Video Cards and Benchmarks

[Skuzzy]

Well,...I have finally gotten overwhelmed by all the subjective rantings on which video card is best for AH and others and decided to run a suite of benchmarks for more objective comparisons.

One of the things that has puzzled me in the past and is a very difficult one to answer is; When does the CPU speed stop being very beneficial to the actual video performance?
Also, when does the monitor become the bottleneck?  And finally, where does the video card start suffering in performance?

So I am putting together an Excel spreadsheet of the above factors.

Here is what I am doing;
1) I will use the 3DMark2000 v1.1 benchmark suite for the benchmarks.  Why?  It has some benchmarks in the tests that can give numbers, not only on triangle performance, but also polygon count and texel performance.  These are all pretty low level, but can show tell-tale signs of what a video card can do best.  I will pretty much ignore the 3DMark value and use the individual test results.

2)  The suite will be run, initially, on processor speeds of 600E and 800E Intel Pentium III Coppermine Processors (Slot 1).  I will also test 667EB, 800EB, and 933EB.

3)  The following graphics modems will be tested:  640x480, 800x600, 1024x768, 1200x1024, and 1600x1024, in 16 and 32 bit modes.

Each suite will be run after a clean reboot.  Only the basic Windows 98SE OS will be installed, along with 3DMark2000.  When changing video cards, a fresh install of the OS will be done to ensure no bits or pieces of the video drivers will still be hanging around.
When multiple drivers for video cards are available, a fresh install of the OS will be done to see if there are any differences and the versions tested will show in the charts.

The tests will be conducted for DX7 and DX8 (if the video card manufacturer can supply DX8 certified drivers).  

All the tests will be run 3 times and each will be averaged to yield the final number.

Where FSAA is available, I will enable each level and test as well.

I will also be defining some standard views in AH to get the frame rates for each of the combinations.

I will be purchasing the video cards from the shelf, instead of trusting the manufacturer to send me what they are currently shipping, so it will take some time to get all this done.

I currently have completed the ATI Rage Maxx at 600E and 800E CPU speeds and will be testing the ATI Radeon 64MB DDR card today.
The GeForce cards are going to be a bit of a pain, as there are a couple of resellers of those cards, so if the community would let me know which you use it would help me decide which card(s) I should look for.
I may or may not do the VooDoo cards.  With them effectively shut down, it may be a waste of time.  I think I may wait until a formal announcement of what is going to happen with them is forthcoming.

Now, the big question:  Would any of you find this information useful?  I plan on posting it on my WEB site at AppLink, if I can get some feedback from this community as to whether or not it would be of any benefit.

Thank you for any feedback.


------------------
Roy "Skuzzy" Neese
President, AppLink Corp.
http://www.applink.net
skuzzy@applink.net

[Toad]

Skuzzy, great idea! Thanks for your work in advance.

Here's one you might look at since it's about the cheapest, basic OEM style GeForce2 GTS card out there:

Evga GeForce2 GTS 32MB, currently on Pricewatch for @ $175 plus shipping.

My system is:

MoBo: Old, early Abit BH-6 running the Celeron through ASUS Slotket

CPU: Celeron 566 Overclocked to 850. (Very stable for last 5 months.)

Ram: 192MB of 100Mhz SDRAM

Video Card: eVGA GeForce 2 GTS, 32 MB AGP

Sound Card: SB Live Retail

Disk Drive: IBM 10GB 7200 rpm EIDE

Joystick: CH Products Force Feedback, CH Pro Throttle, CH Pro Pedals
(attached to SB game port, driver 4.06.711)

OS: Win98  (Version 4.10. Build 1998)

DirectX : 7.0 (4.07.00.0700)

This setup in AH (1600x1200, 16 Bit) gives me
frames in the low 40's looking out the front of a P-51 in flight most times.

I had a "blue flash" problem when I installed the card. Somebody suggested GeForce Tweak to tune the card and that program cured the problem when I enabled all the FSAA options to max and used a beta tweak called vsynchscene. This DID slow the frames down about -15 frames. Before with the random flashes I was getting mid-50's on frames.

Good luck with the tests and thanks!

[BaneX]

Actually I think it would be very useful. Would love to see data on the Diamond cards as well such as the Diamond Stealth Xtreme and the Diamond V770.

Bane
13th TAS

[Ripsnort]

Great idea, Skuzzy, and thks!

[Lephturn]

That would be great stuff Skuzzy.  

Some of the decisions are subjective ones, such as quality.  I find that many times the "numbers" can be misleading, especially when hardware sites do things like quote FPM numbers for Quake 3 at 640x480.  3D Mark is pretty widely accepted and as you mention, give some good details.  This kind of thing is great for figuring out what cards are more or less CPU limited, so we know what card is better for those trying to upgrade a lower spec system.  The numbers are not the only thing I would base my purchasing decision on, but it's a big piece of the pie.

The most widely reccomended card, and the best value right now IMHO, is the GeForce 2 MX.  Although there are many different folks that produce the Geforce cards, I think you can pick one of the common ones like Diamond or Hercules and the others would all be damn close.  An ATI Radeon would also be a good one to use if it is available.  As for the Voodoos... well it is getting to be a moot point.  I think you'll have plenty to do without getting into that.  If you do it at all, it would likely be best to test the Voodoo 3, as there are lots of folks out there with that card.  Those of us with V5's are not going to be looking to upgrade for a while, and they likely won't be selling a whole lot of them at this point, so that doesn't sound like a good time investment IMHO.

