SONY VGN-P and VPCP11 series sub-notebooks

These are cute little devices, which unfortunately contain some serious errors in hardware and software, mostly not documented by Maker SONY. Before diving into critics, let me expressly point out: The VGN-P / VPCP11 are pretty good. Otherwise they would not receive so much of my attention.

As of 2011-06, you find some general doc even in wikipedia, and a lot of reviews with different background in Internet. (Sorry for not listing resources here so far, will be added.)

I do not intent to duplicate wikipedia or any review. I wrote my own log while using the VGN-P, as I do for most devices. Later I extracted parts, translated to English, so it may be useful as reference for others, and me, in the very first place. I will add more research as it is done. Of course, there is even common info, to make this as complete as needed. Hints and additions are welcome!

I go deep into technical details and try to explain reasons. I use to mark (educated) guesses as such.

I use a German WinXP, because I did not got an English one at decision time. This makes it sometimes difficult to find the right English translation for a term used in GUI. I may use the German term, too, or exclusively.

Content (coarse listing)

Features, basic technical data

Differences, model identifiers

The machines differ in:

  1. RAM size (1 or 2 GB)
  2. CPU speed (1.33 to 2.13 GHz)
  3. Mass storage kind (rotating HDD / SSD) and size (60 to 256 GB)
  4. Small but important hardware details (still to be researched / documented)
  5. GSM/UMTS (/not) included, which Maker or Model
  6. pre-installed / shipped Windows (XP, Vista, Win7)
  7. VGA/LAN-Adapter box (/not) included
  8. Headphones (/not) included
  9. color, irrelevant for computational performance
  10. There are 2 generations, with mechanical differences, in Battery, too.
The numbers and letters do not match the same feature set in the different markets. Higher is not necessarily better.

See page devices for different ones sold, and for MODEL identifiers, too.

All second generation devices seem to start with VPCP11 identifiers.

Hardware in details: Harun's VGN-P21Z

USB 2.0 (High speed)

USB 2.0 is handled by a single controller inside US15W, connected to a chip-internal 8 port hub.

The US15W chip supports "USB client functionality" on Port 2. That means with the right software, and cable including 1 or 2 resistors, it should be possible to operate the VGN-P as a USB mass storage device connected to an other computer, or to interconnect (network) with another computer without any active box in-between.

USB 1.1 (Low and Full speed)

USB 1.1 is handled by 3 controllers inside US15W, each connected to an internal 2 port hub, then connected to same external lines (first 6 of 8) of chip as USB 2.0 hub ports.

Both sockets connected to same controller is bad design because it limits the possible data rate by enforcing bandwidth sharing. If you need USB 1.1 full speed full bandwidth for 2 devices, I suggest to put both on the same USB 2.0 capable external Hub. The Hub translates between full and high speed, and both devices are connected to same high speed controller on Host.

PS/2

Keyboard and Control Stick in center of keyboard are connected to US15W by PS/2 internally.

PCI Express

Ethernet (LAN) is via "Marvell Yukon 88E8057 PCI-E Gigabit Ethernet Controller" which only appears when the VGA/LAN-Adapter is plugged in. Connected to PCI socket/bus 1 as device 0. Without VGA/LAN-Adapter, this bus is taken down completely and you see a disturbing red cross in WinXP device manager located at PCI Express Root Port 1. This is no hardware fault.

"2.4 GHz Wireless LAN" is via "Atheros AR928X Wireless Network Adapter" connected to PCI socket/bus 2 as device 0, seems to stay independent from switch WIRELESS=ON and "2.4 GHz Wireless LAN" activated in Wireless Switch Utility.

Future and History

No future, seen by SONY! Last production was second quarter 2010, said SONY sales 2011-07-08. It was second generation, the VPCP11 models. I hope VGN-P and VCPC11 notebooks survive in the hands of users for a long time. I was told SONY laid off large number of development engineers in 2010, the P-series team, too.

SONY wants to sell new devices every 1 or 2 years, although this makes no sense performance-wise and has deadly environmental impact.

Future of interfaces

2011: VGA has been pushed into legacy corner for many years already. It arrives there. Presence is clearly declining, but no reason to worry so far. Technically seen this is ok, because analog RGB signals and digital displays are not the best connection. See here for more.

2011: USB 2.0 should still serve well because it is wide spread, despite USB 3.0 entering market. Do not rely on USB 3.0 products being backwards compatible! This feature is an extra, not naturally present in a product working as 3.0.

2011: Memory Stick Duo is widely available. 2010 announcement by Sony to switch from M.S. to SD-Card for future. There are microSD to M.S. Duo adapters sold for Play Station Portable (PSP). At least the two items I got work with VGN-P21Z and an 1 GByte SanDisk microSD card flawlessly, store data in the usual way, means microSD can be accessed as microSD and you see same data there like in the Memory Stick slot via adapter. These two adapters were branded "Kingston(R)" in white print. They contain a tiny PCB with 2 integrated circuits (chips) and some passive components. But see data rate is not best.

