Original Link: https://www.anandtech.com/show/160

Just about a year ago, around this time, Intel stepped up to the challenge their long time competitor Advanced Micro Devices (AMD) proposed in the microprocessor industry.  For a month prior to Intel's release of their Pentium II processor AMD had taken all of the glory with their high performing and low cost alternative, the K6.  Looking to crush the competition with one swift blow Intel struck back just over 4 weeks later with the even higher performing Pentium II processor, making the statement, as clearly and as well spoken as possible, "Welcome to our world."

Now fast forward to the present day, approximately one year later, Intel had just released the next installments in their Pentium II processor family, the Pentium II 350 and 400 and surprisingly enough, just about a month later, AMD took the world by surprise with the release of their new low-cost, high performing microprocessor, the K6-2.  This time around, with the position of the microprocessor giant in limbo, AMD has proudly held their heads high and managed to respond to Intel's statement made over a year ago with the skillfully placed response: "We're back."

From a performance perspective, AMD was the 'April-Fool' of 1997, their K6 release on April 2nd was plagued by chip shortages and supply problems; missing both the expected shipping dates, and the "fools-day" AMD had a success on their hands but knew more was necessary to combat Intel, especially after the mammoth Pentium II processor made its debut in May of that year. 

AMD let Intel take the early glory this year, in hopes of shaking off the reputation of not being able to deliver on time.  First it was the K5 that was off shipping schedule, then the K6/233 which made its appearance in retail channels over 2 weeks after its release, followed by the 266MHz part that was too little too late.  Ecstatic over their move to the 100MHz Front Side Bus frequency, has Intel become the 'April-Fool' of 1998?  Has AMD managed to breathe life back into the dying Socket-7 market?  Is the threat of the K6-2 as great as many have anticipated it to be, or is the processor's release simply a way of stalling users from completely migrating to the Slot-1 standard?  Let's answer those questions as AnandTech takes a look at the AMD K6-2 Processor...

The Weakness

Not being able to rely on a powerful Floating Point Unit (FPU) for its gaming performance, the sales pitch for the K6 was that it was a high performing business alternative to the Pentium II.  If you wanted a powerful gaming platform you were shyly pointed in the direction of Intel's massive Slot-1.   Luckily AMD's aim to please attitude emerged when they went to revamp the K6 for what was officially intended to be AMD's answer to the Celeron, this processor attained the code-name K6-3D. 

Sounding much like a new video card, very few people really understood the face behind the mask the name K6-3D covered.  What AMD knew, something the rest of the world assumed to be another try at a MMX-like stunt, was that this time around they were prepared to do much more than compete with a crippled cacheless Pentium II.  Instead of focusing on improving the already outstanding Business Application performance of the K6 processor, and instead of simply increasing the clock speed of the processor, AMD decided to confront their weakness head on.  As discussed earlier, this weakness was obviously its poor 3D performance, the market demanded faster 3D performance and in return, AMD provided just that, 3D-Now!   Surprisingly enough, 3DNow! also happens to be the title of AMD's new instruction set incorporated into their highly anticipated sequel to the original K6 microprocessor, rightfully entitled, the K6-2. 

3DNow?

You have probably heard many "techno-geeks" complain about how we don't really need all of this power in our computers today, ever wonder why that topic comes up so frequently?   Consider this, your CPU has the power to process millions of commands in a single second, but how much of your CPU's full potential do you imagine is being completely utilized by simply clicking on the Start Menu?  Or in a more related sense, how much of your CPU's full potential is being used every time you run Quake 2?  The answer is obviously a very limited portion, while there are many interactions taking place between your CPU and on-board cache subsystems, system memory, your video bus, and other such things, your CPU itself, more specifically the FPU is only performing a limited amount of tasks in a highly repetitive manner. 

