Discussion in 'Reviews' started by W1zzard, Oct 21, 2010.
To read this review go to: http://www.techpowerup.com/reviews/ATI/Radeon_HD_6850_CrossFire/
Just finished all 5 of the TPU reviews, all of them are great as always and i have to say i am very impressed with the 6850 and 6870 alone but both of them in crossfire really supprised me.
Under water and overclocked/volted either cards would make for an amazing setup.
Pretty good scaling on these cards, especially if we take into notice, that these are the first drivers. As they release more mature drivers, the scaling will probably even be slightly better. Too bad they don't support 3- or 4-way Crossfire.
Very nice. Price, performance, 1 x 6-pin - could drop in two 6850's and use my current PSU without adapters.
Unfortunately for AMD I'm still too pissed about the new naming to buy anything right now. Perhaps they should take a cue from a company we used to know - ATI!
seems good deal, great work w1z
Awesome review, this just puts alot of pressure on the current GTX460 in SLi, going by the results the 6850 in crossfire out performs the GTX460 in SLi in most of the games tested and at about the same price point here in AUS.
Just a shame they call it a 6850.
It seems like the final Crossfire scaling value is wrong.
(The one in this picture)
In almost every tested game there is an FPS increase that is almost double than the single card excluded some cases (call of juarez, WOW and supreme commander 2).
Excluding Call of Juarez (that is bugged in the drivers) the average increase should be more than 80 % instead of 59% (@fullhd resolution), including Call of Juarez it should be 78-79%.
The graphs should be corrected.
I thought some of the numbers were a bit low but i assumed i may have been missing something and i was too lazy to do the math myself
Yes, that's true.
Techpowerup guys made the same mistake also in the gtx 460 sli review (and I think in other xfire/SLI review).
Here the right values for a 6850 crossfire config:
Scaling @ 1920x1200 Games:
1) Alien Vs Predator : Image
2) Bad Company 2 : Image
3)BattleForge : Image
4)Call of Duty 4 : Image
5)Call of Juarez 2 : Image
-0,9%% !(relevant driver issue)
7)Dawn of war 2 : Image
8)Dirt 2 : Image
9)Far cry 2 : Image
10) HAWX : Image
11) Metro 2033 : Image
12) Riddick : Image
13) Stalker :Image
14)Supreme Commander 2 :Image
15) UT3 : Image
16) WoW : Image
Scaling @ 1920x1200 Bench:
1) Unigine Heaven 2 : Image
2) 3DMark03 : Image
3) 3DMark05 : Image
4) 3DMark 06 : Image
So, Scaling boost with Games (no Bench) @ 1920x1200 in crossfire 2 x 6850 config, (excluding call of juarez 2 for relevant driver issue) is +84,12%!
Scaling Boost including everything: Games,Bench, and also Call of Juarez 2, is: +75,8%
For crossfire in WOW, W1zz has already answered the question, see here -
AMD Radeon HD 6870 CrossFire
Tnx, but this is not an answer to the mistake.
The problem it is completely different.
Based on the review's bench, the final "scaling" boost graph is wrong.
Ahh I see what you mean, but I suppose W1zz had to include all the tests he had run.If he didn't people would complain that he was picking and choosing which results to include.
Maybe these benchmarks could be dropped from future crossfire/SLI reviews.
That does sound much better but i'm still to lazy (or stupid ) to do the math myself so i can't be sure, all i know is with the games i play it is around 95 to 100% scaling so i'm very happy
I've also includeded on the final scaling values ALL the test in the review.
But the final value it is completely different.
No way that it could be 59%.
updated with also bench values.
If your math on the 1920x1200 numbers is right then does that not mean every one of the crossfire scaling numbers for each res and overall are wrong?
Also would it apply to the 6870 crossfire numbers as well?
I don't really know what the numbers in the "performance scaling summary" represent:
But they DO NOT represent the performance boost from one single card to 2 crossfire card.
The correct values are the one i've posted before.
This is starting to get confusing hopefully w1zzard will pop into this thread at some point and let us know a little more.
To be honest though no matter what those numbers at the end of the review mean or are relative to it does not effect me as i have already ordered my 2 6870's and know that at the closest res to what i want to run the few games benchmarked here that i want to play they scale around 95% to 100% and to be honest i would have been really happy with 80%
I hope that's not a fault of my bad english that generate confusion
I'me agree with you: the important values are the single game chart
No definatly not because of your english, it seams pretty much perfect to me.
I have been up all night so i'm probably just confusing myself with numbers as i'm not that great with math
I was just thinking of the individual games and how they effect me and never thought about how relative that was, you are right about how importaint they are, if i played a lot of call of juarez 2 it would almost instantly stop me buying 2 6870's and make me get just the one
Would be great if we see some summary and comparison between GTS 450 SLI; GTX 460 SLI and HD6850; HD6870 CrossFire
That would be quite helpful for a lot of people who can't decide between them yet... plus i hope (looks like 6870 crossfire would come out on top but i can't be sure due to lack of sleep ) it would be something else i can use to make my gamer friends even more jelous
Just for example, in the gtx 460 SLI review, the scaling summary says that for the 1920x1200 resolution, it was 72%, vs the 59% of the 6850 crossfire.
GTX 460 SLI:
But if you made the math yourself, (of if you just see the game performance), you see that the increase of performance with the 6850 CF @ 1920x1200 is about 90% every game.
The correct value is the one i've posted before (around 80%) and NOT 59%.
So, the scaling performance summary for the 6850 crossfire, is wrong. I've not done the math, as i've done for the 6850, for the 460 SLI. I may assume that it is right for the 460 SLI. But for the 6850 CF it is definitely wrong.
When i was reading through the review and looking at the frame rate increases from one to two cards i was expecting the kind of numbers that are shown on the 460 chart you just linked to.
Separate names with a comma.