[rico2001]

Mixed crossfire review: Sapphire 4850 1GB + Sapphire 4870 1GB

Pulled my 4850 X2 and 4870 X2 out of the case and decided to throw my singles in for another crossfire configurations. Lets see how a 4850 + 4870 stack up against the 4850 X2. And yes this will be a pci-e 2.0 8X/8X configurations, so those still believing that misinforming tweaktown article, can see the truth about the bandwidth of pci-e 2.0 8X/8X.

SAPPHIRE Radeon 4870 1GB & SAPPHIRE Radeon 4850 1GB

Installed

Close up

Catalyst Control Center (CCC): Graphics Software info
Driver Packaging Version 8.62-090520m1-081756C-ATI
Catalyst Version09.6
Provider ATI Technologies Inc.
2D Driver Version8.01.01.909
2D Driver File Path /REGISTRY/MACHINE/SYSTEM/ControlSet001/Control/CLASS/{4D36E968-E325-11CE-BFC1-08002BE10318}/0004
Direct3D Version 8.14.10.0671
OpenGL Version 6.14.10.8673
Catalyst™ Control Center Version 2009.0520.1631.27815

Catalyst Control Center (CCC): Graphics hardware info
Primary Adapter
Graphics Card ManufacturerPowered by ATI
Graphics ChipsetATI Radeon HD 4800 Series
Device ID9440
Vendor1002

Subsystem ID0851
Subsystem Vendor ID174B

Graphics Bus CapabilityPCI Express 2.0
Maximum Bus SettingPCI Express 2.0 x8

BIOS Version011.012.000.002
BIOS Part Number113-AX89100-100
BIOS Date2008/12/09

Memory Size1024 MB
Memory TypeGDDR5

Core Clock in MHz750 MHz
Memory Clock in MHz900 MHz
Total Memory Bandwidth in GByte/s115.2 GByte/s

Linked Adapter
Graphics Card ManufacturerPowered by ATI
Graphics ChipsetATI Radeon HD 4800 Series
Device ID9442
Vendor1002

Subsystem IDE810
Subsystem Vendor ID174B

Graphics Bus CapabilityPCI Express 2.0
Maximum Bus SettingPCI Express 2.0 x8

BIOS Version011.004.000.000
BIOS Part Number11X-B50102-X06
BIOS Date2008/07/11

Memory Size1024 MB
Memory TypeGDDR3

Core Clock in MHz625 MHz
Memory Clock in MHz993 MHz
Total Memory Bandwidth in GByte/s63.6 GByte/s

Catalyst Control Center (CCC): CrossfireX configuration

Catalyst Control Center (CCC): ATI Overdrive Configuration

GPU-Z technical info
GPU-Z clock, idle temps., and fan speed info

GPU-Z load temps. info

Testbed:
E8300 @ 3.6G & 4.0G
ASUS P5Q-PRO
6Gb DDR2
ATI Catalyst 9.6
SAPPHIRE Radeon 4850 1GB
SAPPHIRE Radeon 4870 1GB
SAPPHIRE Radeon 4850 X2 2GB

Testing
3DMark 2006
3DMark Vantage
Call of Duty: World at War
Call of Juarez
Crysis (review part 2)
Devil May Cry 4 (review part 2)
Enemy Territory: QUAKE Wars (review part 2)
FarCry 2 (review part 2)
Furmark (review part 2)
Left 4 Dead (review part 2)
Resident Evil 5 (review part 3)
Sanctuary (review part 3)
Street Fight IV (review part 3)
S.T.A.L.K.E.R Clear Sky (review part 3)
X3 Terran Conflict (review part 3)

3DMark 2006:
4850 X2 = 17847
4850 + 4870 = 18803 (5% gain over X2)

3Dmark Vantage: Performance
4850 X2 = P10537
4850 + 4870 = P11034 (4% gain over X2)

