New news on GTX 480 & 470

logitic

Distinguished
Oct 27, 2009
195
0
18,690
http://www.semiaccurate.com/2010/02/20/semiaccurate-gets-some-gtx480-scores/

According to this, things are really not looking good at all.


I don't know how much "Laughabee" hurt the GF100's, but I hope we hear better news soon.
 

logitic

Distinguished
Oct 27, 2009
195
0
18,690
Doesn't sound like it's going to happen any time soon. When the BoM for a 480 cost's more then the 5970 retail. And for something to cost more then a product that it can't beat you know heads are going to roll at NVidia.
 

sabot00

Distinguished
May 4, 2008
2,387
0
19,860


The 4-Series came out in 2008, and went into 2009.
A generation of graphics cards is supposed to last 2 years.
 

AMW1011

Distinguished
From what I see here, we see almost nothing. The performance numbers that they quote are far less than credible, and makes little sense given the supposed specs. 512 SPs should atleast come close to double the performance of a GTX 285, even with low clocks. I'm guessing that the GTX 480 will be about 15% faster than the 5870 and about 10-15% slower than the 5970. I also expect the GTX 470 to match the 5870 pretty closely.

But again, my opinion is based on very little fact, just like the article.

That said, Charlie has been pretty accurate.
 

notty22

Distinguished
Another article where he is going over the same points again. And when he hears things he doesn't like, such as points he can't negatively spin anymore, he *SIGHS* and winks , winks that somebody is lying.
This speed is the first point the sources conflict on, and it could go either way, both sources were adamant on theirs being the correct final clock. *SIGH*.

Does the same, with pricing/performance, he's doing the worrying for Nvidias accountants, that they are going to somehow have to cheat, to sell a product they engineered. And if there are good reviews, they are 'bought'.
Then he's deflecting the Tesselation performance, its one of the core DX11 additions. Its what probably another area Fermi will perform better in DX11/dx10 games vs ATI.
No matter how Fermi turns out, he's got all bases covered, he's going to tell you he told you so.
 

sabot00

Distinguished
May 4, 2008
2,387
0
19,860


Most likely a refresh of the current gen, they won't waste so much R&D for only a year, also they wouldn't have a new process.
GlobalFlo's 28nm process will be ready by the end of 2010 likely but it'll be worse than TSMC's 40nm problems, without a new process they could do an rearrange like
RV670 -> RV770. However RV670 (3870) was RV600 (2900XT) shrunk to 55nm, so it was designed for 80nm. RV770 (4870) took the RV670 & changed it to suit a 55nm process, but Evergreen (RV870) was designed for 40nm, so it's already well very made.
Without a new process they could slightly amp it (5870 -> 5890, 850MHz core to 900-950MHz, 1250MHz RAM to 1300-1450MHz, maybe 1600 SP's to 2000.) but I don't think ATI will call that the 6000 series.
 

AMW1011

Distinguished


I agree here, he seemed too vague and full of excuses. That said, I do believe he is accurate, at least, pertaining to the Unigine Heaven benchmark, as I have read in other sources that the benchmark is poor and wont be a good indicator. nVidia should have a slight to moderate advantage in tessellation, but nothing that ground breaking. His cited performance doesn't make sense given the specs. he stated. The clocks are only slightly lower than the GTX 285, and there are more than twice the SPs so the GTX 480 should be ~twice the performance of the GTX 285 which is right between the 5870 and 5970.



Well that is more or less what I meant, right inbetween the two.
 

logitic

Distinguished
Oct 27, 2009
195
0
18,690
Maybe Johnnie Walker is clouding my judgment, but the more I read about the Heaven Benchmark score for the GTX480 the more I think the 480 may be a beast.

Look at it like this...

Tessellation is basically a quick rendering of group #'s right. And as of now we have only seen games that use Bitmap, clone and other outdated mapping. But were tessellation shines is collision. a k a light source, light shading, group 3D aspect.

So I guess what I am saying is that either ATI is setting a trap and has on purposely governed their own software to embarrass Nvidia on their PR campaign*. Or Nvidia is about to kick some major ass.


*(Clocked down Tessellation to fool Nvidia)
 

sayantan

Distinguished
Dec 9, 2009
692
0
19,060
The article seems to be biased!!!With so much negativity nVidia can never hope to launch them in q2 2010...But according to nVidia CEO Fermi will hit the full stride in q2 2010...which certainly seems contradictory
 

sabot00

Distinguished
May 4, 2008
2,387
0
19,860


What are you saying?
 

Here's a lovely blog that is to Charlie D what Charlie D is to Nvidia.
 
its not out yet and no credible benchmarks so if we just go looking into history we will prolly see the same thing again - Nvidia = a bit quicker, more expensive. ATi = a bit slower, better bang for the buck.

Thermals, and how much more expensive are another thing all together - might be another dust buster FX5800 ultra...
 

noob2222

Distinguished
Nov 19, 2007
2,722
0
20,860

Nvidia is said to be producing only 5,000 to 8,000 of these cards, and it will lose money on each one.
thats full stride? good luck getting one.

The HD 6k is already slated to be a new GPU, most likely ATI's GPGPU. If ATI refreshes the 5000, it will be for the 5890, possibly testing it on the 28nm process, like they did with the 4770. It will be interesting to see who they use for manufacturing, GF is slated to start 28nm in June, while TSMC in q3.