The Official Graphics Card and PC gaming Thread

Bit interested in the 6850 myself. It'sa 5-10% slower than the 5850 at all times but is $100 cheaper than when I bought my 5850s. Does it simply suck less power or what? What's the main thing they changed for Barts?

 

The architecture is redesigned, the 6850 only has 960 shaders versus the 5850's 1440, but produces a very similar amount of rendering power, but with 24W less electrical power (37W for 6870 vs 5870), and thus only one 6-pin connector (still two for the 6870).

Also, if I've seen right, the HD6850 does not have a retailing reference design, so it's potentially going to be less reliable than the other HD5 and HD6 cards.

 

Hmm, doesn't sound like a winner LOL!

 

Have you not seen how much cheaper they are?

 

The HD6950 and HD6970 TDPs seem to have appeared from nowhere, claiming 198W for the 6950 and 220W for the 6970. If true, that will represent serious performance per watt, as it will mean the 6970 will outperform the GTX480 by a country mile yet use less power than its smaller partner the GTX470.

 

<--- Enthusiast

 

If you zoom in all the way Civ V tends to lag a bit for me when there's a bunch of stuff on screen with max settings (iirc) @ 1680x1050 but other than that it's pretty smooth. Not particularly helpful but I can grab some fps data anyone would like it.

(system info)
HD5850
i7 @ 3.8Ghz

 

So the 6850 sounds like a very reasonable card. As far as I've seen it remains consistently the same performance in relation to the 5850, which is a good thing. Not having a reference design can hurt, but I would guess a company like XFX is more than up to the task. Not only that but XFX seem to be becoming the premium ATi brand to have. Reference seems fine for my Sapphire 5850s but if the card has no reference design then Sapphire makes me a bit wary. Interested to see what becomes of it.

What I'm really miffed about is the retarded naming scheme for this series. So the 6800 series is the optimised new version(a la 9800GTX) and the 6900 series will be the new generation performers? Not the most terrible naming convention ever but I was pretty confused when the 6870 was underperforming the 5870.

 

Yeah it's bad of AMD to use a rebrand like that, but since it's not in the slightest the same card, it's semi-forgivable. It means that the low-cost builds that mandate a sub $200 GPU, and a 400W PSU with one 6-pin now get HD5850 like performance.
I have to admit that for the last couple of months the HD5850 and HD5870 have been particularly overpriced, this is obviously why.

 

its really not forgiveable, BUT most people that are going to buy a £150+ gfx card probably research before they buy.

and anyone with a 5850 and 5870, that arnt that knowlegable (read bought a HP etc system with one) probably wont notice even when they need to upgrade, so long as the game plays.

but still its bad, that a 6870 is not better than a 5870, as has been the case since the 3870.

(though the 3870 was a mess in its own right vs a 2900 for performance vs name haha)

man anadtechs numbers are well of everyone else. just checked their review for the 68 series.

even driverheaven (changed their name?!) comply with everyone else.

AHEM anand!



also, 4870x2s going for £130 on ebay recently, its so tempting!

 

Well what have I been telling you about Anand for ages

Driverheaven's tests are usually pretty solid, they just don't test very much at once.

 

The Scaling guide to the HD6000 series

Guru3D to follow.

Legit Reviews
Aliens vs. Predator

HD6850: 1280x1024: 96.5%, 1920x1200: 97.4%
HD6870: 1280x1024: 97.0%, 1920x1200: 97.4%

Batman: Arkham Asylum

HD6850: 1280x1024: 62.1% (227fps), 1920x1200: 90.3% (177fps)
HD6870: 1280x1024: 57.0% (248fps), 1920x1200: 90.5% (200fps)

Just Cause 2

HD6850: 1280x1024: 95.7%, 1920x1200: 95.2%
HD6870: 1280x1024: 98.1%, 1920x1200: 101.5%

Metro 2033 (Since been hotfixed, Catalyst 10.10c)

