DonQ suur videokaartide jõudlustabel (viimati uuendatud 15. märtsil 2008)
DonQ's graphics cards performance table (last updated March, 15th 2008)
Eriline tänu kodanik DazBoot'ile tabeli koostamisel jõu ja nõuga abiks olemise eest!
NB! Käesoleval lehel on levinumate ja uute kaartide koondtabel
(erineva kiiruse ja mäluhulgaga praktiliselt samatüübilised kaardid on toodud ühe reana,
SLI ja OC tulemusi arvestatud pole),
parema ja täpsema võrdluse saamiseks võib tulemusi vaadata ka detailselt ja gruppide kaupa:
üldtabel - kõik kaardid; skoorid on aluseks gruppide moodustamisele
low-end kaardid - tulemus üldtabelis kuni 7000 punkti
mid-range kaardid - tulemus üldtabelis 6000 kuni 14000 punkti
high-end kaardid - tulemus üldtabelis üle 13000 punkti
Kaarditootjate kalduvust erinevate mängude toetamiseks saab vaadata:
mängutabel - ATI ja NVidia tulemused erinevates mängudes
Vanemate kaartide (sh NVidia GeForce 5-6 seeria, ATI Radeon 9 ja X*** seeria, AGP ja mõned muud kaardid) tulemused
on kättesaadavad arhiivilehtedel:
arhiiv - jõudlustabel 4. märtsi 2006 seisuga
arhiiv - jõudlustabel 19. septembri 2007 seisuga
| Legend (...interaktiivne, IE ja FF all mõnikord eriefektidega - katsuge hakkama saada...) | ||||||
| Name | Videokaardi nimi ja tüüp | |||||
| Mem | Videokaardi mälu MB | ATI Radeon | ||||
| Core | GPU (core) kiirus MHz | NVidia GeForce | ||||
| RAM | Videokaardi mälu kiirus MHz | CPU/süsteemi poolt piiratud tulemused | ||||
| Score | Keskmistatud skoor (lõppskoor) | Testitulemuste vahemik (CPU/süsteemi piiranguteta) | ||||
| Res | Testitulemusi kaardil | Lõppskoori asukoht | ||||
| Err | Veategur | SLI ja CF kaartide tulemused | ||||
| Idle | Võimsus tühikäigul W | Load | Maksimaalne tarbitav võimsus W | |||
Special thanks to mister DazBoot to help build these pages!
NB! On this page is displayed unified table, containing results for popular and newer cards
(similar cards are aggregated to one row, SLI and overclocked cards are excluded).
Detailed results for all cards are presented on next pages:
full results - all cards; scores in this table are basis of grouping
low-end cards - score below 7000 points
mid-range cards - score between 6000 and 14000 points
high-end cards - score above 13000 points
Manufacturers support of different games is compared here:
game performance - ATI and NVidia results in different games
Results of older cards (incl NVidia GeForce 5xxx and 6xxx, Radeon 9xxx and Xxxx, AGP and some others) are still
available on archive pages:
archived results - performance table at Mar, 4th 2006 (in estonian only)
archived results - performance table at Sep, 19th 2007
| Legend (...interactive, with some unwanted effects using IE or Firefox...) | ||||||
| Name | Graphics card name and model | |||||
| Mem | RAM size in MB | ATI Radeon | ||||
| Core | GPU (core) clock in MHz | NVidia GeForce | ||||
| RAM | RAM speed in MHz | Results, limited by CPU/system performance | ||||
| Score | Averaged (final) score | Range of test results (excluding limited by CPU/system) | ||||
| Res | Tests count | Final score | ||||
| Err | Error factor | Results of SLI and CF cards | ||||
| Idle | Idle power W | Load | Maximum power consumption W | |||
| OtsiFind | Märgi kaardidMark cards | Lisa märgistusAdd to marked |
