Sdhc - How Fast Is Fast Enough?

[Kramy]

Most of you that replied to this thread - you really have no clue what you're talking about. You're real adamant about your opinions, I'll hand you that. Still, no real world experience then. Amazing.

Alrighty then - lets lay out some facts. I'm not going to bother quoting from the mis-information in the posts above. Not worth it.

:blink:

The worlds fastest consumer SSD transfers data at 260MB/s Megabytes per second. Yes there are some industrial raid ones that are faster - but none of us could afford them.
Engadget - OCZ's Colossus

I think the SSD you're actually referring to is the FusionIO IoDrive. The consumer ones, which can be had for under $1000 (cheaper than the highest end OCZ SSD) have read speeds of over 700MB/sec in real-world scenarios, and 1000+ MB/sec in specific benchmarks.

Of course, it's a PCIe 4x SSD; not a SATA SSD. And it's made by a company that has godly controllers.

A standard notebook HDD - The Samsung Spinpoint M6 500GB drive in my notebook computer transfers data at 57MB/s.
Tom's Hardware 2.5" drive benchmarks

Okay, but you lost me here:

Yes, that is 1/2 the speed of the hotshot high end drive in my notebook computer.

That's a 5400RPM drive. That's not hotshot...

The fastest desktop HDD - The Western Digital VelociRaptor clocks in at 102MB/s. Yes, you could RAID 0 4 of them an reach about 350MB/s.

VelociRaptors have decent access times, but slower sequential read/write speeds than newer 32MB/64MB cache drives.

Some researches found that HDD speeds could be vastly improved by strapping extra cache on them. The future looks bright for HDDs, once they hit 512MB+ of cache.

The class 10 SDHC cards I'm referring to transfer data at 30MB/s. Yes, that is 1/2 the speed of the hotshot high end drive in my notebook computer.
Sandisk 32GB Extreme III SDHC card

Technically these are Class 6 cards sold as "Class 10". Class 10 doesn't actually exist.

It's rather like calling an IoDrive SATA5, just because it's really damn fast. Doesn't make it at all correct.

Follow the Tom's hardware 2.5" notebook HDD list down to where the 3 year old 160GB drives are at and - guess what? 2 years ago the 'normal' HDD had slower average read times than these new hotshot SDHC cards.
Tom's hardware 2.5" HDD listings

3.5" HDDs have been the norm for the recent past. I have some from 7 years ago that outperform yours.

These new SDHC cards are plenty fast for swap - IF the OpenPandora console can take the data rates.

I heard something like 14MB/sec after overhead with the current driver. Any which way - why would you want it as swap? Swap makes linux so horribly slow. I'd rather have that fancy kernel page compression.

So, now that we have established that most of the above posters have no clue what the hell they are talking about. Can we stop speculating and see if just maybe someone with access to the actual motherboards might be interested in answering a question before it gets buried 5 pages deep in crap replies?

Yeah, that seems like a good plan. Maybe whoever first tested the SD hardware can chime in.

Is there the will or the time for anyone to run a set of average read benchmarks on an actual Pandora motherboard to fill us in on if it can handle the throughput from these new class 10 cards? I don't think I'm alone in wanting to get past speculation from the clueless and down to some real facts.

Thank you,

Grench

Class 6.

Sorry if that came across offensive. I'm just bored and had nothing better to do than quote stuff.

Oh, and apparently there's a quote limit of 10 per post.

 

[kmob]

This seems to be going off topic fast. I think the original question was a perfectly reasonable request for some useful practical information.

I'd also like to know what read/write speeds the Pandora is capable of on the SDHC ports. Start a new thread if you want to argue about SDHC certifications or hard drive speeds etc.

 

[Asiyura]

I wonder how most external hard drives would compare, since I'm hoping to use the Pandora plus a portable USB-powered external hard drive as a lighter portable media center (I usually go to other people's houses with my laptop and play shows/movies through svideo.)

I was thinking about doing something similar, but with a usb hdd battery powered, like my old iRiver multi-codec jukebox.
Or a usb powered hdd but with a small battery pack.
I don't want the HDD to drain the battery of the pandora...

 

[zhasha]

In the whole swap debate, you should also consider randomized lookup time. It's usually 8ms for a regular harddisk, where you may get something like 0.1ms on an SSD disk. I'm not sure about SHDC cards but I can imagine they're quite fast.

