VOGONS


The World's Fastest 486

Topic actions

First post, by feipoa

User metadata
Rank l33t++
Rank
l33t++

The World's Fastest 486

I am curious what everyone would consider to be the world's best PCI-based 486 system. Included in the term 'system' would be models/types of,

1) Motherboard/Chipset/Cache
2) CPU
3) Memory
4) Harddrive/Controller
5) Graphics
6) Audio
7) NIC
8) Operating system
9) Other features
like USB, PS/2 mouse port, etc.

The desired application for such a system would be general, such as retro gaming mixed with office-oriented applications. Benchmarks screenshots have been included on page 3 of this thread.

The intent of the question is not so much as an individualised component comparison (like what's the best Graphics card), but more focused on the best combination of items 1-9 that are known to work well together, in the same system. One reason for this particularity is that some graphics cards, for example, don't work with all motherboard/chipsets, or some systems don't work with EDO RAM, or some motherboards don't work with all Cyrix 5x86 features, like linear burst mode.

If possible, try avoid generalisations like 'any motherboard with 4 PCI slots and 1024 KB cache.' I'm more interested in proven fast/stable hardware combinations. Assume the cost for all components is negligable.

I have enclosed the specs for my own system two 486 systems, but at this point, I am more intersted in what other people have succeeded at.

For my system mentioned below, it has been 8 years stable with a Cyrix 5x86-120 and, so far, 8 months stable with a Cyrix 5x86-133/4x. The features which might make this system more optimal would be 1024KB cache (in WB mode) to increase the cacheable limit to 128 MB RAM, an additional PCI slot for a USB card, and perhaps a non-UMC chipset that allowed for graphics cards better than a Matrox Millennium G200-16MB, while still maintaining all other features mentioned below. I have decided to stick with the Ultra2 SCSI controller over the tested SATA controller because I doubt the SATA would be any faster.

While the system works with EDO RAM, I find FPM more stable and allows for faster cache timings with 512 KB cache. Is an AMD X5-200 better if proven long-term stable? This is certainly arguable. Refer to the Ultimate 486 Benchmark Comparison. The system noted below is also stable in W2K sp4 with the SCSI host controller, which not many, if any other, UMC-based motherboards can claim. Biostar really seems to have gotten bus mastering figured out on this board. The M919 and HOT-433, both, seem to have issues with SCSI bus mastering in W2K.

A stable configuration of 66 MHz FSB and a 2X multiplier was recently reported. Such a system may be up to 18% faster than a 33 MHz x 4 system. Refer to this post. Also note that FP_FAST may not be entirely stable with certain graphic cards and GLQuake/Quake II, as noted here.

For more information on modifying your Biostar MB-8433UUD motherboard for CPUs in the 3.45 - 4.10 V range, refer to the PDF in this post.

The system is as follows,

1) Motherboard/Chipset/Cache
● DTK PKM-0033s
AWARD 4.50PG BIOS, 12/11/95
SiS 496 (9618 NV) / SiS 497 (9624 NU)
Modified for native PS/2 mouse support
33 MHz FSB system

1024 KB double-banked Cache, 10 ns TAG (W24512AK-10), 10 ns cache (8 pieces of IS61C1024-10N)
2-1-1 cache wait state and 0/0 for RAM (33 MHz FSB)

● Biostar MB8433-UUD v3.0 or 3.1
AWARD 4.51PG BIOS, 05/20/96
UMC 8881F (9633-EYT) / 8886BF (9631-FXO) Chipset
Socketed Dallas DS12887+ RTC
Modified to accept 1024K external cache
Modified with a variable voltage regulator
66 MHz FSB system

1024 KB double-banked Cache, 10 ns TAG (W241024AK-10), 10 ns cache (8 pieces of IS61C1024-10N)
3-2-2-2 cache wait state and 2/0 for RAM (66 MHz FSB)

2) CPU
● Cyrix 5x86-133 (4 x 33 MHz FSB)
● IBM 5x86c-133 (2 x 66 MHz FSB)

● RSTK_EN = 1
● BTB_EN = 0 (branch prediction is long-term stable in DOS; set BTB_EN = 1 and LOOP_EN=0 in DOS to improve performance)
● LOOP_EN = 1
● LSSER = 0 (optimal performance is when LSSER = 0)
● USE_WBAK = 1
● WT1 = 1 (for hardware stability)
● BWRT = 1
● LINBRST = 1
● FP_FAST = 1 (must be set to 0 with SiS 496/497 chipset)
● MEM_BYP = 1
● DTE_EN = 1
● IORT = 000