Call of Duty: World at War
1,680 x 1,050 4xAA 16xAF, DX9, Maximum Detail
4850 X2 = 129 average fps
4850 + 4870 = 133 average fps (3% gain over X2)

1,920 x 1,080 4xAA 16xAF, DX9, Maximum Detail
4850 X2 = 120 average fps
4850 + 4870 = 125 average fps (4% gain over X2)

Call of Juarez:
1680 x 1050, 4xAA, DX10, highest quality settings
4850 X2 = 61.4 average fps
4850 + 4870 = 63.1 average fps (2% gain over X2)

1920 x 1200, 4xAA, DX10, highest quality settings
4850 X2 = 52.9 average fps
4850 + 4870 = 56.4 average fps (6% gain over X2)

To be continued.......

 

[rico2001]

Crysis:
1600 x 1200
4850 X2 = 33 average fps
4850 + 4870 = 33 average fps

1900 x 1200
4850 X2 = 29 average fps
4850 + 4870 = 28 average fps (3% loss to X2)

2560 x 1600
4850 X2 = 17 average fps
4850 + 4870 = 17 average fps

Devil May Cry 4:
1680 x 1050, DX9, 8xAA
4850 X2 = 170 average fps
4850 + 4870 = 170 average fps

Enemy Territory: QUAKE Wars
1600 x 1200, DX10, highest settings, 16xAF, 4xAA
4850 X2 = 119 average fps
4850 + 4870 = 129 average fps (8% gain over X2)

1920 x 1080, DX10, highest settings, 16xAF, 4xAA
4850 X2 = 110 average fps
4850 + 4870 = 118 average fps (7% gain over X2)

FARCRY 2:
1680 x 1050, DX10, Ultra high settings, 16xAF, 4xAA
4850 X2 = 62.8 average fps
4850 + 4870 = 62.9 average fps

1920 x 1080, DX10, Ultra high settings, 16xAF, 4xAA
4850 X2 = 58.7 average fps
4850 + 4870 = 56.2 average fps (4% loss to X2)

Furmark:
1440x900
4850 X2 = 197 average fps
4850 + 4870 = 185 average fps (6% loss to X2)

1680x1050
4850 X2 = 152 average fps
4850 + 4870 = 143 average fps (6% loss to X2)

Comment: Furmark doesn't like mixed memory type configurations, received same small loss in my "4850 tri-fire vs. 4850 X2 + 4870 tri-fire review.

Left 4 Dead:
1680 x 1050, DX9, 16xATF, 8xAA, V-sync off, Shader: very high, all other settings: high
Difficulty: Expert
Campaign: No Mercy
Mission: 1: apartments
FRAPS settings: 300 seconds
4850 X2 = 128.1 average fps
4850 + 4870 = 129.2 average fps (1% gain to X2)

 

[rico2001]

Resident Evil 5:
1680 x 1050, DX10, 16xAF, 8xAA, motion blur on
4850 X2 = 54.5 average fps
4850 + 4870 = 54.6 average fps

Comment: Cool new benchmark out, great looking, I'm definitely buying RE5. The motion blur on area 3 is killer. Ati said they haven't had a chance to optimize the cat. drivers for this game yet so we can expect better result by the time of official release.

Sanctuary:
1680 x 1050, DX10, 16xAF, 0xAA
4850 X2 = 111 average fps
4850 + 4870 = 113 average fps (1% gain over X2)

Street Fighter IV:
1680 x 1050, DX9, 16xAF 8xAA
4850 X2 = 146 average fps
4850 + 4870 = 146 average fps

S.T.A.L.K.E.R. Clear Sky:
1680 x 1050, DX10, 16xAF, 0xAA
4850 X2 = 54.6 average fps
4850 + 4870 = 58.2 average fps (6% gain over X2)

X3 Terran Conflict:
1680 x 1050, DX10, 16xAF, 8xAA
4850 X2 = 94 average fps
4850 + 4870 = 100 average fps (6% gain over X2)

Overclocking:
In a CF configuration, although the gpus can be clock idependently, the memory is shared and since the memory is shared the speed is dictated by the slowest of the two cards. The amount of memory used in the CF configuration is also dictated by the smallest memory amount of the tow cards. Since I'm running a 4850 and a 4870 together the dual piped DDR3 from the 4850 will be the dictating speed and size for this pairing. Asn since this is the case, I'v decided to leave the memory clock of the 4870 alone b/c changing it will have no affect. Here are my overclocking result for the two cards.

