How much would another 4890 add to a 4870x2 + 4890 31k score.

[tajoh111]

I have a trifire setup consisting of a 4870x2 + a 4890.

I score about 31.1K in 06 with a core i7 @4.66 with ht on.

http://service.futuremark.com/result...eResultType=14

How much would another 4890 add to the score?

The score is made out of the following

sm 2.0 11735
sm 3.0 15620
CPU 8252.

score is 31124.

[TheGoat Eater]

please use proper posting guidelines!
this would be more suitable in a general hardware section or info/support possibly

[tajoh111]

It such an obscure setup that I have a feeling only people that are attempting to break records have tried it.

If you look for a setup like this on HWBOT, you will find 0, so I think the only chance I will get this question answered is in this section. If I search google, zero setup like this will pop up.

Heck, even for those that can't answer my question, I think its interesting to have mixed 3d06 results. As this is a rare setup.

[G H Z]

Hard to say, it depends on the board and whether the drivers work well with the hardware. There are setups posting to hwbot with this config, however they're hard to track down because hybrid CF (like you have) is in the same category as normal CF/CFX rigs.

[chew*]

The answer is very little in 06......you are cpu bottlenecked already....Hybrid crossfire is also not as linear as cards of the same type and clocks...

[tajoh111]

Chew, so you experimented with this setup?

I imagine that with your high overclocked core i7, say 5.3 ghz, you could unleash some of the bottleneck.

Is this setup atleast more powerful than 3 4890s?(1 4870x2 + 2 4890s)

[SNiiPE_DoGG]

well, what youve really got here is 4 4870's because of the way CF works - that said 3 4870's doesn't beat 2 4890's so I dont see why 4 would seeing as ATI's quadfire drivers are the suck.

[chew*]

I experimented with a few hydrid configruations.......None of them did any decent numbers.....Hybrids efficiency isn't that great, Its really cool for the average joe gamer......not so cool for the bencher.

Preety much what snipe said is correct, it will downgrade to your lowest denominator......

I have tested 2x 4890's versus 1 x 4870 x2.....as well......all cards at the max clocks I could get on air, the 4890's won by 1k........2 4870x2's will trump 3 4890's.

Preety sure only thing that could beat 2 4890's is a couple of hipros insanely modded 4870's ( not x2's as they don't clock as good )


Haven't tested with i7, only amd but on LN2, so same difference......

[tajoh111]

well, what youve really got here is 4 4870's because of the way CF works - that said 3 4870's doesn't beat 2 4890's so I dont see why 4 would seeing as ATI's quadfire drivers are the suck.

Atleast for 3dmark, this is not correct.

The 4890 and 4870x2, are clocked independently for some reason. If I run 3dmark06, and keep the 4870x2 at stock and raise the clocks of the 4890, my score will increase compared to everything was stock. That's why I think it has the potential to be such a powerful solution. A 4870x2 is stronger than a 4890 singlely, so if one could do it, a 4870x2 clocked at 860-900 with 2 4890s clocked at 1000, could do huge damage on the orb, with enough cpu power.(but again cpu power is everything in 06).

I thought hybrid crossfire isn't that bad those. I thought the 4890 added almost as much 06 score as a 2nd 4870x2.

[chew*]

Atleast for 3dmark, this is not correct.

The 4890 and 4870x2, are clocked independently for some reason. If I run 3dmark06, and keep the 4870x2 at stock and raise the clocks of the 4890, my score will increase compared to everything was stock. That's why I think it has the potential to be such a powerful solution. A 4870x2 is stronger than a 4890 singlely, so if one could do it, a 4870x2 clocked at 860-900 with 2 4890s clocked at 1000, could do huge damage on the orb, with enough cpu power.(but again cpu power is everything in 06)

Wrong, the higher you clock the cards seperately of each other.....the more you are going to notice a stuttering effect...this stuttering effect is due to the clock speeds not being matched and will most certainly have an impact on your score.

For crossfire best results are obtained with matched core/memory speeds......

8 x costs you nothing on AMD.......I benched 2x 4850's for quite some time on the gigabyte ds4h ( 8x X 8x ) swapping to the asus board with 16x by 16x netted me no gains.....

[tajoh111]

Wrong, the higher you clock the cards seperately of each other.....the more you are going to notice a stuttering effect...this stuttering effect is due to the clock speeds not being matched and will most certainly have an impact on your score.

For crossfire best results are obtained with matched core/memory speeds......

8 x costs you nothing on AMD.......I benched 2x 4850's for quite some time on the gigabyte ds4h ( 8x X 8x ) swapping to the asus board with 16x by 16x netted me no gains.....

I swear those for me that has not been the case. I have clocked them both at 800, and it lowers my score significantly.

Thats not to say my run to run average do change alot, as same clock scores vary from run to run.

[chew*]

It will be more noticable as you eliminate the cpu bottleneck.

[SNiiPE_DoGG]

well if you put a 3rd card in your gonna drop the link to 8x on all slots... that means 4x per GPU on the 4870x2 = not good