Mid-level Graphics Card Shootout

Three video cards compared – Brian

SUMMARY: A three way battle between mid-level cards

[Ring Announcer voice]

“Ladieeeees and Genntlemennn. Welcome to the OCE* Main Event in the Graphics Card Middleweight Division.”

“We have three bouts on the program for tonight, two preliminary matches of which the winner will advance to the finals.”

“The rules are quite simple…each contestant must run with the same parameters. No cards will be overclocked. The AGP bus will be locked at 66MHz. All will use the same hard drive, a Western Digital
WD400BB model, and the same RAM, two sticks of 512MB (1 GB total) Corsair XMS PC3200C2 running in Dual Channel mode. All will use the same software format….Windows XP Pro, SP1, freshly
formatted before installing, MS Direct X 9.0b, and benchmarking programs only will be installed. The hard drive will then be defragmented before running the tests. Unreal Tournament 2003 will be patched to
v2225. All tests will be run at 1024 x 768 resolution, with a refresh rate of 75Hz. AntiAliasing will be off, and Anisoscopic Filtering will only be enabled in the AquaMark 3 test.”

“We have a pair of “rings” set up, one an AMD platform, and the other an Intel. We’ll start off with a look at each of those, and then move on to a closer look at our three contestants.”

[/End Ring Announcer voice]

The AMD System

AMD rig
  • Chieftec Dragon mid-tower case
  • Abit NF7-S v2.0 motherboard
  • AGP aperture in BIOS set to 256MB, 8X mode
  • AMD XP2100+ (Thoroughbred “B”), running at 2200MHz (200 x 11)
  • 2 x 512MB (1GB total) Corsair XMS PC3200C2 DDR, running in Dual Channel Mode, 1:1 ratio (200 FSB)
  • Western Digital WD400BB hard drive (ATA100, 7200 RPM, 40GB)
  • Custom watercooling on CPU and Northbridge, active air cooling on Southbridge and mosfets
  • 420W power supply
  • Windows XP Pro, SP1

The Intel System

P4 rig
  • Raidmax Scorpio 868W mid-tower case
  • Abit IC7 v1.0 motherboard
  • AGP aperture in BIOS set to 256MB, 8X mode
  • Intel P4 1.6A (Northwood), running at 2400MHz (150 x 16)
  • Same DDR as above, running in Dual Channel, 4:5 ratio (187FSB)
  • Same hard drive as above
  • Swiftech MCX4000 w/ 80mm Sunon 39CFM fan on CPU, modified HSF (made from the center slug of another MCX4000) on the Northbridge, passive heatsinks on clock generators, voltage regulator, and mosfets
  • 420W power supply
  • Windows XP Pro, SP1

While the Intel machine owns a slight advantage in overall CPU speed (2400MHz to 2200MHz for the AMD rig), it gets there via a very differing route. The AMD CPU, with it’s unlocked
multiplier, allows for finer tuning of the system. Because of this, I’m able to lower the multiplier, raise the bus speed, and run the Corsair memory used at it’s specified speed of DDR400 (PC3200, or 200FSB).

Where the multiplier of the Intel CPU is locked (this particular model uses a 16x multiplier), the CPU tops out well before 200FSB, and needs a goodly boost of voltage to get to the 150FSB setting used here.
1.75v, to be exact. The Abit motherboard usually undervolts the CPU, so most of the time, it floated in the area of 1.725 to 1.73v. This is borderline dangerously high for this CPU (some might say insanely high).
It normally is set up to run at 144 x 16 (2300MHz) at 1.65v (the motherboard usually reads 1.63ish for voltage there).

Suffice to say, immediately after testing each card, I backed off the FSB and voltage.

The AMD CPU runs using 1.7v to get to it’s 2200MHz overclock. To go faster than this requires the voltage to be raised extremely high as well. Even with the watercooling setup used on it, the
temperatures start climbing fairly quickly above these settings.

Where the memory was run at spec (AMD) or under (187FSB, Intel), I left the memory voltage at default. I did the same for the AGP voltage, where the bus was locked at 66MHz. These two settings apply to
both systems. 187 was the fastest I could get the memory to run in the Intel system. I tried several different ratios in the BIOS, and even with maxing out the voltage on this, I couldn’t get anything higher to boot.
So, I ran it at 187.

I now see at Abit’s website, that the setting’s I wanted might have been limited to me by a BIOS problem, as it seems this issue was addressed in the newest BIOS revision. Unfortunately,
the cards have already been returned to Newegg, so I can’t test that out.

Hindsight truly is 20/20 vision.