Future projects for VGN-P (and VCPC)

Clear statement from SONY sales: SONY does not support internal upgrades. I am used to work without support from original Maker, unfortunately. I share the ideas freely, because I would be happy if someone else brings an idea to life and sells a good product for a reasonable price. I have more good ideas than time in life for realisation. I do not promise to make any of the mentioned things.

HDMI/DVI output

The US15W outputs digital signals at SDVO interface. They are converted to VGA by an external chip. (At least for 1st Gen:) This chip has tiny pins. It can be replaced by a tiny adapter board and a SDVO to DVI or HDMI chip with reasonable effort. See here for missed chance.

Battery discharge while OFF fix

Technically best solution is to find root cause and fix it. Advantage, too: Reduction of Standby current (guess). But guess fix requires difficult soldering at Main board, only to be carried out by really trained people.

Every usual Li-Battery Pack has a MOSFET switch to cut connection as part of protective measures. If we could control this switch, we had a "poor man's solution", could add some tiny mechanical control switch to Battery. But the SONY Battery packs are not made for opening.

There is a work which can be done by every good mechanican: Removing keyboard, replacing starrflex cable to Battery, and milling a small hole into mechanical frame, into Battery bay, just under path of starrflex cable. The hole is for electroic components sitting on replacement starrflex cable, disconnecting plus power line there. Need only good idea how to control, may be via another flat cable, and where to place switch. Another "poor man's solution", because cutting power.

Boot from internal SD-Card and Memory Stick slots

CHAPTER TO BE WRITTEN / LINKED

Faster boot (turn on)

CHAPTER TO BE WRITTEN / LINKED

My SONY experience

I used a PCG-SR11K (Notebook 1.1 kg) for many years, until some important part died 2007 and resisted repair. I did already know: You can take their product as it is, but do not expect any help. Their support will not help you. Do not expect fixes for errors. Sadly, this is still true in 2010 and 2011. Please SONY, change the way you work to the better! I really prefer to write good words over bad ones.

Information is either already publicly available on web, or not available at all to support, according to support responses. What to complain about? Support does not care to research and find information on user's request, even if that info must obviously be available inside company SONY!

Printing "be like.no.other" on product packages is just meaningless.

Issues, faults (Hardware and Software)

User's manual (Benutzerhandbuch)

I studied both German and English variants. They are very well written, even for less experienced users, except for the fact they miss a lot of basic specifications. A lot of hese specs are just nowhere, others are in the "Marketing Specifications". I mean location of antennae, operating temperature, current consumption, current supply of interfaces, ... Asking Support did not provide any additional information.

Power consumption while completely OFF

Done 2009-12-17, repeated later multiple times as part of regular usage.

It draws 10% of full standard Battery capacity within 24 hours of OFF time. This translates to 8.75 mA drawn from Battery (computed, not measured). After 1 week, the Battery is close to being empty, observed. This is not mentioned in original doc. No numbers given by Support on intense requests, only statement "is within specs", but no way to get these specs.

Current measured from power supply to computer, for reference:

The difference is 7.2 mA at 10.5 V (75.6 mW) and matches the observed Battery discharge of 8.75 mA at 7.4 V (64.75 mW) closely, especially considering there is a switching voltage converter involved with its losses.

I consider this a bad design error. Or be a bad guy and call it intentional wear of Battery, to sell more. Have not seen any Notebook before eating that much for nothing. Others last multiple month on Battery while OFF. Even the PCG-SR11K did so. That is why I had no reason to check for this before purchasing VGN-P.

In the mean time I learned by reading and listening, that other current SONY notebooks share the same design error.

SONY support suggested to take out the Battery. This is a bad solution due to extra handling, extra piece to carry or to keep fixed in a pocket, wear of electrical and mechanical connection. It will discharge the tiny clock backup cell.

I tried to turn everything off which could have an impact on OFF current, including the Power options / Lid actions, but got no improvement.

Correction by user - BAD work arounds only

On the road, take out Battery, but see above. At fixed location, keep connected to Mains all the time. The rise of electrical power bill is negligible, but every connected device rises risk of fire. There is extra wear of connector.

Pre-installed operating system Vista

Windows Vista is one of the worst Windows Microsoft made so far. It appears to be slow, contains bad concepts in internal design. SONY was forced to ship Vista with the 2 GB RAM machines by MS licensing restrictions.

Forced? I guess: If anybody has the power to influence Microsoft, it is a large company like Sony. But Sony does not care to even try. Lowest OS license costs are of larger concern than user's productivity. Anyway, I prefer 2 GB RAM and downgrading Win myself over being limited to 1 GB RAM only.