If you want anybody to doll the results up and get them up on a web site for you, I'd be willing to do that.  

------------------
Lephturn - Chief Trainer
A member of The Flying Pigs  http://www.flyingpigs.com
 
"A pig is a jolly companion, Boar, sow, barrow, or gilt --
A pig is a pal, who'll boost your morale, Though mountains may topple and tilt.
When they've blackballed, bamboozled, and burned you, When they've turned on you, Tory and Whig,
Though you may be thrown over by Tabby and Rover, You'll never go wrong with a pig, a pig,
You'll never go wrong with a pig!" -- Thomas Pynchon, "Gravity's Rainbow"

[SOB]

Sounds like a great idea Skuzzy, and I would have loved to have something like this when I upgraded my video.  If you do decide to throw up the Voodoo 5, email me.  I've got an extra card I bought when they dropped in price that I won't be using for a while, and I'd be happy to let ya borrow it if ya need.  


SOB

[Staga]

Skuzzy be sure you won't run out of beer when playing with different cards and especially with different drivers  
When/If I got some free time I'll check what I can do with this Prophet2 MX...

Yesterday I was playing with some different drivers (Official and Beta) and it looks like Prophets own drivers (6.35) are giving best results in 3D-Mark 2000.

[1776]

Hmmmmmmm, prolly when ya get your testing completed a whole new generation of cards will be out!!  I am amazed at the new cards all the time.  I don't have any idea of what card would be the best as I think that that is a moving target!!!

Sounds like you may not be in AH much as you will be scraping knuckles putting cards in and out, heheeeeeeee.  

Good luck in this project!!

[Frosty1]

Sounds pretty kick ass.  

------------------
===>Frosty
====>Exposure2k.com
=====>Frosty@exposure2k.com

[mason22]

skuzzy, awesome man, i can't wait!!

also, if you need ANY help on testing or anything else i will be more than happy to do what i can!! (would like to volunteer!)

i'm one of those slower end CPU with hi-end video card configuration guys:

PIII600E (i have run it at 800, 850, and 950 ALL stable) am currently at 850.
256 pc133 (hitachi)
ASUS p3v4x VIA Apollo Pro133 chipset
Hercules 3D Prophet2 GTS ULTRA 64megs (running AGP of 2x untill i get the reg edit for 4x)
4.3gig Cheetah 10,000rpm SCSI drive

i've been playing around with 3Dmark also, so i am somewhat familiar with tweeks and stuff.


feel free to email if you want jsa@wlion.com



[This message has been edited by mason22 (edited 01-11-2001).]

[bloom25]

One thing to add here.  The GF2 MX is the best value right now, but you can find older GF1 DDR cards for about the same price.  I'd be curious to see how they compare to the MX.  (My bet is a blow out in favor to the GF1 DDR as the resolution gets larger.)  



------------------
bloom25
THUNDERBIRDS

[Staga]

Heres something....

 3DMark results

This chart gives some view for different drivers with "Stock" core175/memory183Mhz and when oc'd to 220/230. Chart

Few more charts: 3DMark Results, Helicopter and Adventure Charts, Texture Chartand Polygon counts

HeresExcel-chart

Staga


[This message has been edited by Staga (edited 01-12-2001).]

[mason22]

<sit rep? punt_MAW   >


how are the test going skuzzy? need any help?

[whels1]

look here for a complete comparision of the Geforce cards from gf1 to the gf2 ultra. http://www4.tomshardware.com/graphic/00q3/000814/geforce2ultra-07.html


whels

 

Originally posted by bloom25:
One thing to add here.  The GF2 MX is the best value right now, but you can find older GF1 DDR cards for about the same price.  I'd be curious to see how they compare to the MX.  (My bet is a blow out in favor to the GF1 DDR as the resolution gets larger.)  

[Skuzzy]

Benchmarks are coming along.  I keep finding things I need to test (i.e. overclocking, different drivers, different CPU's...)
Takes a full day to test each combination.

I currently have tested the ATI Rage Maxx and ATI Radeon 64MB DDR cards in both 600E and 800E processors.  The Radeon has tests using FSAA, w/o FSAA, w/Intel MMX instructions, and w/D3D T&L support.
I also added information regarding what the horizontal and vertical sync rates the video cards use at the different resolutions.  I get this from the monitor OSD menu.
Slow process, but some intersting information is showing up regrading what CPU is overkill and which one isn't.

I am trying to figure out the best screen layout for the WEB page, as there is a ton of info.
Also working on a benchmark film for AH.

I have not acquired the GeForce cards yet, and may take up those that want to contribute a card for the benchmark.  With all the models they have, I could spend a thousand bucks on them alone.

I appreciate the offer for help mason22, but for these benchmarks to have any meaning, they must be run on the same system, with the same software configurations.


------------------
Roy "Skuzzy" Neese
President, AppLink Corp. http://www.applink.net
skuzzy@applink.net


[This message has been edited by Skuzzy (edited 01-19-2001).]