Scientific PC Computing and Benchmark Comparisons
This site has been accessed
times since May 1st, 1997
HX83 Review- New
Benchmarks - Overclock - BIOS
- Memory - System Board
- Storage - Controllers
- Video - Miscellaneous
New Benchmarks and Site Additions
-
Video Card Comparisons: ATI Rage Fury vs 3DFX
Banshee vs Nine Revolution 4 vs Nine Revolution 3D
-
EPoX EP-51MVP3E-M Super
7 AMD K6-2/300 and Intel 233MMX: Main
Page, Setup Notes
-
Benchmarks: EPox K6300 vs. Dell Dimension
PII300, WB97,
Note:
The EP51MVP3 used the Nine Standard driver. The optimized driver
is incompatible with W98. The HOT631 used the Nine optimized driver
in Win95. The optimized driver is nearly 10% faster than the
standard driver. Nevertheless, the EP51MVP matches the PII300 in
high end graphics. The CPU32 marks indicate that the K6300 and the
PII300 are at par. The PII300 floating point marks in WB98 are about
30% higher than the K6300. The marketed system (Dell) does not fair
so well in this contest.
-
Benchmarks: EPoX EP-51MVP3E-M --
vs. HOT631 PII333 and Dell PII300, WB97,
Note: Both the EP51MVP3 and the HOT631 used the Nine Standard
driver and W98. The CPU32 marks indicate that the K6-2/350 (869)
has the edge on the PII333 (829) and crushes the marketed system (733).
At the risk of sounding dramatic -- Socket 7 is alive and well. Not
only is the Dell the slowest of the three, it is, based on successfull
completion of benchmark testing, the least stable.
-
Benchmarks: Tuning
the BIOS -
AGP Aperature Size,
-
Building a Performance Computer: Tips
on components, construction, and software installation (under major
construction)
-
Shuttle HOT631/PII300: Overclocking - 4.5X66(33.0),
4.5X75(37.5), 4.5X83(33.0) - HOT631
Benchmarks February 22, 1998
-
Hardware Locks and
SM/AMI Socket 7 - Possible
Solutions February 11, 1998
-
P6SNE Pro180 WinBench97:
Overclocking:
180MHz (3.0X60) vs. 200MHz (3.0X66) vs. 233MHz
(3.5X66)- 64MEDO NineR3D16M AHA2940W XP332150W FAT32 Plextor
12X OSR2. Note: These results are from the second P6SNE system
I configured for 233MHz. Both of these systems run 24/7 at 233MHz.
They are stable. Running the Pro180 at 233 gives an average performance
increase of about 20%. February 7, 1998.
-
Upcoming 3D Benchmarks: Recently, I built two P6SNE/Pro180
systems. These systems run great at 233MHz. I configured an
HX83/233MMX identical to one of the P6SNE/233 (Pro180) systems with the
exception of motherboard/cpu. This allows meaningful comparisons
between the Pro and PentiumMMX cpus. Anyone building a 3D workstation
on a tight budget ought to consider the Pro180 (if you can find one)!
The P6SNE/Pro233 is not far behind the FX83 PII300 (340MHz) in 3D Marks
(Winbench98) when both are equipped with the Nine R3D16M. Additionally,
recent benchmarks for the 9R3D at another hardware site indicate that this
card is a dog in 3D. This is not supported by my observations.
-
HX83/P233MMX WinBench97:
Overclocking:
3.5X66 (233MHz) vs 3.5X75 (266MHz) -
128MPEDO NineR3D4M Symbios875 ST34501W (Cheetah) FAT32. Note:
Overclocking from 3.5X66 to 3.5X75 was simply a matter of setting the jumpers
and backing off slightly on the memory timing (128M). February
02, 1998.
-
Despite that I have WinBench98, I continue to
use Winbench97 -- Why?
