are buying the access to that weigh-station and you are optimizing the amount of time the truck spends in that station through that access

All external northbridge systems are designed to change the latency communication between the CPU and memory through the northbridge.. How much access and how far is based on the quality of the motherboard and its BIOS. Success is based on the memory product and its ability/stability. The BIOS programmer placed latency shift points in the BIOS based on CPU STRAP and FSB.. at approx 380, 400, 420, 450, 480, 500, etc. When you set the STRAP you have set the math the motherboard uses to set the memory speed and calculate from there..  as you increase FSB past the mentioned points a shift to lower latency takes place across the northbridge based on CAS timing manually set in the BIOS or by SPD if manual timing is not used. Its always best to set the memory timing (primary 4 timings) manually.

That goes back to the formulas posted in this writeup

http://www.thetechrepository.com/showthread.php?t=195


So what you have done is what most never do with thier expensive computer purchase... you have taken true advantage of the purchase you made and are accessing the CPU/memory in the way the manufacture of those expensive products intended you to for the price you paid for that CPU/Memory and motherboard.


In other words... welcome to the "Pro Club"

The difference between men and boys is the price of their toys...  but men know how to use their toys too

all too often people spend the money on the right parts then run around and tell everyone it was not worth it in FSX simply becasue they dont know how to use those expensive parts...

Then there are the self crowned experts who think they know everything... I think you know who I am talking about. People like that are more dangerous to this hobby than anyone else. They perpetuate the myths and as the ball rolls more and more people end up complaining about the application, when what they should be doing is stringing the Internet acquired PhD up by his toenails

Sort of what I figured - After 2 hours of 100% utilization in OCCT - max temp on highest core was 71C.

I seriously doubt that I'll run anything other than a stress test to get it that high.

I did remove, clean and reset the heat block  but still have a 6C difference.  Doesn't seem to much more than an anomaly so I'm not going to sweat it.

BTW, took a very quick flight around LA Basin last night - frames were noticeably higher but what most impressed me was the only popping of textures that I saw were in the *far* distance. Up close and around the a/c was as smooth as silk.

Played around with clocking my EVGA 8800GTS 640 using Riva Tuner. Leaving everything linked I bumped it to 702/1623/793 and it seems happy. 3Dmark gets temps to 70C but nothing higher.

Going over 702 and get allsorts of troubles including system hang.

So far so good.

Thanx for the heads up Nick. Was thinking about it but I'll give it some time. Was able to bump up the 8800GTS a tad more - 702/1728/854 and all is well so I'm good to go for a while.

The only card out there that will beat a 8800GTX which is not 'highly' clocked is a GTX 280

Clocking the numbers requires knowing the exact specs on the procesor, motherboard and memory

I need to know the processor (Q or QX), memory manufacture and model of the sticks... how many sticks and what motherboard

I cant just throw numbers out there.. the value in clocking is not only tied to the proc, its tied to the memory speed and timing ability