Monthly drivers report and popular 3D
accelerators comparison

The i3DSpeed, a project of iXBT.com and Digit-Life.com, is dedicated to informing you about performance of a large number of graphics cards under Windows XP and Vista.

Traditionally, we offer you summaries of performance of the popular graphics cards, and also inform you about the best price/performance deals in the market.

Issue #4 (93)

February 2008

Added test results for RADEON HD 3650, GeForce 8800 GS, GeForce 9600 GT.

In April our list of benchmarks will be updated. Some readers do not like the amount of CRYSIS-based tests, but we believe this game, featuring all current 3D graphics innovations, is an ideal suite for estimating graphics card potential. For this reason we use two CRYSIS levels for benchmarking with DX9 and DX10. According to our experiments, graphics cards perform differently in different scenes.

As more interesting gaming benchmarks arrive, we will be reducing the amount of CRYSIS-based benchmarks (and will gradually stop using the 3Dmark05).

Contents

1. Testbed configurations and test software
2. Tested videocards list
3. i3DSpeed News
4. Test summary diagrams and digest bottom line
5. 3D accelerator rating calculations for the end of the month

1. Testbed configuration

• Intel Core2 Duo (775 Socket) based machine
  • Intel Core2 Extreme QX9650 (3000 MHz) CPU;
  • Gigabyte GA-X38-DQ6 motherboard on Intel X38 chipset;
  • 2 GB DDR2 SDRAM Corsair 1066MHz (CAS (tCL)=5; RAS to CAS delay (tRCD)=5; Row Precharge (tRP)=5; tRAS=15);
  • WD Caviar SE WD1600JD 160GB SATA HDD.

Operating systems:

• Windows XP Professional Service Pack 2, DirectX 9.0c,
• Windows Vista Ultimate 32bit, DirectX 10;

Monitor: Dell 3007WFP (30").

Benchmarks

2. Graphics cards

Below is a list of graphics cards that we tested linked to the respective 3Digest feature description pages.

In the brackets: core clock rate; ROP/TMU, core clock rate; shader unit, effective memory clock rate).

1. PCI-E-ATI RADEON HD 2400 XT 256MB GDDR3 1.2ns (700/700/1600 MHz)
2. PCI-E-ATI RADEON HD 2600 PRO 256MB GDDR2 2.0ns (600/600/1000 MHz)
3. PCI-E-ATI RADEON HD 2600 PRO 256MB GDDR3 1.2ns (700/700/1400 MHz)
4. PCI-E-ATI RADEON HD 2600 XT 256MB GDDR3 1.4ns (800/800/1400 MHz)
5. PCI-E-ATI RADEON HD 2600 XT 256MB GDDR4 0.9ns (800/800/2200 MHz)
6. PCI-E-ATI RADEON HD 3650 256MB GDDR3 1.1ns (725/725/1600 MHz)
7. PCI-E-ATI RADEON HD 2900 PRO 512MB GDDR3 1.0ns (600/600/1600 MHz)
8. PCI-E-ATI RADEON HD 3850 256MB GDDR3 1.0ns (670/670/1660 MHz)
9. PCI-E-ATI RADEON HD 3850 512MB GDDR3 1.0ns (720/720/1800 MHz)
10. PCI-E-ATI RADEON HD 3870 512MB GDDR4 0.8ns (775/775/2250 MHz)
11. PCI-E-ATI RADEON HD 3870 512MB GDDR4 0.8ns (850/850/2380 MHz)
12. PCI-E-ATI RADEON HD 3870 X2 2x512MB GDDR3 1.0ns (825/825/1800 MHz)
13. PCI-E-NVIDIA GeForce 8500 GT 256MB GDDR2 2.5ns (450/900/800 MHz)
14. PCI-E-NVIDIA GeForce 8600 GT 256MB GDDR3 1.2ns (560/1120/1560 MHz)
15. PCI-E-NVIDIA GeForce 8600 GTS 256MB GDDR3 1.0ns (720/1440/2200 MHz)
16. PCI-E-NVIDIA GeForce 9600 GT 512MB GDDR3 1.0ns (650/1625/1800 MHz)
17. PCI-E-NVIDIA GeForce 8800 GS 384MB GDDR3 1.0ns (680/1674/1600 MHz)
18. PCI-E-NVIDIA GeForce 8800 GT 512MB GDDR3 1.0ns (600/1512/1800 MHz)
19. PCI-E-NVIDIA GeForce 8800 GTS 320MB GDDR3 1.2ns (512/1188/1600 MHz)
20. PCI-E-NVIDIA GeForce 8800 GTS 512MB GDDR3 1.0ns (650/1620/1940 MHz)
21. PCI-E-NVIDIA GeForce 8800 GTS 640MB GDDR3 1.2ns (512/1188/1600 MHz)
22. PCI-E-NVIDIA GeForce 8800 GTX 768MB GDDR3 1.1ns (575/1350/1800 MHz)
23. PCI-E-NVIDIA GeForce 8800 Ultra 768MB GDDR3 0.8ns (612/1512/2160 MHz)