Sapphire 4870
gpu: 750 mhz to 810 mhz
memory: 900 mhz to 900 mhz

Sapphire 4850
gpu: 625 mhz to 700 mhz
memory: 993 mhz to 1075 mhz

Here are my results from a couple benches at these overclocked settings.
3DMark Vantage: P11675 (6% gain over stock)
Furmark: 1680x1050 - 263 fps average (3% gain over stock X3)
Sanctuary: 1680x1050 - 123 fps average (8% gain over stock)


Conclusion:
Well another good CF experience, the 4850 and 4870 work well together. The results were as expected, falling between a 4850 X2 and a 4870 X2. Scaling was also very good as expected. Being that all games respond differently to differently configurations, with the mixing of different memory types, I experienced some of the same gains and losses as in my "4850 X2 + 4870 tri-fire" review. I believe some game titles use more of the memory side of the cards than the gpus and vise verse, resulting in some gains in some game titles making better utilization of the increase gpu mhz of the 4870 and some games games not caring at all.

This review shows anyone can crossfire a 4870 with a 4850 with easy and I know this was common knowledge but for those to see with their own eyes and see real results. So those with a 4850 can have no fear in buying a 4800 series card to CF with. I believe the results also put to bed the age old debate of weather pci-e 2.0 at 8X bottlenecks our modern high end cards. In my opinion, the only cards that will have a decrease in performance due to 8X are the 4870 X2 and a highly overclocked 4890. I don't have a 4890 yet, but an eager to find out if my thinking is correct. Again I hope I shed some light on the power and outstanding versatility of Ati crossfireX. Hope you enjoyed.

 

[d3v0]

Pretty sure you can just cut to the conclusion already m8

It's sad that the mixed Xfire does so poorly, I had thought it would do better

 

[flushentitypacket]

Awesome info.

You can Crossfire different card types? What allows you to do that with these cards when it's seemingly not possible with other cards?

Also, just to settle this once and for all in my head, does x8 bottleneck any vid card on the market right now?

 

[I_dalder_I]

Hey we have the same CPU cooler! good show

yeah great review btw rep 4 u

 

[I_dalder_I]

Quote:

Originally Posted by flushentitypacket Awesome info. You can Crossfire different card types? What allows you to do that with these cards when it's seemingly not possible with other cards? Also, just to settle this once and for all in my head, does x8 bottleneck any vid card on the market right now?

YES you can crossfire any same series of ATI card "form 3xxxseries and up"

Only DUEL gpu cards can bottle neck x8

Yes you will die alone

 

[rico2001]

Quote:

Originally Posted by d3v0 Pretty sure you can just cut to the conclusion already m8 It's sad that the mixed Xfire does so poorly, I had thought it would do better

Obviously you have no idea what you are talking about. So far I'm getting scores between my 4850 X2 and my 4870 X2, which is what one could expect when setting up this CF configuration.

 

[skugpezz]

Quote:

Originally Posted by flushentitypacket Awesome info. You can Crossfire different card types? What allows you to do that with these cards when it's seemingly not possible with other cards? Also, just to settle this once and for all in my head, does x8 bottleneck any vid card on the market right now?

you have been on the green side too long, I have been mixing ati cards since the x1650 days, where I had a crossfire configuration consisting of a x1650 and a x1600

 

[Nelson2011]

lol i used to have 4870/4850 crossfire but those were 512mb cards

 

[rico2001]

update: 2nd set of benchmarks added.