 

[fearofshorts]

While we are still off-topic,
in my experiences an SD card's read-speed is not dependent on it's write-speed.
Most, if not all, SD cards have far faster read-speeds than write-speeds. And most cards (be they class 2, class 4 or class 6) have similar read-speeds- a speed much faster than 6MB/s.

And yes, I am interested in the theoretical and practical maximums for the Pandora's SD card bus.
But I do expect them to be reasonably high. I mean, if people can sell SD card readers for one or two dollars with very fast speeds then I doubt the Pandora's one will read or write too slowly.
Of course, I may be wrong about this...

 

[Polossatik]

This seems to be going off topic fast. I think the original question was a perfectly reasonable request for some useful practical information.

I'd also like to know what read/write speeds the Pandora is capable of on the SDHC ports. Start a new thread if you want to argue about SDHC certifications or hard drive speeds etc.

+1 , can someone simply do some basic checking like
* using a bunch of cards, preferably some with like 15Mb/s write (!) speed (or more) and at least 2 identical.
* use h2testw on a pc (or any other tool you fancy) to see what is the write and read speed you get for those cards (might test on different reader also) as a "benchmark".
* then check what is the upper limit when doing a similar write / read test on the pandora itself (suggest to use the same approach as h2testw, generate files when writing them, not copy them from somewhere)
* if a upper limit is established (or not, if the bus seams faster then the cards you have) it would also be very good to have an idea of the troughput of a *copy* from SD to SD (preferable same cards, or at least cards who give the same results, in both slots) and when read/write figures when using USB (mass media attach) and trough FTP.

All cards should be formatted by the SD format tool.

 

[DaMummy]

ok so does anyone have word what the bus can handle? would there be a difference in speed between class 4 and class 10? not trying to waste an additional $100 here...

 

[WizardStan]

ok so does anyone have word what the bus can handle? would there be a difference in speed between class 4 and class 10? not trying to waste an additional $100 here...

Then don't. Wait until you or others actually have one and start testing it. You don't need to have your cards as soon as you receive it. It won't kill you to wait a couple days while you use an older card.

 

[Tom`]

ok so does anyone have word what the bus can handle? would there be a difference in speed between class 4 and class 10? not trying to waste an additional $100 here...

Then don't. Wait until you or others actually have one and start testing it. You don't need to have your cards as soon as you receive it. It won't kill you to wait a couple days while you use an older card.

More verbosely, MWeston stated at one point that the maximum SD speed that could be achieved was 24 MB/s (I can't find the post now - the board only records the last 200 posts by MWeston, and Google is unhelpful, though it does show results in which people discuss that number). Someone else (ED?) stated that he had at least one card that would do 15 MB/s.

More to the point, however, it's impossible to tell whether there will be a speed difference between the two cards - the "class" rating only refers to minimum speed, so it's likely that the class 4 card will read or write somewhat faster than 4 MB/s, and likewise with the other card, but the ratings alone don't tell us anything - it's even possible, if a bit improbable, that the Class 4 card could be faster than the Class 10 card.

 

[cobaltage]

I don't know about these class speeds and hardware limitations, but in my practical experience, if you're running an OS from a flash card, premium flash memory definitely results in faster application loading times. In my old Dell Axim x5v, a 133x SD card was significantly faster than the internal flash memory. 133x SD cards are expensive, though. It seems like one useful measure of SDHC cards of various speeds and classes would be OS boot time.

 

[CoLDSToRAGE]

http://www.sandisk.com/products/imaging/sandisk-extreme-sdhc-cards-

Ouch! Pricey... and possibly overkill?

 

[Nintendo]

I've been using a 32GB SDHC card (speed: class 10) for my Wiz and plan to buy a few more soon just in time for my Pandora to arrive.
It only cost me AU $200. You can fit all the emulators and roms on it, no worries...and still have space for all your music and pr0n.

 

[cobaltage]

http://www.sandisk.com/products/imaging/sandisk-extreme-sdhc-cards-

Ouch! Pricey... and possibly overkill?

If you're talking about price, you should start considering price per gigabyte in your calculations.

 

[DaMummy]

or learn to shop around....

 

[Wolenber]

I ... plan to buy a few more soon just in time for my Pandora to arrive. It only cost me AU $200.

My wallet is jealous of your income.