HD6850: 1280x1024: 33.6%, 1920x1200: 44.1%
HD6870: 1280x1024: 30.0%, 1920x1200: 47.8%

STALKER: Call of Pripyat

HD6850: 1280x1024: 71.4% (139fps), 1920x1200: 95.6% (97fps)
HD6870: 1280x1024: 85.2% (158fps), 1920x1200: 97.4% (113fps)

Starcraft 2: Wings of Liberty

HD6850: 1280x1024: 43.1% (198fps), 1920x1200: 55.5% (173fps)
HD6870: 1280x1024: 30.7% (206fps), 1920x1200: 50.4% (193fps)

TechPowerup [1280x1024 / 1680x1050 / 1920x1200 / 2560x1600]
HD6850CF only

Aliens vs. Predator: 95.5/96.8/97.0/96.5
Battlefield Bad Company 2: 94.8/100.8/100.7/103.6
Battleforge: 97.4/97.0/97.8/98.4
Call of Duty 4: 76.1/82.9/87.2/90.3
Call of Juarez 2: Crossfire not supported
Crysis: 88.3/95.5/96.0/92.9
Dawn of War 2: 67.9/79.0/91.0/100.8
DiRT 2: 76.3/82.3/84.5/92.4
Far Cry 2: 91.0/100.1/100.7/100.2
HAWX: 85.3/95.4/94.9/100.6
Metro 2033: 99.0/95.4/89.7. N/A at 2560x1600 - does not run at 4xAA with 1GB VRAM
Riddick Dark Athena: 77.6/83.0/87.1/92.7
STALKER Clear Sky: 39.7/71.4/94.2/97.5
Supreme Commander 2: 13.2/19.6/25.5/45.0 - CPU limited, results largely irrelevant
Unreal Tournament 3: 18.6/53.3/66.7/90.6 - CPU limited, AvFPS >150 at all resolutions
World of Warcraft: Crossfire not supported, negative scaling


Some really impressive results, even better when I drop this in, from Hexus:


Modern Warfare 2, 30", 4xAA
2x HD6850: Min95, Ave114, 76ºC, 321W, £295
2x GTX460 1GB: Min86, Ave111, 88ºC, 431W, £285

 

Updated:
KitGuru - HD6850 and HD6870 vs GTX460 1GB

Aliens vs. Predator 1920x1080 DX11-TVH-1x16-AO-SCHi-MB
HD6850: Average 38.5->74.9 (94.5%), Minimum 23.2->46.4 (100.0%)
GTX460: Average 37.6->69.3 (84.3%), Minimum 22.1->42.3 (91.4%)
HD6870: Average 47.2->92.1 (95.1%), Minimum 28.4->59.2 (108.5%)

Far Cry 2 1920x1200 DX10-4x16-OO-VegHi-ShadHi-TerHi-GeoHi-PostFXHi-TexHi-ShadoHi-AmbHi-HDR-Bloom-FireVH-PhysVH-RTreeVH
HD6850: Average 62.46->117.53 (88.2%), Minimum 44.17->80.26 (81.7%)
GTX460: Average 70.21->119.63 (70.4%), Minimum 50.71->94.48 (86.3%)
HD6870: Average 71.72->129.42 (80.5%), Minimum 49.63->95.15 (91.7%)

Lost Planet 2 1920x1200 DX11-1x1-MB-ShadoHi-TexHi-RenHi-DX11Mid
HD6850: Average 34.9->64.6 (85.1%), Minimum 24.4->35.3 (44.7%)
GTX460: Average 38.7->75.8 (95.9%), Minimum 19.1->32.5 (70.2%)
HD6870: Average 40.8->80.4 (97.1%), Minimum 28.7->51.2 (78.4%)

Mafia 2 1920x1080 AA-16xAF-ShadoHi-AO-GeoHi-No APEX
HD6850: Average 41.8->71.6 (71.3%), Minimum 24.7->43.5 (76.1%)
GTX460: Average 39.2->67.6 (72.4%), Minimum 23.4->39.3 (67.9%)
HD6870: Average 49.8->82.7 (66.1%), Minimum 29.4->51.2 (74.1%)