Next up, a look at the “contestants”

( * OCE = “OverClockers Entertainment”)

{mospagebreak}

A word on overclocking…

Since the two higher end cards featured here were reviewed (9600 Pro
and 5700 Ultra), I’ve received a number of e-mails asking why I didn’t overclock the cards.

The reasons for this are twofold. first, I wasn’t entirely certain how Newegg would look upon that, where the cards needed to be returned to them after testing.
If I knew they’d have no problems with me winding them up, I would have. I was also hesitant to do so, as right now, both my wife and I are out of work, and if something were to happen
to one of them, I can’t afford to replace it.

It’s been well documented around the internet, that the 9600 Pro does in fact, overclock extremely well. I will admit to this…in the process of installing/configuring the software for the 5700 Ultra,
there is facility included to overclock this card as well. Automatically, in fact…to the highest stable setting. It was kind of a “What does this button do?!?” thing, that I stumbled onto.

I cannot recall the GPU speed, but I do remember the memory clock hitting 1.01GHz (bear in mind, default voltage here) when the program had finished overclocking the card for me.

So I’d be inclined to state the 5700 Ultra might have some good overclocking potential as well. If the memory will bag a 12% or so increase, accidentally and at default voltage, one might expect a good deal
more if you really tried…

Another thought I had on this, was that every card, like every CPU, overclocks differently. Let’s say I managed to overclock the Radeon to XXXMHz core/XXXMHz memory, and the FX5700 to XXXMHz/XXXMHz. There’s going to be someone out there
who’ll say “Dude….ya shoulda pushed _____ card harder. I’ve got one and it’ll do XXX/XXXMHz…way more than you got. If you tried harder, it woulda blown away that other card.”

Running everything at the default settings removes this situation from the equation entirely.

The Cards, and Specs

4200 logo

4200 front

Detailed Specifications

  • Manufacturer: BFG Technologies
  • Model: “Asylum” GF4 Ti 4200
  • Graphics Controller: NVIDIA ® GeForce4 ™ Ti 4200
  • Memory Configuration: 128MB DDR
  • Connectors: VGA out(15 Pin D-Sub), TV-Out(S-Video), DVI Connector
  • Bus Interface: AGP 1X/2X/4X
  • Maximum Resolution: 2048 x 1536 @ 85Hz
  • Video-Out: Supported (TV Tuner and Video-in N/A)
  • Clock Speed: 250MHz
  • Memory speed: 444MHz
  • Operating Systems Support: Windows ® XP/2000/NT/ME/98/95 OSR2
  • 3D Acceleration Features Supports: DirectX ® 8.1, 256-bit 3D and 2D graphics accelerator, NVIDIA ® nfiniteFX ™ II Engine, NVIDIA ® nView ™ display technologies,
    NVIDIA ® Lightspeed Memory Architecture ™ II, NVIDIA ® Accuview Antialiasing ™, Advanced programmable Pixel Shaders

(NOTE: This card is equipped with an “aftermarket” HSF, made by ThermalTake)

GFFX logo

5700 front

Detailed Specifications

  • Manufacturer: Gainward
  • Model: FX5700 Ultra “Golden Sample”
  • Graphics Controller: GeForce FX 5700 Ultra
  • Memory Configuration: 128MB DDR(128 bit)
  • Connectors: VGA Out(15 Pin D-Sub), TV-Out(S-Video), DVI connector (TV Tuner and Video-in N/A)
  • Bus Interface: AGP 1X/2X/4X/8X
  • Maximum Resolution: 2048 X 1536 @ 85Hz
  • Clock Speed: 475MHz
  • Memory speed: 900MHz
  • Operating Systems Support: Windows ® XP/2000/NT/ME/98
  • 3D Acceleration Features Supports: DirectX ® 9.0, NVIDIA CineFX engine, nView Multi-Display Technology, NVIDIA NVRotate, NVIDIA NVKeystone
  • Requires 250W or higher power supply, with an available 4 pin Molex power connector

ATi logo

9600 front

Detailed Specifications

  • Manufacturer: Sapphire
  • Model: RADEON ™ 9600 PRO “Ultimate Edition”
  • Graphics Controller: RADEON ™ 9600 PRO
  • Memory Configuration: 128MB DDR
  • Connectors: VGA out(15 Pin D-Sub), TV-Out(S-Video), DVI Connector
  • Bus Interface: AGP 1X/2X/4X/8X
  • Maximum Resolution: 2048 x 1536 @ 85Hz
  • Video-Out: Supported (TV Tuner and Video-in N/A)
  • Clock Speed: 400MHz
  • Memory speed: 600MHz
  • Operating Systems Support: Windows ® XP/2000/NT/ME/98
  • 3D Acceleration Features Supports: DirectX ® 9.0, the latest OpenGL ® functionality, SMARTSHADER ™ 2.1, SMOOTHVISION ™ 2.1, VIDEOSHADER ™, and ATI’s new FULLSTREAM ™ technology

