Speculations On R400

[Geeforcer]

Why wouldn't ATi clock it accordingly?

Diminishing returns? Requires faster, more expensive RAM (thus obfuscating the low end element)? Requires active cooling (again, pushing prices up)?

There could be all manner of reasons. Why does NVIDIA cripple the MX in its 420 specification? Could it not really do more than the 460 specification? Low end parts have different needs & targets most likely.

But does RV250 use passive cooling? Recent pictures lead me to believe otherwise. In addition, higher core speed would not necessarily translate in high ram speed. Wasn't RV200 clocked at 290MHz?

 

[Dave Baumann]

But does RV250 use passive cooling? Recent pictures lead me to believe otherwise. In addition, higher core speed would not necessarily translate in high ram speed. Wasn't RV200 clocked at 290MHz?

I think I've seen two diffrent images with passive and one with active so far, from the top of my head - but this will be down to the manufacturers. Many 3D processors in the low end don't actually require active cooling but board manufacturers will put them on anyway.

As for the memory, RV200 was a two piped card - if RV250 is a 4 piped card then I'd wager that RV200 would get more benefit from a core/ram clock scew like that then RV250 would.

 

[David G.]

So ... after seeing how RV200 was launched at highyer speeds than R200 we have a precedent ...

I wouldn't be surprised if there will be in fact a R200 on 0.13 micron @ 350-400 Mhz ....

After all ... ATi never let's the old production lines die ... They still make different versions of Rage 128 . I'm not even talking about Radeon or Radeon VE ( 7000 )

 

[RoOoBo]

Yeah, me too, I find it strange that a 55M P4 runs at 2,6GHz while these measly 40-60M GPU's run at 300MHz.

When comparing the complexity of CPUs and GPUs you might want to consider how much of each one's transistor count can be accounted for by its memory caches

In a pipelined design the clock is determined by how much time it requires a pipeline stage to complet its task in a given technology. That means that if the P4 has a greater clock than a GForce each stage does less work (in CPUs the main unit for a stage work is an integer addition).

There are also problems with power, the more frequency the more power it requites: P=V^2*F (more or less).

In any case I think the main reason is that the task a GPU does in a single stage is by far larger than what a CPU does.

 

[MuFu]

Well, from what I know, the RV250 actually yields at ~325MHz, so that is in fact higher.

Why wouldn't ATi clock it accordingly?

I was wondering that too. The architecture was designed for synchronous core/mem operation so perhaps they want to keep memory costs down and leave it at 300/300MHz. Sub-3.3ns memory still carries a huge premium and is only commonly available in the more expensive BGA packaging. I haven't actually seen an actively-cooled card based on the 964 reference design yet; the only one with a HS/fan I've seen is the PowerColor RV25A - which is now the "Radeon 9000 Evil Commando" I think, lol.

There is another RV250-based board in development though. It's designation is "958" and it uses BGA memory (clocked at 300MHz initially). Apparently the second revision is currently being qualified, although I can't really say that's 100% accurate information. I'm sure they'll hold back on it of they see it jeopardising 964 sales in the short term. It's going to cost quite a bit more to manufacture and probably not perform a lot better (I'm thinkin' exactly the same if clocked equivalently - no suprises please ATi!), so will be difficult to justify from a marketing point of view.

MuFu.

 

[phynicle]

damn itz pretty quiet when r300's launch is supposed to be only around the corner

 

[MuFu]

People are sleeeeeeeping......

I have no idea about the launch dates so I can't really comment on this, but since that Flash animation appeared when the ATi offices opened in Toronto, I presume we might see things kick off then.

MuFu.

P.S. In my most recent piece of correspondance with a certain member of the hardware review/news community, he said this...

"Actually I am more than a few steps ahead of you this time. "

...whereas in the past it was more like "dude, where do you get this sh|t?! Thanks...". So I'm guessing the h/w gurus have all been briefed under NDA. Hence no commotion...

 

[Tahir2]

the ATi Radeon 85k and R300 are not running with synchronous buses to mem to gpu.