HSI Vs Rialto Bridge Chip

[IgnorancePersonified]

It's not the latest mtv rock/rapper collaboration but a serious question as too the capabilities and eventual award for "better Tech implementation" for the "competing" chips.
From my limited understanding:
nV's HSi bridge chip is two way - allowing for AGP cores to work on PCIe sockets and vice versa.
ATi's rialto is one way only - PCIe cores can work on an AGP bus. Looing at B3D's latest article it seems the Rialto bridge is quite large - larger than the HSI chip. Both seem to put out appreciable heat but are passivley cooled.
I have no idea of cost for the chip nor ease of manufacturing but was wondering what people's thoughts were on which implementation was "better", more suitable for the transition period, most adaptable, used less power, has lower latency or pentalty, other uses? etc.

Too me the 2way nature, 1st to market and smaller size features of the HSI bridge chip puts it in front...

Cheers

 

[Geo]

Possibly there aren't enough data points to give a valid opinion, but it seems to me that going all the way back to Maxx that ATI has had problems with bridge chips.

 

[kemosabe]

If you're catering to the OEM PCI-E market, it makes no sense to use an AGP to PCI-E bridge as it only increases cost and theoretically hampers performance in those (rare) applications that can benefit from the PCI-E bus.

I think it was more a question of ATI prioritizing those OEM slots right off the bat and thus focusing on ramping their native PCI-E parts. It paid off handsomely initially as it extended the life of their RV350/360 core and catapulted ATI's market share to historical highs, but it could be short-lived as NVDA is well positioned to grab back some 6200/6600 OEM business heading into the spring (hence the urgency for a low-end RV515 ASIC). By contrast, ATI had neglected the retail market and is now poised to recover in that segment with AGP R430 and R480.

Doesn't answer your question about which bridge chip is technologically superior, but in this case I think the strategies were more financially driven.

 

[Xmas]

Comparing pictures, HSI seems a bit longer, but less wide than Rialto.

Both parts seem to get the job done they're made for. We don't know about yields and cost. HSI needs a heatsink, Rialto does not. Then again, Rialto has better
placement from a thermal POV. HSI can run double-speed when paired with an AGP GPU.

 

[Geo]

"more" a question of, possibly, but there seems to be general consensus that it took them longer and was more problematic than they had intended/planned on. I mentioned Maxx because a similar situation happened then, and by the time they had the bridge chip done (to support Win2k/xp) they decided that the market window had closed and it never did see the consumer space (that I'm aware of anyway).

But two data points don't make Spring, or something like that. Tho I would also imagine that the next time they need to develop a bridge chip there will be some preemptive (and understandable) wincing in some quarters.

Edit: Errrm. . maybe it showed up on AIW at some point?

 

[stevem]

...NVDA is well positioned to grab back some 6200/6600 OEM business...

In particular, the demand for native AGP 6200 (NV44a) from AIBs is intense. Retail sales demand is forecast to be just as spectacular.

 

[oddfellow]

There's not enough information available to give a valid answer to the question.

However, since both chips don't offer any actual performance benefits, and since they simply do the job that they're designed to do, without hampering performance, I'd say they're both as good as each other.
They're not designed to out perform each other, they're designed to do a job, and they both do it well.
I guess it's like taking two spoons, and asking which one is better