Correction by user: Install WinXP

WinXP optimized to run only required components works much better than any optimized Vista. I tried that 2008 / 2009 on Samsung Q45 and decided for me: Vista belongs into the trash, not onto a computer. I use WinXP Prof with current updates, because it needs less resources than Win7, and is more open (less digital signature required).

Correction by user: Install Win7

Reportedly Win7 works well on VGN-P, too. I believe these statements after testing Win7, optimized to run only required components, classical appearance, on a more powerful Notebook Samsung Q45. I my test case, Win7 occupied 150 MByte RAM more than WinXP, but was still below 500 MByte.

Instant mode

In this context the term "instant" means to me a single-digit number of seconds, a time not perceived as disturbing wait, not breaking the flow of work.

The "instant mode" starts much too slow (VGN-P21Z original cond.), needs 21 s to be operational, and does not even offer basic organizer functions like address book, calculator, calendar, small notes.

As I suggested to SONY by email 2010-beginning:

These functions can be implemented within less than 64 KByte of x86 machine code. It has been done already more than 20 years ago, like in Atari Portfolio (developed by DIP), and other tools of that time. This code should be put into ROM, like the BIOS, or may even be loaded from harddisk, and could give a true "instant on" experience, together with an optimized BIOS.

I mean less than 3 seconds from pressing the (Instant)On-button to accessing contacts or notes. Only that way the VGN-P21Z may replace an 13 year old handheld like the Casio A-20.

I know the A-20 keeps its 8 MByte main memory powered all the time, nevertheless sleeping more than 1 month on its old 750 mAh Battery. Anyway, 3 seconds to operation can be done from cold off nowadays.

Correction by user - work around

See here for some kind of work-around.

The display and Windows

The text on screen is difficult to read due to its small size. User simply cannot adjust the dots per inch setting to make text larger, because even the lowest useful magnification factor of 1.5 leaves only 512 effective vertical pixels for Windows (XP, Vista, whatever), which is below the 600 pixels (800*600) needed to display the usual dialog boxes and windows. Anyway, factor 2.0 would be even better.

Solution: Add a small piece of software, which implements a panning virtual desktop plus magnification. The pixels presented to Windows (framebuffer) could be 800*768, or even 1600*768, but every framebuffer pixel would occupy 4 pixels on the display (a 2*2 square). The display would track the mouse cursor, showing the interesting part of the frame buffer.

This is no magic. It is done in other products. Later seen: The (very ill) Intel graphics driver can do panning, but it allows it only while mirroring a larger external display. Having more pixels for Windows that way, user can adjust dpi to get larger text.

The magnifier of WinXP (and others) is no solution due to its design.

GlobeTrotter Connect (control software)

Observed and researched 2011 under WinXP, G.T. versions 3.0.0.924 and 3.1.0.1107. After some hours of G.T. running, every screen operation slows down dramatically. Then draw errors occur.

G.T. collects more and more "GDI objects" over time, accelerated by opening and closing the "Einstellungen..." area of it, as shown by TopToBottomNT. The collected objects are Drawing Contexts and Brushes. To much objects slows down all of Windows, leads to errors later.

Correction by user - work around

WinXP frees these objects on G.T. termination. G.T. will break connection when terminated the usual way, without any good reason, but it warns user and gives choice not to terminate. Experience shows it is safe to kill G.T. by Task Manager. This procedure keeps connection. Start G.T. again when you like to see some transfer numbers or want to close connection. Longer experience indicates: Transmission may suddenly stop while G.T. is not running, no packets in and out any more. Start G.T. and transmission continues. Kill G.T. again. This may happen after less than 1 hour, or even more than 4, it seems. Seems to happen sooner on GSM/EDGE connection than UMTS/HSDPA.

Bug reported, more ...

I reported bug directly at support.option.com with Maker of that component, because could not expect any help from SONY support after experience with Battery discharge issue.

When do you fix Globetrotter Connect? After some hours of operation, all screen drawing becomes extremly slow, in all applications, e.g.G.C. "Einstellungen einblenden" / "Einstellungen ausblenden" (German, English "Show settings" / "Hide Settings" needs 1 to 2 seconds, then massive drawing errors on screen. Same with Globetrotter Connect 3.0.0. Reason: G.C. allocates more and more "GDI objects", especially "Drawing Context" (I had more than 2000), and if Settings is shown, a lot of "Brushes" (see less than Drawing Context), too. Happens even when window is hidden, or closed. How to see this? Use e.g. "TopToBottomNT", freeware from http://www.smidgeonsoft.com.
My work around: Make connection, kill G.C. by Task-Manager, because I cannot "Exit" it cleanly without it closing the connection. Later starting it, it sees the established connection correctly. (So there is no reason it wants to close connection during Exit.)
Show / Hide settings, clicked multiple times, seems to accelerate the process of GDI Object allocation.
Happy Debugging and fixing, hope to get a positive response soon.
Best Regards, Harun Scheutzow

option response by email 2011-04-14:
Thanks for using Option products. For all module devices, our resellers and distributors are responsible for the first line support for end users. Please refer your support questions to the supplier from whom you have purchased your device. Thanks for your understanding.
The GlobeTrotter Connect is end of life. So no fixing will happen anymore. The latest revision supporting your product is GlobeTrotter® Connect 3.1 for Modules revision 3.1d0-1107_P

