AMD Radeon HD 6870 CrossFire

[CDdude55]

Originally Posted by jasper1605

Something I've never understood about the framerate deal. Who cares after it passes 60? Monitors only refresh at 60-85hz correct? After 90 the human eye can't tell and it goes into the subliminal message world. Having said that, it does not detract from my previous statement that I want 1,000 fps in solitaire. And 500 fps in farmville.

Some people can actually see a pretty significant difference from going above 60 frames, though i have yet to see it..

[bear jesus]

Originally Posted by jasper1605

Something I've never understood about the framerate deal. Who cares after it passes 60? Monitors only refresh at 60-85hz correct? After 90 the human eye can't tell and it goes into the subliminal message world. Having said that, it does not detract from my previous statement that I want 1,000 fps in solitaire. And 500 fps in farmville.

To me it's about the frame rate drops, if the average is above 60 that's great but if it drops down to like 20fps during cirtain parts of a game then it really messes with the experiance.
As most sites give the average fps the higher the better as it lowers the chance of getting annoying fps drops.

[jasper1605]

Originally Posted by CDdude55

Some people claim to actually see a pretty significant difference from going above 60 frames, though i have yet to see it..

fixed it for ya

edit: good call bear. Didn't think of it that way.

[n-ster]

minimum fps is what is important

[jasper1605]

Originally Posted by n-ster

minimum fps is what is important

indeed. like when my ps3 drops to about 8 fps on super star dust hd in the 3d mode after beating the final planet. I had a lol when I saw how slow it got

[nt300]

Benchmark results are very similar to other review sites, Good Job Wiz, and super fast HD 6870 Crossfire performance, better than HD 5970 monster too

[Tatty_One]

Originally Posted by Dj-ElectriC

The scaling is near perfect

Lol, near perfect at 52% at 19XX res?, try half perfect, or 20% less than GTX460's at the same res (72%). I love the performance figures, especially for some reason with the HD 6850's in crossfire, however I fundemenatally have an issue with any setup that in effect means I flush 48% of the value of the 2nd card down the toilet.

[bear jesus]

Originally Posted by Tatty_One

Lol, near perfect at 52% at 19XX res?, try half perfect, or 20% less than GTX460's at the same res (72%). I love the performance figures, especially for some reason with the HD 6850's in crossfire, however I fundemenatally have an issue with any setup that in effect means I flush 48% of the value of the 2nd card down the toilet.

I think he is refering to the fact that multiple games get 100% scaling and other are at 80/90%+

Really the scaling percentage at the end of the review seams to be much lower than what would be expected when looking at the frame rate increase it gets with most games.
In the 6850 crossfire review comments thread it's been pointed out that the scaling percentage of the games does not match up to the ingame numbers, i really have to wonder how the end percentage is worked out for it to be so much lower than expected when multiple games get upto 90 to 100% scaling

[bear jesus]

Also zero wing did the same for the 6850 review thread, his results are whats quoted below.

"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%"

I fail at working out percentages when it comes to something like this thus why i have not done it myself but it seams that either the numbers on both reviews or wrong or everyone is missing something that was included in the reviews that caused the numbers to be so much lower.

*edit*
i'm assumig the point still stands when you are done panther

[pantherx12]

Sorry for deleting my post, realised I messed up quite badly in a few places,was doing the maths last night as I was falling asleep lol

Gime 10mins and accurate figures will be up.

[HTC]

Originally Posted by Tatty_One

Lol, near perfect at 52% at 19XX res?, try half perfect, or 20% less than GTX460's at the same res (72%). I love the performance figures, especially for some reason with the HD 6850's in crossfire, however I fundemenatally have an issue with any setup that in effect means I flush 48% of the value of the 2nd card down the toilet.

With the 6850, it was pointed out the 19XX res numbers were off, by quite a bit. Let's see if i can add it up:



97.4% increase here.



89.8% here.



99% here.



83.5% here.