3) Memory
● Samsung 128 MB (2x64) CMOS Dynamic RAM w/Fast Page Mode and Parity, 60ns w/gold contacts K4F640411C-TC60 x 8 pieces per stick

A modified BIOS is attached which allows for setting the L2 cache into write-through mode)

4) Harddrive/Controller
● Ultra320 SCSI 146 GB Harddrive (ST3146707LW) - 320 MByte/sec (this is the boot/system/storage drive)
Adaptec 2940U2W, Ultra2-LVD, 40 MHz - 80 Mbyte/sec

5) Graphics
● PCI ATI Rage 128 VR, 32 MB SDRAM
250 MHz RAMDAC, 80 MHz Core, 120 MHz Memory
64-bit Core
Drivers do not function in WinNT 4.0 when a Cyrix 5x86 chip is used.

● PCI Matrox Millennium G200, 16 MB SDRAM
250 MHz RAMDAC, 84 MHz Core, 112 MHz Memory
128-bit Core (dual 64-bit unidirectional buses)
BIOS 3.3.30, Rev0.30 (11/21/00)

● PCI 3dfx Voodoo3 3000, 16 MB SDRAM
350 MHz RAMDAC, 166 MHz Core, 166 MHz Memory
128-bit Core
For use in SiS 496/497 chipsets only
For other fast/overkill graphics combinations on a 486, view this thread

6) Audio
● ISA AWE64 Gold, CT4390 w/28MB RAM
ROM Version 2.08, EMU8000 Sound Engine

7) NIC
● ISA 3Com 3c515-TX 10/100Base-TX
● PCI 3Com 3c905C-TX-M, 10/100Base-TX

Test Results (3c905C):

● Sending a 30 MB file from a PIII to the 486, whereby the PIII initiated the transfer: 1.54 MB/s
● Sending a 30 MB file from a PIII to the 486, whereby the 486 initiated the transfer: 2.37 MB/s
● Sending a 30 MB file from the 486 to the PIII, whereby the PIII initiated the transfer: 3.20 MB/s
● Sending a 30 MB file from the 486 to the PIII, whereby the 486 initiated the transfer: 1.37 MB/s

8) Operating system
● Windows NT 4.0 sp6a, Windows 2000 sp4, Windows 95c, Damn Small Linux (DSL)

9) Other features
● Complete stability in W2K sp4 with all noted hardware
● PS/2 Mouse Port
● External IDE CF-card Reader w/16GB 200X CF Card
(connected to the internal PIO-4 port or 50-pin narrow SCSI connector via AEC7720U - used for testing differing OS's)

● Floppy Drives - NEC 3.5" and 5.25"
● Harman/Kardon HK195 Speakers
● Monitor - NEC 17" AccuSync LCD72VX at 1280x1024 (white colour!)

● Battery Backup - APC Backup-UPS 500
● Laser Printer - Brother HL-5250DN w/NT4.0 drivers
● SCSI Scanner - HP Scanjet 7400c - SCSI (previously UMAX Astra 600S)
● Sony Optiarc SATA/SCSI DVD-ROM (DVD+-RW, CD+-RW, DVD-RAM) connected thru ACARD AEC7732U
● KVM w/audio - IOGEAR GCS614A MiniView, 4-port
● PSU - 250 Watt, unknown
● 2006-era white softkey keyboard
● Linksys WRT54G Router, Hardware Version 2, Software Version 4.21.1
● Linksys EtherFast 10/100 5-port workgroup Switch (EZXS55W)
● VIA Dual PCMCIA External Card Reader (16-bit ISA)

10) Other additions known to work well on newer UMC, PCI-based 486 systems
● ISA U.S. Robotics 5687 v.90 56K modem
● PS/2-to-USB mouse converter for modern mice
● ISA 3Com 3c515-TX, 10/100Base-TX (100Base mode non-duplex)
● PCI Intel Pro/100S 10/100Base-TX
● PCI 2-port USB Host Controller, SIIG Intek 21 (TKP2U022) OPTi 82C861
● PCI Creative DXR2 DVD/MPEG Decoder
● PCI SATA/ATA Host Controller, Promise SATA150 TX2plus (has NT4.0 drivers)
● ACARD AEC7732U Bridge Adapter, SATA (ATAPI DVD-ROM) to 50-pin Ultra SCSI
● ACARD AEC7720U (C1) Bridge Adapter, IDE/ATA (CD/DVD-ROM or harddrive) to 50-pin Ultra SCSI