| Name | Graph | Score | Res | Err | Mem | Core | RAM | Idle | Load | |||||
| HD3870 X2 | 22936 | 124 | 1,30 | 1024 | 825 | 1800 | ||||||||
| 8800 Ultra | 22232 | 83 | 1,14 | 768 | 612 | 2160 | ||||||||
| 8800 GTX | 20474 | 104 | 1,20 | 768 | 575 | 1800 | 70 | 132 | ||||||
| 8800 GTS 512 | 20437 | 108 | 1,19 | 512 | 650 | 1940 | 38 | 99 | ||||||
| 8800 GT | 18664 | 107 | 1,16 | 512 | 600 | 1800 | 35 | 78 | ||||||
| 9600 GT | 16226 | 126 | 1,13 | 512 | 650 | 1800 | 26 | 60 | ||||||
| HD2900 XT | 15802 | 88 | 1,18 | 512 | 750 | 1660 | 72 | 161 | ||||||
| HD3870 | 15626 | 179 | 1,16 | 512 | 775-850 | 2250-2380 | 19 | 81 | ||||||
| 8800 GS | 15582 | 100 | 1,17 | 384 | 680 | 1600 | ||||||||
| 8800 GTS | 14815 | 108 | 1,14 | 320-640 | 512 | 1600 | 56 | 106 | ||||||
| 7950 GX2 | 13387 | 93 | 1,43 | 1024 | 500 | 1200 | 50 | 110 | ||||||
| HD3850 | 13214 | 155 | 1,16 | 256-512 | 670-720 | 1660-1800 | 14 | 63 | ||||||
| X1900 XTX | 13086 | 138 | 1,23 | 512 | 650 | 1550 | 30 | 121 | ||||||
| X1900 XT | 12845 | 106 | 1,16 | 512 | 625 | 1450 | 28 | 109 | ||||||
| 7900 GTX | 12593 | 138 | 1,25 | 512 | 650 | 1600 | 31 | 84 | ||||||
| 7800 GTX 512 | 12460 | 35 | 1,25 | 512 | 550 | 1700 | 29 | 95 | ||||||
| HD2900 Pro | 12286 | 52 | 1,13 | 512 | 600 | 1600 | 71 | 133 | ||||||
| X1800 XT | 9714 | 81 | 1,20 | 512 | 600-625 | 1400-1500 | 30 | 103 | ||||||
| 7900 GT | 9668 | 84 | 1,14 | 256 | 450 | 1320 | 23 | 48 | ||||||
| 7800 GTX | 9305 | 57 | 1,16 | 256 | 430 | 1200 | 29 | 81 | ||||||
| X1900 GT | 8982 | 155 | 1,15 | 256 | 500-575 | 1200 | 25 | 75 | ||||||
| 8600 GTS | 8270 | 86 | 1,13 | 256 | 675-720 | 2000-2200 | ||||||||
| X1950 Pro | 8109 | 43 | 1,18 | 512 | 580 | 1400 | ||||||||
| 7800 GT | 8019 | 116 | 1,12 | 256 | 400 | 1000 | 20 | 57 | ||||||
| X1800 XL | 7966 | 98 | 1,16 | 256 | 500 | 1000 | 27 | 57 | ||||||
| 7600 GT | 6772 | 152 | 1,10 | 256 | 560 | 1400 | 15 | 36 | ||||||
| X1800 GTO | 6598 | 164 | 1,14 | 256 | 500 | 1000 | 25 | 48 | ||||||
| 8600 GT | 6348 | 81 | 1,20 | 256 | 560 | 1560 | ||||||||
| HD3650 | 5972 | 82 | 1,33 | 256 | 725 | 1600 | 8 | 31 | ||||||
| HD2600 XT | 5929 | 121 | 1,25 | 256 | 800 | 1400-2200 | 21 | 49 | ||||||
| 7600 GS | 4762 | 143 | 1,18 | 256 | 400-450 | 800 | 14 | 27 | ||||||
| X1600 XT | 4227 | 137 | 1,15 | 256 | 590 | 1380 | 24 | 42 | ||||||
| HD2600 Pro | 3545 | 99 | 1,75 | 256 | 600-700 | 1000-1400 | 16 | 31 | ||||||
| X1600 Pro | 3272 | 85 | 1,16 | 256-512 | 500 | 810 | 24 | 41 | ||||||
| 8500 GT | 3077 | 67 | 1,24 | 256 | 450 | 800 | 14 | 27 | ||||||
| HD2400 XT | 2715 | 65 | 1,23 | 256 | 700 | 1600 | 7 | 19 | ||||||
| X1300 Pro | 2538 | 83 | 1,36 | 256 | 600 | 800 | 18 | 31 | ||||||
| X1300 | 2020 | 61 | 1,32 | 256-512 | 446-450 | 550-792 | ||||||||
| 7300 GS | 1998 | 46 | 1,24 | 128-512 | 550 | 810 |
Algoritm
Skoori arvutamise aluseks on võetud viidatud artiklite võrdlustestide tulemused.
Erinevad testitulemused on lähendatud (normaliseeritud) algskoorile, minimiseerides tulemuste
ja algskoori kaalutud logaritmide vahelist hälvet tulemuste logaritmide nihutamise teel.
Antud lähenemise eelduseks on videokaartide jõudluste suhte konstantsus erinevates testides.
Kogu protsessi on itereeritud 20 korda, mis peaks olema piisav, kuna skoori muutus
viimaste iteratsioonide vahel on üldjuhul alla 1 punkti.