-2% here: this game clearly doesn't scale @ all.



96.6% here.



80.5% here.



79.2% here.



100.9% here.



97.5% here.



84.6% here.



88.2% here.



70.9% here.



16% only here.



52.4% here.



45.5% here.


Does this not add up to 73.5%, including Call of Juarez 2? 78.5% if not.



Dunno where that 52% came up from but it seems that W1zz's math has had some problems: that or my math sucks.


EDIT

Adding remaining benches to the calculations:



90.7% here.



21.1% here.



22.1% here.



99.3% here.

All benches included, i get an average of 70.7%.

[pantherx12]

I Can't be bothered correcting this entire post, see correct numbers here AMD Radeon HD 6870 CrossFire and
here if you want 3dmark/unigine added to my calculations AMD Radeon HD 6870 CrossFire


Here's my numbers corrected
27.1 - 53.5 AVP 97.41% scaling
52.9 - 100.4 BF:BC 89.79% scaling
42.7 - 85.0 Battleforge 99.06% scaling
115.8 - 212.5 COD4 83.50% scaling
30 - 59 Crysis 96.66% scaling
112.9 - 203.8 DAWN of war 2 80.51% scaling
84.1 - 150.7 dirt 2 79.19% scaling
67.9 - 136.4 farcry 2 100.88% scaling
99.1 - 195.7 HAWX 97.47% scaling
20.1 - 47.1 metro2033 134.32% scaling
55.7 - 104.8 Riddick 88.15% scaling
41.3 - 70.6 stalker cs 70.94% scaling
148.2 - 225.9 UT3 52.42 scaling
67.7 - 98.5 wow 45.49% scaling

81.2 - 94.3 sucom 2 16.13% scaling
96.3 -98.3 COJ2 2.07% scaling

Will confirm average in just a moment : ]

Average excluding sucom2 and COJ2 86.842%

77.12% scaling including the them.


Also if you exclude UT3 and WOW ( tbh honest, why are these games in a crossfire review? lol)

93.15% scaling.


Just in-case I messed up again I used this tool http://www.csgnetwork.com/averagecalc.html and this one http://www.onlineconversion.com/percentcalc.htm to triple check everything.


Correct me if wrong of course!


Does seem wizz's graph is wrong though eh.

[HTC]

Originally Posted by pantherx12

Here's my numbers corrected
27.1 - 53.5 AVP 97.41% scaling
52.9 - 100.4 BF:BC 89.79% scaling
42.7 - 85.0 Battleforge 99.06% scaling
115.8 - 212.5 COD4 83.50% scaling
30 - 59 Crysis 96.66% scaling
112.9 - 203.8 DAWN of war 2 80.51% scaling
84.1 - 150.7 dirt 2 79.19% scaling
67.9 - 136.4 farcry 2 100.88% scaling
99.1 - 195.7 HAWX 97.47% scaling
20.1 - 47.1 metro2033 134.32% scaling
55.7 - 104.8 Riddick 88.15% scaling
41.3 - 70.6 stalker cs 70.94% scaling
148.2 - 225.9 UT3 52.42 scaling
67.7 - 98.5 wow 45.49% scaling

81.2 - 94.3 sucom 2 16.13% scaling
96.3 -98.3 COJ2 2.07% scaling

Will confirm average in just a moment : ]

The one i marked out is wrong: it's 20.1 to 37.1 FPS.

[bear jesus]

The increases within the benchmarks are what got me to order a pair of 6870's (arriving on friday ) if i had only looked at the scaling resluts at the end i would have never once even considered crossfire.

I really want to know why the reviews have it so much lower than everyone is working out, is there something included in the review that everyone is missing?

[HTC]

Originally Posted by bear jesus

The increases within the benchmarks are what got me to order a pair of 6870's (arriving on friday ) if i had only looked at the scaling resluts at the end i would have never once even considered crossfire.

I really want to know why the reviews have it so much lower than everyone is working out, is there something included in the review that everyone is missing?

