Elektro wrote:
Heh Joanna chill :-D I haven't seen any info that it's unstable.
I would call it Marginal... Yeah.. that's most accurate word of that chip. It happens to be best what's available to PPC (with AGP that is) but unfortunately it ain't Great.
I can't tell how many actual problems it has, but considering those two widely announced (this Ram issue and PCI DMA) it's to me quite obvious it's not as good as it's specs would give (their own datasheets mention chip supporting both buffered and UNbuffered rams).
(sorry.. goes technical)...
Ram issue looks too much like weak drivers on bus, they do work on easy load but on serious load they fail. I have seen that 'Alan's positive spin' about 'low power chip' but actually this kind of weaknes comes from design mistakes and/or errors in manufacturing. Never before has some kind of 'energy saving feature' has been used as a excuse for errorneous behaviour.
What I have gathered about DMA.thingy ... It sounds like there is some errors on PCI side state machine logic. It's not biggie itself but some glitches do happen and they cause data corruption and loss. It may work OK on light load but I would be worried to hell on selling those to Server use.
Anyway.. It's still too early to make final judgement on this one. I really would like to see more real and accurate raports what indeed does work ... issues like stability/speed of high.speed high volume USB transfers would be intersting and so are actual CPU usage (including interrupt/kernel %) during Disk IO (both on board Ide and SCSI card on PCI bus).