2011-06-19 I did ask for G.T. source code to fix the problem myself. Let's see. 2011-07-12 Got no response so far, little hope.

GlobeTrotter hardware GSM/UMTS

Observed 2011-06 and before. The MO40x GSM/UMTS thing disappears with the typical sound of USB device connection and reappears after less than 3 s, multiple times a day, but significantly less frequently than once an hour. This of course breaks a connection, but happens without established connection, too. Have to enter PIN again after such event.

So far no root cause known.

2011-06-26: This seems to happen with WIRELESS=OFF, too. At least the sound is audible, but I had no chance to check what is going on at that time.

GSM/UMTS modem needs Battery

The GSM/UMTS modem does just not show up at USB without Battery installed. I dislike this feature, because wanted to keep Battery separate while using VGN-P21Z for extended periods on desk, to save Battery life time by keeping it cooler than it is in a running Notebook.

GUESS: The GlobeTrotter GSM/UMTS modem draws peak currents which cannot be handled by Mains power adapter alone, while VGN-P21Z fully powered.

Sub-sized right Shift key and other keys

Applies to 1st generation

Every "sane" keyboard has Shift keys same size as letter keys, usually even larger. SONY decided to be different for no good reason. Sub-sizing cursor keys (arrows) was a bad decision, too. Result is power user hits the cursor-up key together with Shift, or has to exercise excessive care while typing. Probably no problem for geeks because writing all lowercase anyway.

This illness could easily have been avoided by one of:

Some fixing considerations

All scan codes can be remapped by changing Windows registry. The Fn keys are probably handled at a lower level (still unresearched where exactly, e.g. in keyboard microcontroller, in x86 SM-BIOS).

There is a single key very seldom used: CapsLock. But relocating the up-arrow to there would be very unusual.

Keyboard layout is personal taste and experience, what you are used to have, what you hack on elsewhere. I returned a Psion 5 to store because it missed 2 key colums at right edge, containing the German Umlauts and 3 important other keys. (Display was too dark, too.)

The final fix

... has still to be found!

2nd generation

Looks like SONY listened to user feedback. But did not solve the problem. Made it worse, my opinion. Sub-sizing the 3 keys left from right Shift has been bad idea. Using 2 different sub-sizes is very ill. The super-sized 1 key of 1st gen was not good, but the sub-sized 1 of 2nd gen is even worse.

The pointing stick

Extremly bad handling. Time consuming to hit anything on screen by using it to control cursor. Even worse, while typing and coming too close to it, it moves cursor and even does a click, moving text cursor, too. Result is my typing goes to unwanted place.

Solution during computer development should have been one of (listing reflects personal preference):

Correction by user

If you do not use the stick click feature, disable it in Control Panel of Mouse, on Stick card. The sensivity can be adjusted, too, but I did not try that.

For desktop usage, connect your favorite mouse. On the road, may try some innovative pointing devices connected by Bluetooth.

Major USB slowdown

Observed 2011-06 and all the time before, WinXP. Observed same with a Nokia 6151 mobile phone serving as GSM/UMTS modem connected via USB to Samsung Q45 Notebook running WinXP SP3 already in 2008 or 2009.

This communication USB device does only need to be present to cause slowdown, dramatic reduction of data rates to/from USB mass storage devices connected to same USB controller. No matter whether device is used. I guess this is an error in USB bandwidth allocation by WinXP.

Considering all aspects, there is no need for braking other USB transfers that much. There is no USB traffic without network connection, or any other related activity like establishing or stopping a connection.

Sorry, had no time to research deeper and find fix. Suggest WIRELESS=OFF when you need USB speed.

NOTE: The internal camera does not cause such braking. Only minor slowdown when in use.

RESEARCHING ISSUE

According unreliable (crashes, shows some other parameters wrong) application usbspy.exe V2.2 all 11 end points of MO040x are bulk with bInterval=32. On USB mass storage flash sticks I have seen so far 0, 1, 255. Did some work, no conclusion so far, but bInterval is not the problem. Still in progress ...

