Could the Doom3 demo have used RV250?

[Oompa Loompa]

Just curious. We all know that "next generation" is a marketing term which should logically refer to R300, but in PR-speak could easily mean RV250.

Since RV250 will supposedly be launched at Computex in a couple of weeks, whereas R300 is a Fall product, I think this actually makes some sense. It's just a theory, mind you.

 

[Dave Baumann]

I wouldn't really deem that 'next generation' though. Last year when Radeon 8500 and 7500 were announced and released would you have deemed 7500 next generation?

 

[Oompa Loompa]

I wouldn't really deem that 'next generation' though. Last year when Radeon 8500 and 7500 were announced and released would you have deemed 7500 next generation?

Certainly not. As I said, it would require some "spin" in order to use that definition. But remember that the 7500 & 8500 were launched simultaneously, so there was no issue. It seems clear that the RV250 will precede the R300, however, so it is a possibility this time around.

 

[KimB]

Interesting...possible...but I had thought that the Radeon 8500LE counted as the R250...and perhaps this RV250 is just the AIW version?

 

[Doomtrooper]

Interesting...possible...but I had thought that the Radeon 8500LE counted as the R250...and perhaps this RV250 is just the AIW version?

The Rv250 is NOT a 8500 LE, its a value product based on a the 8500 that is clocked higher and is supposed to be released in June.

 

[Hellbinder]

acctually doom, Its going to get announced in July right? and released sometime after?

 

[Doomtrooper]

I heard rumors of boards to review in June, I may be wrong..I have been wrong before..many times

 

[KimB]

The Rv250 is NOT a 8500 LE, its a value product based on a the 8500 that is clocked higher and is supposed to be released in June.

Why would a tech demo like DOOM3 use a value product?

And besides, JC still says that he prefers the GeForce4 over the Radeon 8500, so I don't know why he'd ever use a derivative of the 8500 for a tech demo.

 

[Pete]

JC said he'd use the fastest product, and he wouldn't play favorites. The RV250 may be a R200 derivative, but it seems it's a faster derivative, not a slower, feature-stripped one.

I wonder if JC had a Parhelia to play with, and if he did, I hope the ATi card was an R300 (otherwise it bodes ill for Matrox). Maybe that should be Chalnath's next e-mail.

 

[Ichneumon]

Heh... This I firmly state without reservations...

Anyone who thinks that Parahelia will long-term hold a candle performance-wise vs. the R300 and NV30 is deluding themsevles.

As matrox themselves stated... its still a 4 pipe part.

 

[Doomtrooper]

The Rv250 is NOT a 8500 LE, its a value product based on a the 8500 that is clocked higher and is supposed to be released in June.

Why would a tech demo like DOOM3 use a value product?

And besides, JC still says that he prefers the GeForce4 over the Radeon 8500, so I don't know why he'd ever use a derivative of the 8500 for a tech demo.

I have VERY good info. it was not a Rv 250

 

[KimB]

JC said he'd use the fastest product, and he wouldn't play favorites. The RV250 may be a R200 derivative, but it seems it's a faster derivative, not a slower, feature-stripped one.

Next generation, to me, is pretty obvious. It must be the R300. I really don't think this needs to be clarified further. After all, just go back and read his reply again: "It shouldn't be surprising that a next-generation card is faster than a current generation card."

A derivative of a current-generation card doesn't necessarily follow this rule. Next generation cards do. I think JC has done everything but jump out and state that it was the R300 (and since ATI didn't say this, JC may not be able to under NDA). I really don't see the need for further clarification.

I wonder if JC had a Parhelia to play with, and if he did, I hope the ATi card was an R300 (otherwise it bodes ill for Matrox). Maybe that should be Chalnath's next e-mail.

Again, I don't see why JC needs to be e-mailed about this. It seems really obvious to me that the Parhelia will be crushed in 3D by the upcoming competition. If you'd like further reasons on this subject, I'd be happy to give 'em.

 

[demalion]

Interesting...possible...but I had thought that the Radeon 8500LE counted as the R250...and perhaps this RV250 is just the AIW version?