Metro 2033 1920x1080 1920x1080 Normal
HD6850: Average 34->63 (85%), Minimum 26->53 (104%)
GTX460: Average 33->59 (79%), Minimum 26->49 (88%)
HD6870: Average 37->67 (81%), Minimum 29->55 (90%)

Resident Evil 5 1920x1200 8xAA Maximum [CPU limits suspected, HD5770CFX ~~ HD6850CFX]
HD6850: Average 85.5->108.2 (26.5%), Minimum 68.3->83.5 (22.3%)
GTX460: Average 78.3->112.3 (43.4%), Minimum 62.7->83.3 (32.9%)
HD6870: Average 97.2->124.9 (28.5%), Minimum 85.2->88.3 (3.6%)

HAWX 1920x1200 8xAA Maximum
HD6850: Average 68->121 (77.9%), Minimum 53->82 (54.7%)
GTX460: Average 65->123 (89.2%), Mimimum 48->83 (72.9%)
HD6870: Average 75->127 (69.3%), Minimum 56->94 (67.9%)

HD6850 overclocking
Default clocks: 775/4000
Stock voltage Overclock limit: 860/unknown
Raised voltage Overclock limit: 950/4532 - both cards tested

HD6870 overclocking
Default clocks: 900/4200
Raised voltage Overclock limit: 950/4560, 940/4600

 

Efficiency of recent graphics cards (single GPUs only)

HD4850: 80% performance, 73% power - 9% gain
HD4870: 100% performance, 100% power - 0% gain
HD4890: 112% performance, 127% power - 12% less efficient
HD5750: 75% performance, 57% power - 32% gain
HD5770: 95% performance, 72% power - 32% gain
HD5830: 130% performance, 117% power - 11% gain
HD5850: 155% performance, 101% power - 54% gain
HD5870: 180% performance, 125% power - 44% gain

HD6850: 150% performance, 85% power - 77% gain
HD6870: 175% performance, 101% power - 74% gain
HD6950: 215% performance, 125% power - 72% gain
HD6970: 260% performance, 140% power - 86% gain


GTS250: 80% performance, 97% power - 17% less efficient
GTX260: 90% performance, 135% power - 33% less efficient
GTX260+a: 100% performance, 121% power - 17% less efficient
GTX260+b: 100% performance, 114% power - 12% less efficient
GTX275: 122% performance, 146% power - 16% less efficient
GTX280: 115% performance, 157% power - 27% less efficient
GTX285: 128% performance, 136% power - 6% less efficient

GTS450: 95% performance, 71% power - 34% gain
GTX460: 125% performance, 100% power - 25% gain
GTX461: 135% performance, 115% power - 17% gain
GTX465: 130% performance, 140% power - 7% less efficient
GTX470: 157% performance, 157% power - identical
GTX480: 190% performance, 197% power - 4% less efficient

 

As of this benchmark, all results are as recorded, not extrapolated, with the exception of the HD6950/6970, GTX580 and dual-GPU setups. Scaling assumed as 80% for Radeon HD5 series and Geforce GTX4 series, 95% for HD6 series.
Results for both 1680x1050 and 2560x1600 for Test Drive Unlimited 2 are estimates. Only 2560x1600 is estimated for Star Wars: The Force Unleashed 2.

