Crazyeyesreaper 6970 series review

[crazyeyesreaper]

Unigine Heaven Bench completed

1920x1200 4xAA 16xAF moderate normal and extreme tessellation runs

 

[Marineborn]

Nice bro, gotta love the 6000 series cards scaling its dam impressive

 

[crazyeyesreaper]

meh scaling is crap right now due to my cpu clock being stock overclocked results will come once i finish the last 3 games for single and dual card ill then do the 3 game comparison to the 5850 xfire after which ill overclock the cpu and the gpus and see what my system at max 24/7 settings is capable of

 

[travva]

nice review reaper!

 

[crazyeyesreaper]

yea its alright everyones really just waiting on me to overclock my cpu and gpus so we can see the real performance numbers on an AMD rig. Tired of seeing intel reviews all the time sure no cpu bottleneck in most situations but there are another 30+ % of the PC world using an amd powered rig so why not fill the gap eitherway im wrapping up stock clock testing today then ill move on to 3 main titles to test with overclocking

Metro 2033

Crysis Warhead

Alien Vs Predator

 

[Lionheart]

yea its alright everyones really just waiting on me to overclock my cpu and gpus so we can see the real performance numbers on an AMD rig. Tired of seeing intel reviews all the time sure no cpu bottleneck in most situations but there are another 30+ % of the PC world using an amd powered rig so why not fill the gap eitherway im wrapping up stock clock testing today then ill move on to 3 main titles to test with overclocking

Metro 2033

Crysis Warhead

Alien Vs Predator

I agree

 

[crazyeyesreaper]

well DMC4 benchs are up DX9 mode for the 5800 vs 6900 and DX10 for 6970 vs 6970 xfire

 

[pantherx12]

Making myself a coffee and I'll get on doing the percentages for you dude.

Do you want a percentage for each thing like min/average/max for each test or just an average?

Can do either or as I'll still need to do roughly same ammount of work.

 

[crazyeyesreaper]

just the avg fps

be carefull with DMC4 tho as theres 4 scenes but go ahead and average out all 4 for general scaling

ill have overclock numbers eventually as well for you to crunch right now im just installing and getting ready to wrap up the last game (The last Remnant)

 

[pantherx12]

Right here you go dude

6970 average crossfire scaling

Heaven moderate : 75.5
Heaven Normal : 86.84
Heaven extreme : 92
Metro test 1 : 69.7
Metro test 2 : 96.15 (personal note for metro, why even turn DOF on! lol)
BBC2 test single player : 58.33
BBC2 test multi : 18.03
dirt 2 : 24.24
AVP : 89.65
Fallout NV: 0 ( due to force v-sync)
Just cause 2: 33.92
l4d2 : 17.39
Mecha rampage : 64.58
Crysis warhead :8.33
DMC4 : sc1 92.18 sc2 84.76 sc3 44.65 sc4 73.38 ( I kept this all sepperate as it's interested to see different scaling in the same game I think lol, scaling is dependant on what's being rendered maybe?)

Bad scaling seems to happen on games the 6970 doesn't do to brilliantly at anyway so hopefully future drivers fix this somewhat.

 

[crazyeyesreaper]

people also need to remember im still at stock clocks

and i can tell everyone here right now

Metro 2033 with DX11 direct compute Depth of Field runs FLAWLESSLY and looks incredible in fast action scenes. This is a title where 6970 muscle allows you to go from unplayable to breath taking by simple tossing another gpu in there + the scaling dosent lie with DoF off im CPU limited DoF on the gpu load goes up so it balances out a bit. It is also interesting to note that by using Depth of Field with Dual Cards the minimum frame rate went above 30fps compared to DoF off, This is a very interesting turn of events mostly due to the fact it is 100% reproducible time and time again.

As for DMC4 Scene 3 is a boss fight and theres not alot going on besides the player and a boss so the lack of render targets seems to make the performance gain minimal as for scene 4 its nothing but a round disc in space with a set # of enemies also not really indicative of a performance change but as wilth all games we have to keep these different scenarios documented for a true idea of what to expect.

Well that concludes the stock clocked benchmarks upnext overclocked results

 

[crazyeyesreaper]

review is done i suppose i could take a few requests to spice things up but otherwise this is is ladies and gents a full amd platform review. I hope you enjoyed seeing what an AMD platform is capable of and truly appreciate that fact that most of you are on Intel i7

 

[Super XP]

Thanks a lot, now I am salivating for a HD 6970 , luckily I’ve had one on order for a few days now and should arrive today.

Nice work man...

 

[crazyeyesreaper]

thanks id stick to 1 card unless you play Metro2033 or the stalker games extremely often tho as xfire scaling on AMD is only around 30-60% at best. The setup is just entirely cpu limited even when overclocked

 

[wahdangun]

great review but shame you can't do eyefinity with it,

btw what is the temp for both card ? and is there any noticeable micro shuttering problem ?