On the hardware side of things, the FPU of a processor is designed to basically handle anything that comes its way, in contrast, on the software side, a 3D game is designed to take all it needs from your system to run and leave it at that.  Whether or not your computer has a FPU capable of calculating a trip to Mars does not matter to a game that simply requires the execution of a few relatively simple FPU instructions.  This is where AMD saw a chance to make an impression that would definitely last, and more importantly a chance to make an impression that would open a window of opportunity for future generations of AMD processors to come.   Instead of concentrating on re-designing the FPU of the K6, why not simply offset the weakness by using a separate set of hard-coded instructions (instructions built into the CPU itself vs software based instructions) to perform common tasks used in 3D games.   This was the hand AMD brought to the table with the K6-2, a winning hand, that is if AMD could pull it off.

The AMD 3DNow! Instructions often fall into the MMX ditch as far as comparisons go, and it should be mentioned here that unlike Intel's 57 MMX Instructions, the 3DNow! instructions actually serve a real world, noticeable, gaming purpose.  It is true that MMX instructions do help in some cases, however the hype surrounding the integration of MMX instructions into the Pentium line of processors blew the realistic impact of the instructions completely out of proportion.   3DNow! exploits a technology known as Single Instruction Multiple Data execution, or SIMD for short.  What this technology does is, as the name implies, it applies a single command (or instruction) to multiple sets of data simultaneously.  A real life example of something like SIMD execution would be much like picking berries.  Instead of picking a berry off of a bush, washing it, then returning to the bush to collect more you could simply collect a handful of berries at once and retrieve them simultaneously.   The key to understanding SIMD is the simultaneous execution of the instruction, therefore saving time.  SIMD is not the only benefit the K6-2 receives from AMD's 3DNow! instructions, however it is definitely a plus.  If you have ever written any code that requires heavy floating point calculations you will know that floating point division is an extremely slow process from the perspective of the CPU.  AMD saw this as a major problem and integrated a reciprocal multiplication function into their 3DNow! instruction set.  For example, instead of dividing a number by 4, using reciprocal multiplication you can achieve the same result by simply multiplying by the fraction 1/4.   While this may not seem like much, you must understand that a processor can multiply faster than it can divide, in this case, instead of dividing by 4 you are multiplying by 1/4, which can make a world of difference in terms of performance on a more complex level.  Since the reciprocal multiplication function is hard-coded on the K6-2 it actually provides a healthy performance increase in many situations.  It is instructions like these that make the 3DNow! instruction set different from the old MMX instructions that drove the market wild what seems like so long ago. 

The Catch

In order to take advantage of the 3DNow! Instructions on the K6-2 you must have software support, that is the catch on which the success of the entire K6-2 platform is based upon.  Without software support, you have nothing more than a K6 running at a high clock speed.  Support for the instructions can come in many flavors, the most intelligent move by AMD was to work with Microsoft on including 3DNow! support in Microsoft's next release of their Direct X API (Application Programming Interface), Direct X 6.0 to be specific.  In that case, any game using Direct X 6 will experience a small benefit, but naturally not the full benefit, from the 3DNow! instructions.

In the event that a game is written to specifically use the 3DNow! instruction set it will experience a definite improvement on a K6-2 system in comparison to a standard K6 system, an improvement that can even eliminate the performance barrier between a Pentium II and a K6-2 system.  Games can, of course, be patched to take advantage of 3DNow! instructions, however as of now Quake 2 is the only game which has been blessed with that.  Don't expect too many older games to be re-worked with 3DNow! support, however you can expect increasing amounts of support in new and upcoming games.  A perfect example of this would be G.T. Interactive's visual wonder, Unreal which has native support for the K6-2 and its 3DNow! instruction set, a feature which is evident in its performance on K6-2 systems.

In order for AMD to achieve the maximum performance out of the aging Socket-7 platform they knew that they would have to initiate their own set of modifications, these modifications to the Socket-7 platform became known as the Super7 standard.