January 31, 1998
-
FX83/PII300/4.5X75(340MHz) WinBench97:
Nine
R3D16M 1.46 Drivers: Optimized vs. Standard
- 128MPEDO NineR3D16M AHA2940UW ST34501W (Cheetah) FAT32.
Note: Not surprisingly, the primary affect of these drivers is on
video performance. The Standard drivers are about 10% slower than
the Optimized drivers(MicroStation excepted). Interestingly, the
Standard drivers give a small, but consistant increment in CPU32.
January 16, 1998.
-
FX83/PII300/4.5X75(340MHz) WinBench97:
BIOS
Settings: DRAM Fast Lead Off
- 128MPEDO NineR3D16M AHA2940UW ST34501W (Cheetah) FAT32.
Note: The DRAM Fast Lead Off bios setting appears to have little,
if any, effect on FX83 performance. In the interest of KISS, I recommend
that it be disabled in similar configurations (ECC PEDO). January
16, 1998.
Overclock
-
P6SNE Pro180 WinBench97:
Overclocking:
180MHz (3.0X60) vs. 200MHz (3.0X66) vs. 233MHz (3.5X66) - 64MEDO
NineR3D16M AHA2940W XP332150W FAT32 Plextor 12X OSR2. Note:
These results are from the second P6SNE system I configured for 233MHz.
Both of these systems run 24/7 at 233MHz. They are stable.
Running the Pro180 at 233 gives an average performance increase of about
20%. February 7, 1998.
-
HX83/P233MMX WinBench97:
Overclocking:
3.5X66 (233MHz) vs 3.5X75 (266MHz)
- 128MPEDO NineR3D4M Symbios875
ST34501W (Cheetah) FAT32. Note: Overclocking from 3.5X66 to
3.5X75 was simply a matter of setting the jumpers. February
02, 1998.
-
HX83
P200MMX:
Over-/Under-Clock - 3.0X66 vs 2.0X83 vs 2.5X83 vs 3.0X75 Winbench97:
1024X768X16@85 64M EDO (2 banks) Millennium 4M Symbios 875 ST34501W (Cheetah
UW) FAT32 MRBIOS 3.41 (55nS) Matrox BIOS 2.5 Matrox driver 36.2 - Note:
Performance at 2.5X83 and 3.0X75 are similar.
-
HX83
- P200MMX: 3X66 (200MHz) vs. 3X75 (225MHz) vs.
2.5X83 (208MHz) Winbench97:
1280X1024 64M Micron EDO (2-32M Simms) Imagine
128-2 (4M VRAM) 2940UW Cheetah ST34501W (FAT32) HX83V1.7.
-
HX83
- P166 (2.5X66) vs. P166 (2X83) vs. P200MMX (3X66)-
Winbench97:
64M Micron EDO (2-32M Simms) Imagine 128-2
(4M VRAM) 2940UW Cheetah ST34501W (FAT32) MrBIOS 3.31 a valient effort
by the P166
-
HX83
- P166 (busMhz/ClkMHz) 66/166 vs. 75/150 vs 83/166
-
Winbench97:
MrBios
3.31 (beta HX83) 64M Micron EDO (2-32M Simms)
Imagine 128-2 (4M VRAM) 2940UW Cheetah ST34501W (FAT32) OSR2 an o'l dog
learns new tricks
BIOS - Type and Settings
-
FX83/PII300 BIOS Settings:
This represents the optimal bios setting for the FX83/PII300 128M Nine
R3D4M Symbios875 Cheetah.October 28, 1997.
-
FX83/PII300/4.5X75(340MHz) WinBench97:
BIOS
Settings: DRAM Speed - 50ns vs. 60ns vs. 70ns (2940UW vs. Fireport 40 revisited)
- 128MPEDO NineR3D16M AHA2940UW ST34501W (Cheetah) FAT32. Note:
These results indicate that the performance decrement for slowing DRAM
Speed from 50nS to 60nS or 70nS is minimal. When overclocking, the difference
between 50nS and 70nS can be dramatic. The system might not boot
at 50nS whereas it could run stable at 70nS. These results also compare
the Symbios875 and Adaptec 2940UW controllers. The Adaptec controller
has a distinct advantage. I compared these two cards previously in
a similar configuration. For this earlier configuration there were
no differences between the cards (Winbench97).