2011-06-27 using Heavy Equipment and some educated guessing. Looks like I found one more fundamental design flaw inside Windows and the drivers supplied by it and third parties. If I had sources, it would be easy to fix, but without sources, as the situation is, it will be much harder. Simply said, a lot of unnecessary polling takes place, eating bandwidth badly for no result. bInterval is not observed for Bulk in endpoints, instead they are polled as often as possible, in every single microframe, means every 125 us, if an in-transfer is made pending by driver software inside Host. I REALLY HATE SUCH EXPERIENCES. DAMN.

SanDisk 4 GByte Memory Stick SDMSPDH-004G-U46: INCOMPATIBLE!

2011-06-18: What appeared to be a hardware-destructive issue resolved to be non-destructive, but most likely be a timing incompatibility between SanDisk M.S (too slow) and Sony SD/M.S.-to-USB-controller, after investing more than 20 hours and even 50 EUR for equipment transport, disassembling a VGN-P21Z and using really heavy measuring equipment: an old but good HP16500B with osci and logic analyzer modules. I stopped research before reaching the final stage of capturing whole M.S. messages, because even that knowledge would not allow me to get what I wanted in the first place, boot from SD-Card or M.S., and not even reliable access to this M.S. in internal slot.

The result for easy digestion: Depending on previous usage, what has been written or deleted, the SDMSPDH-004G-U46 is recognized by VGN-P21Z in its internal M.S. slot or it is not recognized, but brakes and blocks everything badly as long as plugged in. It may be recognized after a few minutes. Writing to SDMSPDH-004G-U46 fails at some point with error message and creates the braking and blocking experience.

This happened with 2 different SDMSPDH-004G-U46 and 2 different VGN-P21Z. Operation of SDMSPDH-004G-U46 in external USB card reader Conrad brand was reliable.

SanDisk technical support was very responsive on web message and email, but first not hitting topic, later refusing to give a vital parameter for operation of an electrical product, the (maximum) operating current it draws. This parameter is not in SanDisk's data sheets. SanDisk support is by custhelp.com, which I see as a bad sign. They are very well to answer the standard question "Is this a genuine SanDisk product?".

Sony support did not respond at all within a whole week! Responded "resend please" on a repeated request, but were of no help then.

Later I found out myself: SONY could fix it, or ask SMSC to do so, see Internal Card Reader.

Not for full usage of its computational power

There is fine print statement on package, explaining a star next to CPU frequency of 1.33 GHz: "Leistungseinschränkungen bei bestimmten Bedingungen möglich" (translated: "Performance limitations possible under certain conditions)"). The user is warned.

The CPU is braked down to 800 MHz quite often and for long periods of time. CPU core is at 50 to 55 degree celsius, much below the permitted Tjmax of 90. But we have to look at the whole system which is quite warm. This seems to happen much more frequently when using an external display 1920*1080*16bit*60Hz. Internal display is off. I reduced already from 32 to 16 bit/pixel without manually observable influence. Ethernet is unused but appears in device tree as soon as Adapter connected.

I operate the VGN-P21Z on a table with 1 cm air gap under it, left and right edge resting for 2 cm width on a book. No fan added so far.

WinXP does not tell you what is going on. (German) Systemsteuerung / System displays a 800 MHz CPU clock in second line only when "CPU-Steuerung=Akku-Lebensdauer" (may be "CPU-Control=Accu-live") is choosen in Power control panels, but definitely not when CPU is braked due to temperature. Need to use some external tool like "Core Temp", which shows temperature, frequency and even CPU load.

I do not demand removal of braking, because this surely had negative impact on life time. I want to note the fact. I want to complain about silent operation - a popup had been wrong, but a place for the user to look for system state should have been provided by maker Sony.

Could it be made better in this size, without fan? Surely, by having a metallic bottom side as heat spreader and good heat conductor, the CPU (and may be other chips) connected to it. If the user wants to avoid hot legs or fingers, he had either to add some insulation, which would cause automatic breaking, or to adjust some software setting to keep CPU and system clocks low.

I guess a more expensive CPU with higher max frequency is of little value, because it will be braked down even more.

See here for more.

VGA/LAN-Adapter (Display/LAN-Adapter)

The cable leaves this Adapter box in a very ill-designed way. The design is good for transportation only, but badly hinders usage of the right USB port. It could have been made better easily.

Considering the short mains cable supplied, the long wire being between Mains adapter and Notebook, I do not believe anybody connects VGA/LAN-Adapter to Mains adapter during usage. As far as I understand SONY doc, the connection is for storage and transportation.

Correction by user - improving it myself

I opened the box. There are the 4 screws you see at bottom, the 4th if cable is not in storage position but pulled out. There are 3 more screws inside. It is easy to rework it mechanically to have the cable leaving the box either in top or bottom half, going in the opposite direction of the VGA cable. Pay attention to fix the cable firmly, to avoid stress on the non-shielded thin wires at the connector to board.