The Super7 motherboard standard calls for the following requirements to be met: A motherboard must...

• support the reliable operation of a Super7 processor, such as the K6-2, at the 100MHz FSB

• support the Accelerated Graphics Port (AGP)

• support Super7 CPUs

The big requirement from the above three is obviously the reliable support for the 100MHz FSB, something only two Socket-7 chipsets have been able to achieve, therefore making up the two ends of the current Super7 market.

ALi Aladdin V

The ALi Aladdin V chipset was the first Super7 chipset announced, in fact it was used internally by AMD for testing in the early developmental stages of the K6-2. While it went through quite a few revisions before finally hitting the market, the Aladdin V chipset is now ready for action. Out of the 10 motherboards AnandTech compared in the Super7 Comparison, only one featured the Aladdin V chipset, the Iwill XA100. While no awards were given out for Super7 motherboards in that comparison, if you're looking for the best overall Super7 Motherboard you will probably want to direct your attention to the ATX Microstar MSI 5169 which AMD used in their K6-2 Evaluation System shipped to AnandTech for testing. The 5169 is a very competitive performer, and although you won't see anything like a 112MHz FSB setting on an Aladdin V board, overall the 5169 is a very well rounded motherboard. Its spacious layout is much like the older Socket-7 AGP motherboards, and the initial setup and configuration of the motherboard isn't plagued by some of the complexities you will find on MVP3 based motherboards. If you are looking for the best overall Super7 motherboard, look no further than the MSI 5169...however if that doesn't suit your needs...

The Low Cost Alternative

You have an AT case, EDO SIMMs, and just enough money left in your upgrade fund for a cheap motherboard and a K6-2. Out of luck? Absolutely not. VIA Technologies has the answer to your prayers with their MVP3 chipset, the second Super7 chipset to emerge since AMD's initial K6-2 announcement. The MVP3 chipset allows for a user to select the option of running the system's Memory Clock at the same speed as the AGP Bus, which is, in most cases, at or around 66MHz. This means that you can re-use your old EDO SIMMs, or non-PC100 SDRAM without having to worry about stability issues with the 100MHz FSB. Considering the price of a decent MVP3 based motherboard is about $90 - $100, and the price of a K6-2 starts at around $160, you can easily have the power of a K6-2 in your system for under $300.

The 300MHz K6-2 will allow you to overclock to a realistic maximum of 350MHz (100MHz x 3.5), while the 336MHz (112MHz x 3.0) setting is still faster, once again due to the increase in FSB frequency.  If you don't mind overclocking, and want to save a few bucks you are better off purchasing a K6-2 266 and clocking it at 300 instead of spending the extra cash on a 300.  Mind you that the 266 won't make it up to 350MHz reliably, so if you really want the most out of your system don't expect the 266 to be able to take you there. 

Of course, like any other manufacturer, AMD does not condone overclocking.  In spite of this, the K6-2 333MHz Evaluation System AnandTech received to perform a full suite of benchmarks on was nothing more than a high yield K6-2 300 clocked at 333MHz.  It would seem odd for AMD to use a 66MHz FSB processor to demonstrate the full capabilities of their latest and greatest processor, which is why AMD didn't use conventional methods to show off their 333MHz K6-2.   Instead of taking chances at running the 300MHz parts at 350MHz, something that cannot be reliably accomplished unless you increase the core voltage of the chip, AMD modified the Microstar MSI 5169 that shipped with the Evaluation System to support a unique 95MHz FSB speed.  Using the 95MHz FSB frequency and a 3.5x clock multiplier the K6-2 was clocked at 332.5MHz, a speed that could not only outperform a K6-2 300 using a 100MHz FSB but one that could perform competitively enough to show off in front of the public eye. 

When?

