My Cyrix 6x86 P200+ & M-Tech Mustang R534 Upgrade
My Cyrix 6x86 P200+ &
The upgrade is complete and I think that I started off the New Year right. Out with the old Cyrix P166+ and Tyan Tomcat I and in with the new Cyrix P200+ and M-Technology Mustang R534 .
I have been receiving a voluminous amount of email from readers who had questions about the Cyrix P200+. Since I did not have any hands-on experience with the P200+ and issues surrounding it, I had a difficult time answering questions and offering advice. As a result, I fired off an email to the folks at Cyrix asking them if they could help me out. They were gracious enough to send me a P200+ and a M-Technology Mustang R534 motherboard. I was ecstatic at the news because the Cyrix P200+ and the M-Technology Mustang R534 is currently one the hottest Cyrix combinations on the market.
The M-Technology Mustang-R534 system board is advertised as being specially designed for Cyrix/IBM 6x86 P200+. It is also Intel MMX, Cyrix M2, and AMD K6 ready.
Highlights:
There are several versions of this board out right now. I received the original R534. The original R534 has 3 PCI/5 ISA slots and a non-switching voltage regulator. The R534E has 4 PCI/4 ISA slots and a switching voltage regulator. The R534G has 3 PCI/5 ISA slots and a switching voltage regulator. Switching voltage regulator generate less heat and can regulate the voltage better than linear voltage regulators.
There is also one other significant difference between these boards. The R534 and R534E have a synchronous PCI bus whereas the R534G has an asynchronous PCI bus. This means that the PCI bus speed of the R534G is always running at 33MHz, but the R534/R534E runs the PCI bus at 1/2 the system clock.
Go to the M-Technology website to get a complete set of specifications.
I received the P200+ first. I opened it up and the first thing I did was turn over the cpu to determine the revision number. Doh! It was a revision 2.6. I was hoping to get my hands on a revision 3.7 or 6x86L.
The motherboard came a few weeks later. The M-Technology Mustang R534 motherboard is one of four P200+ motherboards currently certified by Cyrix. The first thing that I noticed about the motherboard was the size of the heatsink on the voltage regulator. It is massive. I think that it is made of armor instead of aluminum. I promise you will not have any heat problems attributable to the voltage regulator on this motherboard.
The motherboard documentation is average. It is on par with most motherboard manuals that I have seen. The installation information is excellent, but the bios information does not include descriptions for all of the bios options that are available. Check out http://www.mtiusa.com/mupdate.htm to get manual updates. This is where the jumper settings for the 83Mhz bus speed is located.
I set aside an evening to do the upgrade, but it ended up taking much longer. The first thing I did was the backup the harddrive. I took out the Tyan board and I put in the M-Tech board. I reconnected everything, double checked all of the connections to make sure everything was hooked up correctly, and I turned on the power. I made a few bios modifications to get the harddrive and on-board peripherals setup, I rebooted the machine, and Windows 95 was up and running.
Once I knew everything was working correctly, I updated the bios to the latest revision (H_7) that was available on the M-Tech ftp site. Revision E_7 came installed on the board.
I reformatted my C:\ drive and I reinstalled Windows 95 OSR2. It is always a good idea to reinstall Windows 95 after upgrading a motherboard. This will ensure that you eliminate any potential conflicts that could arise because of conflicting registry information. Everything was identified and configured correctly by Windows 95 expect for my Creative Labs 2x Plug & Play CD-ROM. You can tell I play a lot of games.
As a result, I did what any good Cyrix owner would do after they had a system problem. I sent an email to Cyrix tech support informing them that that there was a problem with the Cyrix processor that was causing my plug & play CD-ROM not to work with Windows 95. Wrong!
It never ceases to amaze me that people continually point fingers at the Cyrix processor any time they have a system problem. Intel users do not do this. I have been working with computer hardware for a very long time. I did computer technical support throughout college and rarely is a cpu to blame for system malfunctions unless they overheat. CPUs are rarely defective because they are extensively performance tested before they are ever shipped. If you purchase a Cyrix certified motherboard, use the Cyrix supplied heatsink/fan or something equivalent, and provide adequate cooling conditions you can virtually eliminate any possibility that your system problems are as a result of the cpu or the voltage regulator overheating. This statement is only true, however, for processors that aren’t overclocked.
I had a Cyrix tech support guy tell me people blame the Cyrix processor for every system problem conceivable.
My plug-n-play CD-ROM was causing my system woes. Windows 95 recognized the device and said that it was working correctly, but any time I tried to access it caused Windows 95 to lockup. I ended up having to install the real-mode drivers for the CD-ROM and run it in compatibility mode before I could access it. I tried many different things, but I was never able to get Windows 95 to natively recognize the device.