Testitulemuste arvutamisel on lähtutud järgnevatest üksiktulemuste kaaludest:
CPU/süsteemi poolt limiteeritud tulemused - 0.2
Tulemused resolutsioonil 2048x1536 või rohkem - 0.5
Tulemused resolutsioonil 1024x768 või vähem - 0.5
Süsteemi poolt limiteerimata katsetulemuste arv N antud videokaardil - ln(N)
Kaartide arv konkreetses testis T (sh süsteemi poolt limiteeritud L) - (T-L)/T
K kuud vanad testid - 12/(K+12)
Sünteetilised testid, nagu 3DMark jmt - 0.5
Alglähendiks on valitud aritmeetiline keskmine erinevate artiklite mõningatest testitulemustest
(vajadusel sobiva koefitsiendiga korrutatult). Alustestid on valitud eeldusel, et neis osaleb
maksimaalne arv kaarte ning tulemused ei ole süsteemi poolt oluliselt piiratud.
Seoses 3DMark05 tulemuste kasutamisega alglähendi moodustamisel on võimalik lõppskooride mõningane
kokkulangemine konkreetse kaardi 3DMark05 skooridega (eriti tabeli keskosas asuvatel kaartidel);
selline seos on siiski täiesti juhuslikku laadi.
Märkused
Kaartide järjestust ei tasu võtta puhta kullana. Mõningate kaartide nähtavalt kahtlane asukoht
on tingitud erinevate kaartide osalemises erinevates testides, seega ka erinevatel süsteemidel
ja mitmesuguste draiveriversioonidega.
Ligikaudse hinnangu konkreetse kaardi tulemuste usaldatavusest võib visuaalselt saada kaardile vastava
tulemuste (rohelise) piirkonna laiuse alusel, samuti veateguri järgi (mida lähemal 1-le, seda väiksem
hajumine testitulemustes). Uute (ja eriti SLI) kaartide puhul on tulemused üldjuhul väga hajuvad tänu
süsteemi piirangutele ja erinevate mängude erineva toega SLI kaartidele.
Kaardigruppide kaupa avaldatud tulemused on üldiselt väiksema hajuvusega, kuna sarnase jõudlusega
kaartide käitumine erinevates testides on sarnasema iseloomuga.
Kaartide võimsustarve on esitatud vastavalt reaalsetele mõõtmistulemustele ja ei pruugi
tootjate spetsifikatsioonidega vähimalgi määral kokku langeda.
Algorithms
Cards' rating is computed on basis of results of linked tests and articles.
Different results are approximated (normalized) to initial score, minimizing differences
between weighted logarithms of initial and specific results. It's assumed that cards performance
ratio is constant over different tests.
Entire process is iterated 20 times, score change between last iteration steps does usually
not exceed 1-2 points.
Computations use next weights for particular results:
Results, limited by CPU or system - 0.2
Results for screen resolution 1600x1200 or more - 0.5
Results for screen resolution 800x600 or less - 0.5
Amount of test results N for specific video card, not limited by CPU - ln(N)
Amount of cards T, participating in specific test (incl CPU limited L) - (T-L)/T
K months old tests - 12/(K+12)
Synthetical benchmarks, like 3DMark and similar - 0.5
For first approximation is used arithmetical average of some specific test results from
different articles (multiplied with appropriate coefficient, if needed). Base tests are chosen
to include majority of cards and to not rely on CPU performance.
Although final scores may match 3DMark05 ratings for some cards, this is purely coincidental.
Remarks
Card scores and their relative position is of course not holy truth. A weird position of some cards
is mainly caused by mixing results of very different tests, using different CPU, systems,
driver versions etc.
As a rough guess, accuracy of card score is characterized by width of results bar (green) and
by error factor. What narrower is green area and what closer is error factor to 1, then less is
deviation of test results for particular card. Results of newer and SLI cards are usually very
inaccurate due to the CPU limitations and different games support.
Results in sub-groups are usually more reliable, because similar cards behave more consistently.
Power consumption is measured in lab and is not related to card manufacturers specifications.
Kasutatud materjalid
Reference articles
Tom's Hardware Guide Graphics Cards: VGA Charts VI: PCI Express Update Summer 2005
Tom's Hardware Guide Graphics Cards: VGA Charts VII: AGP Update Summer 2005
NVIDIA's GeForce 7800 GT graphics cards - The Tech report
Radeon X1800XT 512MB reviewed
AnandTech: ATI's Late Response to G70 - Radeon X1800, X1600 and X1300
AnandTech: ATI's X1000 Series: Extended Performance Testing
Video at Digit-Life.com: June 2004 3Digest
Video at Digit-Life.com: 3Digest, August 2005
NVIDIA GeForce 6800 GS 256 M
X-bit labs - Articles - NVIDIA GeForce 6800 GS - The Best of the Holiday Season 2005?
