The Alt.300 Club

Cyrix M II 300 / IBM 6x86MX PR300

Cyrix recently announced the 300MHz M II, the first in a new line of low-cost desktop microprocessors that support a 100MHz bus. The Cyrix M II will replace the 6x86MX microprocessor line and is based on the existing Socket 7 motherboard and chip set architecture.

The first Cyrix M II is a PR300 that runs at 233MHz and utilizes a 66MHz bus. The PR333 will be the first 100MHz M II and run at 250MHz (100MHz/250MHz).

All Cyrix parts that are slower than 233MHz will retain the 6x86MX product name. All parts faster than 233MHz will be an M II. Cyrix doesn't want to associate the name of their fastest processor to Intel's entry level offering.

Please note the M II is not the Cyrix Cayenne. The Cyrix Cayenne is Cyrix's next-generation 6x86MX processor core. The Cayenne will feature a fully pipelined, dual-issue floating point unit and 15 new multimedia floating point instructions to enable the highest-performance 3D graphics, DVD, and 3D audio.

The M II is also not equivalent to the AMD K6-2. The AMD K6-2 is AMD's new name for the AMD K6 3D. The K6-2 is tuned to run graphical applications such as games. It features better floating point performance than standard K6 chips, and is compatible with motherboards featuring a 100 MHz system bus. The boost in floating point performances comes from the addition of 3D Now, a series of 21 new instructions. Cyrix and IDT will eventually adopt the technology. The AMD K6-2 is scheduled to be released on May 28, 1998.

The first M II chips will be made on an older 0.35-micron production process and require a 2.9V core voltage. The M II family will be moved to the advanced 0.25-micron process in May and most likely require a 1.8V core voltage. The chip, which will be released in quantities later in May, will be priced at $180 in 1,000-unit quantities.

After another very public squabble with Cyrix, IBM Microelectronics decided not to follow partner Cyrix Corp.'s lead in renaming its microprocessor line. IBM Microelectronics has decided to build its own brand image further. In addition, IBM has decided that their PR300 will be a 225MHz (75/225) rather than a 233MHz (66/233) part.

This means that the same silicon will have identical PR ratings, but two different product names and two different internal frequencies.

AMD K6/300

Extremely high yields of their new 0.25 micron parts allowed AMD to introduce the K6/300 to the retail channel sooner rather than later. Yields have been so high that AMD is producing more 300MHz than 266MHz parts. As a result, the 300MHz K6 is readily available in the marketplace and the 266MHz K6 is more difficult to find.

Like the K6/266, the 300MHz K6 requires a 2.2V core voltage. This may pose a problem for some older mainboards that do not offer a 2.2V core voltage, but most new boards will support it.

The low core voltage requirement created as a result of AMD's advanced 0.25 manufacturing process enables the K6/300 to run extremely cool. Unlike the K6/233, heat is definitely not an issue for the K6/266 and K6/300 parts.

The K6/300 utilizes a 66MHz system bus and supports clock multipliers up to 5.0x. Many older mainboards will not provide clock multipliers higher than 3.5x and therefore will not support anything faster than a K6/233.

Marketing Magic or Something Real

As an analyst, I feel obligated to point out another inconsistency in the Cyrix/IBM performance rating (PR). I downloaded a complete listing of Performance Ratings from the Cyrix website late last year. Cyrix initially assigned a performance rating of PR266 to the 225MHz (75/225) 6x86MX. My benchmarks below show that a 233MHz (66/233) 6x86MX will produce virtually identical performance to the 225MHz (75/225) 6x86MX.

The internal 6x86MX/MII architecture has not changed, so why has there been a change in the performance rating to PR300? Are we again victims of marketing magic and the performance ratings can change at any time disclaimer that Cyrix/IBM has given us or is the rerating legitimate?

In my opinion, to ensure PR300 performance you must be using a mainboard with a chipset which supports linear burst mode in conjunction with a current Cyrix/IBM optimized mainboard BIOS.  Otherwise, you risk losing performance.

Alternative CPU camp vs. Intel Camp

Here are a couple of points that you should be aware of when comparing the offerings from the Alternative CPU camp against those from the Intel camp:

• The K6/300, Cyrix M II PR300, and IBM 6x86MX PR300 will not outperform a similarly equipped Pentium II 300. AMD, Cyrix, and IBM would like you to think otherwise. Don't let the PR ratings or Megahertz values fool you. The Intel Pentium II is the performance King.
• The K6/300, Cyrix M II PR300, and IBM 6x86MX PR300 will destroy an Intel Celeron on standard business applications because the Celeron lacks an L2 cache. However, the FPU performance of the Celeron still dominates and will send the K6/300, Cyrix M II PR300, and IBM 6x86MX PR300 back to the minor gaming leagues.
• The Intel Celeron is an attempt to move people off of the Socket-7 platform and on to the Slot-1 rather than a serious attempt to build a low cost CPU. Intel won't have a serious sub-$1,000 entry until Mendocino arrives in the fourth quarter of 1998. Mendocino is the code-name for a 300-MHz processor with 128 kilobytes of integrated Level 2 cache that is expected to join the Celeron brand.
• The Alternative CPU camp can’t seem to decide which Intel product poses the greatest threat - the Celeron or the Pentium II. AMD and Cyrix are renaming their products to better associate them with the Pentium II while simultaneously touting benchmarks that illustrate how their products destroy the Celeron.
• Alternative CPUs and Socket-7 mainboards can save you a serious amount of money, but be aware that you are investing in a technology that is close to the end of its lifecycle. Both AMD and Cyrix are looking to migrate from Socket-7 to Slot-1 in 1999. This will limit future upgrade options.

