OC'ing The K6-2



Many of you looking to upgrade to a Super7 system have contacted me requesting information on the overclocking abilities of the K6-2.  It is because of this that I have decided to pursue this undertaking.  The following article will be full of tips and warnings as well as a few benchmarks, but because every system is different, the results you get may vary considerably.
Many of you have successfully overclocked either your Pentium MMX, AMD K6, and even in some cases your Cyrix 6x86's.  In doing so you've gained some knowledge about your systems and in particular:
Processor Cooling Techniques
Core Voltage Capabilities and Settings
Fluctuations in the PCI bus frequency and it's resulting affect on your peripherals
The Stability of the DRAM at higher CPU bus speeds
Having mastered or at least attained a basic knowledge in these four areas you probably pushed the limits of your processors capabilities. Now with Super7 technology pushing front side bus speeds up to and even above 124MHz you may only need to do a little rethinking.  While at speeds up to 100MHz the Super7 platform's PCI clock divider limited the PCI bus speed to 33MHz, increasing the FSB will again cause fluctuations in the PCI bus and at speeds this high you will not only need PC100 SDRAM but you will need high quality 7ns or lower, CAS 2 latency PC100 if you desire to effectively overclock your processor with any degree of stability.
Super7 processors performance relies on the external speed of the CPU to determine how fast it can transfer data to and from the L2 cache. The increase from the old  66MHz external bus speed to Super7's 100MHz bus speed results in a major performance gain that is reduced dramatically when not using the 100MHz setting. For instance, the performance level of the K6-2 300 overclocked to 332MHz  (4.0 x 83MHz) will be less than the performance level at it's rated 300MHz (3.0 x 100MHz) because the 83MHz bus speed affects system performance more than the higher clock multiplier rate.


K6-2 300MHz Processor
As far as I can tell this CPU is sold at just about it's peak performance. In a dozen or so tries I could only get Windows 98 to load only once when running the chip at 100MHz x 3.5 with a high core voltage of 2.5v and as such can't recommend this chip as any kind of overclocking candidate.  It does perform  with a reasonable amount of stability when running the front side bus at 112MHz even at a core voltage of 2.3v (remember, it is important to set the voltage at the lowest possible setting at which the chip will run with stability) but heats rapidly when called upon for floating point intensive applications such as QuakeII; setting off the CPU temperature alarm (66C/151F) before completing even one Massive1 timedemo, using a regular heatsink and fan combination.  Changing the setup to an oversize heatsink and using a case fan instead of one that is mounted to the heatsink itself overcame all heat related concerns as far as my system was concerned though and the CPU ran like a champ from then on.  Running the processor at 333MHz (95MHz x3.5) proved unsatisfactory, only loading Windows about 50% of the time. 


K6-2 333MHz Processor
The K6-2 333 shows a bit more promise than its little brother in the area of overclockability.  Running the chip at 350MHz is easily accomplished with the FSB running at 100MHz and the clock cycle at 3.5 even with the core voltage set at the default 2.2v.  This fact alone leads me to wonder about the recently released K6-2 350MHz chip and its overclocking capabilities.  The chip is also capable of reasonably stable performance at 380MHz (95MHz x 4.0) with a core voltage setting of a medium-high 2.4v.   With the oversize heatsink and case fan combination the processor rarely generated more than 62C, well within its normal operating range.


Setting Up
Unfortunately,  I have yet to receive my rev. 1.1b of the VA-503+ so for most of the tests I used my existing FIC VA-503+ with the "CD' rev. of the MVP3 chipset. The board itself is astoundingly stable at the 112MHz FSB  and has the capability of adjusting the voltage in 0.1v. steps but for some tests I used the Epox 58-MVP3E-M ATX mainboard (w/rev. "CE" of the MVP3 chipset) because of its ability to run the FSB at 95MHz.  Before starting any serious testing I had experimented with using a PCI video card but the processors seem to run quite a bit hotter when testing graphics intensive performance using the  ATI All in Wonder Pro 8 MB PCI graphics card and though I know of no particular reason for this ( and I do intend to follow-up on this ) I decided to use the Viper 330 AGP 4MB card.  Using the AGP card the processors ran at least 8% and as much as 22% cooler when overclocking.  Always starting with the default voltage recommendations, voltages were raised in 0.1v increments when using the VA-503+ but on the Epox board I decided it wouldn't be prudent to exceed the 2.4v. setting as the next step up is 2.8v which would almost certainly mean too much heat and would most likely result in damage to the processors. Thermal paste was used between the processors and heatsink to assist in the cooling process, and the machine was run for a minimum of 6 hours doing general business applications before running any of the benchmarks shown below.  Each benchmark was run three times and then averaged to reflect the scores that you will see below.   Either board was equipped with a single Corsair PC100 64MB DIMM with a CAS latency of 2 and memory accesses within the BIOS were set to the fastest settings available.

ChartObject Winstone 98

As you can see the higher the front side bus the better the performance. The K6-2 300 actually outperforms the the 333 when it runs at 112MHz and the 333 is running at either 95 or 100MHz until you get the 333 up to about 380MHz


ChartObject Winbench 98 CPUmark32

Although the CPUmark32 slightly favors the increased clock rate the FPU Winmark is again higher when the FSB is increased...

 ChartObject Winbench 98 FPU Winmark


Pretty impressive 3D Winbench marks even with a relatively unimpressive video card...

ChartObject 3D Winbench 98

Conclusion:

What it all comes down to is that while you can push more performance out of these processors, it may be worth waiting to see what AMD has up their sleeve with the K6-2 350 although it seems a little pricey.

Before attempting to overclock your processor it is a good idea (unless you have some experience) to read our Overclocking Guide.  Make sure that you understand that AMD does not recommend running your processor at speed other than those specified by the manufacturer.  If you burn up your processor they will not replace it nor will your mainboard manufacturer.  You will be in effect voiding any warranty granted by any of your hardware device manufacturers.

I hope to get ahold of the K6-2 350 soon and will do a full review.  Until then I am happily running my machine at it's recommended bus speed and clock rate and I am happy with the result as well as the peace of mind.  If you like to tinker and don't have any data that you can't afford to lose have at it and have fun...

Last Updated on 9/3/98
By Lyle Boomer


1998/1999/2000
MediaTek All Rights Reserved