General InformationGeneral InformationQ. Why did Cyrix raise the 6x86MX core voltage from 2.8V to 2.9V?
A. Yields my friends its all about yields in the microprocessor industry. It seems that Cyrix and their manufacturing partner IBM are definitely having problems with the 6x86MX in this area. The 6x86MX partners can't produce enough PR233 parts. An indication of this is that they have decided to follow AMD and raise the core voltage of the 6x86MX from 2.8V to 2.9V. A higher core voltage will help them obtain higher yields. The yield does go up with a .1V move of the specification. A rule of thumb is that the higher the voltage the faster the part. A higher voltage simply makes the transistors that make-up the CPU operate faster. This is true until you drive the part to a voltage where the process technology will be pushed beyond its capabilities. Then you have a reliability problem.
Higher yields means that Cyrix/IBM can get more faster parts per silicon wafer. Faster parts can be sold for more money. As a result, higher yields result in a more profitable Cyrix and IBM.
The 6x86MX is built on a two volt process technology that will run fine at 2.9V. The voltage regulators on motherboards are designed to the specifications of the CPUs that the motherboard is designed for. The issue is does the regulator on your motherboard put you in the 2.7-2.8V range instead of the 2.8-2.9V range? If it does, then you would be out of 6x86MX specifications. The 6x86MX data sheet specifications printed by Cyrix say that the 6x86MX core is rated to run between 2.8V and 3.0V.
What does this mean for you and me? Is it a good thing? It is a mixed bag in my opinion. It is a good thing that these chips will be available sooner rather than later, but it will also result in a more limited motherboard selection for the 6x86MX because a lot of motherboards don't support a 2.9V core voltage. The majority of new motherboards do because the AMD K6 has been out for some time, however, most older P55C compatible motherboards do not.
I called Cyrix technical support and they stated that Cyrix guarantees that you can run a 2.9V part at 2.8V and there will not be a problem because the part is in within specifications. This means that the 6x86MX will run fine on a motherboard with a 2.8V core setting. The motherboard Cyrix used for its 6x86MX benchmarks is a M-Tech Mustang R534G. The M-Tech Mustang R534G does not have a 2.9V core setting.
I would like to thank Cyrix and IBM for making this change. It is precisely these type of things that confuse consumers and keep people coming back to my website for help. Cyrix and IBM didn't announce this change. They quietly changed the graphic on their official 6x86MX FAQ to reflect the new core voltage requirement instead. They never have been a talkative bunch. I guess I will have to talk for them.
Q. Has Cyrix made changes to the 6x86MX PR rating since its initial release?
A. Yes, Cyrix has made modifications to the original 6x86MX PR rating that was implemented with the introduction of the 6x86MX on May 30, 1997. Cyrix is planning on adding parts at other clock frequencies and multipliers, but most of the initial clock frequencies and multipliers will be also shipped. As of July 24, 1997, the Cyrix 6x86MX PR166 is now a 133MHz (66MHz x 2) instead of a 150MHz (60MHz x 2.5) as originally introduced. There is also now a 6x86MX PR200 at 150MHz (75MHz x 2) in addition to the PR200 at 166MHz (66MHz x 2.5).
The purpose of the rerating is to increase the availability of higher speed parts and not to mask any problems with the chip. Basically, Cyrix was having to throw away a lot of perfectly good 133MHz parts that they could be selling as PR166s if they would sell them at 133MHz and clock double a 66MHz bus. My own benchmarks prove that this is valid. Cyrix could also call a 150MHz 6x86MX a PR200 instead of a PR166 if they clock doubled a 75MHz bus.
The end result of all of these changes will be a more profitable Cyrix because it increases their revenue per wafer and I suspect that this was a purely a financial decision.
Bus Speed Internal Clock Speed Clock Multiplier PR rating 66MHz 133MHz x 2 PR166 75MHz 150MHz x 2 PR200 66MHz 166MHz x 2.5 PR200 75MHz 188MHz x 2.5 PR233
Cyrix is working with motherboard manufacturers to get the changes implemented in future BIOS releases.
Cyrix is not going out on a limb by themselves on this one. Expect IBM to make modifications to their PR rating system also.
We have already seen a change in the 6x86MX core voltage and PR rating less than two months after it was announced. I told you that Cyrix/IBM sling-shot this product out the door. In my opinion, the end result will be consumers that remain confused by the 6x86MX product offering and the AMD K6 looking like the "safe" alternative processor for the risk averse regardless of the performance figures.
I wonder how long it will take for 6x86MX vendors to sort out this issue. I imagine that one would get lots of different variations of 6x86MX internal clock frequencies and corresponding PR ratings if you started calling around today. I have a hard enough time keeping this stuff straight myself and I keep up with it on a daily basis.
Q. Did Cyrix recall the 6x86MX?
A. Rumors that Cyrix recalled all of its existing inventory from its distributors started flaring in July 1997. Unfortunately, this rumor was true. However, the recall was not caused by a bug in the Cyrix 6x86MX microprocessor. Here is what I have been able to determine.
Cyrix speed qualified the first few batches of 6x86MX parts and shipped them to vendors. Some vendors observed problems with some of the chips at their given internal clock frequency and PR rating. Cyrix determined that certain test tolerances should be tightened and they were. Vendors were given the option of using the chips or sending them back to be retested to the new tolerances. Anyone who wanted to return the chips did so and new chips tested to the new tighter test tolerances were immediately shipped to replace them. The chips in question have already been replaced. Supposedly the quantity of chips returned was small and the retesting simply resulted a few chips being downgraded. However, an official number has not and will never will be released.
