VOGONS


Cyrix MII-433GP Build

Topic actions

Reply 20 of 107, by feipoa

User metadata
Rank l33t++
Rank
l33t++

My WinChip-2A infact works with a 4.5x multiplier as well, unfortunately it didn't overclock well enough at 300 MHz to run Speedsys. 266 MHz seems OK though (75.9 x 3.5 and 66.6 x 4.0).

Attached is the SpeedSys result for 75.9x3.5 (266 MHz). Something doesn't seem quite right with the writing and moving speeds (Write Allocate disabled?). I used a VIA MVP3 but an ALi-V board also indicated similarly odd writing and moving speeds. Notice how SpeedSys doesn't have the correct Extern. clock, it indicates 66 MHz.

Attachments

Plan your life wisely, you'll be dead before you know it.

Reply 21 of 107, by feipoa

User metadata
Rank l33t++
Rank
l33t++

As of yet, I haven't heard anything from Jan about a BIOS for my Biostar socket 7 to support AMD K6-2/3/+ and Cyrix MII-4x CPUs, so this project my be on hold (again) until I hear back from him.

If I do not get a BIOS from Jan, then I will have to cut the comparison up between motherboars. The 50, 55, 60, and some 66 MHz CPUs will need to be on the TX board, while the others on a Super7. I haven't yet decided where to cut things off, perhaps at 266 MHz or 150 MHz? I'm open to suggestions. Due to the possible split-up, I've added Slot 1's up to 600 MHz, which coincides nicely with the AMD K6-III+/600 MHz, or K6-II+/600 MHz.

This set back is unfortunate as I've just begun narrowing down the benchmark programs.

The list has been updated for the maximum number of options.

Last edited by feipoa on 2012-03-05, 18:03. Edited 1 time in total.

Plan your life wisely, you'll be dead before you know it.

Reply 22 of 107, by Anonymous Coward

User metadata
Rank l33t
Rank
l33t

I like the power switch converter. At first I didn't understand the purpose, but I think there must be quite a few older systems where swapping out the power switch is not an option. Especially on the older desktops with the big red rocker switch.

"Will the highways on the internets become more few?" -Gee Dubya
V'Ger XT|Upgraded AT|Ultimate 386|Super VL/EISA 486|SMP VL/EISA Pentium

Reply 24 of 107, by sliderider

User metadata
Rank l33t++
Rank
l33t++
feipoa wrote:
As of yet, I haven't heard anything from Jan about a BIOS for my Biostar socket 7 to support AMD K6-2/3/+ and Cyrix MII-4x CPUs, […]
Show full quote

As of yet, I haven't heard anything from Jan about a BIOS for my Biostar socket 7 to support AMD K6-2/3/+ and Cyrix MII-4x CPUs, so this project my be on hold (again) until I hear back from him.

If I do not get a BIOS from Jan, then I will have to cut the comparison up between motherboars. The 50, 55, 60, and some 66 MHz CPUs will need to be on the TX board, while the others on a Super7. I haven't yet decided where to cut things off, perhaps at 266 MHz or 150 MHz? I'm open to suggestions. Due to the possible split-up, I've added Slot 1's up to 600 MHz, which coincides nicely with the AMD K6-III+/600 MHz, or K6-II+/600 MHz.

This set back is unfortunate as I've just begun narrowing down the benchmark programs.

The list has been updated for the maximum number of options.

Have you tried Badflash?
http://www.badflash.com/

Reply 25 of 107, by feipoa

User metadata
Rank l33t++
Rank
l33t++
sliderider wrote:

Have you tried Badflash?

No. Is there a method to isolate the problem to the BIOS itself without having another backup BIOS for verification? There's other components which can also go bad on a motherboard. So this Badflash company will send me a pre-programmed BIOS replacement chip for $30? I think the cost is too high for not knowing for certain if the issue is BIOS-related. The AZZA board's death did not occur during a flashing operation.

My Biostar 8500TTD board works fine up to 3.5X multipliers (4x for AMD K6). K6-2/3, plus, or MII-4x won't work though. I'll wait 2 weeks to hear from Jan for a BIOS update for these processors on this board. Does anyone know of anyone else who can add this type of functionality to BIOSes?