Star Wars: The Force Unleashed 2
Game engine capped at 30fps
1024x768 M10: Radeon X1800XT/HD2900 series/3690/4650/5570, Geforce 7900GT/8600GTS/9500GT/GT220/430
1024x768 M20: Radeon HD2900 Pro/3850/4670/5570, Geforce 8800GS/9600GSO/GT240/GTS450
1024x768 M30: Radeon HD2900XT/3870/4830/5670, Geforce 8800GT/9800GT/GTS250/450

1280x1024 M10: Radeon X1900XT/HD2900 series/3690/4650/5570, Geforce 7900GTX/8800GS/9600 series/GT220/430
1280x1024 M20: Radeon HD2900 Pro/3850/4670/5570, Geforce 8800GT/9800GT/GT240/GTS450
1280x1024 M30: Radeon HD3870X2/4830/5750, Geforce 8800GTS 512/9800GTX/GTS250/450

1680x1050 M10: Radeon X1950XT/HD2900 series/3850/4670/5570, Geforce 8800GS/9600GSO/GT220/430
1680x1050 M20: Radeon HD2900XT/3870/4830/5570, Geforce 8800GT/9800GT/GT240/GTS450
1680x1050 M30: Radeon HD3870X2/4770/5750, Geforce 9800GX2/GTX260/GTS450

1920x1080 M10: Radeon X1950XT-X/HD2900 Pro/3850/4670/5570, Geforce 8800GS/9600GSO/GT240/430
1920x1080 M20: Radeon HD3870X2/4830/5670, Geforce 8800GTS 512/9800GTX/GTS250/450
1920x1080 M30: Radeon HD3870X2/4860/5770, Geforce 9800GX2/GTX260/GTS450

2560x1600 M10: Radeon HD3870X2/4830/5670, Geforce 8800GT/9800GT/GTS250/450
2560x1600 M20: Radeon HD4890/5830/6850, Geforce GTX275/460
2560x1600 M30: Radeon HD4870X2/5870/6870, Geforce GTX295/470

Test Drive Unlimited 2 Beta - Medium detail, 8xAA
1280x1024 M10: Radeon X1950XT-X/HD2900GT/3850/4670/5570, Geforce 8800GS/9600GSO/GT240/430
1280x1024 M20: Radeon HD3870X2/4830/5670, Geforce 8800GT/9800GT/GTS250/450
1280x1024 M30: Radeon HD3870X2/4850/5770, Geforce 9800GX2/GTX260/GTS450
1280x1024 M40: Radeon HD4890/5830/6850, Geforce GTX280/460
1280x1024 M50: Radeon HD4850X2/5850/6850, Geforce GTX295/465
1280x1024 M60: Radeon HD4870X2/5870/6870, Geforce GTX295/470

1680x1050 M10: Radeon HD2900 Pro/3850/4670/5570, Geforce 8800GS/9600GSO/GT220/430
1680x1050 M20: Radeon HD3870X2/4830/5750, Geforce 8800GTS 512/9800GTX/GTS250/450
1680x1050 M30: Radeon HD4870/5830/6850, Geforce 9800GX2/GTX260-216/GTS450
1680x1050 M40: Radeon HD4850X2/5830/6850, Geforce GTX295/460
1680x1050 M50: Radeon HD4870X2/5850/6870, Geforce GTX295/470
1680x1050 M60: Radeon HD6950/2x HD4890/5830/6850, Geforce GTX295/480

1920x1080 M10: Radeon HD2900 Pro/3850/4670/5570, Geforce 8800GS/9600GSO/GT240/GTS450
1920x1080 M20: Radeon HD3870X2/4770/5750, Geforce 9800GTX+/GTS250/450
1920x1080 M30: Radeon HD4890/5830/6850, Geforce GTX275/460
1920x1080 M40: Radeon HD4870X2/5850/6850, Geforce GTX295/470
1920x1080 M50: Radeon 2xHD4890/5830/6850, Geforce GTX295/480
1920x1080 M60: Radeon 2xHD5830/6850, Geforce GTX580/2x GTX460

2560x1600 M10: Radeon HD3870X2/4830/5750, Geforce 8800GTS 512/9800GTX/GTS250/450
2560x1600 M20: Radeon HD4850X2/5830/6850, Geforce GTX285/460
2560x1600 M30: Radeon HD4870X2/5870/6950/2xHD6850, Geforce GTX295/470
2560x1600 M40: Radeon HD6970/2xHD5830/6850, Geforce GTX580/2xGTX465
2560x1600 M50: 2x Radeon HD5870/6870, 2x Geforce GTX470
2560x1600 M60: 2x Radeon HD6950/3x5850, 2x Geforce GTX480