Other rumors to dispel include a voltage jitter problem at 2.5 clock multiplier. There is no problem with the 6x86MX when using a 2.5 clock multiplier.
Distributors and OEMs may be a little angry and inconvenienced, but end users will be oblivious to it. In my mind this goes a long way to answer the timeless Cyrix vs. IBM quality control argument. It gives credence to the argument that Cyrix's quality control standards are a little on the loose side because they need the cash. Cyrix can't afford to throw away parts and may be willing to allow higher speed tolerance limits when qualifying their parts inorder to sell as many of the faster parts as possible.
For more information, check out my interview with the Brian McWilliams, Editor of PC World Online NewsRadio. I was contacted in regards to the Cyrix 6x86MX recall. Listen to the interview via RealAudio. You can also access the online print version.
Q. How do 6x86MX prices compare to K6, Pentium MMX, and Pentium II prices?
A. In August 1997, a price war began. Intel fired the first salvo and slashed the price of Pentium II and Pentium MMX processors by 10 to 50 percent. AMD promptly retaliated by cutting prices on its K6 chips over 40 percent and escalated the price war with Intel. AMD states that they will maintain about a 25 percent price advantage over Intel at all times. Cyrix, last but not least, was forced to respond. Cyrix 6x86MX prices essential mirror those of the AMD K6. Cyrix will also maintain about a 25 percent price advantage over Intel.
There are a few notables in all of this posturing that I would like to point out.
- The price war that we are witnessing today is much different than price wars of the past. Pricing pressures exist at both the high and low end of microprocessor market. Even if you prefer Intel processors you should be thankful that AMD and Cyrix are doing extremely well. They are forcing Intel to run scared and everyone benefits from the increased competition. We wouldn't be seeing cheaper and faster processors sooner rather than later if it weren't for Cyrix and AMD.
- The differences in price that existed between the AMD K6 and Cyrix 6x86MX when the 6x86MX was introduced no longer exist. Price is no longer a factor when choosing between AMD and Cyrix.
- Cyrix is the big looser in all of this. Cyrix 6x86MX PR233 parts are still not available in large quantities and their margins are already being compressed just a few months after its release.
- Alternatives processors become less appealing to many buyers every time Intel drastically cuts prices.
- Non-MMX CPUs are so cheap that I will be giving them away as party favors.
Q. Which 6x86MX flavor should I buy?
Deciding which 6x86MX processor to buy is once again more confusing than ever. There are currently two different clock multipliers and bus speeds for each PR rating in the 6x86MX lineup of processors.
First of all, in my opinion, I would buy a 6x86MX PR166 before I would buy a 6x86 Classic PR200. My benchmarks show that the 6x86MX PR166 outperforms the 6x86 PR200 by approximately 8.5%. Although the 6x86MX is based on 6x86 core technology, the 6x86MX wrings extra performance out of architectural tweaks designed to let it handle data more efficiently and an increase in the size of the cache memory. The L1 cache has been increased from current 6x86's 16KB to 64KB. You get better performance and compatibility with the next generation of MMX software for just a few dollars more.
If you have the option, choose the 6x86MX that runs at the highest clock frequency at any give PR rating. For example, I would buy the 6x86MX PR200 that runs at 166MHz (66MHz x 2.5) rather than 150MHz (75MHz x 2) because I have the option of running the 166MHz 6x86MX at (83MHz x 2) which would give me the equivalent of PR233 performance. I also have the option of running the 6x86MX PR166 that runs at 150MHz (60MHz x 2.5) at (75MHz x 2) to obtain 6x86MX PR200 performance. I can’t do this with the 6x86MX PR166 that runs at 133MHz (66MHz x 2). Unfortunately, Cyrix finally figured out that they were giving away this extra performance and they are now charging a premium for the faster part. The faster parts can also be much more difficult to find because there are fewer of them available.
Check out my IBM 6x86MX PR200 review to see a benchmark comparison of the 6x86MX PR200 at various internal clock speeds, bus speeds, and clock multipliers.
Q. Where can I find additional information about the 6x86MX?
A. Cyrix has released an official 6x86MX FAQ. It is not very comprehensive, but it is better than nothing for those of you who hunger for 6x86MX information. Check out my CPU & Mainboard page for reviews of the 6x86MX PR200 and 6x86MX PR233.
Q. Is a complete 6x86MX PR rating guide available?
Listed below is the PR-Rating Guide for the 6x86MX that I picked up from the Cyrix website. It is a complete listing of ALL Cyrix PR-ratings, whether parts associated with a particular PR-rating are in production or not.
6x86MX™ PR-Rating Table
PR Rating Bus
(MHz)CPU Core
(MHz)Clock
Multiplier6x86MX-PR133 50 100 2x 6x86MX-PR133 55 110 2x 6x86MX-PR150 60 120 2x 6x86MX-PR150 50 125 2.5x 6x86MX-PR166 66 133 2x 6x86MX-PR166 55 138 2.5x 6x86MX-PR166 50 150 3x 6x86MX-PR166 60 150 2.5x 6x86MX-PR200 75 150 2x 6x86MX-PR200 55 165 3x 6x86MX-PR200 66 166 2.5x 6x86MX-PR200 60 180 3x 6x86MX-PR233 75 188 2.5x 6x86MX-PR233 66 200 3x 6x86MX-PR233 83 166 2x 6x86MX-PR266 66 233 3.5x 6x86MX-PR266 75 225 3x 6x86MX-PR266 83 208 2.5x
