Why are ebook readers so slow?

[ilovejedd]

Originally Posted by Brianrh

I have a Sony PRS300 and a Kobo Touch and to load or change books seems to take so long, especially on longer books. I have kept the number of books stored on the device down to about forty and keep them in the main memory. Books of over 1000 pages can take ages into minutes for the Sony.

Is this normal for ebook readers? I thought they had microprocessors as powerful as mobile phone ones so why should they be any slower?

The Kobo Touch has a Cortex A8 800MHz CPU, which is just a little bit slower than a Samsung Nexus S. Aside from the Kindle Fire/Nook Tablet, that's the fastest CPU I've seen on ebook readers. If you're having issues opening a file on the Kobo (3 minutes sounds abnormal), it's possible that there might be some corruption or weirdness going on. Try to re-download the epub or do an epub-to-epub conversion via Calibre.

Older readers have much worse CPU. The Kindle 3g/Wi-Fi circa 2010 models sport a Freescale i.MX35 (ARM1136JF-S core, 532MHz) CPU. That's around 11% slower than my LG Thrive (AT&T version of the Optimus One) and the Thrive already crashes on Mantano Reader quite frequently particularly when there's a large number of books in the library.

Readers aren't built to be number-crunching powerhouses. They tend to have CPU and RAM that's just good enough for the task. According to the MR Wiki, the PRS-300 has a Freescale i.MXL MC9328MXLVP20 (ARM920T core, 200MHz) CPU. Yes, it's going to be slow.

[Edward M. Grant]

Originally Posted by ilovejedd

According to the MR Wiki, the PRS-300 has a Freescale i.MXL MC9328MXLVP20 (ARM920T core, 200MHz) CPU. Yes, it's going to be slow.

It's not so long ago that a 200MHz RISC CPU would have been considered blindingly fast. Back in the mid-90s I had a Sun workstation with a 40MHz CPU and 32MB of RAM and never complained about the speed.

You're right that e-reader CPUs are chosen to be 'just fast enough', but I still don't understand why they seem so slow on an absolute level. They're basically just uncompressing a .zip file, reading the HTML it contains and formatting that for display, which seems easy enough... the Sun seemed to do it faster with a CPU running at a tenth of the speed.

[BeccaPrice]

I think we've established that, in general, ebook readers open books relatively quickly, and that there is probably an issue with the specific ereaders that prompted the OP.

[HarryT]

With an ePub file, the computationally intense process, when opening a book, is creating the "parse tree" for each "flow" (each separate HTML file) of the book. This can take quite a while for a book with a complex structure. It doesn't happen with Mobipocket because, unlike ePub, Mobi is "stateless".

[Edward M. Grant]

Ah, I didn't realise that epub was much more complex than mobi. From what I'd seen, the formats all seemed pretty similar.

[HarryT]

Originally Posted by Edward M. Grant

Ah, I didn't realise that epub was much more complex than mobi. From what I'd seen, the formats all seemed pretty similar.

They are both HTML but, because ePub supports CSS (and Mobi doesn't), the ePub reader has to maintain state information, and to do that it has to construct a parse tree from the CSS. That can be quite a computationally intensive task.

[ilovejedd]

Originally Posted by Edward M. Grant

You're right that e-reader CPUs are chosen to be 'just fast enough', but I still don't understand why they seem so slow on an absolute level. They're basically just uncompressing a .zip file, reading the HTML it contains and formatting that for display, which seems easy enough... the Sun seemed to do it faster with a CPU running at a tenth of the speed.

Maybe you've just forgotten how slow it actually took.

I do still have a first-gen iPhone and webpage rendering on that thing takes ages compared to the iPhone 4S. Same with downloading epub files using Stanza (Stanza explodes epubs directly after downloading instead of during opening to shorten epub load times).

[Freeshadow]

Exactly, why oversize the CPU and drastically raise power consumption?

[HarryT]

Originally Posted by Freeshadow

Exactly, why oversize the CPU and drastically raise power consumption?

That's the reason that so many readers use CPUs like the ARM Freescale CPUs which dramatically reduce their clock speed (and power usage) when idle. An eInk device typically spends the overwhelming majority of its time sitting around doing nothing.

[Polarpenguin]

Originally Posted by BeccaPrice

I think we've established that, in general, ebook readers open books relatively quickly, and that there is probably an issue with the specific ereaders that prompted the OP.

Have to agree I have a 2.5 yr old PRS-300. Have just timed a variety of books, the longest to load from scratch (never been opened before) was a Lrf book, which timed at roughly 45 secs. Epub & Pdf were instant.

Originally Posted by HarryT

I've never known a book to be slow at opening on the Kindle. I just tap the name in the library and it opens instantly.

On the older PRS-505 and PRS-300 there can be a delay in initially opening a book if it's in LRF format, because the device paginates the book on initial opening, which can take a little while. But that shouldn't happen with an ePub book.

Exactly my experience.

[Brianrh]

Right I reconverted the offending epub and it was 4kb bigger and 5 pages longer but it still took an age to open. I rather feel that the file is corrupt or has a problem.

The book is Why Didn't They Ask Evans by Agatha Christie and is only 140 pages long. One strange thing I did notice is that although it is Georgia font, the start of the book is normal font but somewhere in the middle it changes to italic for the rest of the book.

Your comments on general opening times for other ebook readers seem to be what I find most of the time but I do seem to have a few books which really dawdle.

Thanks to you all for your help and information. I know more about epub construction and ebook reader characteristics now.

Cheers.

Brianrh

[Belfaborac]

Just out of interest, would anyone happen to know what CPU hides within the PRS-950?

[HarryT]

Originally Posted by Belfaborac

Just out of interest, would anyone happen to know what CPU hides within the PRS-950?

The PRS-650 uses a 260MHz Freescale iMX31 CPU, so it's a reasonable guess that the 950 has the same.

[ilovejedd]

Originally Posted by HarryT

The PRS-650 uses a 260MHz Freescale iMX31 CPU, so it's a reasonable guess that the 950 has the same.

Actually, top clock on that is 532MHz, I believe, similar to the Kindle c.2010. Whether it's locked at 260MHz or if it downclocks to 260MHz at idle (which is most of the time), the PRS+ devs would probably know.

The PRS-950 might have the Freescale iMX35. I seem to recall reading somewhere that the iMX31 and iMX35 are nearly identical except the iMX31 doesn't have networking capabilities.

[Belfaborac]

Thanks guys!