If I can't get an update, I've decided to cut the 686 comparison in half whereby CPUs up to 3.5x/266 MHz are tested on a 430TX board and the rest on a super7 board. I'll double-up on the 266 MHz cpus on the super7 board to maintain a sense of continuity. Unfortunately, my 430TX board has 512 KB of cache and the Super7 has 1024 KB of cache.

Plan your life wisely, you'll be dead before you know it.

Reply 26 of 107, by kool kitty89

User metadata
Rank Member
Rank
Member

A note for the Ultimate 686 Benchmark Comparison:

If possible, it would be good to include an FIC VA-502 board in the comparison, especially for testing the 6x86MX PR200 chips.
Given Red Hill's CPU and motherboard museums on this:
http://redhill.net.au/b/b-98.html

If you ran a 6x86MX-200, the VA-502 was simply the best board you could buy at any price. Those two made a magic combination. On the test bench with an MX-200, the 502 out-performed boards costing nearly twice as much, and on a VA-502 the 6x86MX-200 comfortably beat not just its direct competitors, the Pentium MMX and K6 Classic 200s, but the 233 MMX as well.

http://redhill.net.au/c/c-9.html#mx-200

Our favourite combination was an MX-200 and an FIC VA-502 motherboard. For some reason unknown to us, this particular pairing was phenomenally fast — much faster than the same chip in different boards, or the same board with different chips — had bulletproof reliability, and provided all of that for surprisingly little cost.

They don't mention use of the 502 with later Cyrix chips, and I have no idea how those might fare, but it would definitely be interesting to try. (either the performance wasn't notable for whatever reason, or 502s were scarce by the time later 6x86s/MIIs arrived -FIC discontinued them rather abruptly in favor of the 503, so use with later Cyrix chips would have been mainly limited to CPU upgrades)

Reply 27 of 107, by feipoa

User metadata
Rank l33t++
Rank
l33t++

My budget for this project is spent, but if you or anyone else has a FIC VA-502 they'd like to donate to the cause, I will gladly accept.

Alternately, if someone who has a FIC VA-502 wants to test a Cyrix 6x86MX chip using the ensuite of programs selected for this comparison, I can compare the results with what my Biostar MB-8500TTD and FIC PA-2013 get. If the results are similar, then we are justified in using the selected boards for this project.

I have attached the PDF test matrix for this project. The test matrix includes the entire configuration of CPUs to be tested, which boards will be used for which CPUs, which benchmark programs will be used, and spaces to input data. EDIT: test matrix has been removed.

The test options for Speedsys have been greatly expanded due to variability in min/max/ave yields. I'll have to decide which Speedsys test rows are relavent after having tested the CPUs.

I have no idea how long this project will take. With infrequent and casual testing, expect completion in 6 months to 2 years.

Last edited by feipoa on 2012-07-26, 04:11. Edited 7 times in total.

Plan your life wisely, you'll be dead before you know it.

Reply 28 of 107, by kool kitty89

User metadata
Rank Member
Rank
Member

OK, that makes sense. I'll definitely keep an eye out for a VA-502. For that matter, I'll keep a look out for the Iwill P55XPlus, which is what IBM used to benchmark the 3x83 MHz PR333. (and is also supposedly one of the few boards that are solidly reliable at 83 MHz)

One more note, and not so much a performance benchmark-related one, but it would be interesting to find out how tolerant (in general) Cyrix MI/MII chips are to FSB overclocks (not core overclocks). If a large number of 66/75/83 MHz rated parts actually run fine at 100 MHz (or possibly higher), it would be odd that Cyrix didn't release 100 MHz rated models sooner than the PR366. (obviously, it wouldn't make sense prior to the availability of super 7 boards in 1998, but having at least a 2x100 MHz part in mid 1998 should have been very significant)

The 5x86 certainly seems to handle FSB overclocking well.

Reply 29 of 107, by feipoa

User metadata
Rank l33t++
Rank
l33t++

Testing for long-term stability isn't so easy, nor is it a quick process, unfortunately. The best method I've found for this is to run the system routinely for everyday use. I may eventually retire my Cyrix 5x86-133/4x box from everyday use and use a MII @ 300 MHz, which would make for good stability conditions. If I do end up doing that (could be years away), I could easily test for FSB tolerance on the MII-400GP at 95 vs. 100 MHz. Seeing how the MII-400GP isn't too hard to find (I have found 3 working units), the knowledge of stability with a 100 MHz FSB would be good for retro system builders.