Why is the Adaptec is faster in this configuration? January
12, 1998.
-
FX83/PII300/4.5X75(340MHz) WinBench97:
BIOS
Settings: DRAM Integrity Mode (ECC) - Disabled vs. Parity vs. ECC2
- 128MPEDO NineR3D4M AHA2940UW ST34501W (Cheetah) FAT32. Note: There
is debate on various news groups about whether ECC is implemented by a
particular bios. News posts stated that there is a drop in performance
when enabling ECC on a system that supports ECC. These marks indicate
that there is little, if any, disadvantage to enabling ECC1. In fact,
these results are in the opposite direction. What do they mean?
December 5, 1997.
-
FX83/PII300/4.5X75(340MHz) WinBench97:
BIOS
Settings: DRAM Fast Lead Off - 128MPEDO NineR3D16M AHA2940UW
ST34501W (Cheetah) FAT32. Note: The DRAM Fast Lead Off bios
setting appears to have little, if any, effect on FX83 performance.
In the interest of KISS, I recommend that it be disabled in similar configurations
(ECC PEDO). January 16, 1998.
-
FX83/PII300/4.5X75(340MHz) WinBench97:
BIOS
Settings: USB Passive Release - Enabled vs. Disabled - 128MPEDO
NineR3D4M AHA2940UW ST34501W (Cheetah) FAT32.
November 2, 1997.
-
FX83/PII300/4.5X75(337MHz) WinBench97:
BIOS
Settings: DRAM Integrity Mode (ECC) - Disabled vs. Parity vs. ECC2
- 128MPEDO NineR3D4M AHA2940UW ST34501W (Cheetah) FAT32. Note: There
is debate on various news groups about whether ECC is implemented by a
particular bios. News posts stated that there is a drop in performance
when enabling ECC on a system that supports ECC. These marks indicate
that there is little, if any, disadvantage to enabling ECC1. In fact,
these results are in the opposite direction. What do they mean?
December 5, 1997.
-
HX83/200MMX:
Turn Around Insertion - 128M PEDO 3X75 Millenium 4M 3.63 ST34501W
(Cheetah) 1024X768X16@85 Millennium 4M Adaptec AHA2940UW ST34501W (Cheetah
UW) FAT32AMI BIOS 1.75 Matrox BIOS 2.5 Matrox driver 3.63 Winbench97Note:
These results are for PEDO. Turn Around Insertion should be enabled for
stable operation when using EDO. In older versions of the AMI BIOS, enabling
Turn Around Insertion caused instability with PEDO.
-
HX83
P200MMX
3X75 (225MHz): AMI 1.75 vs MrBios 3.41 at 64M and 128M PEDO - 16 Bit depth
1024X768X16@85 Toshiba PEDO Millennium 4M Symbios 875 ST34501W (Cheetah
UW) FAT32 Matrox BIOS 2.5 Matrox driver 3.61 - Winbench97
Note: In general AMI 1.75 offers a performance edge over MrBios 3.41. My
observations are that MrBios 3.41 is easier to configure and is more stable.
However, this stability might be simply due to undertuning. These benchmarks
address the observation that large amouts of installed ram often require
slowing the memory timing. With 128M PEDO, the MrBios BIOS and the AMI
BIOS are nearly equal.
-
HX83
P200MMX
3X75 (225MHz): AMI 1.75 vs MrBios 3.41 at 64M and 128M PEDO - 08 Bit depth
1024X768X16@85 Toshiba PEDO Millennium 4M Symbios 875 ST34501W (Cheetah
UW) FAT32 Matrox BIOS 2.5 Matrox driver 3.61 - Winbench97
-
HX83
P200MMX
3X75 (225MHz): AMI 1.75 vs. MrBios
3.41 (Evaluation Only): 1024X768X16@85 HX83/512 3X75=225MHz
96M EDO AMI HX83V1.75 2940UW ST34501W (Cheetah UW) FAT32 - Winbench97
Note: The MrBios configuration runs stable and fast "out of the box."
-
HX83
- 3X66 (200MHz) P200MMX - MrBios
3.31 vs. AMI HX83V1.6 vs. AMI HX83V1.7:
64M
Micron EDO (2-32M Simms) Imagine 128-2 (4M VRAM) 2940UW Cheetah
ST34501W (FAT32) HX83V1.7 improved the manual settings. - Winbench97
-
HX83
- AMI STE5021 (HX83) vs. AMI STE S6828T (P5STE)
vs. MrBios 3.31 (beta
HX83): Comparisons are for Pentium 166
performance. 64M Toshiba PEDO (2-32M Simms) Imagine 128-2 (4M VRAM)
2940UW Cheetah ST34501W (FAT32) OSR2 - Winbench97
-
HX83
- P166: 83MHz bus 166MHz CPU - MrBios
3.31 vs. AMI STE S6828T vs. AMI
STE 5021 64M Micron EDO (2-32M
Simms) Imagine 128-2 (4M VRAM) 2940UW Cheetah ST34501W (FAT32) OSR2 features
and performance - Winbench97
Memory - Type and Amount
-
HX83-
P200MMX 3X75 (225MHz): 32M (2X16) vs. 64M (2X32) vs. 96M (2X32 + 2X16):
1024X768 MrBIOS 3.31-55nS Micron EDO Imagine 128-2 (4M VRAM) Adaptec 2940UW
Cheetah ST34501W(FAT32) OSR2 Note: There is no penalty for increasing memory
beyond 64M. Simarly, there is no penalty for populating both Banks 0 and
1 (cpu16 = 490, Business Graphics = 90) - Winbench97
-
P5STE/P166(166Mhz) WinBench97:
Single
Tag on HX Chipset: 64M versus 128M of EDO Memory - SuperMicro
P5STE/P166, 64/128MPEDO, Millennium 4M, Symbios 825 ST32430N. Note: These
marks show the distinct performance decrement when exceeding 64M of RAM
in a single tag board. Most notable is a near 40% drop in performance
for CPU Marks. November 11, 1997.
-
P5STE - DRAM Above 64M (P5STE with dual tag):
64M (2-32M SIMMS) vs 128M (4-32M SIMMS) Toshiba ECC PEDO - Wenbench97
-
HX83
- Micron EDO vs. Toshiba PEDO: Comparisons
are for Pentium 166 performance. The Micron EDO is faster, but not by much.
Memory timing data are shown. Included hardware is the Cheetah and the
Imagine 128-2. - Winbench97
System Board and CPU
-
FX83 Diary: New comments regarding
the PII300 installation, the Nine R3D, and memory termination for the Pro
CPU. October 27, 1997.
-
FX83/PII300/4.5X75(340MHz) WinBench97:
PPro200/512
vs. PII300/512 - 128MPEDO NineR3D4M AHA2940UW ST34501W (Cheetah)
FAT32. Note: Consider the price of the PII300, the opportunity
to OC the bus, and the performance. There is no apparent reason to buy
the more expensive PPro200/512. The exception might be in a server or an
SMP application. November 2, 1997.
-
FX83/Pro200/512
vs HX83/200MMX
Winbench97:
1024X768X16@85 Millennium 4M 3.62 Adaptec AHA2940UW ST34501W (Cheetah
UW) FAT32 (FX83 Diary) - Note: These
benchmarks indicate that the Pro200/512 and the Pentium 200MMX offer similar
performance in OSR2. However, my personal observations are that, for large
image files in either Photoshop or Picture Publisher, the Pro200/512 has
a distinct advantage. I would not argue price/performance considerations
in favor of the Pro as a workstation CPU. Nevertheless, having access to
both systems, I prefer the FX83 combination for image processing. Modified
July 27th, 1997.
