VOGONS


486 mobo + 586 chip

Topic actions

Reply 100 of 148, by retro games 100

User metadata
Rank l33t
Rank
l33t

Interesting. However, I did a quick search on Vogons for benchmark data for the Cyrix 120 chip, and discovered that it wasn't all that impressive. At the moment, having looked at some of this old test data on Vogons, and reading feipoa's post above, I am a bit puzzled.

Reply 101 of 148, by Tetrium

User metadata
Rank l33t++
Rank
l33t++
retro games 100 wrote:

Interesting. However, I did a quick search on Vogons for benchmark data for the Cyrix 120 chip, and discovered that it wasn't all that impressive. At the moment, having looked at some of this old test data on Vogons, and reading feipoa's post above, I am a bit puzzled.

Perhaps the pne who tested his Cyrix 5x86 (overclocked to 133Mhz I might add) hadn't enabled it's special features?

And for anyone wanting to try the Cyrix @133Mhz, it is believed the IBM 100's are better overclockers then the Cyrix 100's, so try to find an IBM part.
And idd, Cyrix's over 100 are hard to find for Socket 3, and I have no idea in as much you can enable it's special features or not.
Perhaps it was broken on only some of it's CPU's?

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

Reply 102 of 148, by udam_u

User metadata
Rank Member
Rank
Member

@feipoa
During my tests even Cx5x86@150MHz wasn't faster than am5x86@180MHz (; I know that Cx5x86 has some unique features which might work to his advantage but your results are too good. Please check your system using 3dbench, speedsys or mdk_dos_benchmark. Here you are links:
http://www.mediafire.com/?879qjmlsohshz8p
download.php?id=8162
http://motherboards.mbarron.net/download/dostools/sst470.zip

Thanks!

Reply 103 of 148, by Tetrium

User metadata
Rank l33t++
Rank
l33t++

What I find weird is, how come using the same FSB results in slower access times for his RAM?

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

Reply 104 of 148, by udam_u

User metadata
Rank Member
Rank
Member

What I find weird is, how come using the same FSB results in slower access times for his RAM?

Maybe his motherboard takes advantage from cyrix linear burst mode?

Reply 105 of 148, by feipoa

User metadata
Rank l33t++
Rank
l33t++

I have attached some screen shots of SpeedSys and Cachechk.
[NOTE: filenames of the first and second cachechk image are slipped. I can't seem to correct it -- perhaps some bug with this forum's software.]

The measured cpu clock speed from Cachechk 4 does not agree well with that of Cachechk 7. Cachechk 4 reports the correct speed at ~120 Mhz, whereas Cachechk 7 reports it as ~107 Mhz. Both agree well on the L1/L2/RAM speeds though. Has anyone else noted this with Cyrix cpus?

SpeedSys reports the cpu speed at 109 Mhz. I suspect the speed discrepancies are due to some kind of software timing algorithm used to measure the speed. I've run a dozen bench utilities, some say 89 Mhz, some say 120 Mhz, 107 Mhz, 109 Mhz. This occurrence is not isolated to this particular motherboard/cpu combination. Identical hardware on my Biostar MB-8433UUD ver. 3 shows similar cpu clock reporting conflicts. Nonetheless, the data from the actual tests is what is important.

For the poster of the AMD 200 Mhz cpu, could you also include the sub-screen on the memory speed tests, its option M on SpeedSys. I would be very interested if you can stably run Windows NT 4.0 on that system.

I get a 3dBench score of 76.9 frames/second. I am not sure if this is good or bad. I do not play games.

I use this computer as my daily E-mail machine, web browser (IE6, Opera8.5, and Firefox2), mp3 player, MS Office, LAN server and HTTP server. Several of the nexgen enhancements will need to be enabled if your motherboard doesn't natively enable them. For example, enabling Fast FPU Exception Handling reduced cpu consumption by 15% when playing unreduced 128kbit mp3s in Winamp 2.85 with the waveout Outout (best not to use DirectSound output on a 486 as it is much more cpu hungry). The FPU in the Cyrix 5x86 with enhancements on is about that of the Pentium Overdrive 83. That was really the only advantage in the iPOD-83. Have you compared your AMD 200 Mhz with the FPU of the iPOD83? That would be an interesting showdown.

Here are my general machine specifications:

CPU:
Cyrix 5x86 – 120 (G5F9550C)
16 KB Unified Write-Back Cache (L1), 4-way Associative
246 MB/s, 4.3 ns/byte, 4 μs/kb
64-bit internal data bus, 32-bit external data bus, 32-bit address bus
80-bit floating point unit with 64-bit interface (x87 instruction set), FPU auto-idle
Parallel execution of integer and FPU operations, Six-stage execution pipeline, Decoupled load/store unit, Memory management unit

Cyrix 5x86 CPU Enhancements Enabled:
Load Store Reordering (LSSER)
Pre-fetch Buffer Loop (LOOP_EN)
Return Stack (RSTK)
Write-Back Cache (USE_WBAK)
Write-Through Region 1, 640KBytes - 1 MByte (WT1)
Burst Write Cycles (BWRT)
Linear Address Burst Mode (LINBRST)
Memory Read Bypassing (MEM_BYP)
Directory Table Entry Cache (DTE_EN)
Fast FPU Exception Handling (FP_FAST)
[The only enhancement feature I cannot get working in either WinNT or Win98 is Branch Prediction]

Register Bits: LSSER, USE_SMI, USE_SUSP, WT1, LOCK_NW, USE_WBAK, MAPEN0,
(default) SMM_MODE, LINBRST, DET_E, MEM_BYP, CLK1, CLK0
Register Bits: LOOP_EN, RSTK_EN, BWRT, FP_FAST
(enabled)

Motherboard:
Biostar MB-8433UUD v3.0
PCI 2.1 Compliant
Award 4.50PG, UUD960326S, 26 March 1996

BIOS Enhancements Enabled:
Write-back L1 Cache Scheme
Early Cache Write Mode
Video/System BIOS Cacheable
HOST-to-PCI Post Write, 0 WS
HOST-to-PCI Burse Write
PCI Bus Park Option
PCI Post Memory Write
Burst Copy-Back Option
HOST Clock / PCI Clock, 1:2/3

Chipset:
UMC 881E/886B
UM8881F (9639-EYA)
UM8886BF (9640-FXA)

Timing:
L2: 2-1-1-1, DRAM: 0WS / 0WS (if using EDO, 3-2-2-2 needed for stability)

Cache (L2):
UMC 512 KB, 4 chips × 128 KB
96 MB/s, 10.9 ns/byte, 11 μs/kb

Each chip reads (UMC):
UM61 M 1024K-15
9615L M1L68

Tag reads (Winbond):
W24257AK-15
94261

Memory:
256 MB Fast-Page Mode (60 ns), 4 modules × 64 MB, 72-pin, Gold contacts (60 ns)
8 × Samsung 007 (K4F640411C-TC60) chips per stick
4 × LGS GM71C16100CJ6 chips per stick
55 MB/s, 19.0 ns/byte, 20 μs/kb

Graphics Card:
(PCI) Matrox Millennium G200, 16 MB SDRAM
MGA-G200 Graphics Chip
250 MHz RAMDAC Speed
84-90 MHz Core Speed
112-120 MHz Memory
64-bit Memory Interface
128-bit Core (contains dual 64-bit unidirectional buses)
BIOS Version 3.3.30 (21 November 2000), Revision 0.30

SCSI Card:
(PCI) Adaptec 2940U2W
Ultra2 LVD
40 MHz
80 MByte/sec

Network Card:
(PCI) 3Com 3c905C-TX-M
10/100Base-TX

Sound Card:
(ISA) Creative Labs AWE64 Gold, 4 MB

Modem Card:
(ISA) U.S. Robotics 5687
V.90 56K

Harddrive:
Seagate ST3146707LW
Ultra320, 146 GB
320 MByte/sec

DVD-ROM:
Toshiba SD-M1401 SCSI-2 Ultra

Floppy Drives:
NEC 3 ½, 5 ¼,

Mouse:
PS/2 Logitech MouseMan (3-button)

Monitor:
Planar LCD PL150M

Battery Backup:
APC Backup-UPS 500

Operating System:
Windows NT 4.0 (Build 1381: Service Pack 6)

Laser Printer:
Brother HL-5250DN
10/100Base-TX, USB 2.0, Parallel ECP/EPP
Up to 30 pages/min - B/W
1200 dpi × 1200 dpi
NEC VR5500 - 266 MHz
SDRAM: 32 MB / (544 MB max)

Speakers:
Harman/Kardon HK195

Keyboard:
Microsoft RT2300

KVM:
IOGEAR GCS614A MiniView, 4-port KVM Switch w/Audio

Power Supply:
250 Watts, Generic

Attachments

  • Cachechk4-M919.jpg
    Filename
    Cachechk4-M919.jpg
    File size
    89.82 KiB
    Views
    3383 views
    File comment
    Cachechk v4 M919, Cyrix 5x86 120
    File license
    Fair use/fair dealing exception
  • Cachechk4-Biostar.jpg
    Filename
    Cachechk4-Biostar.jpg
    File size
    65.29 KiB
    Views
    3383 views
    File comment
    Cachechk v4 Biostar MB-8433UUD ver. 3, Cyrix 5x86 120
    File license
    Fair use/fair dealing exception
  • Cachechk7-M919.jpg
    Filename
    Cachechk7-M919.jpg
    File size
    85.04 KiB
    Views
    3385 views
    File comment
    Cachechk v7 M919, Cyrix 5x86 120
    File license
    Fair use/fair dealing exception
  • SpeedSys-M919-1.jpg
    Filename
    SpeedSys-M919-1.jpg
    File size
    133.34 KiB
    Views
    3386 views
    File comment
    SpeedSys M919 Cyrix 5x86 120
    File license
    Fair use/fair dealing exception
  • SpeedSys-M919-2.jpg
    Filename
    SpeedSys-M919-2.jpg
    File size
    73.67 KiB
    Views
    3386 views
    File comment
    SpeedSys M919 Cyrix 5x86 120
    File license
    Fair use/fair dealing exception

Reply 106 of 148, by retro games 100

User metadata
Rank l33t
Rank
l33t

feipoa, thanks a lot for posting all of this info! 😀 For comparative information, I posted my test results for the Biostar MB-8433UUD on this Vogons thread here. If you scroll down that page to my second post which begins "I have just returned to this Biostar mobo".

I didn't know about the "M memory" option for Speedsys, however luckily you can get an idea of these values by looking at the Speedsys graph on that webpage. The memory values seen in the "M memory" section are values taken from this graph. They are taken from the graphical lines on the Speedsys screenshot before they "dip downwards" for the very first time. Also found on this post is a Cachechk(7) .txt attachment, and I also mention my 3DBench (1.0c) score, and a few other bits of info.

I have tested a POD83, but when I overclocked it to 100 MHz, I could not get Quake 1 stable in DOS. That's a good DOS stability test, and so I decided to go back to the AMD chip. The next time I mess about with the Biostar mobo, I'll get it running in Windows 98 (I don't have NT), and make sure it all works without any problems by running Prime95 and SuperPi, and other Windows based benchmarks. I'll post back with the results, but that might not be for a while now, as I've got some other mobos to test first.

Thanks for posting the test results of your Cyrix chip, it was interesting to read.

Reply 107 of 148, by udam_u

User metadata
Rank Member
Rank
Member

Thank you for your results! Now everything is clear.

The measured cpu clock speed from Cachechk 4 does not agree well with that of Cachechk 7. Cachechk 4 reports the correct speed at ~120 Mhz, whereas Cachechk 7 reports it as ~107 Mhz. Both agree well on the L1/L2/RAM speeds though. Has anyone else noted this with Cyrix cpus?

SpeedSys reports the cpu speed at 109 Mhz. I suspect the speed discrepancies are due to some kind of software timing algorithm used to measure the speed. I've run a dozen bench utilities, some say 89 Mhz, some say 120 Mhz, 107 Mhz, 109 Mhz. This occurrence is not isolated to this particular motherboard/cpu combination.

You are right, my cx5x86@120 was also recognized as 109MHz and 5x86@150MHz as 136MHz. [:

Cyrix 5x86 CPU Enhancements Enabled: Load Store Reordering (LSSER) Pre-fetch Buffer Loop (LOOP_EN) Return Stack (RSTK) Write-Bac […]
Show full quote

Cyrix 5x86 CPU Enhancements Enabled:
Load Store Reordering (LSSER)
Pre-fetch Buffer Loop (LOOP_EN)
Return Stack (RSTK)
Write-Back Cache (USE_WBAK)
Write-Through Region 1, 640KBytes - 1 MByte (WT1)
Burst Write Cycles (BWRT)
Linear Address Burst Mode (LINBRST)
Memory Read Bypassing (MEM_BYP)
Directory Table Entry Cache (DTE_EN)
Fast FPU Exception Handling (FP_FAST)
[The only enhancement feature I cannot get working in either WinNT or Win98 is Branch Prediction]

On the other hand for me only branch prediction worked correctly. All my tries to enable additional options except BP ended BSOD in windows. However I didn't try other options with BP turned off because I didn't have time to check all combinations. I simply accepted that BP is the most important feature. Your 3Dbench score is not the best. As far as I remember my result on 5x86@120 (2x60MHz) oscillated around 90fps.