Drivers used

BFG Tech (nvidia) GF4 Ti 4200 : nVidia Detonator v43.03
Gainward (nvidia) GF FX5700 Ultra : nVidia Detonator v53.03
Sapphire (ATi) 9600 Pro : ATi Catalyst v3.10

Benchmarks used

  • 3D Mark 2000 (DX7)
  • 3D Mark 2001 SE (DX8)
  • 3D Mark 2003 (DX7, 8, and 9)
  • PC Mark 2002 (System)
  • PC Mark 2004 (System)
  • Unreal Tournament 2003 Benchmark (DX8)
  • AquaMark 3 (DX9)
  • Shadermark 2 (DX9 programmable shaders)

Let’s go “ringside” for the first round

{mospagebreak}

In our first round, let’s see how the GF4 Ti 4200 stacks up against the GF FX5700 Ultra.

gloves57x42

* DING! *

Benchmark scores

All benchmarking was done at 1024 x 768, 75 Hz. The default settings were used with each individual program. For example, 3D Mark 2000 runs at a 16 bit depth, with triple buffering,
16 bit Z-Buffering, and Texture formatting. 3D Mark 2001SE uses 32 bit depth and 24 bit Z-Buffering for default settings. Whatever came up when the programs were first run, is how they were run.

I ran each of them a minimum of three times thru, and kept the best (highest) score. The ShaderMark v2.0 benchmark doesn’t give a “total” score like the rest I used, but rather a
complete breakdown of each individual test within the benchmark itself. A screenshot of this in Notepad from each system is included below. The 5700 Ultra didn’t have support for several
of the tests in ShaderMark 2, looking closely at the screenshots will show numbered lines missing. Those tests were skipped due to non support. The Ti 4200 would not run this benchmark at all.

Scores in Italics represent the highest in that test, between the two cards and systems

~~ Card ~~> —————— Benchmark

Ti 4200 AMD system

Ti 4200 Intel system

FX5700 AMD system

FX5700 Intel system

3D Mark 2000

14535 Marks

12312 Marks

14971 Marks

13013 Marks

3D Mark 2001SE

12112 Marks

10675 Marks

13778 Marks

12924 Marks

3D Mark 2003

1426 Marks*

1566 Marks*

3892 Marks

3895 Marks

Aquamark 3

15454

15473

29182

29163

Unreal Tournament 2003 Flyby

135.580

131.859

166.534

162.392

Unreal Tournament 2003 Botmatch

70.345

61.662

73.835

62.393

PC Mark 2002 CPU

6781

5949

6774

5945

PC Mark 2002 Memory

5431

7697

5759

8188

PC Mark 2002 HDD

696

675

670

668

PC Mark 2004

3704

3151

3878

3277

Shadermark 2

Would not run

Would not run

See below

See below

(* NOTE: The Ti 4200 would not run “Game 4 – Mother Nature”, and the Feature Test “Pixel shader 2.0” in 3D Mark 2003)

5700 ShaderAMD

FX5700 ShaderMark v2.0, AMD system (above), and Intel (below)

5700 ShaderIntel

(Missing (not supported) were test numbers 1, 4, 8, 19 through 23)

It’s interesting to note, that the FX5700/AMD combo swept all but four of the benchmarks above. The four it missed on were (the combo that led in parentheses):

  • 3D Mark 2003 (FX5700/Intel)
  • PC Mark 2002 CPU test (Ti 4200/AMD)
  • PC Mark 2002 HDD test (Ti 4200/AMD)
  • PC Mark 2002 Memory test (FX5700/Intel)

It’s also noteworthy, that in the DX7 and DX8 tests, the Ti 4200 stood up fairly well against the FX5700. Only when DX9 tests were used, did it tail off significantly,
or refuse to run the tests, due to lack of support. Also noteworthy is that while most of the scores between the two systems are fairly close to each other, one score
stands out from that form, and is a total rout. The Memory Test score from PC Mark 2002. In this test, the Intel machine repeatedly led by a wide margin, in every test, with every card.

“And the winner is:”

Moving on to the finals, in a split decision, will be the FX5700 Ultra.

