Friday, December 29, 2006

Overclocked at 3.0GHz

I ran Memtest86 overnight for over 10 hours and 33 minutes without any errors.

I restarted into BIOS and checked the hardware monitor:

Motherboard temperature: 31C/87.5
CPU temperature: 45.5C/113.5F
CPU Fan Speed: 1776 RPM

All systems go for overclocking.

I then changed CPU Frequency from 266 MHz to 300MHz and verified that the DRAM Frequency automatically followed at DDR2-600 MHz staying at the 1:1 (DRAM:FSB) ratio.

I saved this profile to O.C. Profile 2 in BIOS, saved BIOS and restarted Nighthawk.

Computer "Properties" in Windows XP MCE 2005 (Rollup 2) showed:

Intel(r) Core(TM) 2 CPU
6700 @ 2.66 GHz
3.00 GHz, 2.81 GB of RAM
Physical Address Extension

and Everest Ultimate confirmed:

Maximum Clock: 3800 MHz
Current Clock: 3000 MHz
CPU Overclock: 12%
Memory Bus: 300 MHz
DRAM:FSB Ratio: 1:1

I allowed Nighthawk to sit for awhile and come to an idle state with the following readings:

Motherboard temperature: 27C/81F
CPU temperature: 44C/111F
CPU Fan Speed: 1776 RPM
GPU temperature: 51C/124F
Disk temperature: 23C/73F

The first benchmark I ran was the Everest Ultimate Cache & Memory Benchmark. There were significant improvements in all results except for two: Memory latency went up from 71.3 ns to 76.1 ns and memory read went down from 6781 MB/s to 6178 MB/s. At the same time, memory write went up from 4818 MB/s to 5419 MB/s and memory copy went up from 5399 MB/s to 5654 MB/s. The L1 and L2 results all improved.

The next benchmark I ran was 3dMark06. After overclocking, Nighthawk scored 8772 3DMarks, improving on the 8536 3DMarks scored before overclocking.

There was an improvement of 309 Marks in the CPU score, but only 7 and 3 Marks improvement in the SM 2.0 and 3.0 scores, respectively. Based on these results, my initial conclusion is that the Core 2 Duo e6700 CPU running at stock 2.66 GHz speed is not holding back the Nvidia 8800GTS video card at all. Hmmm. I may have to look into overclocking the video card also!

The last benchmark I ran were 2 instances of SuperPi Mod1.5 XS, primarily to see what impact 100% CPU utilization would have on the CPU temperature. The screen capture on the right was taken after the completion of the 20th loop for each instance of SuperPi. I was relieved to see the CPU temperature not go higher than 49C/120F @ 100% CPU usage by 2 instances of SuperPi.

From a performance improvement perspective, SuperPi showed none. The 1st instance showed an improvement of 2.031 seconds but the 2nd instance was worse by 2.048 seconds.

All in all, the 3 benchmarks show a slight performance gain from this initial step at overclocking. The proof will come when the final stable overclock is reached.

I really wanted to see how Nighthawk performs from a temperature perspective when both the CPU and GPU are running at 100% utilization after overclocking. I downloaded the 3 Nvidia GeForce 8800 Tech Demos: Adrianne, Box of Smoke and Froggy. While the Adrianne demo is visually stunning and stimulating, the demo stresses the GPU but not the CPU (only an average of 30% CPU utilization). The same applied to the "Box of Smoke" demo (average 5% CPU utilization.)

On the other hand, the "Froggy" demo uses both the CPU and GPU simultaneously at 100% so I ran it for one hour. The highest temperature readings during this period were:

Motherboard temperature: 30C/86F
CPU temperature: 51C/124F
GPU temperature: 65C/149F

After this run, the temperatures quickly settled down to their "idle" temperatures reported earlier in this post (see above.)

I will be running Nighthawk at 3.0GHz for about a week before I overclock it to 3.2GHz. This will give the components in Nighthawk time to get acclimatized to running at this faster speed.

No comments: