d3v0

Greetings all. My name is d3v0 and I am a graphics card addict. Now that we have that out of the way, let's talk about this thread and it's implications. Over the past 6 months I have purchased, benchmarked, and resold 16 different graphics cards, at a relatively small cost (I lost the cost of shipping on nearly every card) and was able to punch out a multitude of statistics relative to each.

The name if my system came to be not at its conception but rather once I noticed what it ended up becoming: a test bench. This system is so common around sites like ours and a medium-high overclock on quite possibly the most cost-efficient processor on a very cost-efficient overclockable mainboard made this project very important to me and to you, the consumer. I am a quasi-enthusiast, but like most of you I also live on a budget and cost-efficiency is incredibly important to me. The GURU3D VGA charts are weak and while they offer a mass of data, most of it is apples and oranges with the powerful driver releases as of late. However, they can be commended for putting the information out there, as I have quoted it numerous times and I thank the guys at GURU3D.com for putting this idea in my head.

All of these graphics cards were tested with IDENTICAL drivers. I did this to ensure the most fairness and to elude as many confounding variables as possible. **Please note that in Far Cry 2 I used the 180.60 beta drivers for Nvidia due to 177.92's lack of support).

Each Nvidia card was tested using the Xtreme_G 177.92s for Vista x64
Each ATI card was tested using the Xtreme_G 8.12s for Vista x64

I decided upon the Xtreme_G drivers due to their popularity with OCN and similar websites with consumers who can appreciate the minor boosts they offer over regular drivers. When I started testing in October 2008 the 177.92's were fastest for Nvidia. When I started testing ATI cards in November, the 8.12's, despite some criticism, outperformed my 8.11s.

I went through an OCD type process when swapping cards:

1.) Uninstall XG via Ccleaner
2.) Uninstall Display via CCleaner
3.) Reboot into safe mode and run driver sweeper, sweeping Nvidia/ATI display.
4.) Insert new card and reboot. Load drivers.
5.) Upon restart, run PerfectDisk, anti-virus, and spyware cleaner.
6.) Reboot into windows, close all processes on system tray ONLY and make sure my desktop was set to display at "check for best appearance" (this was to ensure a realistic quality to the benches, not like a suicide bench)
7.) Run all stock benchmarks. Save for nvidia's FC2, which I did after all other stock and overclocked benches were done.
8.) Use a combination of ATI tool and Crysis GPU benchmark looped to find the HIGHEST STABLE possible overclock on the card. Some of these I could really push (3850/70) and others didn't budge (19xx series). I also made sure to denote the overclock as pertaining to its actual strap on most GPUs(found in rivatuner monitoring). Factory overclocked GPUs (BFG 9600GT and 9800GTX+ were set to their reference stock clocks for the "stock" tests.)

Another note that is terribly important, I pencil vmodded ANY video card with an easy vmod and chose its stable overclock AFTER the pencil vmod as its final overclock. Most of us own a pencil and a multimeter (if you don't they're <$20 at Home Depot) and can perform a pencil mod with little trouble if temperatures are good. This is another cost effective improvement that I would suggest to anyone

Pencil Mods Performed:
HD4850: 1.31vgpu
HD3850: 1.417vgpu
HD3870: 1.48vgpu
8800GT: Bios 1.1Vmod
8600GT: Pencil mod ineffective - Temperatures were awful
7950GX2: Pencil mod performed, ineffective.

Big ups to the HD3000 series for being so amazingly vmoddable/overclockable (and running cool to boot!)

Attached Files

GPU benchmark battery.xls (33.5 KB, 441 views)

__________________

d3v0 - 16 Graphics Cards - Old/New Stock/Overclocked/Vmodded: Extensive Benchmarks

d3v0

Let's move along now and meet the cards and their overclocks.

__________________

d3v0 - 16 Graphics Cards - Old/New Stock/Overclocked/Vmodded: Extensive Benchmarks

d3v0

Benchmarks - I originally had never planned to make such a hulking monster of a spreadsheet. It just sort of formed over a while of OCD personal testing. Sharing this with the community was always something I had really wanted to do, but how could I present this in the proper format? Well 3DMarks are an easy start. What games? I only had three with built in benches, Crysis, Far Cry 2, and FEAR. All of them have their quirks. I will go into detail about them later on in the thread. Just know that I added Far Cry 2 and 3DMark2006 a month after I began this project so the 8800GT is missing from the two of them.

3DMark2001 SE:
Settings: All default.
Scoring: 3DMarks overall

3DMark2003:
Settings: Default
Scoring: 3DMarks overall w/ CPU score noted (on spreadsheet only)

3DMark2005:
Settings: 1680x1050, rest default
Scoring: 3DMarks overall w/ CPU score noted (on spreadsheet only)

3DMark2006:
Settings: Default
Scoring: 3DMarks overall and SM 3.0 w/ SM 2.0, CPU score (on spreadsheet only

3DMark Vantage:
Settings: Performance Preset
Scoring: GPU overall score w/Pxxxx and CPU score (on spreadsheet only)

Furmark:
Settings: 1680x1050, 60s Benchmark
Scoring: o3Marks w/ Min/Max/Avg scores (on spreadsheet only)

Crysis GPU Benchmark:
Settings: 1680x1050, All High, 2xAA, 64-bit, DX10, 9AM
Scoring: Avg FPS w/ Min/Max (on spreadsheet only)

FEAR Benchmark:
Settings: 1680x1050, 4xAA, 16xAF, All Maximum
Scoring: Avg FPS w/ Min/Max (on spreadsheet only)

Far Cry 2 Benchmark:
Settings: Ranch Small, 1680x1060, 4xAA, DX10, Vsync Off, Ultra settings
Scoring: Avg FPS w/ Min/Max (on spreadsheet only)

Methodology and choices of settings:

I have a few reasons for running the benchmarks at the posted settings. 3DMark2001, 2003, 2006, and Vantage are all very commonly listed benchmarks here, and I felt that giving their default scores would best fit my purposes here, with the "wow my score is 10 points higher lololol?!" etc etc.