I decided that I would call M-Technology and put their technical support to the test. Maybe someone there could suggest something else to try. I dialed the the number and I spoke to a real person. I asked for support and I was immediately forwarded to another real person. No holding and no voice mail. I couldn’t believe it. I did this more than once at different times of the day to make sure that it wasn't an isolated incidence. That is my kind of technical support. Bravo M-Technology! They also have made a M-Tech FAQ available.
I spoke to a super nice technical support guy. He suggested changing a few property settings for the CD-ROM, but it did not resolve the problem.
In my
opinion, the BIOS is still very immature because it is a new board based on the new SIS
5511 chipset. In the short time that I have owned the board there have been two (H_7 &
I_7) bios upgrades released.
M-Technology was having to release bios upgrades every few weeks to fix problems after the board was initially released. Here is a case study. Using bios version E_7, Windows 95 would not recognize my plug-n-play CD-ROM. After I upgraded to bios version H_7, Windows 95 recognized my CD-ROM, but Windows 95 would lock up anytime I tried to access it. After I upgraded to bios version I_7, the CD-ROM works perfectly and I dumped the real-mode drivers that I was having to use previously. CD-ROM problem solved.
Thankfully, the frequent BIOS updates have finally subsided. BIOS updates are now very rare.
Confident that I could proceed forward, the next thing I did was start tweaking the BIOS to increase performance. In case you do not know, the default memory and cache settings on most motherboards, in terms of performance, are some of the worst settings that you can choose. The default options exist to guarantee that the board will function properly regardless of the speed and type of memory and cache that you have. It sacrifices performance for functionality. I always immediately disable the auto memory setting.
The performance of the motherboard is highly depending on the memory timing and cache settings that you can adjust in your BIOS setup. The memory and cache settings on my Award BIOS are under the chipset menu option. There are also many memory options related to timing that I would of liked to have seen, but were not available. Once the bios setup menu has been accessed, you want to look for options that can be figured out using common sense. Options such as internal cache - enabled or system speed - fast. Other things to look for are wait states. In general, for optimal performance you should keep most of the values as low as possible. If you select a value that is too low you can cause a system crashes/hang or your system won’t boot at all. Simply readjust your setting and everything will run as before. The best way to find out how aggressive you can be with your memory settings is to start by reducing one option at a time, note the effect, and if it doesn't cause any systems problems then change another.
I was able to set all but one setting to its lowest value with my 60ns Micron EDO RAM.
Most people do not know what each bios setting means. That is perfectly understandable considering they are not always intuitive. Check out the Bios Survival Guide for help or the System Optimization Page for more BIOS resources. Tom's Hardware Guide also has a bios guide available.
The M-Technology Mustang R534, which is based on the SIS 5571 chipset, is officially capable of and has a documented 83MHz bus setting. Did I try it? You bet I did. What would the P rating be for a Cyrix 6x86 166MHz (clock-doubled 83MHz)?
Unfortunately, my system repeatedly locked up. Honestly, I have read only a handful of posts and emails from individuals who have been able to overclock a Cyrix 6x86 P200+. Intel owners are having much better success with the 83 MHz bus than Cyrix owners. M-Tech confirmed that this was true when I specifically asked them about it.
Many Cyrix owners that have tried the 83MHz bus speed and were unsuccessful placed the blame on their slow 60ns EDO memory. They also seem to think that SDRAM will solve the problem. I am also left wondering if my memory isn't fast enough. However, I doubt that this would make a difference in my case. My P200+ simply did not like being overclocked. This is not surprising considering my P200+ is an older revision 2.6.
Please email me if you have a success story so I can pass it along.
Benchmarks were performed on the following system. Here is a summary features:
I prefer to use the Winbench 97 and WinStone 97 benchmarks developed by the Ziff-Davis Benchmark Operation (ZDBOp). These benchmarks offer a variety of tests based on today's most popular business applications. The goal of these tests is to measure the level of performance typical users will achieve while performing common operations with leading business applications. Real world performance is something that I can and like to identify with.
I will briefly describe the purpose of each benchmark before I give the results. It is important to gain an understanding of what the scores represent otherwise they are just meaningless numbers.
WinBench 97 offers two processor subsystem performance tests: the CPUmark32 and CPUmark16. CPUmark scores reflect the speed of a PC's processor subsystem- which includes the CPU, secondary cache, and system RAM. The CPUmark32 is a 32-bit test that reflects the processor usage of 32-bit applications; CPUmark16 mirrors the processor behavior of 16-bit applications. Focus on the CPUmark32 because most applications today are 32-bit.