I reworked mine so the cable leaves in bottom half just next to the post inside used for the screw fixing the cable shield connection. One of the posts used for fixing the cable in old position by a shell I had to remove. The result looks like manufactured that way.

Final critics

The whole SONY box is stylish design, but bad in volume and weight. All of bottom half is only to store the cable, which is has no need for this protection. 15 mm length is only for connection to Mains adapter. Best solution had been to produce a Y-cable. For the VGA branch it is straight forward, because no electronic components are involved. The magnetics for Ethernet could be molded into Ethernet socket or put into computer (there is still room inside). Such a cable does not need any extra protection during transport.

The VGA/LAN-Adapter of earlier VGN-UX series clearly shows SONY was able to produce a lean and not port blocking device in the past. The connector to computer looks the same, judgeing from a distant photo. It may work for VGN-P, too, NOT TRIED SO FAR. VGN-UX is 10 / 100 MBit only, which might limit function with VGN-P to 100 MBit, if it works at all. VGA and LAN socket are on "wrong" side for VGN-P, but this does not matter that much because cable is long enough on its own to leave room at USB socket, and furthermore can easily be bend by 45 to 90 degree.

Final critics: Opportunity missed: DVI/HDMI

VGN-P's connector should not carry analog VGA signals as it does, but digital SDVO signals instead. The SDVO-VGA-converter chip inside VGN-P should be in this Adapter here. This had been allowed DVI or HDMI output as an external accessory, had reduced VGN-P production costs a little, too. Very bad design decision by SONY. May be they intended to re-use an existing Adapter in first place, like from VGN-UX series. See for ideas.

Final critics: Ethernet (and PCI express)

Why is the Ethernet controller (interface from PCI express to Ethernet) placed inside VGN-P? It generates significant heat while in use, see ####. The PCI express signals could have been brought out at VGN-P's port instead of Ethernet. VGN-P's production cost had been reduced. A major source of heat had been removed from VGN-P, placed into the Adapter. A major opportunity for extensions (pheripherals) had been opened by having a 1 lane PCI express port, anything from external graphics cards, over various kinds of mass storage interfaces, and even any kind of bridges to other busses.

External boot (USB)

It seems sometimes the external USB device is not recognized at all. This causes trouble, and needs special consideration and re-try while experimenting, wanting reliable statements. To proceed, power-cycle either the USB device by temporarily unplugging, or the whole system. Ctrl+Alt+Del is not sufficient.

BIOS R2073U3 does not try to boot from internal slots for Memory Stick or SD-Card. I tried it 2011-06-19 in all thinkable combinations without any success. This means it treats this USB mass storage device specially, causing inconvenience for users, without any good reason. See here for considered improvement.

You have to enable external USB boot in BIOS setup and set the "Boot Priority" as desired. BIOS R2073U3 used for latest tries 2011-06-19. R2060U3 may have more issues with non-present media, unsure.

The left USB is consulted first, then the right. All USB mass storage devices are checked. The first bootable found is taken. Any first sector containing last 2 bytes 0x55 0xAA is considered valid for booting, say old PC docs. Garbage inside may easily crash (hang) computer without giving any message. I tried HDDs, a lot of USB flash sticks, multi slot card readers, even a 1.44 inch floppy TEAC FD-05PU, up to 3 levels of USB hubs inserted. CD/DVD have been used by others. A device with a non-bootable medium inside is skipped, e.g. a USB flash stick containing 0-bytes.

A device without a medium - this is still not completely clear. Experience is it is skipped completely. For multi slot devices like typical card readers the slots are scanned beginning with 1st (by USB view). On the first slot without medium inside the whole device is skipped. I tried this with Conrad card reader, SD-card in or not in 1st slot and CompactFlash in 2nd slot. TEAC floppy was skipped without disk inside.

Slot order by USB view is not related to slot mechanical position on device. It uses to be same as order of drive letters in Windows unless user has remapped them.

Bad visibility of card activity LED

The heading says it. In normal sitting position (like shown in user's manual!) visibility of LEDs for WIRELESS, Power, Battery state, HDD activity is still sufficient. The card activity LED is invisible. Could easily have been made better.

Note: Having the blinking HDD LED in a not too prominent position like it is is very well, because a prominently positioned LED would disturb by attracting attention.

No reset (key)

This complaint applies to a lot of modern PC. While this hard reset is a last resort, seldom needed in normal usage, it nevertheless avoids a power cycle, which is stress for all components.

Holding power key for 5 s forces power off. With little extra effort, a 2.5 s hold then release could have forced a hard reset. Or better for the impatient experimenter: "instant on" key plus short touch of power key.

