Unigine Heaven 2.0 is out

[spinejam]

spinejam: the top benchmark you posted shows you running in dx11 w/ tessalation off. From your signature, I thought you were running a 4870x2. I'm confused as to how you were able to run it with that setting.

that was the default DX option and i set AA & AF myself.

(tessellation disabled)

 

[HurleyBird]

well, heaven 2.0 has a massive boost in performance on the dragon scene, but overall has more content and doesen't really get higher avg. FPS (although, to me it does feel a lot more smooth than 1.0 overall)

Nvidia, on the other hand, used version 1.1 in their benchmarks. It seemed like the content was pretty much the same as 1.0, but performance was probably improved over 1.0

The only question is, how much of an improvement? If the FPS improvement in the dragon scene with 1.1 is similair to the improvement in 2.0, tesselation performance is very similair between 5870 and Fermi and Nvidia has been been blowing a lot of hot air. I'm guessing Fermi will still have better tesselation performance on average (it better given that extra billion transistors), but I doubt it will be as huge as Nvidia has been hyping it up to be.

 

[Apocalypse23]

5870 stock Extreme tessellation 2xAA 16xAF @e8400-3.9ghz

5870 stock Normal tessellation 2xAA 16xAF @e8400-3.9ghz

 

[blanketyblank]

Not sure if it's just random chance or not, but I've run the test a few times now and when I overclock my card my minimum drops to 10.2 though I do get a faster average overall. 31.2 with minimum of 10.2 when I overclock just my memory so 850/1300. 31.5 when I overclock just my core so 900/1200 also with 10.2 minimum.
No overclocking at all gives me 30.9 with my minimum above 15.

Can someone else try this benchmark at stock and with an OC and see if their minimums change? I know it's just a single datapoint in the benchmark, but it seems odd to me why it seems to correspond to overclocking.

 

[FiLeZz]

Do I win anything for my scores?

 

[happy medium]

Not sure if it's just random chance or not, but I've run the test a few times now and when I overclock my card my minimum drops to 10.2 though I do get a faster average overall. 31.2 with minimum of 10.2 when I overclock just my memory so 850/1300. 31.5 when I overclock just my core so 900/1200 also with 10.2 minimum.
No overclocking at all gives me 30.9 with my minimum above 15.

Can someone else try this benchmark at stock and with an OC and see if their minimums change? I know it's just a single datapoint in the benchmark, but it seems odd to me why it seems to correspond to overclocking.

Just re-ran with a lower overclock

@ 870,1320 I got

avg 29.8
MIN 14.3
max 79.3

Compared to my 890 ,1325
30.2
6.5
81

My minimums more then doubled?

 

[Apocalypse23]

Do I win anything for my scores?

Maybe add some AA to those benches, perhaps also try extreme tessellation, so we can compare scores

 

[Dark4ng3l]

Just re-ran with a lower overclock

@ 870,1320 I got

avg 29.8
MIN 14.3
max 79.3

Compared to my 890 ,1325
30.2
6.5
81

My minimums more then doubled?

Yeah I get similar bizarre results with my 5770. I redid some tests with the original version and I don't see anything like that.

Also I see quite a bit of difference between runs. I can benchmark it 5 times and get significantly different results. It seems to vary about +- 7%(would need to run it another 10-15 times to see the real dispersion). I can't remember a benchmark being that inconsistent on my system.

 

[blanketyblank]

Yeah I get similar bizarre results with my 5770. I redid some tests with the original version and I don't see anything like that.

I agree there's something very strange about this test. It doesn't seem to be random chance though I'm not going to run the benchmark like 100 times to rule that out, but the minimum seems to be affected by even the most trivial OC.
I don't think I'm getting memory errors with just 50 Mhz OC, but I guess it is possible.

 

[happy medium]

Just re-ran with a lower overclock

@ 870,1320 I got

avg 29.8
MIN 14.3
max 79.3

Compared to my 890 ,1325
30.2
6.5
81

My minimums more then doubled?

Now with no overclock

Avg24.8
min 12.7
max 66.4


minimums are still double of my highest overclock?
Wierd!

It's not memory errors. My memory overclock was only 5 mhz different in my above test.

Edit: tried it with a 900 core overclock and no memory overclock...
29.1
14.3
76.6

This test sux! Memory makes no difference?

 

[FiLeZz]

Maybe add some AA to those benches, perhaps also try extreme tessellation, so we can compare scores

It would only let me run in my lcd's aspect so I did 1920x1200

It also only lets me do 8x aa maybe thats a crossfire x limitation but here are the results.

So this all settings to MAX at this res.

 

[Rubycon]

236MB? Are you kidding?
That's a few Led Zeppelin albums right there.

 

[FiLeZz]

236MB? Are you kidding?
That's a few Led Zeppelin albums right there.

there is a zepplein in this one too.. LOL

 

[soccerballtux]

236MB? Are you kidding?
That's a few Led Zeppelin albums right there.

what's your limit on bandwidth?

 

[Executor_]

Any performance comparisons for Heaven 1.0 vs 2.0?

 

[1h4x4s3x]

Executor, 1st post.

Some interesting insight on the changes from 1.0 to 2.0 and also some info on 1.1.

 

[mhouck]

that was the default DX option and i set AA & AF myself.

I didn't think that 48xx could run dx11. So can dx10 cards run dx11 as long as the dx11 specific features are turned off? i.e. metro 2033 can be run in dx11 just w/ tessalation and depth of field off w/ any dx10 card?

not saying that you would want to, just that it would be possible...?

 

[MrK6]

Now with no overclock

Avg24.8
min 12.7
max 66.4


minimums are still double of my highest overclock?
Wierd!

It's not memory errors. My memory overclock was only 5 mhz different in my above test.

Edit: tried it with a 900 core overclock and no memory overclock...
29.1
14.3
76.6

This test sux! Memory makes no difference?

Sounds like your previous overclock wasn't stable. Also, it's entirely possible you're GPU limited, not vRAM limited.

Looks interesting, I'll definitely give this a run when I get home.

 

[RussianSensation]

I didn't think that 48xx could run dx11. So can dx10 cards run dx11 as long as the dx11 specific features are turned off? i.e.

I just tried it on my card and it runs Unigine 2.0 with DX11 no tesselation. However, it likely actually defaults to DX10 once you are running the program. The benchmarks with 4AA/16AF 1080P are identical whether I select DX10 or 11. It's fair to say 48xx series cards can't do DX11 or ATI would have labelled them as such

 

[Rubycon]

what's your limit on bandwidth?

No transfer limits just speed can be slow. I did finish the download just now.

 

[SSChevy2001]

I just tried it on my card and it runs Unigine 2.0 with DX11 no tesselation. However, it likely actually defaults to DX10 once you are running the program. The benchmarks with 4AA/16AF 1080P are identical whether I select DX10 or 11. It's fair to say 48xx series cards can't do DX11 or ATI would have labelled them as such

Parts of DX11 API can run on DX9, DX10, and DX10.1 hardware. The 4800 series is able to using DirectCompute 4.1 and Multithreaded Rendering from the DX11 API, beside all of features from the DX10.1 API.

 

[Voo]

All the people that get lower min. rates with overclocks probably just overclocked their memory too high and the controller has to resend faulty data..
Just overclocking till you get a bluescreen won't work with a correct GDDR5 implementation that well

Though that on the fly error detection makes it kinda hard to find the correct limit - benchmarking after each step sounds especially tedious.

 

[Meelis S]

Hi

Here is mine