 

[rico2001]

update: 3nd set of benchmarks and conclusion added.

 

[gsk3rd]

rico, thank you for all the benches and testing you do. to have most of all the high end ati cards at your disposal must be helpful. +rep.

 

[rico2001]

Quote:

Originally Posted by gsk3rd rico, thank you for all the benches and testing you do. to have most of all the high end ati cards at your disposal must be helpful. +rep.

Thanks, gsk3rd. Just stay one level under the latest card and find good deals on the net when you can for the others. I don't think I'll get a 4890 until the 5800 series gets really close. As you see in my sig, I'm looking for sapph. cards under the level of 4850.

 

[pipnasty]

great review! i don't know why many people are complaining. +rep

just a question, did you experience any microstuttering? or is microstuttering only limited to dual gpus on a single PCB?

also, i have a 4850 1GB card (oc'ed to 750/1100 - max stable) and I've been thinking of adding a 4870 1GB for hybrid crossfire. but you mentioned that memory downclocks to the weakest of the two (DDR3 of 4850), so would it be more cost efficient (since 4850 is cheaper) to get another 4850 1GB (and overclock it) rather than a 4870 1GB?

 

[rico2001]

Quote:

Originally Posted by pipnasty great review! i don't know why many people are complaining. +rep just a question, did you experience any microstuttering? or is microstuttering only limited to dual gpus on a single PCB? also, i have a 4850 1GB card (oc'ed to 750/1100 - max stable) and I've been thinking of adding a 4870 1GB for hybrid crossfire. but you mentioned that memory downclocks to the weakest of the two (DDR3 of 4850), so would it be more cost efficient (since 4850 is cheaper) to get another 4850 1GB (and overclock it) rather than a 4870 1GB?

Thanks, glad you like the review.

No microstuttering on this CF config. The only time I experienced ms was with the 4850 X2 alone. Was in Left 4 Dead and was easily fixed by turning off vertical sync (v-sync) in the game options. In all my other CF confgs, haven't seen any ms.

Yes with a mixed CF of 4870 + 4850, you can think of it as (1) 4850 gpu at 625 mhz + (1) 4870 gpu at 750 mhz, both working together under 1 Ghz of DDR3 running at 993 mhz. Pairing the two isn't a waste if you have a 4850 and come across a 4870 for cheap but yes it make more sense to just get a 4850 to CF with.

The memory issue with pairing a 4850 with a 4870 is a bandwidth issue. The memory of the 4850 is dual piped at 993 mhz = 63.6 GB/s bandwidth and the memory of the 4870 is quad pipe at 900 mhz = 115.2 GB/s bandwidth. And there is the bottleneck. You can never overclock the memory of the 4850 enough to match the stock bandwidth of the 4870. You would have to get the DDR3 of the 4850 up to roughly around 1700 mhz.

 

[pipnasty]

Quote:

Originally Posted by rico2001 Thanks, glad you like the review. No microstuttering on this CF config. The only time I experienced ms was with the 4850 X2 alone. Was in Left 4 Dead and was easily fixed by turning off vertical sync (v-sync) in the game options. In all my other CF confgs, haven't seen any ms. Yes with a mixed CF of 4870 + 4850, you can think of it as (1) 4850 gpu at 625 mhz + (1) 4870 gpu at 750 mhz, both working together under 1 Ghz of DDR3 running at 993 mhz. Pairing the two isn't a waste if you have a 4850 and come across a 4870 for cheap but yes it make more sense to just get a 4850 to CF with. The memory issue with pairing a 4850 with a 4870 is a bandwidth issue. The memory of the 4850 is dual piped at 993 mhz = 63.6 GB/s bandwidth and the memory of the 4870 is quad pipe at 900 mhz = 115.2 GB/s bandwidth. And there is the bottleneck. You can never overclock the memory of the 4850 enough to match the stock bandwidth of the 4870. You would have to get the DDR3 of the 4850 up to roughly around 1700 mhz.

thanks for that clear explanation