-
HX83
vs. SuperMicro P5STE: Comparisons are
for Pentium 166 performance. The HX83 appears to have a graphics advantage
(Business Graphics WinMark = 58.8). Included hardware is the Cheetah and
the Imagine 128-2. - Winbench97
-
Memory Lane: HX83/P200MMX
vs P5STE/P166 vs CWA/P90 vs P5MP3/60: A brief look at the history
of Pentium computing. - Winbench97
-
SuperMicro P5STE/512: Memory Timing:
Manual Settings (fastest) vs. 60ns vs. 70ns vs. 70ns + NAD Disabled P166
64M Toshiba PEDO 128-2 (4M) 2940UW ST32550W (FAT32) OSR2 (Win95) - Winbench97
-
SuperMicro P5STE/512: Overclocking
from 166 MHz to 180 MHz P166 64M Toshiba PEDO 128-2 (4M) 2940UW ST32550W
(FAT32) OSR2 (Win95) - WinBench
97 (this is not an effective overclocking)
Storage Devices
-
HX83/200MMX:Cheetah
(ST34501W) vs. Barracuda Ultra (ST32171N) vs. Barracuda Wide (ST32550W)
Winbench97:
Note: Also shown are the effects of benchmarking an empy drive vs 50% capacity.
As expected the Cheetah wins hands down. The newer 50 pin Barracuda beats
the older 68 pin version. August 20th, 1997
-
SuperMicro P5STE/512Cheetah Benchmark:
Seagate ST32155W (5400RPM UW Hawk) vs. Seagate ST32550W (7200RPM Wide Barracuda)
vs. Quantum XP32150W (7200RPM Wide Atlas) vs. Seagate ST34501W (10000RPM
Ultra Wide Cheetah) - Winbench97
-
SuperMicro P5STE/512: Seagate ST32155W
(5400RPM UW Hawk) vs. Seagate ST32550W (7200RPM Wide Barracuda) vs. Quantum
XP32150W (7200RPM Wide Atlas) vs. Seagate ST34501W (10000RPM Ultra Wide
Cheetah) - Winbench97
-
SuperMicro P5STE/512: Seagate ST32155W
(Hawk UW) vs. Quantum XP32150W (Atlas W) vs. Seagate ST32550W (Barracuda
W) (P5STE/512 P166) - Winbench97
-
These are the same charts as the Millennium, Mystique, and Imagine 128-2
comparison data.
-
SuperMicro P5STE/512: Iomega Internal
SCSI JAZ vs. Seagate Barracuda ST32550W (what a race!) - Winbench97
SCSI Controllers
-
FX83/PII300/4.5X75(337MHz) WinBench97:
Diamond
FirePort 875 versus Adaptec AHA2940UW - 128MPEDO NineR3D4M AHA2940UW
ST34501W (Cheetah) FAT32. November 1, 1997
-
HX83
- P200MMX 3X75 (225MHz): SymbiosUW vs Adaptec
2940UW: 1024X768X16 AMI HX83V1.7 96M Micron EDO Imagine 128-2
(4M VRAM) SymbiosUW/AdaptecUW Cheetah ST34501W(FAT32) OSR2 Note: These
data indicate similar performance for the Symbios and Adaptec controllers.
HX83 performance is shown in a highly tuned configuration. - Winbench97
Video Accelerators
-
FX83/PII300/4.5X75(340MHz) WinBench97:
Nine
R3D16M 1.46 Drivers: Optimized vs. Standard
- 128MPEDO NineR3D16M
AHA2940UW ST34501W (Cheetah) FAT32. Note: Not surprisingly,
the primary affect of these drivers is on video performance. The
Standard drivers are about 10% slower than the Optimized drivers(MicroStation
excepted). Interestingly, the Standard drivers give a small, but
consistant increment in CPU32.
