1x HD6950 > 2x HD4870 CF?

[OOZMAN]

A buddy of mine got another 4870, we benched using 3DMark vantage, his GPU score was P17000, mine was P22000 with my 6950. I he doing something wrong? I thought for sure 2x 4870s would be better. Thoughts? Opinions?

 

[trickson]

A buddy of mine got another 4870, we benched using 3DMark vantage, his GPU score was P17000, mine was P22000 with my 6950. I he doing something wrong? I thought for sure 2x 4870s would be better. Thoughts? Opinions?

Sounds about right to me .

 

[Volkszorn88]

A buddy of mine got another 4870, we benched using 3DMark vantage, his GPU score was P17000, mine was P22000 with my 6950. I he doing something wrong? I thought for sure 2x 4870s would be better. Thoughts? Opinions?

Not only is the 69xx 2 gens newer, you forget that scaling for 4k series isn't great. You're almost better off having just 1 card.

But now with the AMD 6k and 7k series, scaling is BEYOND beautiful. Same goes to the newer nvidia cards.

Just newer tech out performing older tech, nothing strange about that.

 

[lilhasselhoffer]

Crossfire was a relatively new technology with the 4xxx series. It was implemented without any refinement, and the hardware/software combinations really not up to snuff.

Yes, the 4xxx series and 6xxx series are both based upon the same manufacturing technology. No, this does not mean they will perform the same. New instruction sets, revised structures, and upgrading levels means a 48xx series crossfire wouldn't even be reasonable to compare to a 59xx series card. This doesn't even count a two generation leap, where software and hardware have matured.


So, these results seem reasonable. Two generations, to say nothing of the performance level disparity, worth of improvements means the 6950 single card should be at or well above the performance of a CF with two older cards.

 

[erocker]

Crossfire was a relatively new technology with the 4xxx series. It was implemented without any refinement, and the hardware/software combinations really not up to snuff.

Not really. Though it didn't have scaling as good as the 5 and 6 series, it was a good improvement over the 3xxx series and prior. The current implementation of CrossFire started with the x1950 series so CrossFire wasn't really that new when the 4 series was released.

 

[renq]

Yes, the 4xxx series and 6xxx series are both based upon the same manufacturing technology.

No, the 4xxxx series (with the exception of the 4770 which was AMD's first try at 40nm) is based on 55nm tech, and the 6xxx is based on the 40nm tech

 

[GotNoRice]

A buddy of mine got another 4870, we benched using 3DMark vantage, his GPU score was P17000, mine was P22000 with my 6950. I he doing something wrong? I thought for sure 2x 4870s would be better. Thoughts? Opinions?

Yeah that's about right.

Here is my score from back when I was only running a single 4870x2. I scored right around 17k, although it might have been held back to an extent by the Q9650 @ 4.4Ghz I was running at the time.

I disagree, however, with the idea that these cards have bad scaling. The 6xxx series may have made scaling improvements but that doesn't retroactively mean the 4xxx series is any worse than it was before- it scales just fine. The 4xxx series has also continued to benefit from new drivers and new caps just as newer hardware has.

Scaling might not be as good as 6xxx series or 7xxx series, but it's good enough for 2x 4870x2 to produce a score that would give a typical 7970 a run for it's money...

 

[renq]

Scaling might not be as good as 6xxx series or 7xxx series, but it's good enough for 2x 4870x2 to produce a score that would give a typical 7970 a run for it's money...

In 3dMark, yes. In games- depends. 4000 series doesn't support DirectX11, 3D (HD3D) and some more features/technologies present in newer GPUs. Also not all games benefit from CF, regardless of the GPU generation

 

[OOZMAN]

Yeah that's about right.

Here is my score from back when I was only running a single 4870x2. I scored right around 17k, although it might have been held back to an extent by the Q9650 @ 4.4Ghz I was running at the time.

http://gotnorice.no-ip.org/q9650hd4870x2.png

..

He is running a P2 X4 940 unlocked to 5 cores and 3.9ghz... So that seems right

 

[LifeOnMars]

5 cores? wow that's a unique chip he has there

 

[renq]

5 cores? wow that's a unique chip he has there

Yep, AMD has only five Thubans with disabled cores called Zosma- 650T, 840T, 960T, 960T BE and 970BE, it's possible with these to unlock an additional core to make 5 cores
http://en.wikipedia.org/wiki/List_o...rs#.22Zosma.22_.28E0.2C_45_nm.2C_Quad-core.29

 

[LifeOnMars]

Ok but you said he had a 940?