The question as to when 333MHz parts will begin to ship has been brought up, right now you should expect to see 266 and 300MHz units in decent quantities, with prices at reasonable levels for the processors considering the demands.  Don't be surprised if AMD decides to forego the production of the 333MHz K6-2 in favor of a 350MHz part that could run at a true 100MHz FSB speed.  It is doubtful that too many people would purchase a 333/66 K6-2 over a 300/100 K6-2 if they didn't plan on overclocking the chip to 350/100 and beyond.

AMD plans on eventually taking the K6-2 line up to 400MHz, early indications show that a 400MHz K6-2 could easily topple the balance of power between Intel and AMD...the question is when AMD will release the K6-2 400.  Hopefully we will not see a repeat of the K6/266 disappearing act of Q4-97 with the K6-2 400...

• K6-2 clocked at 333MHz (95MHz FSB)

• Microstar MS-5169 Aladdin V Motherboard w/ 512KB L2 Cache

• 64MB PC100 SDRAM

• IBM Ultrastar Ultra-Wide SCSI 3 HDD (9GB)

• Plextor UltraPlex 32X SCSI CD-ROM Drive

• Adaptec 2940 Ultra Wide SCSI PCI Controller Card

• Voyetra Montego PCI A3D Sound Card

• Diamond Viper V330 AGP Video Card

The Pentium II comparison system differed only in terms of the processor and motherboard in which case the following components were used:

• Pentium II clocked at 300MHz (66MHz FSB), 333MHz (83MHz FSB), 400MHz (100MHz FSB)

• Shuttle HOT-641 Intel BX Motherboard

The following drivers were common to both test systems:

• nVidia Drivers v4.10.01.0236 (Diamond Viper V330)

• DirectX 6 v5.1

• AMD 3DNow!/3Dfx Quake 2 Driver

All tests were run at 1024 x 768 x 16-bit color

NOTE: Higher Scores are Better

As far as Raw FPU power is concerned, the Pentium II has the title locked down, however overall the K6-2 333 gives the more expensive Pentium II 333 a run for its money. 

NOTE: Results in Million Pixels per Second - Higher is Better

As you can clearly see, the K6-2, in combination with DirectX 6 does have a significant advantage over the Pentium II platform in Direct Draw situations.  Using identically configured systems, the K6-2 consistently provided benchmark numbers equal to or greater than those of a Pentium II - 400 under the Direct Draw tests of Winbench 98.

Video Playback seems to fare quite well on the K6-2, it is interesting to note that Intel's Indeo Codec performs horribly on the K6-2 in comparison to the Pentium II while the Cinepak and MPEG-1 tests revealed that the K6-2 performed just as well if not better than the Pentium II at an equivalent clock speed.  The Pentium II - 400 seems to be the best option here, although not by a tremendous margin, definitely not worth the price difference between it and a sub-$200 K6-2.

Wait for V-Sync was Disabled

Memory Clock was set at 93MHz

NOTE: The 3DNow! Q2 Drivers only effect Quake 2 performance and do not help/hurt performance in any other games

There is obviously a considerable improvement over the K6-2's performance under Quake 2 when using the 3DNow! Quake 2 Drivers provided by AMD.  These drivers should be making their public release some time in the next week.  The drivers are still not taking complete advantage of the K6-2's potential, and there is still room to improve on the performance end of things.   With Quake 2 as a working example, a bit of hard work can definitely pay off with the K6-2, all that is required is software support, and if the demand is there in the market, the software support will follow. 

Older Glide games, that won't take advantage of the K6-2's 3DNow! Instructions perform just as poorly on a K6-2 as they would on an original K6, in comparison to a Pentium II.  It is doubtful that patches will be written for older games, so depending on what games you like to run the most you may be out of luck.  Even games that do take advantage of the DirectX 6 engine fall behind the Pentium II in terms of performance as can be seen by the Forsaken benchmarks (although they are all over 98 fps, which is beyond smooth). 

Product: AMD K6-2 Processor
Vendor: Treasure Chest Computers
Website: http://www.tccomputers.com/