Snipped DEAD link
Hahahahaha... that's funny.
Snipped DEAD link
Hahahahaha... that's funny.
I previously asked this question on comp.system.ibm.pc.storage and got zero answers, but since this newsgroup contains a large number of computer savvy folk with the right sort of kit and an inclination to test things as opposed to believing the manufacturers marketting hype. Here is a repost:
I wonder if anyone can point me at reliable benchmarks of measured flash memory performance?
Or recommendations based on first hand experience with the larger drives. 8GB, 16GB.
I found the following URL with some older drives in the 1-2GB range tested with h2benchw 3.6, but AFAICT there is very little else out there with real measured performance figures. I am interested in buying the largest USB flash drive I can get consistent with still getting around 20MB/s data rate for read and the lowest possible random access seek time. I don't care much about the write time as I only expect to write the data once. Cheaper is obviously better. (and I am hoping it will be a lot quieter and cooler than a 10k rpm spindle)
The benchmarks that I have found are summarised on Toms Hardware page:
All of the drives reviewed there are too small for the data I need to store. My bog standard no-name 1GB manages 12MB/s read rate and that is too slow for what I want to do. It is also much too small. The seek time is also important as I expect the vast majority of accesses to be uncorrelated and random.
I was considering the Corsair 8GB or 16GB basic blue units, but I was a bit alarmed at the apparently slow ~30ms random access time shown for their smaller drives cf others like Kingston, Crucial and Memina in at around
For low volume or a one off:
Scavenge a case from a small-capacity unit. Lots of 8M, 16M 32M, 64M,128M, 256M units that were sold and/or given away as convention freebies floating around doing nothing much these days.
Chop off a chunk of a butane lighter (empty, pick up on side of road) and use hot glue, silicone or epoxy to mount/encapsulate.
Use hot glue straight up to encapsulate, perhaps pulling a mold from one you like the shape of with RTV silicone or plaster - or just get artistic in your application of heat to achieve the shape you want with no mold at all. Cut a pair of sides from a wooden tongue depressor and make a hot-glue ice-cream sandwich around the board, as one approach.
Following a typical commercial design - Buy suitable metal tubing (round) and squish to (oval) - then all you need is end plugs - again, epoxy, silicone or hot glue should work in low volumes.
-- Cats, coffee, chocolate...vices to live by
It is NOT "marketing hype", dipshit. There are many many reviews on the product, AND I own one, and it DOES run JUST THAT FAST.
It's the flash voyager GT, idiot, not the standard voyager.
It is a new product, and is NOT listed in Tom's diagram!
There are plenty of real world test reviews by many of the computer mags out there. Just go look. Sheesh! You obviously didn't look close enough to see that it is not even the same product as that you gave a diagram for.
I don't have a problem with the read write speeds for large files. I can well believe them.
What worries me is that random access speeds shown for the smaller Corsair products in the only online review made with a reliable benchmarking tool. I am sure that for large files a GT will max out around 30+MB/s (and the cheaper one 20MB/s). The big question is what is their seek time for random access. 30ms is glacially slow!
I want to know the measured seek time for small sector random access. A number that I have not seen quoted in any reviews online, in the trade press or in any of the comics. If you can point me at one reviewing 8GB flash drives properly then I would be very interested. I am in the UK so US magazines are rare but can be found if necessary.
Failing that how about proving your drive performance with one of the standard benchmarks :
h2benchw 3.6 (German magazine) and HDTach 2.xx are both linked from that page - either would do. The number I am most interested in is the seek time. Bandwidth for large sequential transfers is not in doubt.
As I pointed out in the original post I was alarmed by the huge apparent seek times of the smaller Corsair drives and I have not been able to find any competent reviews at all of larger USB flash drives. There are endless reviews of classical disk platters (where manufacturers publish performance & avg seek time anyway). But USB flash drive manufacturers specs are deliberately vague about these important parameters.
I am now toying with the idea of putting together an ad hoc RAID0 array using a pair of the cheapest possible USB drives. Effectively doubling their bandwidth although others trying this approach have encountered some snags.
Regards, Martin Brown
HDTach is proprietary, and not available, and only points to their site. (it even says so).
H2Benchw results are: (and my USB 2.0 port is a very early implementation) Note also that this benchmark is meant for real hard drives and that the read/write operations performed on USB flash drives differ from that of the sequencing used on real hard drives, regardless of file system.
Capacity: CHS=(1007/255/63), 16177455 sectors = 7899 MByte
Interface transfer rate w/ block size 128 sectors at 0.0% of capacity: Sequential read rate medium (w/out delay): 24185 KByte/s Sequential transfer rate w/ read-ahead (delay: 2.91 ms): 23365 KByte/s Repetitive sequential read ("core test"): 24343 KByte/s Sequential write rate medium (w/out delay): 19985 KByte/s Sequential transfer rate write cache (delay: 3.52 ms): 20598 KByte/s Repetitive sequential write: 10124 KByte/s
Sustained transfer rate (block size: 128 sectors): Reading: average 23852.3, min 23173.4, max 24381.6 [KByte/s] Writing: average 21728.6, min 12899.1, max 22507.6 [KByte/s]
Random access read: average 0.91, min 0.45, max 2.50 [ms] Random access write: average 38.04, min 1.38, max 120.98 [ms] Random access read (
Odd. I admit I didn't follow the link, but since early versions of HDtach are given away on shovelware CDs I had assumed their online site would also include that basic version. Thanks for posting the results. Most reassuring!
Indeed, but since it hammers the drive pretty effectively it is a lot better than trusting the manufacturers figures.
And this is for the Corsair GT model with a nominal bandwidth capability of 30+MB/s?
Tends to suggest your USB 2.0 implementation is throttled back somewhere. A full link at 480Mbps should be good for nearly 60MB/s before saturating. Anyone have the basic red Corsair model nominal20MB/s xfer rate (or even an el cheapo brand) to see how it compares?
So chances are 20MB/s is a realistic performance. I wonder if the 16GB unit is the same speed?
Thanks for your help. Regards,
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.