Dunno, really: maybe we're doing it wrong? My math skills aren't very good but i think i did this particular math properly.

[bear jesus]

Originally Posted by HTC

Dunno, really: maybe we're doing it wrong? My math skills aren't very good but i think i did this particular math properly.

I really don't expect w1zzard to have got something wrong thus why i want to know if we are all missing something as multiple people are all working the numbers out to be higher than is listed on the scaling page of the review.

[pantherx12]

cheers for pointing that out, must of typod whilst editing and missed it.
we working things out then lol


Okay with that in mind, excluding the games that don't scale well ( ut3 and below)
89.01% scaling
with wow/ut3 83.28 % scaling

Including all games 74.015% scaling.

[HTC]

Originally Posted by pantherx12

cheers for pointing that out, must of typod whilst editing and missed it.
we working things out then lol

No prob.

Just thought of something: what if he's including the 3DMark benches in his average increases?

[pantherx12]

Originally Posted by HTC

No prob.

Just thought of something: what if he's including the 3DMark benches in his average increases?

Adding that to calculations now : ]

[Tatty_One]

yes I make it 73.5 ish, although I have rounded down to 100% on those couple that said slightly over 100, not quite sure how you can get 2 cards in crossfire performing more than twice as well as a single card. it does seem strange how the calculation is made, although I am assuming the process would have been the same for the 460's in SLi.

[pantherx12]

Including all bench marks ( unigine and all 3d marks) scaling is 70.8685%

Although to be honest, considering games scaling so well and 3dmark doesn't to me shows it's no longer a relevant benchmark.


These are the numbers used for calculation
% scaling
97.41 AVP ( all these ones are in same order as tests)
89.79
99.06
83.5
96.66
80.51
79.19
100.88
97.47
84.57
88.15
70.94
52.42 UT3
45.49 WOW


16.13 sucom 2
2.07 COJ2


90.65 3d04
21.1 3d05
22.1 3d06
99.28 Unigine

[discy]

Originally Posted by bear jesus

Even though the scaling is not that awesome in many games the fps are still great and makes me wonder how well a pair of these do with eyefinity.

Will TPU be able/willing to do some eyefinity benchmarks with crossfire 6870 eyefinity?

http://www.widescreengamingforum.com...eatured_Review

[nt300]

Originally Posted by HXL492

crossfire scaling is not too impressive...
No wonder they kept it at a maximum of 2 cards lol

Umm, Crossfire scaling for the HD 6870 and HD 6850 is much better than Evergreen. Most if not all benchmarks including other sites they score almost double performance in Crossfire vs. the single HD 6870 card.

[zekans]

Originally Posted by Vipeax

What happened here (note the 5970 & 6870 Crossfire performance)?

GTX 460 review:
http://tpucdn.com/reviews/MSI/GTX_46..._2560_1600.gif

6870 Crossfire review:
http://tpucdn.com/reviews/ATI/Radeon..._2560_1600.gif

I believe the drivers are at fault, as far as I know these benchmarks where Crossfire destroys FPS were done with Catalyst 10.10 whilst the others were 10.7 or something.

Sadly I cant get 10.5-10.9 working on my new 6870 crossfire setup so I am stuck with rather bad FPS in WoW because of it. Hopefully AMD/ATI will pull the finger out and fix these drivers before I end up returning these and getting a 460 SLi setup.... if that turns out to be better I won't be returning to ATi anytime soon :S.

10.11 better fix it :/

[W1zzard]

Originally Posted by zekans

I believe the drivers are at fault, as far as I know these benchmarks where Crossfire destroys FPS were done with Catalyst 10.10 whilst the others were 10.7 or something.

my 6850 and 6870 numbers were done on the 10.10 driver. if the 10.10 driver were magically doing something for cf then you would see that in the scores, which you apparently dont

we are using wow dx11 in this round of benchmarks, so maybe that caused the problem