CliePet

I'm not surprised by the results, and a little disappointed.
The updated CF driver is essentially just as fast as the 1.0 driver for CF transfers, and only a little faster with the cache turned on for a very limited set of cases.

The 50% speedup is bogus claim IMHO, at least based on my measurements.

> "On some cards this will boost performance of the card by 50% or more "...

See comments below regarding VFS Mark.

----
I'll only report my 'best' case CF card, a 30X Transcend CF card. Other cards (like the 1GB SanDisk) are appreciably slower, and in some cases can be slower than normal memory sticks.

----
re: Raw Read/Write speed

For raw reads (no caching) that means my best case is still around 900->925 kB/sec for reads, and 500->525 kB/sec for writes.

The difference between the 1.0 and 1.1 numbers for raw transfers is essentially the same (within the sampling error)

If you turn on caching in the driver, my measured raw read speed goes up by about 10%. I suspect that's because my test program is still hitting cached data. Write speeds are unaffected, as you would expect

So my best raw numbers (with caching): 1020 kB/sec read, 520 Kb/sec write -- a little bit better, but not 50%

----

re: VFS Mark

From the announcement, the 50%+ speed increase claimed for 1.1 is a based on caching and using the VFSMark program.

In my case, My VFS number went up from around 150 to 208 !

On the face of it, that looks like a very good speedup, so why am I complaining ? ->

The problem is the VFSMark benchmark - a very bad benchmark.

To get that 30% speedup in my case, two tests were much faster - the ones that relied on repeated small reads. The rest were essentially the same.

File read went up by about 30% (half the test is repeated small reads)

DB Import went up by a whopping 190% - which totally screws up the benchmark total (just an average of the sub-tests)

Ie. if you added caching on top of the existing MS driver, you would see similar bogosities.

---
They optimized the program to improve a benchmark number, not the technology going to the CF storage.

The VFSMark is a poor benchmark for measuring VFS statistics to begin with, but becomes even more useless if the VFS driver is exploiting flaws in the testing procedure.

I'm not saying that is bad. It is common-place in the industry (company optimizes for benchmark, benchmark publishers counter with more realistic benchmarks, repeat cycle).
Unfortunately the CLIE, all we have is VFSMark :-<

-------

So my conclusions:

Take the "50%+" speedup with a *big* grain of salt.

Don't trust VFSMark numbers anymore. With caching on, what little validity the VFSMark number had before has been thrown even further out the window.

Base your own conclusions on real measurements within your own real applications.

Of course you may or may not see much difference in actual use -- and probably many people don't really need the extra speed.

All IMHO.

Pre-emptive comment: If you want to ***** at me for saying something critical about "the driver", feel free to do so, but be sure to have some measurements to back it up.

Alistar

Well, I actually see a general speed increase.
With the 1.0 driver I could not use the auidoplayer in teh background becasue everything would else would run to slow.
With the 1.1 driver I can now run audioplayer in the background and play games with very little slowdown.
And this is with the cache turned off.

ayasin

I realize that many people don't like VFSMark, but I like it for the reason that instead of testing raw hardware speed it does things that a real application would do. These include creating fiels, writing to files, and reading from files. These things can benefit significantly from driver improvements showing real world benefits even though the raw transfer speed remained relatively constant. This is especially true if you use programs like PowerRun to load and unload programs.

__________________
Start by doing what's necessary, then what's possible and suddenly you are doing the impossible.
- St. Francis of Assisi

CliePet

> I realize that many people don't like VFSMark, but I like it for the reason that instead of testing raw hardware speed it does things that a real application would do....

True, but if you read my experience above, the VFS mark number improvements are not typical of real applications.

For my CLIE, only two of the tests are faster - and one disproportionalty so that it greatly skews the results.

These cases are not all that smart (doing things in small blocks) - not typical for what real applications do.

----
ayasin,
the 1.1 announcement mail mentions VFSMark improving from 112 to 180.

Can you please post the report summaries for this case (ie. what you get when you "Save Results") to provide more useful information ?