 

[crazyeyesreaper]

no micro stuttering evident or any hints there off and temps do get toasty around 85-90c per card when there overclocked stock clocks they tend to hover around 85-88c max

 

[Lionheart]

no micro stuttering evident or any hints there off and temps do get toasty around 85-90c per card when there overclocked stock clocks they tend to hover around 85-88c max

Great review did you notice any micro stuttering in Metro 2033? As from my experience I got alot of micro stuttering in that game with my previous HD5870 Xfire setup and HD5970 too

 

[crazyeyesreaper]

no micro stuttering man thing was smooth as can be i can honestly tell you with every bell and whistle turned on the game ran smooth and was just a Joy to play my 5850s were never up to the task i can assert that 6970s in Dx11 no DoF gave me a 60-70% performance improvement from 5850 xfire to 6970 xfire and the smooth gameplay reflects that. The only gripe i have is I have to mod my case side panel for some fans to help force fresh air between the gpus so i can have piece of mind. Otherwise im extremely pleased with the 6900 series the experience here at launch with them is better then my past trials with the 4870x2 and my 5850 xfire setup.

And it seems on an AMD rig apples to apples with a gtx 580 the 6970 seems to be 100% equal to a 580 on i7 the 580 pulls ahead rather interesting tid bit ive found tonight

 

[Lionheart]

no micro stuttering man thing was smooth as can be i can honestly tell you with every bell and whistle turned on the game ran smooth and was just a Joy to play my 5850s were never up to the task i can assert that 6970s in Dx11 no DoF gave me a 60-70% performance improvement from 5850 xfire to 6970 xfire and the smooth gameplay reflects that. The only gripe i have is I have to mod my case side panel for some fans to help force fresh air between the gpus so i can have piece of mind. Otherwise im extremely pleased with the 6900 series the experience here at launch with them is better then my past trials with the 4870x2 and my 5850 xfire setup.

And it seems on an AMD rig apples to apples with a gtx 580 the 6970 seems to be 100% equal to a 580 on i7 the 580 pulls ahead rather interesting tid bit ive found tonight

Good to hear bro, Im gonna wait and see if MSI bring out any of their twin Frozr II coolers for the HD6970 and I will most likely pick up two of em

 

[crazyeyesreaper]

well its official in heaven Bench and Alien vs predator same systems the 580 is slower then a 6970 stock gpu vs stock gpu 1on1 with a 965be rather interesting since on i7 the 580 tends to be a solid 10% or more faster

 

[Super XP]

Very well writen review,
Interesting, hooked up my new Sapphire Radeon HD 6970, big and heavy, dam didn't know it was sooo big and super long. Anyway for some reason STEAM does not recognize the card yet, so the performance in Left 4 Dead 2 is sort of off. If I leave eveything at default no problem, but once I touch something it seems to stutter a little. That STEAM screen also keeps popping up.

Any thoughts?

 

[crazyeyesreaper]

install cat 10.12a and steam will ask if they can monitor your hardware or some such make sure steam is off during the driver install and once u start steam it should ask about your hardware other then that i cant tell ya whats up as L4D2 worked just fine for me besides the whole cpu limited thing lol

 

[Thatguy]

well its official in heaven Bench and Alien vs predator same systems the 580 is slower then a 6970 stock gpu vs stock gpu 1on1 with a 965be rather interesting since on i7 the 580 tends to be a solid 10% or more faster

I can't say I am surprised by this. Not that I think the I7 clock for clock really beat the AMD phenom II in short pipeline workloads, becuase it actually doesn't. My thoughts are instructions and code optimizations as far as where the driver complier offload data. Does anyone know here what in house code compiler that nvidia uses ?

I'd be curious to see the cpu load between the 2 cards on both systems and see what they look like.

 

[crazyeyesreaper]

atis driver sucks bollocks in terms of cpu usage and overclocking dosent help examples i5 750 @ stock will give you better xfire scaling then any amd CPU overclocked to the limits but SLI scaling is the same on both Phenom II / i7 for the most part basically in this case Nvidias driver is more agnostic to each system so the fact the 6970 pulls ahead is a bit of a shock but dual card wise 580s and 570s would walk all over 6900 series xfire on an amd rig

 

[Thatguy]

atis driver sucks bollocks in terms of cpu usage and overclocking dosent help examples i5 750 @ stock will give you better xfire scaling then any amd CPU overclocked to the limits but SLI scaling is the same on both Phenom II / i7 for the most part basically in this case Nvidias driver is more agnostic to each system so the fact the 6970 pulls ahead is a bit of a shock but dual card wise 580s and 570s would walk all over 6900 series xfire on an amd rig

Actually your seeing just the opposite. The nvidia driver is not "system agnostic" as witnessed with your I7 testing.

GPU compiler loads are all very short pipeline work, the fact that a I7 scores better smells bad.

for a number of reasons, a long pipeline takes more time to run a cycle. it also get into resource contention more easily.

a shorter pipeline is better for graphics compiler loads.

all of this is in thoery not practice BTW, theres a million factos there.

I wish I had acess to some of these cards so I could watch running code in the kernel and see if anything pops up that might be of interest.


ehhhhem.

http://www.theinquirer.net/inquirer/news/1567108/intel-compiler-cripples-code-amd-via-chips


theres actually a few good site showing the actuall problems BTW if you look around and the performance differences.