I am still not sure how stable the IBM 5x86c is at 60/66 MHz. It is in my back-up rig, which doesn't see a whole lot of action. I am currently running it a 66x2 (133 MHz) and 3.70 V. I do get hang-ups about 1 in 10 times when I play Subspace for an hour. The 1:10 ratio seems to be the same whether branch prediction is enabled or not. I do have plans to run the system slower at 60x2 (120 MHz), 50x2 (100 MHz), and 33x3 (100 MHz) to determine if the occasional hang-ups are due to the CPU's PLL multiplier FSB limit, the total internal CPU speed, or the chipset's/RAM's inability to cope with the high speed. Unfortunately, no matter how much juicy truth I uncover about these little speed demons, it won't bring the chips back into production.

My spare time efforts are now focused on the 686 benchmark comparison. I just re-uploaded the test matrix, I've completed data for 3 CPUs. My plan is for 1 CPU per day, no testing on the weekends (wife gets mad), and obviously no testing while out of town. I may squeeze in more CPUs/day if the conditions permit. The data is the easy part, after that is making sense of the data, normalising, making pretty charts, and documenting my findings in text.

Plan your life wisely, you'll be dead before you know it.

Reply 30 of 107, by feipoa

User metadata
Rank l33t++
Rank
l33t++

My RISE CPU has arrived, however the screen doesn't turn on when it is placed into an Intel 430TX board. Does anyone know if they will work in a 430TX board?

For the working RISE chips that LZF is selling, he mentioned that he tested them in a VIA MVP4 board. I didn't buy one that he particularly tested, but I suppose I should also test my RISE in a VIA MVP3 board before writing it off.

EDIT: The RISE 266 didn't work in any of my socket 7 motherboards: ALi-V, MVP3, 430TX. Either my RISE is a dud, or the chip only works in an MVP4, which is unlikely.

Plan your life wisely, you'll be dead before you know it.

Reply 31 of 107, by kool kitty89

User metadata
Rank Member
Rank
Member
feipoa wrote:

My RISE CPU has arrived, however the screen doesn't turn on when it is placed into an Intel 430TX board. Does anyone know if they will work in a 430TX board?

For the working RISE chips that LZF is selling, he mentioned that he tested them in a VIA MVP4 board. I didn't buy one that he particularly tested, but I suppose I should also test my RISE in a VIA MVP3 board before writing it off.

EDIT: The RISE 266 didn't work in any of my socket 7 motherboards: ALi-V, MVP3, 430TX. Either my RISE is a dud, or the chip only works in an MVP4, which is unlikely.

What settings are you using for the RISE? (the RISE 266 should be a 200 MHz part -rated for 100x2)

Reply 33 of 107, by sliderider

User metadata
Rank l33t++
Rank
l33t++

These boards were tested with the Rise chip

Chaintech 6BTM and Chaintech 5AGM2 mainboards;

in this review here, http://www.xbitlabs.com/articles/cpu/display/ … -mp6.html#sect0

so others besides the one the seller told you about should work as long as they supply the correct voltage.

Reply 34 of 107, by feipoa

User metadata
Rank l33t++
Rank
l33t++

I do not think the RISE is so motherboard fussy. It is supposed to be Pentium pin compatabile. Unfortunately, I do not have any of those boards.

Well, I told the seller I need a working piece, he refunded my money and asked to send the CPU back. When going to bid on the "working" RISE pieces, I find that I have been blocked from bidding. I am probably going to call him on this; seems awefully silly to block me on this trivial matter. This would be the first bad chip I have received from this seller.

Plan your life wisely, you'll be dead before you know it.

Reply 35 of 107, by sliderider

User metadata
Rank l33t++
Rank
l33t++
feipoa wrote:

I do not think the RISE is so motherboard fussy. It is supposed to be Pentium pin compatabile. Unfortunately, I do not have any of those boards.

