AMD RV770 refresh -> RV790

Yep. It seems close to a 850/1000-1100 HD4870.

 

Maybe so, but if I do a search on google, almost all tested HD4870 systems gets around 8500+ marks. I'm confused...

 

We get a GPU score of 7715 points:
- HD 4870 1G
- Core 2 Quad Q9450 @ 3.6 GHz
- Catalyst 9.1

So the 9801 points for HD 4890 are a really good result. 27 percent higher score!

 

9801?

 

The biggest (and probably the best) benchmark scores database in the world.

3DMark Vantage

http://www.hwbot.org/listResults.do...r=true&filterBlocked=true&sli=false&limit=100

3DMark06

http://www.hwbot.org/listResults.do...r=true&filterBlocked=true&sli=false&limit=100

 

Yes, I mean the GPU score. You can better compare the GPU score than the overall score to see differences between graphic cards.

 

10436 for a stock GTX285:

http://www.hardware-infos.com/tests.php?test=56&seite=4

Jawed

 

Of course... Being a bit slow today.

 

So basically... if the leaked benchmark is valid you get the following GPU scores:

HD4870: 7715
HD4890: 9801
GTX285: 10436

And as KonKort already pointed out, HD4890 scores 27% higher than HD4870. It cannot be explained by a 13% core speed increase... There's more to this puppy than clocks!

 

Yes, AMD had many months to optimize the architecture. They have done a good job, if you find in more applications this speed bump. But I am skceptic.
I think in average the chip will not run faster than 20 percent, but this is - of course - more than a clock increase.

 

Isn't it wonderful how history keeps repeating itself?

 

And GTX280: 9367.

Anyway, we have to wait for real benchmarks and real reviews to better understand the entity of performance improvements clock to clock vs RV770, if any. Synthetic benches have proven to be very misleading in the past...

 

Sorry, you could not compare the results. In our reviews we bench in 1280x1024, 1xAA/1xAF, too, but we set all other options to Extreme.

I have a new score for you. With the same setting of OC Heaven HD 4870 1G scores 8696 points. So it is an increase of 13%. We use a Core 2 Quad (3 GHz).

 

Really should know better, never to spend any time with Futuremark numbers.

Jawed

 

Damn, it's just a clock increase.

 

So true.

 

they still hold some weight. Graphics cards that win in 3dmark will also win in some games. I think in general people are too quick to dismiss 3dmark.

 

Full rate FP16 filtering again?

 

Even if, I don't think it would make a difference in 3DMark 06, since it doesn't use very much FP16-filtering, being able to run very decently on the X1K-Series, which do not support this feature.

 

Tell that to the OEMs. Many love 3DMark(s)...and pleasing OEMs is important.