Windows Vista ReadyBoost Research - Page 3

That's the point- if there were enough virtual memory via ReadyBoost, would these apps NOT have to resort to writing dedicated scratch files and instead use the virtual cache area? I don't know since they may be hard coded to do so and thus gain no benefit. Categorically saying they won't for an entire group of apps without testing is premature in my opinion.

Same goes for games such as a few titles from LucasArts I'm looking at right now on my hard drive. I see a series of 50 Kb to 400 Kb .dll, .png and .wav files for their star wars titles that my kids play (me too, lol). Looks like basic geometry is contained in large .LVL files that would not benefit with sizes of several Mb. However, when these games cycle through a playlist of repeating maps, I wonder if caching the smaller files might benefit map load times.

For instance, our old WinXP laptop w/5400 rpm hard drive almost always loads into maps in Star Wars Battlefront 2 a complete spawn period behind other players whereas our 7200 rpm desktop usually loads into the first spawn wave. This is a big disadvantage for capturing initial command posts since the game's preset 15 second spawn period is significant in lost time if you miss a spawn wave (each map is usually 5 mins or so). Could Vista with ReadyBoost cut the time just enough to make it into play on the first spawn? We'll see when I can test it with and without ReadyBoost. Our newly arriving laptop w/7200 rpm drive may be fast enough without ReadyBoost anyway, but we'll soon see. Even slightly speedier loads can result in significant benefits for time critical gaming.

I'm not expecting performance to for instance double of course, but "significant" effects will vary per machine config, the applications usage, and on a personal perception basis too. On the gaming map load, if a fraction of a sec results in hitting the first spawn wave then that's quite significant for avoiding loss of territory. If apps load any faster at all in our multi-user switching usage at home, I'll personally be happy for that alone. On our 2.5 yr old laptop w/XP, it can take over a minute since I just timed it at 1:05 to switch among accounts and I keep it tuned pretty well. I'd be happy to cut that even just a bit each time since I'm impatient, lol.

Being a numbers nerd, here's more math on estimating "crossover data size" for minimum acceptable flash device speeds for ReadyBoost. This is the point I'm defining where a hard drive and flash device both fetch a given data amount at the same elapsed time.

Using the ReadyBoost threshold of 2.5 Mb/s rated speed I've seen mentioned in multiple places and an estimated average of slower flash cards' access speeds of 3 ms:

Hard Drive: 11 ms Latency to move head, Transfer = 30 Mb/s avg at 5400 Rpm
Slower Flash Unit: 3 ms Latency to access data, Transfer = 2.5 Mb/s avg

Crossover Point = 22 Kb where they take same amount of time to read data (11.7 ms).

This is much lower than a hi-speed flash unit's crossover value of 310 Kb per the example I calculated in a previous post. What this means in real world effects I don't know, but I thought I'd share how much difference the flash device itself can make on a low level "data bucket" basis. For me, this reinforces how much older flash devices and/or device readers that barely pass the acceptance test may be limiting ReadyBoost benefits.

Heh- I looked at a lot of machines over several weeks before picking the E1705 and now I can hardly wait another couple days to get it. hehe

Not trying to stray too far off-topic, but since it's somewhat related I wonder how progress is getting along with the new hybrid hard drives with integral flash memory? That seems to be taking a ReadyBoost like approach but to a new level of integration. I've not been keeping up regarding how much flash memory are the manufacturers typically going to integrate?

Hey Funky1, Dang, I wish I'd seen the X4 card version before I just bought the X3 version. If it's not too much more, I'd probably have bought the X4 for the chance to potentially double the speed to 40 Mb/s. That's getting well into typical laptop hard drive speed ranges for single drives. Note that total performance also depends on your reader speed and whether of course it's compatible with your card to start with. I've seen charts of compatibility lists for readers versus cards and there seem to be several combos that don't work shown here: http://www.hjreggel.net/cardspeed/special-sd.html My X3 card arrived yesterday and testing in my desktop's internal Dynex USB 2.0 reader showed a rock solid 20 Mb/s read speed over the entire card capacity with average access of 0.8 ms for the SanDisk Extreme 3 SD card. My desktop reader at least makes maximum use of the card's speed and thus the card does seem to perform as advertised. The X3's rapid access time is one of the major performance features I was after. So if the X4 maintains that same access or better, then it sure seems like a good candidate to me with its even faster sequential rate... IF IT's COMPATIBLE with your reader. Keep in mind that if the flash device cost gets near that of a Ram upgrade, you're certainly better off expanding your Ram first. My E1705 laptop has just arrived today at home (I'm itching to leave work early now, lol). One of the first things I'll do is test the X3 card inside the latop's media reader to see if the reader/card are compatible and if the reader is a good performer or not compared to my desktop. I used the HD Tach 3 freebie test software below for XP, although it doesn't support write tests unless you register: http://www.majorgeeks.com/HDTach_d672.html ... or at the author's web site here: http://www.simplisoftware.com My card just fired right up in the desktop reader and away it went with no issues. I downloaded and installed the HD Tach software, popped in my SD card, started up HD Tach and ran the test. Your card should be a lettered device choice in the selection list of devices to test. Select the Long Benchmark test for the most consistent results, and it should take just a few minutes. You of course can go ahead and test your hard drive(s) too while you're at it.

I did a quick test of the SanDisk Extreme 3 SD card in my new E1705's media reader with Vista Ultimate and it accepted and enabled the card for ReadyBoost. I'm still running benchmarks on it, but it looks like the E1705's media reader is at least 30% slower than my XP desktop's internal reader.

