As Ramhound points out in the his comment, basically all of these will affect it.
PDFs are rendered on the fly, and that's why they are so resource intensive. This high resource use is a trade-off to one of the main points of PDF -- that the file size will be smaller, for easier portability. Since size (emailing them, storing them for a long time, etc.) doesn't seem to be a concern for the task at hand, then using PDFs is probably not a great choice.
Depending on the size and complexity of your PDFs, you may not even be able to get a computer that can render them as fast as you require.
So, trade off CPU and RAM usage for disk usage...
To do this, don't use PDFs; use pre-rendered images of the pages and just flip through those. the files will be significantly larger on disk, but will display much quicker, as no real-time rendering will be needed.
Using a basic 2D graphics engine (many to choose from) you should be able to cobble something together (in the language of your choice) with which you could easily achieve 60FPS, even on "low-end" GPUs.
Alternatively, consider using plain-text instead of PDFs or graphics, as you can EASILY render full pages of text faster even than it can be displayed.
1All of the above; Its entirely depends on the size of the file. – Ramhound – 2014-10-16T12:17:08.363
Well, is there any particular order(importance wise)? Should a cluster of 4 core i5 quad-core 4GHz/16Gb RAM be sufficient to handle 40 such instances running all at once? Usual file size is around 10MB. – rinfinity – 2014-10-16T12:29:53.037
I don't like hypothetical questions. – Ramhound – 2014-10-16T12:31:19.113
1
I can't read a page in 0.1 seconds. Perhaps you should tell us why you need to do this? See What is the XY problem?
– DavidPostill – 2014-10-16T12:33:10.333For a research project in human learning and cognition we need to display PDFs to subjects with pages flipping at a speed such that one page of the pdf is displayed for a time of <0.10s.(and we need 36+ instances of such pdf readers running at once) – rinfinity – 2014-10-16T12:35:49.590
@Ramhound "I don't like hypothetical questions." Really? isn't the (implicit) statement that "this is a hypothetical question" a hypothesis in itself? :) BTW, For a research project in human learning and cognition we need to display PDFs to subjects with pages flipping at a speed such that one page of the pdf is displayed for a time of <0.10s.(and we need 36+ instances of such pdf readers running at once) – rinfinity – 2014-10-16T12:41:23.113
@DavidPostill Now that you know the 'X', do you think you can help me with it? When you can't answer a question,'question' the question. Right? BTW, you CAN read a page in 12 ms- folks at MIT say that. :) – rinfinity – 2014-10-16T12:52:37.683
It's "hypothetical" (for lack of a better term) until you share with us why you want to do it. With out the context of Why, it's often hard to give a proper answer to How. Especially considering how often we run into the XY Problem around here. "How can I cut off my own arm without bleeding out?" "Why?" "Because I have a string tied to my finger that I want to get rid of." "Answer: Just cut off the string, much safer". – Ƭᴇcʜιᴇ007 – 2014-10-16T13:13:40.390
Also, what system (specs) have you tested this on already? What were the results? – Ƭᴇcʜιᴇ007 – 2014-10-16T13:21:39.723
1Well, I should've expected that. We call it the "Scientist-Engineer Divide". A scientist won't discard any no. of Ys (because they don't need to bother about practical utility). So, in essence, if you asked me "How can I cut off my own arm without bleeding out?". I'd answer that as well AND ask for the "why" too. :) – rinfinity – 2014-10-16T13:31:19.593
We've a Core i7 4GHz. quad core/ 16GB RAM/ Pdf stored on SSD. I haven't seen any page taking longer than 3s in the worst case. – rinfinity – 2014-10-16T13:35:26.403
"I'd answer that as well AND ask for the "why" too", asking "Why" isn't an answer. ;) "We've a Core i7 4GHz. quad core/ 16GB RAM/ Pdf stored on SSD. I haven't seen any page taking longer than 3s in the worst case", that's pretty current sounding tech and it's 30x too slow (3s when you want 0.1s). What would you expect to replace it with that's "faster"? :) – Ƭᴇcʜιᴇ007 – 2014-10-16T20:00:44.020
Not a replacement but an addition- an ikea helmer cluster of (say, 6 core i5 quad core 3.1GHz) http://helmer.sfe.se/ [3 s is the worst case, and was a quick and dirty approximation]
– rinfinity – 2014-10-16T23:48:36.847