1680x1050 is the most common gamer's resolution as well. Thus, when adding 3DMark05 I decided to put it at that resolution. Sure it skews the CPU score a bit I think but the results were poignant and not nearly as inflated as the stock 3DMark2001 examples, for example where my Q6600 carried the results fairly heavily.

I believe that playing games at max settings is the best way to immerse oneself in said game, especially in shooters with mediocre plotlines (don't get me wrong, I spent months on Chrono Trigger/FFVI on my SNES fairly recently). So wondering why I would run Crysis @ such ridiculous settings on an 8600GT/X1950PRO was to visualize the real difference between one of these cards and one of the newer/faster ones ($30 more for an 8800GT and it is 3-4x as fast in this game, for example). FEAR scales so well that running at max was quite enjoyable, even the X1950PRO could maintain decent frame rates. Obviously I was defaulting DX10 settings on the newer GPUs, but a few of the GPUs did not support it and I was forced to run the game in DX9 mode, slightly skewing the results, but not really by much. GPUs that old are simply going to fall by the wayside due to the older technology anyway even in DX9 mode, so the results are quite appropriate.

By this point, if you are still reading then I am disappointed in you. Check the spreadsheet!

__________________

d3v0 - 16 Graphics Cards - Old/New Stock/Overclocked/Vmodded: Extensive Benchmarks

d3v0

Now it's what we've all been waiting for. The graphs. These are going to be a work in progress, because right now I will be posting MASSIVE graphs in descending order, 4 GPUs (stock and overclocked) at a time. It is just too cumbersome to build a chart with 32 sets of data (16 GPUs stock and overclocked) and post it as a single image. Feel free to PM me with ideas! I don't mind editing this. I am hoping for a sticky :P

TEST SYSTEM:
Gigabyte P35-DS3L Rev. 2.0
Intel Core 2 Quad Q6600 @ 3600mhz (400x9)
G.Skill DDR2 1000mhz 5-5-5-15
WD Raptor X 150b w/OS, benches and games
WD500AAKS for media files (not utilized in benches)
Corsair HX620
Vista Ultimate x64

ALL Nvidia Cards tested with Xtreme_G 177.92 Vista x64 Drivers (except Far Cry 2, where 180.60 was used)

ALL ATI Cards tested with Xtreme_G 8.12 Vista x64 Drivers

GREEN = STOCK NVIDIA
DARK GREEN = OVERCLOCKED NVIDIA
RED = STOCK ATI
DARK RED = OVERCLOCKED ATI


3DMark2001 SE Results:




3DMark 2003 Results:




3DMark05 Results (NOTE: 8800GT and 7950GX2 NOT tested):




3DMark06 Results:



3DMark Vantage Results:

__________________

d3v0 - 16 Graphics Cards - Old/New Stock/Overclocked/Vmodded: Extensive Benchmarks

d3v0

Furmark Results (NOTE: 3870X2 not tested):



Crysis Results:



F.E.A.R. Results:



Far Cry 2 Results (NOTE: 8800GT and 7950GX2 NOT tested):

__________________

d3v0 - 16 Graphics Cards - Old/New Stock/Overclocked/Vmodded: Extensive Benchmarks

d3v0

Reserved for final thoughts and price vs. performance comparison (To be performed at a later date, at a better hour than 4 a.m.


ENJOY EVERYONE!!! Ask any questions right here in the thread and I will answer honestly to the best of my ability about my adventures.

ALSO: Let me know if you like the fill effect of the FC2 charts, as is fairly obvious, I am not a huge chart maker. But I have imporved (you should have seen the alpha and beta versions of this post) and am willing to further update at the behest of OCN

__________________

d3v0 - 16 Graphics Cards - Old/New Stock/Overclocked/Vmodded: Extensive Benchmarks

Lxcivic2k1

Wow, talk about detailed. Very nice work! I'm surprised that just the OC on the 3870 took it up 2k in 3D06 and 10 fps in FEAR.

__________________
PSN: JeffyBondaris
Games owned: GTA:IV and Burnout Paradise
PC Games:Mainly Play CSS, TF2 but I also have CoD 4, UT3, DoD:S, Crysis, Grid
Xfire: Silentkillerstl Steam: Neverending

d3v0

Well the 3870 overclock was from 769mhz to 932mhz

EDIT: Added SPREADSHEET

__________________

d3v0 - 16 Graphics Cards - Old/New Stock/Overclocked/Vmodded: Extensive Benchmarks

H3||scr3am

insanity, but great, solid and thorough testing great job

__________________

I am 91% addicted to Counterstrike. What about you?
OCN CS:S GG: overclock.nuclearfallout.net:27015
OCN TF2: 68.232.173.130:27015 BRAND NEW 500FPS!!!

My Role in this community is to make your gaming experience on our gameservers enjoyable
please feel free to PM me with your suggestions and issues.

d3v0

Let me know if you think anything needs worked on!

__________________

d3v0 - 16 Graphics Cards - Old/New Stock/Overclocked/Vmodded: Extensive Benchmarks