Reply 108 of 148, by feipoa

User metadata
Rank l33t++
Rank
l33t++

My 3Dbench results aren't spectacular because I am using a 1:2/3 Host-to-PCI clock divider in the BIOS for enhanced system stability in WinNT4. When it is set at 1:1, I get a 3Dbench score of 91. Also, a Matrox G200 graphics card isn't the best 3D performer; it is know for its 2D capabilities (all I use it for).

As for the speed results of Cachechk7 and perhaps Speedsys, it might be also due to the PCI clock divider and how these programs measure the frequency. For example, 2/3 * 40 Mhz * 4 = 106.8 Mhz. This is what Cachechk7 reports. I'm not sure why it multiplied it by 4 though.

I get a similar 3Dbench score of 71.4 fps on the M919 with identical hardware, however this bios doesn't have a PCI clock divider, which is curious. You'd think it would be 91 fps, unless the BIOS is doing some auto division without me knowing it.

Reply 109 of 148, by feipoa

User metadata
Rank l33t++
Rank
l33t++

I ran Speedsys on my Biostar MB-8433UUD v3.0 as well. It performed much better than the PCchips M919. All memory results now outperform the AMD at 200 Mhz with a stock Cyrix 5x86 - 120 Mhz and a Matrox G200. Here is a summary of all tests I have run:

Speedsys:
Benchmark number: 65.49
L1: 172.98 MB/s
L2: 67.92 MB/s
Memory throughput: 47.97 MB/s

L1 Read: 224.66 MB/s
L1 Write: 75.49 MB/s
L1 Move: 218.80 MB/s
L1 Average: 172.98 MB/s
L2 Read: 87.66 MB/s
L2 Write: 75.21 MB/s
L2 Move: 40.89 MB/s
L2 Average: 67.92 MB/s
RAM Read: 50.44 MB/s
RAM Write: 75.48 MB/s
RAM Move: 18.01 MB/s
RAM Average: 47.98 MB/s

PcpBench: 8.8 [EDIT: 9.7 when using Diamond VLB and M919]
3DBench (1.0): 90.9

CTCM7: 32MByte, Cache Area (should go to 64Mbyte when I put 512KB back in, or 128Mbyte in Write-Thru mode)

MAXUM Processor clock speed test shuffles between 121 Mhz and 104 Mhz

Cachechk v4
L1: 246.3 MB/s
L2: 95.8 MB/s
RAM: 55.1 MB/s
RAM access, read: 76 ns
RAM access, write: 51 ns
120.5 Mhz

Attachments

  • SpeedSys-Biostar-1.jpg
    Filename
    SpeedSys-Biostar-1.jpg
    File size
    158.42 KiB
    Views
    3320 views
    File comment
    Biostar MB-8433UUD v3.0
    File license
    Fair use/fair dealing exception
Last edited by feipoa on 2011-03-08, 17:05. Edited 1 time in total.

Reply 110 of 148, by retro games 100

User metadata
Rank l33t
Rank
l33t

If you're interested in the following type of experiment, please try and get your Biostar Cyrix rig to either match or beat these Biostar AMD rig results:

PcpBench = 12.5 (3x multi, 60 MHz bus speed = 180 MHz clock speed)
3DBench 1.0c = 118.1 (4x multi, 50 MHz bus speed = 200 MHz clock speed)
SpeedSys score = 74.99 (4x multi, as above)
Quake 1 fullscreen timedemo demo1 = 18.3 (4x multi, as above)

You'll find that it doesn't make much difference what PCI video card you use. More or less any 2MB - 8MB well known VGA card from roughly the mid to late '90s era should be sufficient.

Reply 111 of 148, by feipoa

User metadata
Rank l33t++
Rank
l33t++

Do you have the Quake 1 source?

I doubt my system can outperform the AMD-x5 200 Mhz on all accounts, except for memory throughput and probably most or all 2D-based office apps. I've only been experimenting with tweaking for 1 day, so who knows what might come up. My goal is more for speed and long-term stability. I've had NT4 uptimes of over 5 months with heavy daily use.