Archive (graphics cards, information on which is no longer updated)

3. February 2008 news

Added a number of new graphics cards (marked "NEW").

• 3.1. Latest drivers by the end of the reporting period:
  
  
  • NVIDIA ForceWare for all NVIDIA cards:
    
    
    • Windows XP: 169.28 beta (for up to 9xxx series);
    • Windows Vista: 169.28 beta (for up to 9xxx series);
    • Windows XP: 174.16 beta (for the 9xxx series);
    • Windows Vista: 174.16 beta (for the 9xxx series);

    
    

  • ATI Catalyst for all ATI cards:
    
    
    • Windows XP: CATALYST 8.2
    • Windows Vista: CATALYST 8.2
      

  
  
  

• 3.2. This month we used 169.28 and 174.12 (GeForce 9600GT) drivers for NVIDIA-based charts; 8.2 drivers for ATI-based charts.
  
  

4. Summary charts of graphics cards performance with the latest drivers available in February 2008.

All willing to see the complete results in Excel 2003 format can download this RAR 3.0 or ZIP archive.

If you are familiar with 3D graphics, you will be able to draw all the conclusions yourself after looking at the charts provided. Those interested in our judgement of test results should see reviews dedicated to specific GPUs (see our 3D graphics section). And for newbies and those who has just started choosing a graphics card we'll provide some more detailed explanations.

First of all, you should take a look at our brief references of modern graphics card series and their respective GPUs. Note clock rates, shader support and pipeline architecture.

Reference for ATI RADEON X1300-1600-1800-1900-x, HD 2900 series

Reference for NVIDIA GeForce 7300-7600-7800-7900-8800 series

Secondly, in our 3D graphics section all newbies can read about the basics of 3D (this knowledge will be required anyway, because you'll surely try to adjust some game settings like textures, lighting, etc.) Besides, newbies should read separate reviews dedicated to new products. There are just two major GPU vendors: ATI (recently merged with AMD under its brand) and NVIDIA. Thus, all information is generally separated in two parts. We also publish monthly i3DSpeed articles (one of which you are reading now) that sum up all comparisons of graphics cards for various price ranges. In the February 2007 issue we analyzed dependence of modern graphics cards on processors without antialiasing and anisotropic filtering. Similar tests with AA and AF were conducted in March.

Thirdly, let's see the charts and make conclusions.

