Chip said:
Tom` said:
Unfortunately, I can't find the post, because for some reason MWeston's post history apparently only includes either a certain number of posts, or only posts back to a certain time. Google was similarly unhelpful - maybe problems caused by the board upgrade.
To avoid the cutoff, you have to do an advanced search and specify a date range. I just set it to between 2003 and now.
MWeston said:
24MB/s based on a 48MHz clock, 4 bit SDIO
Thank you! Finally - a real data point. VERY good find.
Looking at SDHC card benchmarks on:
Tom's SDHC benchmarks
It looks like:
Class 4 card average read speeds top out around 15MB/sec
Class 6 card average read speeds top out around 19.6MB/sec
So, in theory, the limiting factor with these would be the cards themselves.
There is no 3rd party benchmarking data yet that I can find for the new class 10 cards. However, SanDisk is claiming 30MB/sec reads/writes:
SanDisk 32GB SDHC class 10 card
We would not likely be able to top out that card and it would be limited to the board throughput. Still - this MAY be 20+% faster 24MB/sec bus with 30MB/sec card = <24MB/sec VS using a 19.6MB/sec card.
All of this winds up being theoretical though since nobody has tested a high end class 6 card, let alone the new class 10 cards in a Pandora.
We do know, now, that there is a cap. However, the claims that the cap would limit the card to Class 2 or Class 4 performance strikes me as bunk. That would be 10-15MB/sec not 24MB/sec. In theory the class 6 cards would be appropriate and about 30% faster than the class 4. Whether or not any speed gains could be seen from using a class 10 card... it all depends on how much of the 24MB/sec bus we can actually use. If we can get the whole thing, then a class 10 card potentially being 20% faster than a class 6 -might- make sense to some people.
This -does- shoot down the RAID0 for performance thing, unless you're using class 4 or slower cards to begin with.
We didn't all buy our Pandora with our lunch money. I would like to purchase a set of cards that top out the performance of the handheld - if the gain going from a class 4 to class 6 or class 10 is negligible, I'll save my money for the Pandorappstore.
Anxiously awaiting benchmarks - but we're going to need some standardized way to do it.