Well, I told the seller I need a working piece, he refunded my money and asked to send the CPU back. When going to bid on the "working" RISE pieces, I find that I have been blocked from bidding. I am probably going to call him on this; seems awefully silly to block me on this trivial matter. This would be the first bad chip I have received from this seller.

This is why I don't like that particular seller. He does say in his listings "for collection only" in red letters, which to me does not imply any sort of warranty at all. You are buying it as a display piece and not as a working part, is what that says to me. He probably went through with the refund just to avoid getting a neg, but don't count on him removing you from the blocked bidder list.

Reply 36 of 107, by feipoa

User metadata
Rank l33t++
Rank
l33t++

He's been good to me in the past, perhaps he's been overwhelmed with grandeur lately. I hope he doesn't turn into the Chinese version of Wiredfor*.

To me, "collection only" does not mean "wall hanger."

Last edited by feipoa on 2012-08-01, 13:59. Edited 1 time in total.

Plan your life wisely, you'll be dead before you know it.

Reply 37 of 107, by feipoa

User metadata
Rank l33t++
Rank
l33t++

After speaking with this particular seller on the phone, he claims that eBay automatically puts a time limit on when a user can re-bid when a previous auction has been "cancelled." He said to try again in a few days, but I am still banned. He also doesn't reply to my inquiries, particularly my request for his return address in Chinese characters.

I'm now 22 CPUs into this benchmark comparison, which is about 1 CPU/day. I've added a few more socket 3 processors, just for comparison, for example, an Intel DX4-133 (2x66), AMD X5-133/160, POD83/100, and IBM 5x86-100/120/133. High FSBs used whenever possible. I've updated the test matrix, but I have a couple of questions from you guys about the benchmarks.

Do you know the dominant test types for the benchmarks listed below? I have encluded my best guesses, in order of most dominant. Options are ALU, FPU, GPU, RAM throughput (including L2 cache), and PCI BUS. Some of them are obvious, but for others I may need to look at the documentation. My guesses are based loosely on the data tabulated so far. I'd include any of the above noted features if they contribute to 25% or more of the total score.

Symantec Sysinfo v8.0
ALU

PiDOS
FPU

Landmark v2.0
ALU - Integer
FPU - Floating-point
PCI BUS - Video

Bytemark v2, 32-bit DOS
ALU - Numeric Sort
ALU - String Sort
ALU - Bitfield
ALU - FP Emulation
FPU - Fourier
ALU - Assignment
ALU - IDEA
ALU - Huffman
FPU - Neural Net
FPU - LU Decomposition
ALU - Integer Index (% of Pentium 90)
FPU - Floating-point Index (% of P90)

Roy Longbottom Dhrystone v1.1
DHRY1OD (VAX MIPS Rating)
ALU

Roy Longbottom Whetstone
FPU - WHETCOD, MWIPS
FPU - N1, Floating Point (MFLOPS)
FPU - N2, Floating Point (MFLOPS)
ALU, FPU - N3, If Then Else (MOPS)
FPU - N4, Fixed Point (MOPS)
FPU - N5, Sine, Cosine (MOPS)
FPU - N6, Floating Point (MFLOPS)
FPU - N7, Assignments (MOPS)
FPU - N8, Exp, Sqrt, etc (MOPS)

Speedsys v4.78
ALU - Total Score
PCI BUS - Video Memory Bandwidth
RAM - System Memory Bandwidth

ALU - L1 Cache
RAM - L2 Cache
RAM - Memory Throughput

Cachechk v7.0
ALU - L1 Cache
RAM - L2 Cache
RAM - Memory

3Dbench v1.0c
ALU, GPU

Doom v1.9s timedemo1
ALU, GPU

Pcpbench v1.04
ALU, GPU

Quake v1.06 timedemo1
FPU, GPU

SuperPi v1.1
FPU

Ziff-Davis Winbench96
ALU, RAM - CPUMark32 v1.0
ALU, GPU - Graphics WinMark v1.0

Ziff-Davis Winbench99
ALU, RAM - CPUMark99 V1.0
FPU - FPU WinMark99 v1.1

Ziff-Davis 3D Winbench97 v1.0
FPU, GPU, RAM

WinTune98
ALU - Integer (MIPS)
FPU - Floating Point (MFLOPS)
GPU, MMX - Video 2D (Mpixels/s)
GPU, MMX - Direct3D (Mpixels/s)
GPU, MMX - OpenGL (Mpixels/s)
RAM - Memory