K6/300, Cyrix M II PR300, and IBM 6x86MX PR300 Performance

We know how the K6/300, Cyrix M II PR300, and IBM 6x86MX PR300 compare against the Intel Pentium II and Celeron, but how do they compare against each other?

Benchmarks were performed on the following system. Here is a summary features:

• IBM 6x86MX PR300 (75/225) and AMD K6/300 (66/300)
• M-Tech R581A Rev B2, Bios R581_0005
• 32 MB 10ns Macrotron SDRAM
• 1MB Synchronous Pipeline Burst Cache
• STB Velocy 128 4MB AGP (1,024-by-768, 16-bit color, small fonts, 75MHz refresh rate)
• Quantum Fireball ST 3.2 GB Ultra-ATA Harddrive
• Windows 95 OSR2.1 and Directx 5.0
• Default BIOS settings
• Linear-burst Mode Enabled for 6x86MX and M II tests

The recently released M-Tech R581A R581_0005 BIOS properly identified the 225MHz and 233MHz 6x86MX as a PR300.

The IBM 6x86MX PR300 (75/225) that I received for benchmarking/testing ran overclocked at 233MHz (66/233). As a result, I can tell you what to expect from the Cyrix M II PR300.

Cyrix decided to fall back to the standard 66MHz bus for their M II, whereas IBM decided to stick with the program and remain at 75MHz. The end result from a performance standpoint is moot. The performance of the Cyrix M II PR300 and IBM 6x86MX PR300 is virtually identical.

If your mainboard doesn't support a 75MHz bus speed, the Cyrix M II PR300 or K6/300 are your only choices. Also given the fact that majority of IBM parts are sold in Europe, the IBM 6x86MX PR300 is a more difficult option for consumers in the U.S.

The K6/300 wouldn't post in my revision B2 M-Tech Mustang R581A. From what I have been able to determine, M-Tech made a change in the early B2 revisions that takes care of the voltage supply problem that has been seen by some users like myself when operating the R581A in the low 2V core voltage range. Once again I have fallen victim to early technology. Supposedly, this problem has since been resolved.

The only other mainboard that I have in my possession that meets the K6/300 core voltage requirement and has a 4.5x clock multiplier is an AOpen AX5T-3.1. Unfortunately, this board is based on a TX chipset and doesn’t offer linear burst mode support. The BIOS is also very old. It is dated on 11/97.

Because the AOpen AX5T-3.1 doesn’t meet the criteria that I described above for optimal PR300 performance, I am not going to publish the AMD K6, Cyrix M II, and IBM 6x86MX benchmarks produced on my AOpen AX5T-3.1.  Instead, I am seeking a resolution to my R581A problem through my vendor so that I can generate a series of K6/300 benchmarks on the R581A and therefore guarantee an accurate comparison against the M II and 6x86MX.

Conclusions

The K6/300, Cyrix M II PR300, and IBM 6x86MX PR300 each have there own unique strengths and weaknesses. None can satisfy every user's needs. After determining how much you are willing to spend, it is very important to figure out what kinds of applications are most important to you. This information should help you make a buying decision. 

Here are the pros and cons of each CPU:

Cyrix M II PR300 / IBM 6x86MX PR300

Pros:

• You will mostly likely be able to save some cash and use your existing Socket-7 mainboard because the M II / 6x86MX PR300 uses a 2.9V core voltage. A 2.9V core voltage jumper setting is commonly found on existing socket-7 mainboards.
• The CPU offerings from Cyrix and IBM cost less and drop more frequently than the CPUs from AMD.
• Tremendous business and Internet application performance

Cons:

• Manufactured using an older 0.35 micron process.
• Floating point performance is dismal at best. Your friends with Intel processors will make fun of you. A fast 3D video card is essential if you plan on doing a serious amount of gaming with this CPU.
• You can't buy it. Cyrix/IBM have a well documented history of announcing a product with very little product in stock.

AMD K6 300

Pros:

• Manufactured using an advanced 0.25 micron process.
• The K6/300 generates little heat in part because of its low 2.2V core voltage.  The result is a CPU that is very overclockable.
• The best overall performing Alternative CPU.
• You can buy one now.

Cons:

• The K6/300 is currently very expensive. It currently retails for approximately $300.00. It is also very likely that you may have to buy a new mainboard because yours doesn't provide 2.2V core voltage support. You may be better off waiting until the 100MHz Super Socket-7 mainboards become generally available. These mainboards will provide you with a wider array of future CPU options.
• Floating point performance still sucks, but it is better than the M II and 6x86MX.