Attachments

  • Filename
    Biostar-MB-8433UUD-BIOS_and_Manual.zip
    File size
    2.65 MiB
    Downloads
    658 downloads
    File comment
    Jumpers are on pages 17-20. BIOS settings pages 21-24. 1024K cache mod on pages 8-11. RTC mod pg 13. VRM mod pg 4.
    File license
    Fair use/fair dealing exception
  • Cache_and_RAM.jpg
    Filename
    Cache_and_RAM.jpg
    File size
    152.03 KiB
    Views
    52318 views
    File license
    Fair use/fair dealing exception
  • Cyrix_5x86-133.jpg
    Filename
    Cyrix_5x86-133.jpg
    File size
    228.49 KiB
    Views
    52317 views
    File license
    Fair use/fair dealing exception
  • Biostar_MB8433-UUD.jpg
    Filename
    Biostar_MB8433-UUD.jpg
    File size
    316.3 KiB
    Views
    52317 views
    File license
    Fair use/fair dealing exception
  • Twin_Towers.jpg
    Filename
    Twin_Towers.jpg
    File size
    111.34 KiB
    Views
    52317 views
    File license
    Fair use/fair dealing exception
Last edited by feipoa on 2016-06-12, 00:37. Edited 46 times in total.

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

Reply 1 of 742, by Tetrium

User metadata
Rank l33t++
Rank
l33t++

I remembered 1 thread on a German forum, someone had build a 486 with USB, SATA and a 200Mhz 5x86

Heres the thread: http://www.planet3dnow.de/vbulletin/showthread.php?t=388817
He messed up when he forgot to do his research and bought an Intel Overdrive instead of a Pentium Overdrive and spend hours getting XP to install.

If your German isn't as good, then translate using google

Whats missing in your collections?
My retro rigs (old topic)
Interesting Vogons threads (links to Vogonswiki)
Report spammers here!

Reply 3 of 742, by Tetrium

User metadata
Rank l33t++
Rank
l33t++

Works for me though

Alternative is to do a google search for: Mein 486er @ 200MHz,128MB RAM,SATA,USB 2.0

Whats missing in your collections?
My retro rigs (old topic)
Interesting Vogons threads (links to Vogonswiki)
Report spammers here!

Reply 4 of 742, by sliderider

User metadata
Rank l33t++
Rank
l33t++
Tetrium wrote:
I remembered 1 thread on a German forum, someone had build a 486 with USB, SATA and a 200Mhz 5x86 […]
Show full quote

I remembered 1 thread on a German forum, someone had build a 486 with USB, SATA and a 200Mhz 5x86

Heres the thread: http://www.planet3dnow.de/vbulletin/showthread.php?t=388817
He messed up when he forgot to do his research and bought an Intel Overdrive instead of a Pentium Overdrive and spend hours getting XP to install.

If your German isn't as good, then translate using google

How did he get XP to install to a 486 Overdrive or 5x86? XP requires a Pentium.Those guys who got it running at 8mhz couldn't get it to boot at all on a 486.

Reply 5 of 742, by Tetrium

User metadata
Rank l33t++
Rank
l33t++
sliderider wrote:
Tetrium wrote:
I remembered 1 thread on a German forum, someone had build a 486 with USB, SATA and a 200Mhz 5x86 […]
Show full quote

I remembered 1 thread on a German forum, someone had build a 486 with USB, SATA and a 200Mhz 5x86

Heres the thread: http://www.planet3dnow.de/vbulletin/showthread.php?t=388817
He messed up when he forgot to do his research and bought an Intel Overdrive instead of a Pentium Overdrive and spend hours getting XP to install.

If your German isn't as good, then translate using google

How did he get XP to install to a 486 Overdrive or 5x86? XP requires a Pentium.Those guys who got it running at 8mhz couldn't get it to boot at all on a 486.

