Eyefinity Comparison 6950 CF -> 7970 CF

[MeMeMeMe]

I had an MSI Twin Frozr III 6950 Crossfire setup, and I've just swapped out my graphic cards for a HIS HD 7970 IceQ X² Turbo Crossfire setup.

I ran some benchmarks before the change and repeated them after. My system is not overclocked at all, and probably is a bit overburdened with unnecessary services, etc. It's the increases, rather than the absolute numbers that are most relevant, I think.

In all games, unless otherwise stated, Resolution 5760x1200, all settings maximum except AA which will be noted.

Aliens vs Predator 3
AAx4
6950CF: 37.8
7970CF: 61.5
Increase:+63%

Just Cause 2
AAx4 /AAx8
Dark Tower
6950CF: 46.2 / 39.6
7970CF: 91.1 / 71.1
Increase:+97%/+80%

Desert Sunrise
6950CF: 60.1 / 51.6
7970CF: 118 / 93.7
Increase:+96%/+82%

Concrete Jungle
6950CF: 34.1 / 30.3
7970CF: 63.7 / 57.8
Increase:+87%/+91%

Metro 2033
DX11, H, AAA (no DOF)
6950CF: 34
7970CF: 54.5
Increase:+60%

DX11, VH, AAA (no DOF)
6950CF: 24.5
7970CF: 40
Increase:+63%

DX11, VH, 4xAA (no DOF)
6950CF: 17.5 (actually, much worse than this, it only finished the run about half the time I tried)
7970CF: 39
Increase:+123% (and more, because it actually finished every time)

Dirt 3
All max, AAx4
Minimums/Averages
6950CF: 46.9 / 39.7
7970CF: 94.4 / 78.9
Increase:+101%/+99%

Unigine Heaven 3
same settings as in the heaven benchmark thread: 1920x1080, 4xAA, tesselation normal, etc.
Showing Score (FPS Minimum, Average, Maximum)
6950CF: 1943 (32.9, 77.1, 151.4)
7970CF: 3381 (35.0, 134, 253)
Increase: +74%

The 7970s love Just Cause 2 and Dirt 3!

 

[Tonester0011]

Were these 6950>70s? Or just bog standard non flashed 6950s?
If the former, bloody decent upgrade. If the latter, still huge jumps.

 

[MeMeMeMe]

Oh yes, temps and noise.

The 6950 Twin Frozrs were not quiet at first, but I applied a custom fan profile in afterburner, which made them incredibly quiet. Sometimes I'd come in, forget my PC was on, click the power button and realise I'd just forced a reboot.
Even in game, they got noisier but it was plenty bearable.

The ICEQ's aren't as quiet, but are acceptable. I did fiddle with a custom fan profile, but found I couldn't make them any quieter than the default setting, which is still much quieter than previous cards I've used - just a tiny bit noisier than my Twin Frozrs.

The temps though are in a whole different league. With my fan profile, the top twin frozr's idled around 50-60, and ramped up to the high 70s under load. The bottom one idled around 40, and went up to the same high 70s under load.

The IceQ 7970s idle under 40, both of them. And under load, I haven't seem them go over 60 very often, only under a very heavy benchmarking session, and haven['t gone as high as 70. I'm seriously impressed.

I wonder what they can do overclocked.

 

[MeMeMeMe]

Were these 6950>70s? Or just bog standard non flashed 6950s?
If the former, bloody decent upgrade. If the latter, still huge jumps.

Bog standard 6950s, but the twin frozrs are really good 6950s.

 

[cjph]

Nice comparison - are you using the latest AMD drivers, 12.11 beta ?

And did/do you ever experience micro stutters ?

cj

 

[MeMeMeMe]

I'm using the 12.10 drivers, which I had on both sets.

With the 6950s, I *may* have experienced microstutter when playing Crysis 2, I'm not sure. It may just have been the game was too demanding at the settings I was running. I didn't notice it in other games. I usually played game a notch below maximum settings, at high rather than very high, and with AA no higher than x2, to ease the workload.

I haven't noticed any microstutter with the 7970s, but I haven't used them in actual gaming much yet, I've mainly been benchmarking

I'm looking forward to trying out the 12.11's, but I'll wait till they are out of beta. They seem most useful to BF3 players, and I dont play that.

One thing I meant to mention: In metro 2033, there is zero performance difference in going from AAA to AAx4. The AAx4 setting brought the 6950s to their knees, but the 7970s don't even notice it. I ran the benchmarks several times to test this wasn't a glitch, but it was completely repeatable.

 

[MeMeMeMe]

If the former, bloody decent upgrade. If the latter, still huge jumps.

Something that just occurred to me.

I am very pleased with my upgrade, HOWEVER:

My old cards were bought for £400 in May or June 2011. These cards cost £65-ish, 18 months later, and only provide roughly 70-80% average increase. That's really not very impressive, in terms of advancing tech.

Still, it's nice to be able to crank settings up to max on eyefinity

 

[Lokken86]

I'm using the 12.10 drivers, which I had on both sets.

With the 6950s, I *may* have experienced microstutter when playing Crysis 2, I'm not sure. It may just have been the game was too demanding at the settings I was running. I didn't notice it in other games. I usually played game a notch below maximum settings, at high rather than very high, and with AA no higher than x2, to ease the workload.

I haven't noticed any microstutter with the 7970s, but I haven't used them in actual gaming much yet, I've mainly been benchmarking

I'm looking forward to trying out the 12.11's, but I'll wait till they are out of beta. They seem most useful to BF3 players, and I dont play that.

One thing I meant to mention: In metro 2033, there is zero performance difference in going from AAA to AAx4. The AAx4 setting brought the 6950s to their knees, but the 7970s don't even notice it. I ran the benchmarks several times to test this wasn't a glitch, but it was completely repeatable.

Feel free to add at least another 10% to most of those gains

 

[Tonester0011]

Something that just occurred to me.

I am very pleased with my upgrade, HOWEVER:

My old cards were bought for £400 in May or June 2011. These cards cost £65-ish, 18 months later, and only provide roughly 70-80% average increase. That's really not very impressive, in terms of advancing tech.

Still, it's nice to be able to crank settings up to max on eyefinity

The price is moot, however a 70% gain in one generation is good.
It's a damn sight more than the original quotation (some people on this forum) of 15% upgrade only. If you take overclocking potential into it as well, even better.

That's my opinion obviously =)

 

[MeMeMeMe]

Yeah, I'm quote happy with the performance increase, and had a nice windfall so the cost wasn't an issue. I just wondered if this was normal - I had the feeling that with most gens you saw a 70-80% increase for roughly the same price. Maybe gains are slowing, or my assumption was wrong.

I am looking forward to seeing what more i can get with overclocking, and new drivers.