I ran the CPUmark tests with linear burst mode off and off to see if linear burst mode really does increase overall system performance. I also ran the the tests after running Mikael Johansson's very popular 6x86 optimizer to see if it increased system performance. Many readers have stated that this optimizer definitely benefited them.
| Performance Comparisons | CPUmark16 | CPUmark32 |
|---|---|---|
| Linear Burst Mode - Off | 356 | 376 |
| Linear Burst Mode - On | 370 | 380 |
| Linear Burst Mode - On & 6x86opt074 | 370 | 389 |
Linear burst mode does increase system performance and the 6x86 optimizer definitely helps my system deliver a higher CPUmark 32 score. This optimizer can be found on my files and utilities page. Please read why these optimizers exist before deciding to use one.
Next I ran the Winstone 96 and Winstone 32 benchmarks that I used to benchmark my 6x86 P166+ and Tyan Tomcat I. I wanted to get to an idea of how much faster the scores from my new system are compared to my old ones. I reproduced the testing environment and nothing has changed between the two setups except for the processor and motherboard.
| Performance Comparisons | Winstone 96 | Winstone 32 |
|---|---|---|
| 6x86 P166+ & Tyan Tomcat I | 71.2 | 106.1 |
| 6x86 P200+ & M-Tech Mustang R534 | 79.3 | 125.9 |
The Winstone scores were substantially higher. Benchmarks that I ran show that this is attributable to the much faster Cyrix 6x86 P200+ rather than the SIS 5571 chipset.
I also ran the Business Winstone 97 benchmark. Business Winstone 97 gauges the performance of an entire PC. Business Winstone 97 measures 32-bit system performance by running actual 32-bit mainstream applications under Windows 95 or Windows NT.
Business Winstone 97 scales the overall score to the overall score of the base machine, which by definition is set to 10.0. The base machine for the business benchmarks is a 66MHz 80486-based Gateway PC with 16MB of RAM, running at 1,024x768 resolution and 16-bit color.
On December 3, 1996, PC Magazine presented the ultimate PC buyers' guide! They have reviews of 51 200-MHz–class PCs based on Intel's Pentium and Cyrix's 6x86 processor. They determined the average Business Winstone 97 scores for a Intel P200 to be 34 and a Cyrix 6x86 P200+ to be 35. My system is between the averages at 34.6.
Note: In order to accurately compare my Business Winstone 97 scores to those captured by PC Magazine, I configured my system as they did when the tests were performed. All systems tested were configured with 1.6GB or larger hard disks, 16MB of RAM, 1,024-by-768 resolution with 64K colors, small fonts, and a refresh rate of 75. If you run the Winstone 97 with a system configured differently then you can't compare your scores with these.
| Performance Comparisons | Business Winstone 97 |
|---|---|
| The average Intel P200 tested by PC Magazine | 34 |
| Cyrix 6x86 P200+ & M-Tech Mustang R534 | 34.6 |
| The average 6x86 P200+ tested by PC Magazine | 35 |
My system ranked at the lower end of the pack of 7 Cyrix 6x86 P200+ systems that they tested in the review.
The R534 has been having many problems with SDRAM. This problem has been reported with many different speeds and manufacturers of SDRAM. Revision H_7 was released to alleviate some of them, but unfortunately the problems still persist. Overall, EDO is much more stable on this board. Until the problem is resolved by M-Tech hopefully through a BIOS upgrade, I wouldn't rush out to buy SDRAM to replace your EDO. SDRAM is more expensive and is currently not producing significant performance increases when used with the R534.
John Howland has graciously offered to provide me with benchmarks and some of his experiences with this board and SDRAM. John sells this board. He is providing these results for me because I do not own any SDRAM. Here are his comments.
Bryan,
Below are the benchmarks for EDO & SDRAM on a M-Tech R534 along with some comments. Still doing some "playing around" with 83MHz
Benchmarks were performed on the following system. Here is a summary features:
- BIOS H_7
- IBM 6x86 P200+ 3.3v (Cyrix equivalent 2.7 ~ 3.7)
- 16MB memory
- WD 31600 HD
- STB Lightspeed 128
- No optimization programs used
Cachechk 6 60ns EDO 10ns SDRAM L1 627.8 627.9 L2 279.1 279.1 Main 123.5 146.2 Read 68 57 Write 132 101 (Following @ 1024x768-64K 60Hz)
Winbench 97 60ns EDO 10ns SDRAM CPUmark16 365 369 CPUmark32 371 381
Performance Comparison 60ns EDO 10ns SDRAM Business Winstone 97 31.7 31.2 SDRAM - Does it work & is it worth it on a M-Tech R534?