He didn't 😜

He tried and kept failing, thinking he had bought a Pentium overdrive instead of the regular 486 Overdrive he ended up with.

Whats missing in your collections?
My retro rigs (old topic)
Interesting Vogons threads (links to Vogonswiki)
Report spammers here!

Reply 6 of 742, by sliderider

User metadata
Rank l33t++
Rank
l33t++
Tetrium wrote:
sliderider wrote:
Tetrium wrote:
I remembered 1 thread on a German forum, someone had build a 486 with USB, SATA and a 200Mhz 5x86 […]
Show full quote

I remembered 1 thread on a German forum, someone had build a 486 with USB, SATA and a 200Mhz 5x86

Heres the thread: http://www.planet3dnow.de/vbulletin/showthread.php?t=388817
He messed up when he forgot to do his research and bought an Intel Overdrive instead of a Pentium Overdrive and spend hours getting XP to install.

If your German isn't as good, then translate using google

How did he get XP to install to a 486 Overdrive or 5x86? XP requires a Pentium.Those guys who got it running at 8mhz couldn't get it to boot at all on a 486.

He didn't 😜

He tried and kept failing, thinking he had bought a Pentium overdrive instead of the regular 486 Overdrive he ended up with.

I am impressed that he managed to find a 5x86-160. Where the heck did he get that from? I thought 150 was as high as they went and that even that one was extremely rare. A 160 must have been nearly impossible to get, unless it's a remarked fake.

Reply 7 of 742, by noshutdown

User metadata
Rank Oldbie
Rank
Oldbie

some amd5x86-133 are capable of running at 200, although i believe running at 60*3 would be faster(if your mainboard and ram can cope with that).
well those cpus may be hard to find today, but some may have a lot of them left from last century...

Reply 8 of 742, by Tetrium

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

How did he get XP to install to a 486 Overdrive or 5x86? XP requires a Pentium.Those guys who got it running at 8mhz couldn't get it to boot at all on a 486.

He didn't 😜

He tried and kept failing, thinking he had bought a Pentium overdrive instead of the regular 486 Overdrive he ended up with.

I am impressed that he managed to find a 5x86-160. Where the heck did he get that from? I thought 150 was as high as they went and that even that one was extremely rare. A 160 must have been nearly impossible to get, unless it's a remarked fake.

It is mentioned in the wiki though, although wiki also mentions the Cyrix 5x86-133 probably doesn't exist at all (me an feipoa have some really good "shit" in the house!!! 😁 😁 😁 😜).

And AMD 5x86 seems an easier overclock so 200Mhz is possible (though no guarantees it'll reach 200Mhz stably).

And a shame he never took a pic of the AMD 160 though. My guess he bought it for an inflated price as he seemed to be willing to spend big time.
If you got the big money (or want to spend big money), theres a lot less limits to what one can purchase, especially if you want it yesterday 😜

Whats missing in your collections?
My retro rigs (old topic)
Interesting Vogons threads (links to Vogonswiki)
Report spammers here!

Reply 9 of 742, by feipoa

User metadata
Rank l33t++
Rank
l33t++

Tetrium, thank you for the post; it was, at times, very interesting, amusing, and even sadening to read. His test methodolgy may not have been all hammered out very well. In the end he sold his working mobo/AMD-ADZ-160-200 system (on eBay!), perhaps out of frusturation, and got slammed with an Intel DX4 Overdrive when he thought it was a geniune Pentium Overdrive.

He was plagued with compatability issues using a Voodoo4-4500-32MB RAM in Windows ME, Windows ME gave him a BSOD every hour, he couldn't get Win2000 working correctly, and was stubborn not to use NT4 due to the lack of sufficient USB support. He could never get his PCI SATA/ATA card to boot using it as the system root; he had to use the onboard CPU-hogging PIO4 as his system drive, which, he deduced, gave him terrible network performance on his ISA card, at 250 KB/s (FYI, I can routinely get around 900 KB/s on an ISA network card).