Moving on to our second matchup…

{mospagebreak}

In our second round, let’s see how the GF4 Ti 4200 fares against the Radeon 9600 Pro.

gloves96x42

* DING! *

Benchmark scores

The same formula was used in the benchmarks here, and throughout all tests, as outlined on Pg. 3

Again, as noted before, the Ti 4200 would not run “Game 4 – Mother Nature”, and the Feature Test “Pixel Shader 2.0” in 3D Mark 2003, and would not run the Shadermark 2 benchmark at all. The 9600 Pro
was able to run the Shadermark 2 tests in their entirety.

~~ Card ~~> —————— Benchmark

Ti 4200 AMD system

Ti 4200 Intel system

9600 Pro AMD system

9600 Pro Intel system

3D Mark 2000

14535 Marks

12312 Marks

14539 Marks

12987 Marks

3D Mark 2001SE

12112 Marks

10675 Marks

11691 Marks

11270 Marks

3D Mark 2003

1426 Marks*

1566 Marks*

3371 Marks

3374 Marks

Aquamark 3

15454

15473

26334

26328

Unreal Tournament 2003 Flyby

135.580

131.859

148.785

145.832

Unreal Tournament 2003 Botmatch

70.345

61.662

72.781

61.730

PC Mark 2002 CPU

6781

5949

6793

5900

PC Mark 2002 Memory

5431

7697

5837

8302

PC Mark 2002 HDD

696

675

669

677

PC Mark 2004

3704

3151

3925

3337

Shadermark 2

Would not run

Would not run

See below

See below

9600 ShaderAMD

9600 Pro ShaderMark v2.0, AMD system (above), and Intel (below)

9600 ShaderIntel

This time through, there was again a dominant combo, that being the 9600 Pro/AMD pairing. It also led all but four of the tests, those being:

  • 3D Mark 2001 SE (Ti 4200/AMD)
  • 3D Mark 2003 (9600 Pro/Intel)
  • PC Mark 2002 HDD test (Ti 4200/AMD)
  • PC Mark 2002 Memory test (9600 Pro/Intel)

I think I see a pattern emerging here

In the DX7 and DX8 tests here, the Ti 4200 had another strong performance against the 9600 Pro. It just barely missed eeking out a win in the 3D Mark 2000 runs, again ran well in the UT2003
Botmatch, and took 2nd and 3rd (out of four scores) in the PC Mark 2002 CPU tests.

And again, when DX9 tests were run, the Ti 4200 trailed by a much wider margin, due to the lack of DX9 and pixel shader support. Also, the scores system to system remained close, with the
exception of the Memory Test score from PC Mark 2002, which was again dominated by the Intel system.

“And the winner is:”

Moving on to the finals, in a split decision, will be the Radeon 9600 Pro.

Moving on to our final matchup…

{mospagebreak}

[Ring Announcer’s voice]

“Ladieeeees and Genntlemennn. Welcome to the Chammmmpeeeeenship Round of the OCE Main Event in the Graphics Card Middleweight Division.”

“In the black PCB with white trim corner, weighing in at 400MHz, fresh off it’s split decision victory over the Ti 4200, the Sapphire Radeon 9600 Pro!”

“In the red PCB with red trim corner, weighing in at 475MHz, also fresh off it’s split decision over the Ti 4200, The Gainward GF FX5700 Ultra!”

“Ok boys….I want a clean fight. No hitting below the heatsink. When the bell rings, come out swingin’.”

[/End Ring Announcer’s voice]

gloves96x57

* DING! DING! DING!*

Benchmark scores

The same formula was used in the benchmarks here, and throughout all tests, as outlined on Pg. 3

Again, as noted before, The 5700 Ultra didn’t have support for several of the tests in ShaderMark 2, and looking closely at the screenshots will show numbered lines missing.
Those tests were skipped due to non support. The 9600 Pro was able to run the Shadermark 2 tests in their entirety.

~~ Card ~~> —————— Benchmark

FX5700 AMD system

FX5700 Intel system

9600 Pro AMD system

9600 Pro Intel system

3D Mark 2000

14971 Marks

13013 Marks

14539 Marks

12987 Marks

3D Mark 2001SE

13778 Marks

12924 Marks

11691 Marks

11270 Marks

3D Mark 2003

3892 Marks

3895 Marks

3371 Marks

3374 Marks

Aquamark 3

29182

29163

26334

26328

Unreal Tournament 2003 Flyby

166.534

162.392

148.785

145.832

Unreal Tournament 2003 Botmatch

73.835

62.393

72.781

61.730

PC Mark 2002 CPU

6774

5945

6793

5900

PC Mark 2002 Memory

5759

8188

5837

8302

PC Mark 2002 HDD

670

668

669

677

PC Mark 2004

3878

3277

3925

3337

Shadermark 2

See below

See below

See below

See below

9600 ShaderAMD

9600 Pro ShaderMark v2.0, AMD system (above), and Intel (below)

9600 ShaderIntel

5700 ShaderAMD

FX5700 ShaderMark v2.0, AMD system (above), and Intel (below)

5700 ShaderIntel

(Missing (not supported) were test numbers 1, 4, 8, 19 through 23)

[Ring Announcers voice]

“Ladieeeees and Genntlemennn. We’re awaiting a final decision from the judges. This fight was not a “knockout” for either card, and went the full eleven rounds. We’ve got
another “split decision” here folks. Let’s go over to the scorer’s table and see who won.”

