With 3GHz in th bag, my next overclock attempt should take my system to the dizzy heights of 3.2GHz.

So into the BIOS again and set the FSB to 356MHz which will take the CPU frequency to the magic 3,204MHz. I've left the vcore at 1.3v to see if I can get away with it for now although it's highly unlikely this will be enough juice for the CPU to cope. We'll see. Again, I've had to change the DRAM frequency to accommodate the increased FSB, so I've changed the DRAM frequency to run to the nearest frequency to 800MHz at 712MHz. This means a memory divider of 1:1 so the memory is now running synchronously with the system bus. Once I've established a stable and decent clock speed, I look at way of optimising the memory.

Surprisingly, the system did boot into windows and I managed to run the CPC benchmarks. However, the results where less than they were at stock settings, which suggests that something was not quite right. This is possibly the system dropping the frequency to enable the application to run, albeit at a slower speed. Speedstep maybe? Anyway, the overclock proved too much for 3DMark06, which caused a system restart. So, as predicted, I increased the vcore to 1.35v, which had the same outcome. Only 1.4v seemed to do the trick. CPU-Z clearly shows that the vdroop is making it's presence known, as the VID is showing 1.3v. However, the system booted and all the benchmarks ran with no grazed knees.

NB - GPU overclocking disabled for CPU benchmarking

3DMark 06 benchmarks :

Overall score = 12,641  (SM2.0 = 5,070,  SM3.0 = 5,293,  CPU=4,447)

CPC benchmarks :

GIMP Image editing = 1,090 |  H. 264 Video encoding = 1,715 | Multi-tasking = 798 |  Overall = 1,201

Company of Heroes Performance test :

@ 1600x1200 highest settings

Average Frame Rate = 98fps

Maximum Frame Rate = 247fps

Minimum Frame Rate = 50.1fps

Ok the overall scores are looking good so let's see how stable this really is.

Yet again, as per previous overclocking attempts, I have ran a stress test using Orthas Prime which was monitored using CoreTemp. Full details of my torture testing can be read here.

The Gromac test ran for best part of the day with no errors. However, the Blend test, which stresses both CPU and Memory, produced an error almost immediately. Strangely, this only occurs on core #3 if all core cores are under load during the blend test. Running instances of the blend test on core #3 on its own, or in separate pairs or even on three cores is fine. I'm wondering if this is the Northbridge. So I increased the NB from 1.25v to 1.4v but this made no difference. This is strange because everything appears to be stable bar this one test. I will have to research this and come back to this latter but since the Gromac test completed, I'm going to accept the the settings.

CoreTemp reported the following temperatures:

Core #0 idle/load = 32/57 | Core #1 idle/load = 32/57 | Core #2 idle/load = 32/56 | Core #3 idle/load = 32/57

So despite the Blend test failure, it's another successful overclock. Since all other tests appeared to be working fine and the fact I've played about 4 hours of Gear of War without a single glitch, I think it's safe to say that the system is stable as it will be but as said, I have some investigating to do.

800MHz (or 33%) is a massive over clock above the stock speed of 2.4Ghz. The system is stable and the temperatures aren't that bad either. I'm sure there's enough room for some more. At 3.2GHz, the system should be happy with running with a FSB of 400MHz by dropping the CPU divider to 8. (8x400). There's a possibility to reach higher clock speeds such as 3.4GHz although things might be getting a bit toasty. There's a possibility to increase overclock the DRAM speed to say to 854MHz using a 5/6 ratio with the current settings. I will certainly look into these but for now, we'll enjoy 3.2GHz for a quad core CPU costing less then £180.

Here's a summary of the changed settings:

JumperFree Configuration Settings

FSB Frequency :  356MHz

vcore : 1.40v

DRAM frequency : 712MHz  (1:1 Synchronous)