 

[Artas1984]

First off all you guys are talking garbage "crossfire was relatively new in HD4800 series" - i recall doing benchmarks with CrossFire HD3870, and there were a few games where scaling was 100 %. Scaling was good even with the first CrossFire cards, CrossFire X1950 Pro 512 setup bested the 8800GTS 320 in many games, in higher resolutions already.

I have done many many tests with AMD cards, and can say that CrossFire HD4870 can be EQUAL to single HD6950. 3Dmark Vadtage is BS, that does not do any justice to CrossFire scaling.

That fact of the matter is: know your game! For instance in older games like Bioshock, Serious Sam 2, Necrovision, Jericho, FEAR 2 the HD4870 CrossFire beats the HD6950 by a small margin. In more recent titles CrossFire HD4870 just does not scale that right.

 

[JATownes]

First off all you guys are talking garbage "crossfire was relatively new in HD4800 series" - i recall doing benchmarks with CrossFire HD3870, and there were a few games where scaling was 100 %. Scaling was good even with the first CrossFire cards, CrossFire X1950 Pro 512 setup bested the 8800GTS 320 in many games, in higher resolutions already.

I have done many many tests with AMD cards, and can say that CrossFire HD4870 can be EQUAL to single HD6950. 3Dmark Vadtage is BS, that does not do any justice to CrossFire scaling.

That fact of the matter is: know your game! For instance in older games like Bioshock, Serious Sam 2, Necrovision, Jericho, FEAR 2 the HD4870 CrossFire beats the HD6950 by a small margin. In more recent titles CrossFire HD4870 just does not scale that right.

IDK if I agree with this. I went from a pair of 4850s to a single 6870 and the 6870 definitely gave better frame rates in most games, including quite a few older games.

 

[Artas1984]

And i have HD6870 too, and it is better than CrossFire HD4850 indeed.

But we are talking about CF HD4870 VS HD6950, why are you sticking here your HD6870?

 

[JATownes]

Because the 6950 is significantly faster than the 6870, and considering the 6870 trades blows with 4870x2 (and beats is more often than not), it stands to reason that the 6950 should definitely be faster 4870x2, since the 4870 is only slightly faster than the 4850.

I was just taking issue with this statement:

I have done many many tests with AMD cards, and can say that CrossFire HD4870 can be EQUAL to single HD6950. 3Dmark Vadtage is BS, that does not do any justice to CrossFire scaling.

Just trying to help OP with some comparison.

Edit: Here is a good link for you OP: 4870x2 vs 6970 vs 6850. Figure the 6950 to be ~10% slower than the 6970 that is shown.

 

[Artas1984]

PERHAPS I WAS NOT CLEAR, i will repeat - CF HD4870 beats the HD6950 in a minority of games by a small margin, particularly in 2-3 year old titles; in the majority of scenarios, however, HD6950 will come on top.

And...

HD4870 is ALOT faster than HD4850. Just the mere memory bandwidth is TWICE bigger. Sure, in 800X600 resolutions the difference between them will be the difference between core clock, but try 1920X1200 with 4XAA and HD4850 will choke to dust and HD4870 will pull out 2X more frame rates.

I am recommending HD6950 of course, don't get me wrong.

 

[JATownes]

I meant no offense. I agree that at high resolution the 4870 is quite a bit faster than the 4850. I was just attempting to help the OP by offering a personal viewpoint regarding going from the 4000 series to the 6000 series.

I am with you in recommending the 6950.

 

[OOZMAN]

Ok but you said he had a 940?

960.. watever.

He planned to unlock it to 6 cores and 4.5ghz, but he could only do 5 cores and 3.9, he's quite naive, he still believes his setup is better for gaming than mine.

Thanks to everyone for your input, once again I have learnt much!

 

[LifeOnMars]

No probs, and yes your rig is substantially better ( I myself am going back to sandy bridge tomorrow) I have had a SB setup before and it rocked, unfortunately I had to sell at the time due to losing my job but things are back on the up and I can now provide my gtx 560ti with the minimum frame rates it deserves