BTW: the VFS mark "average" is conceptually incorrect (ie. it is an average of speedup ratios - which is not the way time works in this universe. It should at least be an average of their reciprocals or something smarter...)

----

If you are seeing 50->60% speed improvements across the board then that would be very interesting.

If however, like me, you are seeing almost all the speedup in one case (DB Import), then my original statement stands
(ie. adding a cache makes this already dubious test all but completely useless)

fnm

I was very dissapointed. After installing the new driver I am experiencing too many craches and most application are performing weird. For an example. I cannot see my picture using AcidImage. I tried Slow mode, Slow mount, cache on, masquerade, etc. without success. I cannot hear my mp3s. The only workaround at this moment is unchecking the Cache on, but then again I don't see any improvement in speed. Does anyone experiencing problem with the cache on?

Thanks.

PS. By the way I went back to version 1.0 and everything is working fine again.

MikeLang

Me too. Using 1.1, I couldn't run AcidImage anymore. As soon as it started, 2 folders (Palm/Images/ClieAlbum, 1 on each card: CF and MS) were listed (just as 1.0 driver) but as soon as I clicked on the one on CF, my NX-60 hung and I had to do a soft reset with the push button. It worked with the 1.0 driver.

Bobbyn

For some reason the update has not been a good improvement... But the again "who said the new means better"???

I've just installed it, and after only one hour I had to soft reset my clie 6 times due to crashes on the Audio Player and the Clie Files...

That's it! It crashed again! I'm going back to 1.0.0...

I have an NX60... I didn't notice anything good out of the update...

__________________
bobbyn

greg elmassian

yep, the update made it so cliefiles won't show the msaudio directory with all the mp3 files, but it WILL play them!!!

back to 1.0

Greg

Quetsch

I also got a strange behaviour with Cache option activated.
Using Audio Player causes a serious error and a reset, stored images on the card look corrupted. This is only with a Transcend 30x 1 GB CF Card with activated cache. Without caching it works fine.
Another CF card, an extrememory 128 MB card works fine with activated cache...

MikeLang

I'm curious why nobody saw anything from the beta testing? Next time I think Eruware really needs to find somebody with a PhD in beta testing to do the beta testing... :-)

Yes, I'm going to back 1.0 too until there is something else new...

greg elmassian

it seems obvious that the beta testers were a small community... i thought there was a statement somewhere that there were 100 beta testers, but i can't believe it, no one tried a 1 gig sandisk? one of the largest manufacturers of cf..

if you read the various threads, you will see that the sandisk 1 gig has weird problems, like directories not showing under cliefiles, etc..

i recommend the eruware follow other software developers, to have a stable version on the web site, and a beta also, with the appropriate caveats...

greg

n2ifp

That was the number of testers&nbsp;initially, now it's only a handful. The biggest card I have is 256MB and the CF Driver for me works pretty good. Issues seem to develop with cards 512MB and larger.

73

__________________
Cheers!
Larry Winward
T|T3 + SE T610 + SE HBH-65 = BT Heaven!

n2ifp

There is nothing wrong with objective criticism, it's all in the approach

If one can conduct themselves such as yourself in a responsible professional manner with supporting facts&nbsp;is one thing. When someone starts off with this sux, that sux, and everyone sux, to express only anger, then it's another matter.

__________________
Cheers!
Larry Winward
T|T3 + SE T610 + SE HBH-65 = BT Heaven!

KenSong

I too am having problems if I have Cache checked. iSilo documants in CF will not scroll and refresh properly. In TomeRaider, index filter don't work. Again, my pdb is in CF. I am using an NCP 256 MB card.

In both instances, deactivating Cache resolves the problem. The only 2 feature I appreciate with this update is the HiRes icon and it not defaulting to CF launcher when mounting.

__________________
Kenny Song
Treo 680 + 8GB SDHC<LifeFlash+4GB SD< Palm<Clie<Prism<Palm<Newton<Sharp