So is R580 supposed to be 48 pipes or 48 ALUs?

Heres alittle more info on the R580. This is a X1900XT with a core or 500 and mem at 600.
http://www.ocworkbench.com/ocwbcgi/newspro/viewnews.cgi?newsid1137593522,11212,,

Link fixed

 

Link not working for me. But that's hella impressive for 500/1200. Also, some tidbits from the manual are up - http://www.dailytech.com/article.aspx?newsid=344

 

The link is fixed now. The motherboard is a ECS KA1 MVP RD480.

 

I'm glad to see that the number of threads hasn't been reduced - it's still 128 threads per shader unit. With 4 shader units and 48 fragments per thread, that's 24576 fragments in flight

That's one hell of a lot of memory for all them registers.

All we need now is a die shot!

Jawed

 

So why does the ATi PDF say 512 threads total? Lol, nevermind 128x4 = 512

 

He means 128 threads per "processing core" (for want of a better term, seeing as quad doesn't work) - there are 4 cores working on different threads, so in total there 512 threads.

 

The AA results for X1900 look quite nice. Comparable to 7800GTX SLI in FEAR.
Interesting that the AA results show dominance, but not with no-AA. Is this due to the memory controller or scheduler, rather than fragment pipes?

 

Just hope my faith is restored.

 

Doesn't it make sense that a better indication of the R580s performance will be with future pixel shader intensive titles such as those based on UE3.

 

Ii cant find the other posts covering the dailytech artcle.

 

Starting here: http://www.beyond3d.com/forum/showpost.php?p=677984&postcount=746

 

looks like it wasn't worth waiting for right now if you already have a GTX or x1800XT

 

This is true can always test it out too when the r580 comes out since that Intel demo is out there

 

Yeah, seems very odd...
as it follows from their numbers 1900xt gains only 7,5% from gtx256 score at that settings.
Though looking over xbitlabs and FS reviews, 1800xt beat gtx256 by a margins of 24% and 38% respectively
( the scores are: 69 - 86 and 58 - 80 resp.).

 

This is more about what other teams are willing to do that what the chargers are willing to do.

It will be interesting to see how this thing turns out. If the numbers hold up to what we have seen so far, and the G71 is a 32/64 monster card then its going to be major overkill compared to the 1900.

If its more like 24/48 then it could be more competitive but i would still give the lead to the G71 in most cases.

This will be fun to see how this turns out.

 

hahahaha that post makes no sense.. and i cant edit it. I accidently grabbed the last line of the spell check i did for a chargers.com post.

 

Guess we'll wait a bit longer to see what those percentages look like in real world numbers.

http://www.hardwareoc.hu/index.php/p/news/cid/1/y/9617.html

 

I really hate when they truncate graphs. But its nice to see the x1900s with CF are playable at max resolution in everything.

 

I was looking at those scores again. And I'm finding some of the information their suspect. In my experience SLI rarely ever aproaches 80% performance scaling. But those graphs sure would lead you to believe that SLI is giving an 80%-90% bonus over a single graphic card.

Or am I reading that all wrong? I certainly have never experienced anything like that.

 

I find those numbers highly suspect.

For starters some of those have Corssfire giving 100% or greater performance. No way. Next the opengl numbers dont seem to reflect ATi's recent driver improvements.