 

[yannv]

AFAICT there's a lot of discussion of possible bitrates and not much else. The SDHC spec classes are rated in MB/s minimum write, yet some manufacturers used classes that weren't yet specified (class 10?) and many use markings that vary in meaning between manufacturers.
The Pandora itself, by information earlier in this thread, has been measured at 15MB/s and could peak at 24MB/s theoretically.
Lots of discussion on whether it'd be usable for swap, disregarding that the point of swap is to free up actual RAM - thrashing is horrible for performance, but paging out some memory that's currently not needed means that much becomes available. It could help, won't be necessary mostly, and you sure don't want a high swappiness setting (yes, an actual setting in Linux).

My own choice of first card (I have not selected a second) for the Pandora was based largely on a Tom's Hardware article, "13 SDHC Memory Cards Reviewed". Specifically, the most interesting measure to me is total access time. I'm expecting to do quite a bit of random access, for instance compiling which reads and creates lots of small files, and then both seek time (fundamentally lower on SDs than HDDs) and command rate matter a lot. These were tested under the heading "Access Time, I/O Performance".

Does anyone know of more tests of this type?

 

[Grench]

AFAICT there's a lot of discussion of possible bitrates and not much else. The SDHC spec classes are rated in MB/s minimum write, yet some manufacturers used classes that weren't yet specified (class 10?) and many use markings that vary in meaning between manufacturers.
The Pandora itself, by information earlier in this thread, has been measured at 15MB/s and could peak at 24MB/s theoretically.
Lots of discussion on whether it'd be usable for swap, disregarding that the point of swap is to free up actual RAM - thrashing is horrible for performance, but paging out some memory that's currently not needed means that much becomes available. It could help, won't be necessary mostly, and you sure don't want a high swappiness setting (yes, an actual setting in Linux).

My own choice of first card (I have not selected a second) for the Pandora was based largely on a Tom's Hardware article, "13 SDHC Memory Cards Reviewed". Specifically, the most interesting measure to me is total access time. I'm expecting to do quite a bit of random access, for instance compiling which reads and creates lots of small files, and then both seek time (fundamentally lower on SDs than HDDs) and command rate matter a lot. These were tested under the heading "Access Time, I/O Performance".

Does anyone know of more tests of this type?

I wish Tom's would do an update with some of the new cards coming out. In lieu of that, I've been trying to figure out where they would fall by comparing the data from a few stand alone reviews.

Here is a stand-alone on the Patriot 32GB Class 10.
http://www.ccereviews.com/reviews/patriot-memory-lx-series-32gb-sdhc-memory-card/

Another stand alone (not so well done) for the Sandisk 32GB Extreme III Class 10
http://www.legitreviews.com/article/1172/1/

Still wish THG would get going on a refresh'd listing.

 

[fraseyboy]

Personally, I want to know whether it would be faster to have a Class 4 SDHC or a USB flash drive...

 

[Pleng]

It depends on the speed of the flash drive!

The flash drive will almost certainly suck more battery power, and be bulkier, though.

 

[kmob]

As has been said in various other threads on this subject. You need to take into account what you are going to be using your card for and tailor its performance to that use. Focusing on the minimum write speed (which is the only thing that SDHC class ratings relate to) as the most important factor when choosing your card is a bit foolish. Min/max read speed, max write speed, access time, reliability etc could all be more important depending on your use. Also, we will have two slots...

Think about your usage scenarios, and don’t pay more than you need to when it won’t bring any benefit.

If you’re loading a lot of read only content (video, music, eBooks, ROMs etc) which all have low read speed requirements (for example the max bit rate of a 1080P Bueray video is only 40Mb/sec or 5MB/sec, and DVD is 1.2 MB/sec) and you’ll probably only be writing the cards sporadically and from your PC/Mac only. So a cheap and cheerful “low class” card would be fine for that. If you’re doing more heavyweight use (running another OS, or on-Pandora development comes to mind) where read speed and access time are significant then spend more to get more.

That said, from the (now old) famous Tom’s hardware 13 SDHC card review their top SDHC card (bearing in mind the age of the review) hit 18.6 MB/s write and 19.7 MB/s read transfer rates – which makes its claims of 133x seem justified (class 6 is “only” 40x). So if the Pandora does a real world 24 MB/s writes per SDHC slot that’s pretty impressive, and you would need a card that delivers 160x to achieve that.

We still are missing some real world performance numbers on read/write per slot and read/write to both slots at the same time though – but then again discovering this sort of thing is part of fun