VGA signal quality not high

The VGA signal quality is not low, but it is far from being high. It is reasonable to work with a 1920*1080*60Hz external display, but there are small shadows trailing vertical edges. This is a typical sign of impedance mismatch and reflexions somewhere on the analog transmission line.

Shielding of VGA cable is much less important than impedance match for R,G,B lines is my experience after more than 5 years in Graphics Card business, and having a sharp eye on all these things anyway. SONY did a poor job here. Inside VGN-P, the VGA signals run over starrflex PCB, which are probably anything but not 75 Ohm, long distance from Comm board, where they are generated, to socket. The external cable from VGN-P to adapter box is for sure anything but not controlled 75 Ohm.

The external cable could be improved with reasonable effort, but the even longer starrflex internally would be more expensive to make, so I do not try and have to take the shadows as they are. Anyway, before modifying cables, had to check the R,G,B filter network at chip outputs, because it can not only act as a low pass filter, but cause impedance mismatch, too, and measure starrflex cables. See for some details.

Dirty display, by keyboard

All dirt from keys (finger fat) stamps onto display. In the long term, in dirtier environments with abrasive dust, the display may even become scratched. It is due to insufficient distance between the display surface and the key's top surfaces.

Correction by user - work around

The user may insert a thin and clean sheet of paper or tissue to fix this. But it is very inconvenient to keep the sheet with you and handle it during mobile usage. Unless the sheet is very thin, it must not cover the small posts, especially the one at display lower edge center position. Pay attention not to use an organic reactive material, like some plastics, or printed tissue, which may solve display surface by emitting organic solvents (I guess display surface is Plastics). It may work well to glue flexible tissue to top-edge of display, so it can be flapped in protective position, or flapped out for work, giving a nice flag. Natural tissue like cotton needs regular washing, paper needs exchange, because fat penetrates it over time.

Considerations

A lot of notebooks made 1998 to 2011 share this problem, even much larger ones. It is all the same bad design. The small rubberish posts often present for distance are no solution, if not accompanied by other means. VGN-P has 3 of these at the base (front of keyboard), 3 at display (left, right, bottom).

The Maker could easily correct it by adding 1 or 2 mm distance, and making the display (especially cover frame) sufficiently stiff, by using these mm in the frame. I am sure most users will accept a device 2 mm ticker and a few gramm heavier without complaints.

It has been made better: The year 1997 WinCE handheld Casio A-20G does not have this problem. It is even a little smaller in ground area than VGN-P. Other devices from same past years were made as well.

VAIO Event Service does bad things

2011-06-25 I researched one more time the issue: Why is WinXP not truely silent when it should be idle, because there is no work to do? One of the answers follows. It is not WinXP's fault.

"VAIO Event Service" handles the function keys for e.g. adjusting display brightness. Means you cannot kill it without loosing something. My very target was VESMgr.exe V2.3.0.4130 (file info), Version.txt in folder says "Version=2.9.01.03180", as installed during downgrade to WinXP.

Every 15 s, it tries to open for write (CreateFile()) the "C:\log\VESSuPerform.txt", but usually receives a "PATH NOT FOUND" error on my computer. If this folder exists on your computer, you will get the file there, and see it does not contain anything useful, only lines like "2011:6:26 14:33:48 00000968 VES_SERVICE_MSG_PM". The complete path + file name appears as ASCII in "VESSuPerform.dll" only. Most probably used from there, because elsewhere not even partial traces of file name.

2011-07-02 update: This seems not to happen today. Something has changed. No NAS? CPU cold today because Ethernet off? Only external display since boot? No Battery charging? I do not know. Task-Manager open did nothing, Energieoptionen open does nothing. Heating by my gemul8r.exe alone no success, Ethernet controller Activate, now 58 celius and now it brakes to 800 MHz. But still no log at all, have C:\log\ existing. Later more frequent clock changes, but no log. Even Camera Activate again, no log. Unplugging Mains adapter seems to cause first entry. But there need to be more to cause all 15 s entry, could reproduce it, was 12 s now, later between 12 and 18 s, is it communication with tray icon for Battery? Once started, it seems to continue, even while Battery charging and no icon in tray. But interval now 30 s. No more log entries after Battery LED went off, charging completed. Mains adapter unplugged again, 2 entries in distance 30 s, but then no more, although running on Battery. May depend on some info update, still not known which parameter change causes this. On Mains adapter (un)plug, may even get 2 entries in same second.

Interesting: During startup, it looks for file "C:\BCF.txt".

TO DO: Either patch it to kill this activity completely, or redirect to "nul:" or a RAM-Disk. But currently I do not put more efforts into it, because it is still a little unclear when a log entry is tried. You may either Terminate the Service or kill VESMgr.exe in Task-Manager, copy VESUPerform.dll, find the ASCII string (see above) in it, replace it by "nul:" followed by a 0-byte, and Start the Service, or restart your computer.So far I do not recommend this.

