R580 and Crossfire R580 Early 2006

I have a feeling they will want more money for the r580 initially, so that may stem consumer ire.

 

Well as long as they are able to use those cards to their full potential now when the R580 isn't available I see no reason to regret that purchase.

 

I don't think X1800XT owners will be to upset considering they can upgrade their cards to a XTPE with just a bios update. It's not a R580 but a nice free upgrade none the less.

 

Well 7800GTX 512 owners shouldnt be upset either since the PE seems to be on par with it.

 

[size=-3]Except for that 50% price discrepancy, eh?[/size]

 

Hey, I just came home from there! But all I got was a USB drive enclosure.

 

[size=-3]In some cases it's nearly 100% price discrepancy, with available 512Mb 7800 GTX going for $1K and X1800XTs going for $500.[/size]

 

the RV530 / Radeon X1600 is really a 4 pixel pipeline GPU / card but it has
12 pixel shader units (3 per pixel pipeline) it has 4 texture units (1 per pixel pipeline) feeding into 4 ROPs.

oh and it has 5 vertex shaders which seems really beefy concidering its only a 4 pixel-pipeline GPU. I am hoping that Nintendo-ATI 'Hollywood' for Revolution is based on at least the RV530 with Flipper functions for compatibility. A few years ago I was hoping GCN2 would be based on R500 or R600 but these dreams have since been shattered, and now I'm hoping for a modestly powerful RV530........

 

currently I can manage to get my XL to render a 70,000 triangle model at around 1700 frames per second (a really small window , so I seriously doubt that the vertex shading capabilities of the x1k family are going to be a bottleneck, especially as the r580 will be ~35% faster again clock rate wise. Also the only performance hit you get from 'moving' verticies around with the likes of physics is usually from an increase in batch numbers because of a larger number of small objects, but... A) you can use instancing in this case, and B) from what I remember x1k is *extremly* good at handling large batch numbers (I was recently testing 10,000 batches at ~100 fps but that was most definitly cpu limited).

Short answer, I may well be forced to sell my XL ;P

 

So what do you gain by adding ROPs? I mean the number of pixels on a screen is fixed to the resoluton, so I don't see why you'd need to output more pixels.

 

More ROPs gives one the opportunity to read/write/blend more values in a buffer per clock cycle - thereby reducing the amount of time it takes to completely read/write/blend a frame.

 

There are people still buying 512's today at that price discrepancy. Why would they all of a sudden regret it next year when they don't mind now?

But right now $250 difference looks like pocket change compared to the kind of cash being spent this holiday season It's all relative.

 

Thanks!!!!!

Another question?
If Nvidia will have more pipelines and ATI will have more Pixel Shader units. Who will have more advantage ATI or Nvidia????? like maybe OpenGL. Is More Pixel Shader units only good for Direct3D.

Because the whole story from very beginnen being told ATI R580 will have 24-32 pipelines and now it will end up with only 16 pipes.

 

Higher resolution, higher FPS.

 

It's good if you have more shaders, be it via D3D or OGL.

You were listening to the wrong story (from sites like VR Zone and The Inq), but expecting doubled pipes was understandable. Initial rumors were vague, but talk of "extreme" pipes on R520 hinted at it maybe being 32 pipes, as R420 was rumored to be 8 "extreme" pipes but turned out to have 16. But the first "solid" rumors showed R520 to have 16 pipes and R580 to have 16 "full" pipes (shader, texture unit, ROP) but basically 48 shader pipes. While it made sense to think R520 may have 24 or 32 pipes, like G70, once RV530 was announced simultaneously with R520, the 24-32 pipe rumors no longer made sense.

 

If the story is true that R580 is 16 pipelines having 48 Pixel Shader units should it double the score what R520 gets.

My wonder question is R580 is going to be same as the time Radeon 9700PRO gave Nvidia a trouble and they did not had an answer to R300.

 

To Whom It May Concern,

Please notice that it's clearly said "30" and not "90". It's said 'Thirty' and not 'Ninety'. In fact it settles pretty well with the circulating Jan 24th date for the X1900 series.

Itai @ HWzone.co.il

 

As godfather of the "30" school of philosophy, I would of course like to agree with you.

However, there is the little problem in your analysis that Dec 21 + 30 will not get you to January 24th. So, then, Dave Orton would have to have been speaking allegorically/poetically, using the more euphonic "30 days" as a stand-in for the much more prosaic "next month". Fans of the beauty and grace of R300 have often observed that Orton must have the sensibilities of a poet (observers of the X800 line have been known to put forward alternative theories, however), but I rather doubt we'd want to go there using this particular snippet of a quarterly financial call.

 

Exactly. IMO, Dave meant "30" as in "next month". The other option is that he doesn't know math that well

Anyway, I'll say again that the "30 days" as in "next month" suits well with the the circulating Jan 24th date for the X1900 series (a date that we also heard).

 

The Inq's take - http://www.theinquirer.net/?article=28540. They seem to think R520 will be killed off.