AnandTech: NVIDIA GeForce 7800 GTX 512: More Than Just More Memory
AnandTech: AT Exclusive Preview: The GeForce 7800 GS Hits the Scene
X-bit labs - Articles - Modern Games and Modern GPUs: The Grand Clash
X-bit labs - Articles - S3 Chrome S27 Graphics Processor Review: Worthy Performance for Its Class?
3Digest: January 2006 (Monthly drivers report and popular 3D accelerators comparison)
3DNews: Результаты тестов GeForce 7300GS от ASUS
CoolTechZone.com - NVIDIA 7800GS AGP: Rekindling the Last Generation Standard
EVGA e-GeForce 7800 GS CO Superclock Review
iXBT: 3DGiТоги (Итоги по 3D акселераторам): Март 2006 года
Red Hot VGA Charts: SLI, Crossfire and AGP Graphics Added | Tom's Hardware
iXBT: 3DGiТоги (Итоги по 3D акселераторам): Август 2006 года
AnandTech: Mid Range GPU Roundup - Summer 2006
Сравнительное тестирование 45 видеокарт PCI-E, III квартал 2006 - 3DNews
X-bit labs - Articles - Seven Games and One Week: Games vs. GPUs
iXBT: 3DGiТоги (Итоги по 3D акселераторам): Апрель 2007 года
iXBT: 3DGiТоги (Итоги по 3D акселераторам): Август 2007 года
i3D-Speed 02'08: Февраль 2008 года
Biostar GeForce 9600 GT 512 MB V9603GT52
CrossFire X explored - The Tech Report
AnandTech: CrossFireX Arrives: First Look at 3 and 4 GPUs in 2 Card Setups
Videokaartide võimsustarve
Power consumption
X-bit labs - Articles - Power Consumption of Contemporary Graphics Accelerators. Part II: NVIDIA vs. ATI
X-bit labs - Articles - The Grand Clash for Watts: Power Consumption of Modern Graphics Cards
The Truth About Graphics Power Requirements V2
Muud viited videokaartide ja jõudluse kohta
Other links to graphic cards information
Laptop 3D Mark tulemused
Video Card Information, Reviews, and Specifications - Welcome To GPUReview.com!
Kopeerimine ja tagasiside
Kõik käesoleva(te)l leh(ted)el toodud tulemused on vabalt kasutatavad mistahes mitteärilistel eesmärkidel.
Käesoleva(te)le leh(ted)ele on lubatud viidata vaid koondtabeli aadressi ja käesoleva lehe nime alusel,
aadressi lõpus võib kasutada keeleseadet (hetkel #EST või #ENG, vaikimisi eesti):
<a href="http://www.stv.ee/~donq/videocards.htm#EST">DonQ suur videokaartide jõudlustabel</a> <a href="http://www.stv.ee/~donq/videocards.htm#ENG">DonQ's graphics cards performance table</a>
Foorumite postituste tekstides (mitte signatuurides jm staatilises osas!) on lubatud viidata ka üld- ja grupilehtedele,
samuti kasutada viidet koos kaartide märkimisvõimalusega (aadressi järel #kaart1,kaart2 ...).
Lehe nime asemel võib kasutada ühte järgnevatest banneritest; teistsuguste bannerite kasutamine ja kujundus
tuleb eelnevalt kooskõlastada lehe autoriga:
Tagasiside autorile lehe sisu, vigade jm kohta, samuti ideed ja ettepanekud võib saata aadressil don q at stv ee.
Lehe autor ei võta endale mitte mingisugust kohustust tagasisidele reageerida.
Copyright © 2005-2008 by DonQ
Copying and feedback
All data on these pages is freely available for any noncommercial use.
Linking to these pages is allowed only by address and name of this page; address can be followed with
initial language identifier (currently #ENG and #EST are supported):
<a href="http://www.stv.ee/~donq/videocards.htm#EST">DonQ suur videokaartide jõudlustabel</a> <a href="http://www.stv.ee/~donq/videocards.htm#ENG">DonQ's graphics cards performance table</a>
In forum postings (but not in signatures and in other static content) it is allowed to link to other pages
on this site, so can be used links with marking specific cards (page address, followed by #card1,card2 etc).
Instead of page title any of next banners can be used; using other images must be agreed before with author
of this site:
Any feedback about page content, errors etc; also all ideas and suggestions can be emailed to don q at stv ee.
Author of this site doesn't take any liability to respond to such feedback.
Copyright © 2005-2008 by DonQ