The RV250 is, on paper (based on a leak that fairly accurately described the RV200 and R200), "Value" like the RV200 (7500) is value compared to the R100 (7200): higher yields due to similar design and transistor count of lower "number" EDIT: on more compact design process/EDIT, yet clocked higher and with some tweaked extra features thrown in. The "AIW" part is the other chips on the board, not the graphics core.

I say this to clarify your current misconception of it, not because I think it was anything but an R300 at E3 (though my baseline of confidence in assumptions about how reasonable his comments are has been shaken a bit). This is especially true since a reasonable assumption is that the (rumored?) NV28 is the "tweaked GF4" nVidia tried to get up and running there (and the RV250 is by absolutely no stretch of the imagination more "next generation" than the NV28, or even an NV25 if you consider the GF3 and 8500 the same generation, as Carmack has indicated he does).

 

[demalion]

Heh... This I firmly state without reservations...

Anyone who thinks that Parahelia will long-term hold a candle performance-wise vs. the R300 and NV30 is deluding themsevles.

As matrox themselves stated... its still a 4 pipe part.

Heh, I don't know, depends on what you mean by "Hold a candle". I'd say the 8500 "Holds a candle" to the Ti4600...and with all that bandwidth, are you really sure that the Parahelia will fare any worse compared to the R300? Then again, some of the hints you are dropping...you know a bit about the actual R300....but you don't know how the Parahelia will look in the real world (even though the core will be around 220, there is still room for the memory clock to be pretty high).

If you respond with confidence in light of this, then...you must know something juicy...what will it take to bribe you? <-note the tasteful jiggling lol.

 

[RussSchultz]

If the RV250 is a die shrink (.18->.13)of the R200, it definately could be called "next generation". It would be theoretically 50%-100% faster than its larger sibling and use 50-100% less power at the same speed even though it has the exact same netlist and feature set.

It also costs about half as much to mass produce (excluding NRE).

I think its completely conceivable that the RV250 could beat a GF4-Ti (by a decent margin, even) for this demo.

 

[Dave Baumann]

If the RV250 is a die shrink (.18->.13)of the R200

Not sure if thats a typo or not Russ, but R200 is .15um.

 

[RussSchultz]

It was supposition. If its .18 to .15, then its only a 25-50% increase/reduction instead of a 50-100%.

 

[Althornin]

It was supposition. If its .18 to .15, then its only a 25-50% increase/reduction instead of a 50-100%.

No no no.
It would be .15 to .13 as r200 is already .15
So if rv250 is a die shrink of r200, it would have to be to .13

 

[RussSchultz]

Aaah, duh.

 

[Ichneumon]

Heh... This I firmly state without reservations...

Anyone who thinks that Parahelia will long-term hold a candle performance-wise vs. the R300 and NV30 is deluding themsevles.

As matrox themselves stated... its still a 4 pipe part.

Heh, I don't know, depends on what you mean by "Hold a candle". I'd say the 8500 "Holds a candle" to the Ti4600...and with all that bandwidth, are you really sure that the Parahelia will fare any worse compared to the R300? Then again, some of the hints you are dropping...you know a bit about the actual R300....but you don't know how the Parahelia will look in the real world (even though the core will be around 220, there is still room for the memory clock to be pretty high).

If you respond with confidence in light of this, then...you must know something juicy...what will it take to bribe you? <-note the tasteful jiggling lol.

Oh, IMO the parahelia is definately going to be a far better card than the 4600 or 8500 for numerous reasons... My statement was more in response to some peoples strange idea that it will be able to consistantly compete (on all levels) against the R300 and NV30 (though its sounding more and more like the NV30 is farther out than a lot of us expected, even knowing the NV30 was delayed some).

There's still a lot of questions about what trade-offs ATI might make in the pro vs. consumer r300 boards for various reasons that you wouldn't historically associate as issues with an ATI designed part, so everything I (may) know/have heard is still potentially up in the air. Matrox does have a lot of bandwidth designed into the Parahelia... however, consider everything you've already heard about the R300, and add to that the fact that Matrox isn't the only company that can put together fast ram and fat memory interfaces on their cards... and ATI has 3rd gen HyperZ now as well. It will be exceedingly fast.