[/End Ring Announcers voice]

“Round by Round” breakdown

(System the highest score was achieved with in parentheses.)

  • 3D Mark 2000 : FX5700 (AMD)
  • 3D Mark 2001 SE : FX5700 (AMD)
  • 3D Mark 2003 : FX5700 (Intel)
  • Aquamark 3 : FX5700 (AMD)
  • UT2003 Flyby : FX5700 (AMD)
  • UT2003 Botmatch : FX5700 (AMD)
  • PC Mark 2002 CPU : 9600 Pro (AMD)
  • PC Mark 2002 Memory : 9600 Pro (Intel)
  • PC Mark 2002 HDD : 9600 Pro (Intel)
  • PC Mark 2004 : 9600 Pro (AMD)
  • Shadermark 2 : 9600 Pro (AMD)

[Ring Announcers voice]

“We have a winner folks. In a split decision, winning by a score of six rounds to five, our new Chammmmpeeeeen is the Gainward GF FX5700 Ultra!

[/End Ring Announcers voice]

Final thoughts next

{mospagebreak}

CONCLUSIONS

These scores are what I was able to achieve, running these cards at default speeds. Naturally, overclocking them would get higher scores for all of the cards used here.

What I was looking for was a “baseline”, to judge by.

There are no guarantees in overclocking. You might get a card that does very well, or one that doesn’t. If I had one card here that did, and two that didn’t, it could/would skew the results considerably.
Then I’d have people crying “Foul!”

These are the scores I got….your mileage may vary. Your system likely varies from mine, and if nothing, I’ve proven that two systems using several common components can realize drastically different results.

The FX5700 did in fact win 6 of 11 of the benchmarks run. Some slightly, some by a larger margin. It fared better in the DX8 and DX9 graphics tests. Conversely, the 9600 Pro dominated the “System” tests, and the
pixel shader specific “Shadermark 2” tests.

So it is also apparent, that the video card you use can and will significantly effect your score in the Futuremark PC 2002 and 2004 tests.

This is also, in part, why the FX5700 emerged “victorious” here. This was a graphic capabilities test, with system benchmarks added, to see how much different video cards effected those scores. But primarily a
graphics test. And in those, the FX5700 clearly led.

However…

In all honesty, the GF4 Ti series cards still hold up fairly well, despite their limitations. The Ti 4200 card discussed here is an older AGP 2.0 (4x) able card, with a fairly (comparatively)
slow GPU clock speed and memory speed. This card is towards the “low end of the food chain” in the Ti Series family. The 64MB version of this same card actually uses a faster memory speed
(500MHz to the 444MHz of the card used), and also features an 8.0GB/second memory bandwidth for the 64MB card versus 7.1GB/second for the 128MB card

This card still has no problems however, running UT2003 with all of the eye candy on it’s highest setting, and very infrequently comes across anything that slows the framerates to a crawl. But it does
happen. I’ve been playing the latest “TOMB RAIDER” (“Angel of Darkness”) using this card, and it plays very well with it. So do a lot of other games.

Is the Ti 4200 the “best thing since sliced bread”?

Hardly.

But it’s still a pretty tasty “dinner roll”, that’s only just starting to go stale. =)

It held up very well in the DX7 and DX8 benchmarks, and some of the system tests. Even winning two tests against each of the DX9 cards.

So, unless you require the pixel shaders, and DX9 support, do you really need to upgrade today to a card that offers them? There isn’t a glut of DX9 specific games on the market (yet), and
the two highly anticipated games that will (most likely) require an upgrade (DOOM III and Half Life 2) still have yet to be released as I write this.

Which will come first, those games, or the dawning of the new “PCI Express”, which when implemented, will require an entirely different graphics card format, and new cards for everyone?

These are decisions you and your checkbook need to figure out. Obviously, if money were no object to you, you’re reading this on a machine that’s (likely) already got a 9800 Pro or FX5950 Ultra installed
in it. Those who don’t, and are weighing the pros and cons of the semi-major upgrade to their system that a graphics card is are the ones this article is for.

Cheers!

Email Brian

Be the first to comment

Leave a Reply