Very High Detail, 8xAA
1280x1024 M10: Radeon HD2900XT/3870/4830/5670, Geforce 8800GT/9600GT/GT240/430
1280x1024 M20: Radeon HD4850X2/5830/6850, Geforce 9800GX2/GTX260/GTS450
1280x1024 M30: Radeon HD4870X2/5850/6870, Geforce GTX285/460
1280x1024 M40: Radeon HD6950(spec)/2xHD5830/6850, Geforce GTX295/470
1280x1024 M50: Radeon HD5970/HD6970(spec)/2x6850, Geforce GTX480
1280x1024 M60: 2x Radeon HD5870/6870, 2x Geforce GTX465

1680x1050 M10: Radeon HD3870X2/4830/5750, Geforce 8800GT/9800GT/GT240/GTS450
1680x1050 M20: Radeon HD4850X2/5830/6850, Geforce 9800GX2/GTX260-216/GTS450
1680x1050 M30: Radeon HD5870/6950/2xHD4890, Geforce GTX295/460 1GB
1680x1050 M40: Radeon HD6970/2xHD5830/6850, Geforce GTX295/480
1680x1050 M50: 2x Radeon HD5870/6870, Geforce GTX580/2x GTX460
1680x1050 M60: 2x Radeon HD6950/3xHD5850, 2x Geforce GTX470

1920x1080 M10: Radeon HD3870X2/4770/5750, Geforce 8800GT/9800GT/GTS250/450
1920x1080 M20: Radeon HD4870X2/5850/6850, Geforce GTX275/460
1920x1080 M30: Radeon HD6950(spec)/2x5830/6850, Geforce GTX295/470
1920x1080 M40: 2xRadeon HD5870/6850, Geforce GTX580/2x GTX460
1920x1080 M50: 2xRadeon HD6950(spec)/3x5850, 2x Geforce GTX470
1920x1080 M60: 2xRadeon HD6970(spec)/3x5870, 2x Geforce GTX480

2560x1600 M10: Radeon HD4850X2/5830/6850, Geforce 9800GX2/GTX260/GTS450
2560x1600 M20: Radeon HD6970/2x HD5830/6850, Geforce GTX295/480
2560x1600 M30: 2x Radeon HD6950/3x HD5850, 2x Geforce GTX470
2560x1600 M40: 2x Radeon HD6970/4x HD5850, 2x Geforce GTX480
2560x1600 M50: 3x Radeon HD6970, 2x Geforce GTX580/3x GTX480
2560x1600 M60: 4x Radeon HD6970, 3x Geforce GTX480



Chalk and cheese here, the force unleashed 2 can be maxed out on a 30" monitor with a single sub £180 graphics card.
Test Drive Unlimited 2 though, you're looking at 4 of AMD's as-yet unreleased future flagship, and that's assuming that quad crossfire scaling is nearly as epic as the HD6800's dual GPU scaling. It's going to take at least 180% to pull that off.
Meanwhile a 'mere' Tri-SLI combo from nvidia would seem to work if the scaling is good, but that's still almost 900W of TDP from either side, and at least £1000 from nvidia, probably £1500 from AMD.

 

The price we pay for bleeding edge technology LOL!

 

Yeah but a 4-figure sum for a racing game? Unusually excessive. Especially since the first TDU didn't even look that good.

 

its a racig game, why would AA even be needed, everything is blured as it is, esp 8xaa.

 

Actually aliasing was a real problem in the original Test Drive Unlimited, far moreso than most other racers I've played. To be fair, a properly designed racing game really does still need AA applied, as not all the textures are in motion!

 

600$ is almost too much for me :S