To summarise, this is what he ended up getting stable (but for how long?):

1) Motherboard/Chipset/Cache
SOYO SY-4SAW2
SIS Chipset
256 KB single-banked Cache

2) CPU
AMD X5-160ADZ overclocked to 200 MHz

3) Memory
112 MB (4x32) FPM (something broke and it reports 112 MB instead of 128 MB)

4) Harddrive/Controller
Onboard PIO-4 IDE, 8.4 GB

5) Graphics
PCI Elsa Winner 2000-8MB

6) Audio
Unknown

7) NIC
ISA 3Com Etherlink III

8) Operating system
Windows 98SE

9) Other features
PCI 4-USB Controller
USB Optical Mouse
NTFS Read/Write Support in 98SE
PCI SATA/ATA Controller - Via VT6421A (secondary, storage only)
2 TB SATA harddrive (secondary, storage only)

To critique this system:

1) I'd want to know the long-term stability of the AMD X5 @ 200 MHz

2) 256 KB cache is too low for the amount of RAM. Unfortunately the cache looks soldered, but difficult to tell. Maybe he can upgrade to 512 KB and put his cache in WT mode?

3) I'd swap out the RAID and put in a 10/100 network card. RAID in Win98SE for a data server? Ugh!

4) He toyed with the idea of using an ATI Rage 128 - 32MB but sold his system before any attempts were made. I'd have liked to be the eBay buyer of that system!

5) AMD X5-200, assuming the cache/memory wait states are the same as in the motherboard used in the Ultimate 486 Benchmark Comparison, he'd have an ALU/FPU/OVERALL score of 167/81.7/103.8, respectively. Pretty impressive, considering the Overall score beats the Cyrix 5x86-133 by 9%, although the FPU is still behind by 4%. For ALU, he would be the reining champion. Unfortunately, it is unlikely that the AMD X5-200 can be run at 2-1-2 cache and 0/0 RAM settings, so take anywhere from 1-15% off these numbers.

6) After this read, I'm motivated to update the Ultimate 486 Benchmark Comparison to include derived CPU data for AMD X5-180, AMD X5-200, and Cyrix 5x86-150.

My guess he bought it for an inflated price as he seemed to be willing to spend big time.

That sorta goes in line with my thought that there are a few retro geeks out there with deep pockets jacking up the eBay prices.

It is mentioned in the wiki though, although wiki also mentions the Cyrix 5x86-133 probably doesn't exist at all

I can assure everyone that they exist and work. I've run some long-term stability tests on mine at 3.7 V, and the system hasn't crashed -- its been going for 2 weeks straight. My big debate now is 64MB vs. 128MB RAM with 512KB WB cache. For few open windows, 64MB feels faster, but for a modest amount of use, 128MB RAM feels faster, esp. with the Cyrix 5x86-120 because of the faster memory FSB. Anyway, this is off-topic.

Any other "World's Fastest 486" contenders? I know Udam_U got an AMD X5-180 running on a HOT-433 and rg100 got an AMD X5-200 runnong on a Biostar MB8433-UUD v3.1, but I don't have the whole system picture. Udam_U got his Cyrix 5x86-100 running at 150 MHz, but again, no full sytem specs and benchmark evidence.

Last edited by feipoa on 2011-07-12, 20:59. Edited 3 times in total.

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

Reply 10 of 742, by nemesis

User metadata
Rank Member
Rank
Member

I'm kicking myself now for this, I let a Cyrix 5x86 133 get by me for an extremely low price because I looked it up and the site said that it was probably a fake. Sounds like there's a pretty good chance it was real now. But my question is, are they really overclockable? Or did I miss that somewhere? 😖
EDIT: Sorry for the unnecessary post, I just found the info I was looking for rigth in front of me.

Reply 11 of 742, by feipoa

User metadata
Rank l33t++
Rank
l33t++

To be honest, I haven't experimented much with the overclocking my Cyrix 5x86-133/4X out of fear of killing it. Please refer to the warnings in the Cyrix literature about heat and voltage for these processors! I did a quick test at 3.7 Volts and 160 MHz, but the screen stayed blank. I did a quick test with a 120 MHz Cyrix once for 150 MHz operation at 5V, and the screen turned on, but it wasn't stable. The Cyrix 5x86-100/4X will run at 133 MHz, but mine was not long-term stable. I forget what voltage I used, probably not high enough.