• 4.1. S.T.A.L.K.E.R.
  • All resolutions on one page, No AA, No AF
    • 1024x768
    • 1280x1024
    • 1600x1200
    • 2560x1600

  • All resolutions on one page, No AA + AF 16x
    • 1024x768, AA+AF
    • 1280x1024, AA+AF
    • 1600x1200, AA+AF
    • 2560x1600, AA+AF

  
  
  

• 4.2. Call Of Juarez
  • All resolutions on one page, No AA, No AF
    • 1024x768
    • 1280x1024
    • 1600x1200
    • 2560x1600

  • All resolutions on one page, AA 4x + AF 16x
    • 1024x768, AA+AF
    • 1280x1024, AA+AF
    • 1600x1200, AA+AF
    • 2560x1600, AA+AF

  
  
  

• 4.3. CRYSIS DirectX 9.0 (high), RESCUE
  • All resolutions on one page, No AA, No AF
    • 1024x768
    • 1280x1024
    • 1600x1200
    • 2560x1600

  • All resolutions on one page, AA 4x + AF 16x
    • 1024x768, AA+AF
    • 1280x1024, AA+AF
    • 1600x1200, AA+AF
    • 2560x1600, AA+AF

  
  
  

• 4.4. CRYSIS DirectX 9.0 (high), HARBOR
  • All resolutions on one page, No AA, No AF
    • 1024x768
    • 1280x1024
    • 1600x1200
    • 2560x1600

  • All resolutions on one page, AA 4x + AF 16x
    • 1024x768, AA+AF
    • 1280x1024, AA+AF
    • 1600x1200, AA+AF
    • 2560x1600, AA+AF

  
  
  

• 4.5. CRYSIS DirectX 10.0 (Very High), RESCUE
  • All resolutions on one page, No AA, No AF
    • 1024x768
    • 1280x1024
    • 1600x1200
    • 2560x1600

  • All resolutions on one page, AA 4x + AF 16x
    • 1024x768, AA+AF
    • 1280x1024, AA+AF
    • 1600x1200, AA+AF
    • 2560x1600, AA+AF

  
  
  

• 4.6. CRYSIS DirectX 10.0 (Very High), HARBOR
  • All resolutions on one page, No AA, No AF
    • 1024x768
    • 1280x1024
    • 1600x1200
    • 2560x1600

  • All resolutions on one page, AA 4x + AF 16x
    • 1024x768, AA+AF
    • 1280x1024, AA+AF
    • 1600x1200, AA+AF
    • 2560x1600, AA+AF

  
  
  

• 4.7. Company Of Heroes
  • All resolutions on one page, No AA, No AF
    • 1024x768
    • 1280x1024
    • 1600x1200
    • 2560x1600

  • All resolutions on one page, AA 4x + AF 16x
    • 1024x768, AA+AF
    • 1280x1024, AA+AF
    • 1600x1200, AA+AF
    • 2560x1600, AA+AF

  
  
  

• 4.8. 3DMark05 MARKS
  • All resolutions on one page, No AA, No AF
    • 1024x768
    • 1280x1024
    • 1600x1200
    • 2560x1600

  • All resolutions on one page, AA 4x + AF 16x
    • 1024x768, AA+AF
    • 1280x1024, AA+AF
    • 1600x1200, AA+AF
    • 2560x1600, AA+AF

  
  
  

• 4.9. 3DMark 06, Shader Model 2.0 MARKS
  • All resolutions on one page, No AA, No AF
    • 1024x768
    • 1280x1024
    • 1600x1200
    • 2560x1600

  • All resolutions on one page, AA 4x + AF 16x
    • 1024x768, AA+AF
    • 1280x1024, AA+AF
    • 1600x1200, AA+AF
    • 2560x1600, AA+AF

  
  
  

• 4.10. 3DMark 06, Shader Model 3.0 MARKS
  • All resolutions on one page, No AA, No AF
    • 1024x768
    • 1280x1024
    • 1600x1200
    • 2560x1600

  • All resolutions on one page, AA 4x + AF 16x
    • 1024x768, AA+AF
    • 1280x1024, AA+AF
    • 1600x1200, AA+AF
    • 2560x1600, AA+AF

  
  
  