I'd be interested to know how your system stability in Windows98SE or NT4 is at 200 Mhz. What applications tend to be more problematic, particularly what activity that leads to a BOSD?

Since you are using the same motherboard as I am, I am wondering if you have found a way to set the L2 cache to Write-thru instead of the default of write-back? This will allow for more RAM being cacheable by L2. Have you tried 512 KB cache? I just took mine out for some tests. What version of the MB-8433UUD do you have? I have a 3.1 and a 3.0.

Have you tried overclocking a Cyrix 5x86-120 in the Biostar MB-8433UUD?

Reply 112 of 148, by retro games 100

User metadata
Rank l33t
Rank
l33t

I don't have the Quake 1 source, only the freely downloadable shareware version 1.06. I'll dig out this board again soon, and mess about with it @ 200 MHz with Windows 98 SE. I'll run some Windows benchmarks, including Prime95 and SuperPi, and let you know what happens.

BTW, I don't have NT. Also, I haven't yet experimented with setting the mobo's L2 cache to write-through, instead of write-back. Also, I only use 1 stick of 32MB EDO RAM at the moment. I have revision 3.1 of this Biostar mobo, and I currently don't own a Cyrix 120 chip.

Reply 113 of 148, by udam_u

User metadata
Rank Member
Rank
Member

My 3Dbench results aren't spectacular because I am using a 1:2/3 Host-to-PCI clock divider in the BIOS for enhanced system stability in WinNT4. When it is set at 1:1, I get a 3Dbench score of 91. Also, a Matrox G200 graphics card isn't the best 3D performer; it is know for its 2D capabilities (all I use it for).

I made tests with divider set to 1:1/2 -> PCI worked at 30 MHz which is 4MHz higher than yours.

Why do you use 1:2/3 divider with FSB set to 40MHz? I think that 1:1 is much better in this case - PCI performance is slightly improved and most peripherals work correctly at this speed ("gentle overclocking" he he he). If I were you I would use 1:2/3 divider with FSB set to 50MHz and 1:2 with FSB set to 60 or 66MHz. In all other cases, I would use 1:1 divider.

Reply 114 of 148, by feipoa

User metadata
Rank l33t++
Rank
l33t++

Windows NT 4.0 will crash during boot at 1:1 with a 40 Mhz FSB and a Cyrix 5x86. I have not yet isolated exactly what is causing the crash. It may be that the Matrox G200, Adaptec Ultra2 SCSI, or the 3Com 3c905C-TX-M PCI card refuses to run at 40 Mhz, or a combination of these and particular Bios settings

I don't have another PCI graphics card at the moment to begin a lengthy troubleshooting process, although I do have a 1 MB ISA one which may get me started.

Reply 115 of 148, by feipoa

User metadata
Rank l33t++
Rank
l33t++

Thought I'd post the face to those specs.

Attachments

  • Biostar-Cyrix_1.jpg
    Filename
    Biostar-Cyrix_1.jpg
    File size
    63.11 KiB
    Views
    3215 views
    File license
    Fair use/fair dealing exception
  • Biostar-Cyrix_2.jpg
    Filename
    Biostar-Cyrix_2.jpg
    File size
    63.31 KiB
    Views
    3215 views
    File license
    Fair use/fair dealing exception

Reply 117 of 148, by feipoa

User metadata
Rank l33t++
Rank
l33t++

What does your 486 case look like? I doubt you can get still get this on. Its pretty modern looking for a 486 case. There isn't a model number on it. I bought it new in Jan. 1997 with my 486.

I am also looking for another case for my second 486, the m919 I have setup on the table. Unfortunately for you, the LCD display can only make a "1" as the first digit so you wouldn't be able to put 200 Mhz.

These cases have become scarce on eBay. All I can recommend is scouting the thrift stores and computer junk stores.

Reply 118 of 148, by retro games 100

User metadata
Rank l33t
Rank
l33t
feipoa wrote:

What does your 486 case look like? I doubt you can get still get this on. Its pretty modern looking for a 486 case. There isn't a model number on it. I bought it new in Jan. 1997 with my 486.

Unfortunately, I don't have any AT cases! I just put my mobos on a piece of chipboard, for testing. I've got years more testing to do, before I start building my retro machines.

feipoa wrote:

I am also looking for another case for my second 486, the m919 I have setup on the table. Unfortunately for you, the LCD display can only make a "1" as the first digit so you wouldn't be able to put 200 Mhz.

Hehe, they didn't think about that. 😉