Sandra99
ALU - Dhrystone (MIPS)
FPU - Whetstone (MFLOPS)
ALU, MMX - MMX/3DNow: ALU Integer (it/s)
ALU, MMX - MMX/3DNow: FPU Floating Point (it/s)
RAM - Memory: ALU Bandwidth (MB/s)
RAM - Memory: FPU Bandwidth (MB/s)

PassMark v4.0
GPU - 2D Graphics Mark
RAM - Memory Mark
ALU, FPU - Math Mark
ALU, FPU - Math Max MFLOPS
MMX ALU, FPU - MMX Mark

ALU - Integer Addition
ALU - Integer Subtraction
ALU - Integer Multiplication
ALU - Integer Division
FPU - Floating-point Addition
FPU - Floating-point Subtraction
FPU - Floating-point Multiplication
FPU - Floating-point Division

3DMark99Max
MMX FPU, GPU - Graphics 3DMarks
MMX FPU - CPU 3DMarks

Final Reality v1.01
GPU, FPU - Direct3D - Overall Marks
FPU - Software - Overall Marks
GPU, FPU - Direct3D - 2D
GPU, FPU - Direct3D - 3D
GPU, FPU - Direct3D - Bus
FPU - Software - 3D

MDK Performance - Direct3D
GPU, FPU

MDK Performance - Software
FPU

Quake2 Timedemo1 - OpenGL
GPU, FPU

Quake2 Timedemo1 - Software
FPU

Last edited by feipoa on 2012-03-27, 06:04. Edited 1 time in total.

Plan your life wisely, you'll be dead before you know it.

Reply 38 of 107, by kool kitty89

User metadata
Rank Member
Rank
Member
feipoa wrote:

Testing for long-term stability isn't so easy, nor is it a quick process, unfortunately. The best method I've found for this is to run the system routinely for everyday use. I may eventually retire my Cyrix 5x86-133/4x box from everyday use and use a MII @ 300 MHz, which would make for good stability conditions. If I do end up doing that (could be years away), I could easily test for FSB tolerance on the MII-400GP at 95 vs. 100 MHz. Seeing how the MII-400GP isn't too hard to find (I have found 3 working units), the knowledge of stability with a 100 MHz FSB would be good for retro system builders.

I was thinking about this again (in the context of my overclocked bus suggestion).

Short of actually doing comprehensive testing for stability and performance, couldn't you at least do some quick preliminary tests for basic stability (ie will with system at least boot and maybe run some basic tests/benchmarks -like speedsys, etc), at least for curiosity purposes. (I've seen almost no reference to attempts to run older 6x86 and MX processors at 100 MHz FSB -plenty of mention of there being only a slim margin for core speed OC, but little to no mention of bus OC for the older pre-2.2v parts)

Reply 39 of 107, by feipoa

User metadata
Rank l33t++
Rank
l33t++
kool kitty89 wrote:

couldn't you at least do some quick preliminary tests for basic stability (ie will with system at least boot and maybe run some basic tests/benchmarks -like speedsys, etc)

You could, but it wouldn't prove long-term stability. I have found that CPUs can pass all the benchmarks, pass Windows98 installation, and still fail in the long-run. Have you been looking for a SS7 and a Cyrix MII? Perhaps you could run this test. I'm busy with the 686 benchmark comparison and don't want to go too far astray.

I'm going to try and finish up a Cyrix 6x86MX-200 MHz before heading off to bed. I have found that one of these older Cyrix 6x86MX's won't even run reliably at its rated speed. It was a Cyrix 6x86MX-233PR (75x2.5, 2.9V). It ran fine at 133 and 150 MHz for all the benchmarks, but started failing on some Windows benchmarks at 166 MHz. I had to replace it with a black-top IBM 6x86MX-233PR (66x3.0) to run the 166 and 200 MHz tests.

EDIT: Test matrix has been updated for 24 tested CPUs.

Plan your life wisely, you'll be dead before you know it.