Since HD Tach doesn't run under Vista, I'm trying out SiS Sandra version 11 Lite for benchmarking and it can be fetched here:

http://www.sisoftware.co.uk/index.ht...buy&langx=en&a=

Sandra has a test mode wherein it does a series of read/writes at various file sizes and measures the operations per minute speeds under load. It has several reference device data curves for comparison and shows a Peak Extreme SD card kicking some serious butt on speeds, but I don't know if it's compatible with Dell's readers or not. An entry at the RB compatibility site (http://www.grantgibson.co.uk/misc/readyboost/) shows a similar Peak device is acceptable for Vista. I've no idea regarding the readers utilized for the reference devices shown within Sandra.

Dell E1705 Media Reader Performance is Slow

As I was afraid might be the case with my brand new E1705 I got this week, the internal media reader does not perform up to par with SD cards. Using a combination of ATTO Disk and SiS Sandra benchmarking software, here's the speeds compared to an external reader and a thumb drive I have:

E1705 Internal Media Reader w/SanDisk 2Gb Extreme 3 SD Card:
Read @ 256Kb Files = 14.9 Mb/s
Write @ 256Kb Files = 3.8 Mb/s

External SanDisk USB 2.0 Reader w/SanDisk 2Gb Extreme 3 SD Card:
Read @ 256Kb Files = 19.4 Mb/s
Write @ 256Kb Files = 16.0 Mb/s

SanDisk 2Gb USB 2.0 Micro Cruzer Thumb Drive in USB port:
Read @ 256Kb Files = 13.6 Mb/s
Write @ 256Kb Files = 6.8 Mb/s

Another Possible Option I've not tested but saw at the store:
PCMCIA Adapter (or ExpressCard for the E1705) for media cards
-Would "hide" a flash card similar to internal laptop reader for klutzes like me

The internal laptop reader with the Extreme 3 SD card performs poorly on writes compared to an external reader as shown above using the same exact card- 3.8 vs 16.0 Mb/s. Its read speed is slower too but not to a horrible degree with 14.9 vs 19.4 Mb/s. The thumb drive performs fairly well on reads although slower for writes compared to the external reader, and is also accepted by Vista as "ReadyBoost-able".

Thoughts:

- ReadyBoost does accept the internal reader with SD card and activates on it, but I've not had a chance to see what real world effects it has on my laptop responsiveness yet. To what degree does write speed vs read speed come into play with ReadyBoost- unknown. Cost: $51 after rebate.

- The SanDisk external reader is a slim ImageMate 5-in-1 unit (model SDDR-99) about the size of a cigarette pack but slimmer that I got at BestBuy for $21. It's thin enough to velcro to the back of the lid and cable into one of the back USB ports without too much bulk, so I may do that for the better performance. Cost: $21 + $51 = $72

- A thumb drive can now be had with 2 Gb on the cheap these days although I'm bad about bending or breaking the things when I forget they're inserted and bang them against something. My personal preference would be for an external reader velcroed to lid rather than leaving a thumb drive inserted that we'll (me or my kids) eventually break. (I've killed two thumb drives so far as well as a USB port w/service send-in repair reqd, lol). Cost = $29

- It's much more convenient to have the SD card almost totally tucked away in the internal reader other than about an eigth of an inch. Dangit, if it weren't for the issue of the much slower write speed it'd be an easy decision in favor of an SD card in the internal reader for me! I forgot already and set the laptop on its right side last nite looking for the bottom express tag number and it didn't hurt the card while resting on it- so it's robust enough to stand up to my forgetfulness so far. I assume the E1705 write speed will be slower with other cards too, but that's unknown at this point. I'd say it's probable for other media cards too, but can't verify that.

Hmmmmm....... lol.

I'll be doing timed testing of how much benefit the internal SD card does over the next several days, so more to follow...

Cstrading- Thanks a bunch for the info, I'll try that this afternoon and post any speed effects from optimizing it for performance. Here's the results for the block sizes used by ReadyBoost (see end of post): Before Setting to Performance Mode for SD Card/Internal Reader: Read @ 512KB: 15.01 MB/s Write @ 4KB: 2.57 MB/s After Setting to Performance Mode for SD Card/Internal Reader: Read @ 512KB: 14.25 MB/s Write @ 4KB: 2.60 MB/s Before Setting to Performance Mode for SD Card/External Reader: Read @ 512KB: 19.39 MB/s Write @ 4KB: 5.13 MB/s After Setting to Performance Mode for SD Card/External Reader: Read @ 512KB: 19.44 MB/s Write @ 4KB: 5.15 MB/s Seems like a very minimal effect from changing to Performance Mode for the SD card in either the internal reader or the external one. I've now found 3 different versions of performance requirements listed for an acceptable ReadyBoost device: 2.5MB/s for random 4K reads; 1.75MB/s for 512K random writes 3.5MB/s for random 4K reads; 2.50MB/s for 512K random writes 5.0MB/s for random 4K reads; 3.00MB/s for 512K random writes; access time < 1 ms They're at least all consistent in terms of the read and write block sizes, lol. As I understand, the read block size corresponds to Windows' page file increments. If the most stringent set listed last (5.0/3.0 read/write speeds) is a true threshold, then my SD card inside the internal reader would not qualify. It would though for the other two, even if barely for the 2.5 MB/s write speed with 4KB blocks. And since ReadyBoost DOES accept my SD card/internal reader, then apparently the last set of 5/3 MB/s isn't valid for the shipping Vista Ultimate version. I'll have to double check at the smaller read block size and the larger write block sizes for the thumb drive I have.