In reality, and for the really determined individual, you can probably OC a Cyrix 5x86 to 150 MHz, but you'd need to put in place a variable resistor on your motherboard's voltage regulation circuit to find the optimum voltage for which the CPU didn't overheat/crash. I'd buy the most expensive heatsink compound you could find for that extra 5% of cooling power, and a somewhat rare 3-clipped Socket 370 heatsink/fan. I say rare because I've only ever tested one 3-pronged clip style that will clip successfully onto a socket 3 ZIF socket. I have 2 now. Most will initially grab, but then slip off. An extra pivot point is needed in the clip. Unfortunately, most motherboards have capacitors and other SMD parts in the way, except for the MB8433-UUD.

You'd also need a batch of Cyrix 5x86 CPUs to determine which was the best overclocker. If I were to g uess, it would be one of the Cyrix 5x86-133/4X/3.6V CPUs. I have a 3.7V variant. But perhaps also a Cyrix 5x86-120/4X or 5x86-100/4X. You'd also need a motherboard with a manual FSB-to-PCI divider option, particularly 1/2 and 2/3.

EDIT: It turns out that the IBM 5x86c chips for the Thinkpad TP365E overclock better than expected.

IBM-5x86c-Thinkpad_TP365E_Mod_1.jpg
Filename
IBM-5x86c-Thinkpad_TP365E_Mod_1.jpg
File size
150.89 KiB
Views
44788 views
File license
Fair use/fair dealing exception
IBM-5x86c-Thinkpad_TP365E_Mod_2.jpg
Filename
IBM-5x86c-Thinkpad_TP365E_Mod_2.jpg
File size
171.91 KiB
Views
44788 views
File license
Fair use/fair dealing exception
IBM-5x86c-150-Speedsys_1.png
Filename
IBM-5x86c-150-Speedsys_1.png
File size
12.44 KiB
Views
44788 views
File license
Fair use/fair dealing exception
IBM-5x86c-150-Speedsys_2.png
Filename
IBM-5x86c-150-Speedsys_2.png
File size
9.92 KiB
Views
44788 views
File license
Fair use/fair dealing exception
Last edited by feipoa on 2014-09-20, 01:04. Edited 1 time in total.

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

Reply 12 of 742, by Tetrium

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

To be honest, I haven't experimented much with the overclocking my Cyrix 5x86-133/4X out of fear of killing it.

Actually, for that reason I haven't even bothered to test mine until I get the attic sorted out.

feipoa wrote:

If I were to g uess, it would be one of the Cyrix 5x86-133/4X/3.6V CPUs. I have a 3.7V variant. But perhaps also a Cyrix 5x86-120/4X or 5x86-100/4X. You'd also need a motherboard with a manual FSB-to-PCI divider option, particularly 1/2 and 2/3.

Afaik there were very few production runs of this CPU (the Cyrix-133) and initially had 3.6v printed on them. Later they raised it to 3.7v due to stability issues so my guess would be that (even though usually lower voltage means a better part) because the production run was so short and the 3.6v parts were actually made before the 3.7v part, overclockability (if they have any that is) would be the same.

Btw, mine is one of the earlier 3.6v parts (and I actually wanted a later one 🤣, sure backfired 😜)

Whats missing in your collections?
My retro rigs (old topic)
Interesting Vogons threads (links to Vogonswiki)
Report spammers here!

Reply 13 of 742, by feipoa

User metadata
Rank l33t++
Rank
l33t++

Added Speedsys images of the Cyrix 5x86-133/4x on the SiS496 chipset vs. the IBM 5x86c-133/2x on the UM8881 chpset.

Added images of the MB-8433UUD with 1024K cache modification.

Added an image of the DTK PKM-0033s with PS/2 mouse modification.

Cx5x86-133-2x-MB-8433UUD.png
Filename
Cx5x86-133-2x-MB-8433UUD.png
File size
12.96 KiB
Views
44788 views
File license
Fair use/fair dealing exception
Cx5x86-133-4x-DTK-PKM-0033s.png
Filename
Cx5x86-133-4x-DTK-PKM-0033s.png
File size
13.5 KiB
Views
44788 views
File license
Fair use/fair dealing exception
MB-8433UUD_1024K_Mod_1.jpg
Filename
MB-8433UUD_1024K_Mod_1.jpg
File size
84.27 KiB
Views
44788 views
File license
Fair use/fair dealing exception
MB-8433UUD_1024K_Mod_2.jpg
Filename
MB-8433UUD_1024K_Mod_2.jpg
File size
124.06 KiB
Views
44788 views
File license
Fair use/fair dealing exception
DTK_PKM-033S_with_PS2.jpg
Filename
DTK_PKM-033S_with_PS2.jpg
File size
280.69 KiB
Views
44788 views
File license
Fair use/fair dealing exception
Last edited by feipoa on 2014-09-20, 00:44. Edited 3 times in total.