Yes it does work, I've been able to run SDRAM with all BIOS versions, but had to turn LinBurst = Off until recently. But you need very high quality SDRAM with a minimum speed of 10ns to do it, 6 - 8ns would be even better - if you can find it. According to M-Tech, you should have SDRAM that matches the speed of your L2 cache for maximum speed, I've been forced to run all my test with "SDRAM WR Retrieve Rate" set at X-2-2-2, with faster SDRAM X-1-1-1 could be used and the performance gain would show. I've had no stability problems and have been able to Boot @ 83MHz with SDRAM & run for a short time, but not with EDO (I've found after much testing that it's the CPU that crashes the system, not the SDRAM). As you can see above, SDRAM does have a faster memory speed, but it doesn't show in application benchmarks. Is it worth it - Not for overall system speed, but if you plan to run @ 83MHz it should help with stability. My advice - wait for the speed to come down (along with the price) before buying, unless you just want the latest & best now. One of the (if not the only) major complaints I've received from customers of the R534 is with memory - a lot of 60ns EDO just can't quite handle 75MHz, 10ns SDRAM should solve that problem and give you the option of running @ 83MHz in the future. I haven't checked with M-Tech as to why yet, but I've had no problems using "Read Prefetch Memory RD" with SDRAM - they do recommend it be disabled (slightly faster when enabled). Best SDRAM I've found is Kingston with 10ns SEC chips - expensive, but worth it.
Couple of notes on the R534 - Rear Heatsink can get very hot (I've seen almost 200 deg. F), it's normal & built to handle it (an extra case fan can drop that by 40 deg F). It's very sensitive to memory quality at higher speeds (even 66MHz). Is not the fastest (I'd put it at average) P200+ MoBo, but very stable. With the earlier BIOS I'd seen no advantage to using an optimization program.
Thanks John for your comments and your opinion. I appreciate your hard work.
There is a mailing list setup by Dennis Brenner dedicated to the Mustang R534 motherboard. This list is a great way to share information, post questions, and help solve problems. M-Technology support representatives also hang out here.
To subscribe, send email to the following address listproc@hallux.medschool.hscbklyn.edu with the following command in the body of the message:
subscribe r534 email_address
for the regular subscription or
subscribe r534-digest email_address
for digest mode (2 msgs per day)
1. The 6x86 P200+ is definitely faster and my system is much more responsive than the one based on the P166+. It is up to you to decide if the extra increase in performance is worth the additional cost. Feel comfort in knowing that there is a noticeable difference in speed between the two if you decide to fork out the extra cash.
2. This motherboard is a great choice for those of you who are wanting a 6x86 P200+ or the possibility of upgrading to it or the M2 in the future. Also rest assured that there will not be heat issues associated with the voltage regulator on this motherboard.
3. The Mustang R534 is truly a Cyrix optimized board. I really like the official support of linear burst mode, SDRAM, and the 83MHz bus. This is a great combination. Linear burst mode does increase 6x86 performance over those boards that do not offer it. All of you overclockers will be foaming at the mouth if you can get a hold of a 6x86 P200+ that you can overclock to 83MHz.
4. Certified by Cyrix.
5. Excellent technical support from the manufacturer. This is rare in the motherboard world.
1. SDRAM support is one of the primary reasons that so much interest has been generated around this board. To bad it currently doesn't work flawlessly with all SDRAM. This is a big deal and a big negative. I recommend using EDO.
2. There are many bios options that I would of like to have seen but were not available. What good is fast EDO memory if your BIOS won't allow you to take full advantage of it. If you are like me and you want to tweak everything then you may be disappointed.
3. While the SiS chipset allows for the PCI bus to run asynchronously (not tied directly to the system clock), M-Tech has implemented it synchronously on the R534 boards. This has been confirmed with M-Tech. This means that contrary to popular belief (which I believed as well), the PCI bus is NOT locked at 33 MHz, but instead runs at 1/2 the system clock. As a result, you are overclocking the PCI bus at 75MHz and 83MHz bus speeds. This has been an issue with most overclocked Pentium boards because of the Pentium architecture's synchronous operation with the PCI bus.
Overall, with the current bios the Cyrix P200+ and M-Technology motherboard is a good, but not a great performer. I am sure that there are other Apollo VIA-x chipset and VLSI Lynx based boards that will run the Cyrix 6x86 P200+ as fast or faster than the M-Tech R534. Intel chipsets outperform them all, but Intel does not officially support a 75MHz system bus. The true strength of this motherboard lies in its feature set rather than performance.
Feedback is welcome as always. Please send comments. Contact my sponsor Real World Technologies if you are interested in this motherboard. They have it available.