• 4.11. Archive (information from the previous i3DSpeed issues absent in this one)
  
  
• 4.12. Notes on graphics cards performance with the latest drivers:
  
  
  • If a particular card is not in the summary, i.e. its results are either 0, or 1, it means the given benchmark hasn't work or has worked incorrectly with this card.
  • If you imagine a line connecting ends of columns representing test results on a chart, you can judge if a game is critical to a card or a CPU. If this curve is "convex" and many cards perform similarly, it means at this resolution and under this load performance is bottlenecked by system resources, not graphics card potential. If, vice versa, the curve is "concave" and performance drops quickly as card potential reduces, it means the game puts heavy load on graphics cards and depends exclusively on their potential.

5. 3D accelerator ratings calculations for the end of the month

Calculations were based on gaming benchmark results as well as the results from 3DMark06 new tests (ATI CATALYST 8.2, NVIDIA ForceWare 169.28/174.12).

All willing to conduct their own rating calculations with own values are invited to download this Excel XP table in RAR 3.0 or ZIP format.

The method of rating calculation is described here.

Calculations were conducted considering the following assumptions:

1. Percentage of performance and quality demand ratio:
   • performance priority - 60%
   • quality priority - 40%

2. Only the HIGHEST resolution available in games was used to calculate prospect ratings.
3. The usability rating indicates performance and features of a card (numerator) related to its price (denominator).

According to your wishes, we separated ratings into two groups: the first gathers all benchmarks, and the second contains only the gaming tests without m3DMark05/06. Disbelieving the synthetic tests, many think the second group is of higher importance.

Usability rating:

GENERAL prospect rating

#	Card	Usability rating	Prospect rating	Approx. price
01.	HD 3870 X2 2x512MB, 825/825/1800	506	1220	618
02.	8800 Ultra 768MB, 612/1512/2160	463	1219	680
03.	8800 GTX 768MB, 575/1350/1800	526	1153	542
04.	8800 GTS 512MB, 650/1620/1940	654	1145	441
05.	8800 GT 512MB, 600/1512/1800	800	1081	325
06.	HD 3870 512MB, 850/850/2380	671	1013	345
07.	9600 GT 512MB, 650/1625/1800	977	1002	230
08.	8800 GTS 640MB, 512/1188/1600	536	973	391
09.	HD 3870 512MB, 775/775/2250	608	957	340
10.	8800 GS 384MB, 680/1674/1600	799	928	261
11.	HD 3850 512MB, 720/720/1800	691	907	276
12.	HD 2900 PRO 512MB, 600/600/1600	799	849	209
13.	HD 3850 256MB, 670/670/1660	689	833	240
14.	8800 GTS 320MB, 512/1188/1600	492	819	346
15.	8600 GTS 256MB, 720/1440/2200	571	575	150
16.	HD 2600 XT 256MB, 800/800/2200	495	522	140
17.	HD 3650 256MB, 725/725/1600	513	500	124
18.	HD 2600 XT 256MB, 800/800/1400	576	471	100
19.	8600 GT 256MB, 560/1120/1560	467	467	114
20.	HD 2600 PRO 256MB, 700/700/1400	527	443	93
21.	HD 2600 PRO 256MB, 600/700/1000	512	410	76
22.	8500 GT 256MB, 450/900/800	424	371	69
23.	HD 2400 XT 256MB, 700/700/1600	388	339	66

We express gratitude to the following companies:
TUL (PowerColor),
Chaintech,
Zotac,
BFG and Mikhail Proshletsov,
EVGA and Victoria Nikolaeva,
Foxconn and Alexander Troukhachev,
MSI,
Gainward,
HIS and Vitaly Milov,
NVIDIA,
AMD,
Sapphire Technology,
ASUStek.

PSU provided by TAGAN

Monitor provided by NVIDIA

Andrey Vorobiev (anvakams@ixbt.com)

February 12, 2008