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

Reply 14 of 742, by feipoa

User metadata
Rank l33t++
Rank
l33t++

Afaik there were very few production runs of this CPU (the Cyrix-133) and initially had 3.6v printed on them. Later they raised it to 3.7v due to stability issues so my guess would be that (even though usually lower voltage means a better part) because the production run was so short and the 3.6v parts were actually made before the 3.7v part, overclockability (if they have any that is) would be the same.

Refer to my follow-up post in cpu-world.com. That wasn't so the case, as I've found out. My Cyrix 5x86-133 is from 49th week, 1995 and it has 3.7 V written on it. From that massive eBay Cyrix 5x86-133 auction, cpushack mentioned that all the CPUs were from the 2-7th week of 1996, where some had 3.6 V and some 3.7 V.

It may be that Cyrix stamped the top after characterisation? Makes sense to me. Lets assume that a Cyrix 5x86-133 wasn't stable at 3.6V 133 MHz, but it was at 120 MHz, so they stamp it as Cyrix 5x86-120/4X, same for the Cyrix 5x86-100/4X, but for the 100/4X, they ensured it was stable at 3.45 Volts. Or perhaps they'd get a few samples made and later stamp on it how the engineering sample comes out prior to each production run?

I used to think the way you do, but it wouldn't make sense to stamp them 3.7 V, later 3.6 V, then 3.7 V, unless they had some really high hopes afterwards. If that were the case, the first one (mine), would read 3.6 V as well. Unless, of course, the characterised sample was the oddball good one stable at 3.6 V.

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

Reply 15 of 742, by Tetrium

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

Is an AMD X5-200 better if proven long-term stable? This is certainly arguable.

Sorry for not giving you any more 200Mhz systems, but I think there were simply very few ever made.

And on top of that, we only figured the PCI divider out for ourselves quite recently, basically a few months or so before you joined.

The few people who did manage to get it running at 200Mhz would possibly not admit to having stability problems, if they had them.
I really don't know which boards had proper bus mastering.
Could you tell when the Biostar was made?
It can read EDO, so it should be 1995 or newer.
Could you check the year+week code on the PCB? It's usually printed on the backside of the PCB and is VERY handy in figuring out when a board was probably produced ;D

Whats missing in your collections?
My retro rigs (old topic)
Interesting Vogons threads (links to Vogonswiki)
Report spammers here!

Reply 16 of 742, by Tetrium

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

Afaik there were very few production runs of this CPU (the Cyrix-133) and initially had 3.6v printed on them. Later they raised it to 3.7v due to stability issues so my guess would be that (even though usually lower voltage means a better part) because the production run was so short and the 3.6v parts were actually made before the 3.7v part, overclockability (if they have any that is) would be the same.

Refer to my follow-up post in cpu-world.com. That wasn't so the case, as I've found out. My Cyrix 5x86-133 is from 49th week, 1995 and it has 3.7 V written on it. From that massive eBay Cyrix 5x86-133 auction, cpushack mentioned that all the CPUs were from the 2-7th week of 1996, where some had 3.6 V and some 3.7 V.

It may be that Cyrix stamped the top after characterisation? Makes sense to me. Lets assume that a Cyrix 5x86-133 wasn't stable at 3.6V 133 MHz, but it was at 120 MHz, so they stamp it as Cyrix 5x86-120/4X, same for the Cyrix 5x86-100/4X, but for the 100/4X, they ensured it was stable at 3.45 Volts. Or perhaps they'd get a few samples made and later stamp on it how the engineering sample comes out prior to each production run?

That's how it usually works. A chipmaker produces a batch of chips and chips with similar theoretical performance are often specced differently (take for example the K6-III+ line, they go like:400Mhz 1.6v 450Mhz 1.7v 500Mhz 1.8v etc)

feipoa wrote:

I used to think the way you do, but it wouldn't make sense to stamp them 3.7 V, later 3.6 V, then 3.7 V, unless they had some really high hopes afterwards. If that were the case, the first one (mine), would read 3.6 V as well. Unless, of course, the characterised sample was the oddball good one stable at 3.6 V.