January 16, 1998.
-
HX83/200MMX:Matrox
Millennium drivers - 3.63 vs. 3.70 Winbench97:
128M PEDO 3X75 Millenium 4M 3.63 ST34501W (Cheetah) 1024X768X16@85 Millennium
4M Adaptec AHA2940UW ST34501W (Cheetah UW) FAT32AMI BIOS 1.75 Matrox BIOS
2.5 Note: These results show that there is about a 10% decrement
in performance in the 3.70 drivers relative to the 3.63 drivers. August
16, 1997
-
HX83
P200MMX
3X75 (225MHz): Millennium Driver 3.61 vs. 3.62 at 08, 16, and 32 bit Color
Depth: MrBIOS 3.41 128M Toshiba PEDO Symbios 875UW ST34501W
(Cheetah UW) FAT32 - Winbench97
Note: The increased performance for 3.62 over 3.61 is 5-7%. This is similar
to that observed for 3.62 over 3.41 (not published at this site). The 128M
PEDO configuration failed at 3X75 with a 2940UW. For this particular configuration,
you need the Symbios 875.
-
HX83
P200MMX
3X75 (225MHz): Millennium Bios 2.2 vs. 2.5 at 08, 16, and 32 bit Color
Depth: 1024X768X-08,16,32@85 HX83/512 3X75=225MHz 96M EDO 2940UW
ST34501W (Cheetah UW) FAT32 Note: In terms of speed, this is little difference
between BIOS versions - Winbench97
-
HX83
P200MMX
3X75 (225MHz):Millennium Vs Nine 128-2 at 08, 16, and 32 bit Color Depth:
HX83/512 3X75=225MHz 96M EDO AMI HX83V1.75 2940UW ST34501W (Cheetah
UW) FAT32 Note: These data use the new 3.61 Matrox drivers - Winbench97
-
HX83
P200MMX
3X75 (225MHz): Millennium 4M Vs. Nine 128-2 4M: 1024X768X16@85
HX83/512 3X75=225MHz 96M EDO AMI HX83V1.75 2940UW ST34501W (Cheetah UW)
FAT32 - Winbench97
-
SuperMicro P5STE/512: Millennium 4M
vs. Mystique 4M vs. 128-2 4M VRAM (1024X768X32@85) (P5STE/512 P166) - WinBench97
-
SuperMicro P5STE/512: Millennium 4M
vs. Mystique 4M vs. 128-2 4M VRAM (1280X1024X16@75) (P5STE/512 P166) -
WinBench97
-
These are the same charts as the Cheetah comparison data
-
SuperMicro P5STE/512: Number Nine
Imagine 128-2 Driver Benchmarks: Versions 309 vs. 416
Optimize vs. 416 WHQL (P5STE/512 P166) - Winbench
97
Miscellaneous
-
FX83 Diary: New comments regarding
the PII300 installation, the Nine R3D, and memory termination for the Pro
CPU. October 27, 1997.
-
HX83
P200MMX
3X75 (225MHz): WinBench97 - Ver. 1.0 Vs. Ver 1.1: 1024X768X16@85
HX83/512 3X75=225MHz 96M EDO AMI HX83V1.75 2940UW ST34501W (Cheetah UW)
FAT32 - Winbench97 Note:
Ziff Davis states that Disk WinMarks for WinBench97 Versions 1.0 and 1.1
are not comparable. Generally, Ver. 1.1 gives higher disk marks.
-
Hardware Locks and SM/AMI Socket 7 - Possible
Solutions
January 23, 1998
-
Diary of a Virus: The FORM.C
virus vs. my HX83 (this is not a benchmark)
-
Despite that I have WinBench98, I continue to
use Winbench97 -- Why?
January 31, 1998
This site is optimized for 1280X1024 resolution
Email: lae2@psu.edu
http: http://www.personal.psu.edu/lae2