Woops, I missed that thread, I forgot your 133 isn't from that lot.
Nope, that would indeed make no sense, at least not if they were send from the factory that way.
Theres the possibility your chip was send back as defective/malfunctioning, retested and re-stamped with 3.7v instead of 3.6v as it would work that way.
Intel did a similar thing with their first batches of 386's (from before they called them 386DX) when they found out some didn't work in 32 bit properly

Whats missing in your collections?
My retro rigs (old topic)
Interesting Vogons threads (links to Vogonswiki)
Report spammers here!

Reply 17 of 742, by feipoa

User metadata
Rank l33t++
Rank
l33t++

It was from this thread,
http://www.cpu-world.com/forum/viewtopic.php?t=15889

We may never really know for sure about the true intent of the 3.6V vs. 3.7V markings. Another user on the cpu-world forum, Windmiller, has a Cyrix 5x86-133/4x/3.6V from the same batch as mine, 49th week 1995.
You can refer here,
http://www.cpu-world.com/forum/viewtopic.php?t=9089

With all this shuffling around of core voltages (for the same production run even), it may be true that Cyrix discovered some stable at at 3.6V and some at 3.7V, and not necessarily intended them all to be run at 3.7V. I wonder what criteria they used to establish 'stability'?

Perhaps anything stable at a voltage higher than 3.7V, they renamed as a 100/120/4X unit. I have a 100/4X from 2nd week of 1996, but there is no voltage printed on it. In any case, if I had to choose, I'd pick the 3.6V version over the 3.7V version.

P.S. This thread is still accepting participants in the World's Fastest 486 contest, the prize being personal satisfaction.

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

Reply 18 of 742, by retro games 100

User metadata
Rank l33t
Rank
l33t

feipoa, you're definitely the 486 champ around here. You've done some awesome research in to this. Congratulations! 😀 Getting 200MHz out of my 486 was a hoot. I've talked about it elsewhere on Vogons, and so I'll skip the details. I'll very quickly summarise some thoughts:

1+3) Used 32MB, 256KB cache (WB). Seemed fine. Upgrage option = double this amount (but not tested).
2) CPU may have been "magic".
4) ISA EIDE IO controller seemed to work OK, with a CF drive.
5) G200 worked well. (Thanks for that tip!)
6) Audio, again not tested. Something old and crusty (ISA) should be OK.
1+7) 50MHz bus, divider 1:1. Wonder if any NIC would work at that speed? Unlikely / not tested.
8 ) Definitely Windows 95. Wouldn't consider anything later.
9) No USB tested. PS/2 worked OK.

BTW, "3x & 60 = 180MHz" got a good PcpBench score, so that's another option, but would probably make things more unstable in the long term. And that's the problem I have with my set up, compared to yours. I haven't tested my 200 MHz machine long enough. Not nearly long enough. I'd say it was switched on for no more than 10 hours in total. But during that time, the benchmark scores I got were "out of this world". I vaguelly recall doing a stress test too - with a whole bunch of tests running at the same time in Windows 98, including SuperPi, Prime95, Sandra, and WinTune. All tests worked, no problems.

Reply 19 of 742, by feipoa

User metadata
Rank l33t++
Rank
l33t++

rg100, your results were most promising! You have some data to show for it, unlike that German dude. 32MB cached RAM in 98SE is just not ideal, and Win95C should be better. 32MB in NT4.0 would be a headache with what I have running. It would be interesting to benchmark the PIO-4 CF with my Ultra2-LVD SCSI, just to have the numbers. I can do that when I have time.

Your results, with the addition of a second 200 MHz sighting, prompted me to create some AMD X5-180, X5-200, and Cyrix 5x86-150 scores in the Ultimate 486 Benchmark Comparison. The charts are now updated (in colour now!) and so is all the text and pdf. If I had to do it all over again, I'd probably use the Biostar with 64MB RAM, 512KB cache so that all the RAM is cached. That way, the scores are more comparable to other people's systems who have all cached RAM.
The Ultimate 486 Benchmark Comparison

So far, I'd say our two systems are pretty good contenders, at least as close as I've seen so far. I'm hoping for some others out there, preferably some non-UMC-based successes!

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