SONY marketing specs partially wrong (SATA or PATA)

The question is the interface to HDD / SDD, old style P-ATA or the more future-safe S-ATA (Serial ATA), and how it is done.

VGN-P530H

The VGN-P530H is advertised in US (still current html page and pdf file as of 2011-07-01) as having Serial ATA interface (SATA) for its 60 GB HDD. This is wrong. I had a P530H in my own hands: HDD is Toshiba MK6025GAL, same P-ATA as the MK8025GAL, and of course no Serial ATA chip seen by software anywhere in system.

VGN-P799L

2011-07-03: Got info from an US computer dealer, who has access to the top-of-line model VGN-P799L equipped with 256 GB SSD. He writes it is P-ATA. What does SONY US store web SPECIFICATIONS tell me same day, in two non-adjacent chapters, but on same page?:

Solid State Drive

Storage

Clearly self-contradictory info on a single page. The mksp pdf mentions only Serial ATA, is completely wrong, but consistent.

Or call it Marketing. Advertising by buzzwords, even if technically wrong.

VGN-P15G (some EXPLANATION)

The asian model with 64 GB SSD. Again, web claims Serial ATA. Marketing of SONY Hong Kong responded "the model VGN-P15G were preinstalled Serial ATA SSD memory". After seeing a disassembly picture on Japanese page, this appears to be true. There is a SONY-made PATA (Host) to SATA (SSD) adapter in the cable (replacing most of cable length) from Host to SSD visible. This adapter might collide due to height with other than original SSD.

That means this model uses the same PATA interface of US15W chip as others. There is a SATA driver on SONY HK support web, but it is for Intel Motherboard chipsets which have SATA included and do not work with Atom Z5xx CPU, and clearly are not used in VGN-P. I guess it is a mis-placement.

Experiments and results

Standby enter and exit time

Wakeup from any key, including power key, to password request is short, a few seconds. 2011-06-25 WinXP SP with 3 open apps (Internet Explorer 8 multiple windows with local content, UltraEdit V15.20, Adobe Reader 10.0.1), external display, LAN=deactivaed, WIRELESS=ON WWAN: 8.5 s, on next try only 7.5 s. From password to usable desktop - no delay.

2009-12-25: Time to enter standby is more than 5 s. 2011-06-25: Experienced 8 s, on next try only 3.5 s. Does probably depend a lot on state of system, whether code for function already loaded or not.

Standby power consumption

2009-12-25: Discharges Battery from 100% to 96% within 2 h 15 min.

2009-12-26: Discharges Battery from 100% to 73% within 14 h 05 min. This means approx 40.5 mA current.

Current and Power consumption

Done 2010-01-14. Current measured from power supply to computer by Multimeter Voltcraft VC220 in 200 mA or 20 A range.

The OFF computer emits a little peeep sound, means a switching voltage converter is active, if Mains adapters is connected or Battery is inserted. There should be no real need fro this.

2011-06-22..23: A lot of measurements, currently only posted in the Raw Logs from Lab, will be re-worked in near future, to be easier to understand and of general use to save power where possible. I started already writing the conclusions, without giving exact numbers, for good reason, see log.

Why save power?

The answer is obvious while running on Battery, isn't it? While running on Mains saving power means a cooler VGN-P and less likely slowing CPU due to high temperature.

Worth to be turned off to save power

A good notebook should turn off all non-needed power eaters without user intervention. But VGN-P running WinXP SP3 does not. You can help manually, by using "Geräte-Manager" (Device Manager?) to "deactivate" individual devices. If you need these devices again, you have to "activate" them.

If you are using an external display and the VGA/LAN-adapter, but no Ethernet, deactivate "Marvell Yukon ..." Ethernet controller. This helps a lot.

As long as you do not have an active GSM/UMTS (WWAN) connection, either move the WIRELESS switch to OFF position (will disable WLAN and Bluetooth, too), or use the Wireless switch panel to turn WWAN off.

While running on Battery, reducing display brightness is useful, unless it reduces your productivity. On Mains, reduction is no use, because the heat is dissipated mainly in display, where it does not disturb.

No use turning off

I did not observe any significant reduction in current consumption after "deactivate" of the following unused devices: Camera, Atheros WLAN, Bluetooth.

Camera "deactivate" is still good to get a little more USB bandwidth for USB mass storage devices.

Feeling the temperature

While operating it, WWAN UMTS/HSDPA connected, with 1 cm air below, the left 1/3 (CPU board) and center 1/3 (Comm board) feel warm to hot, but the right 1/3 (HDD) feels cooler than the other 2. Good to know the HDD does not get full heat.

Data rate / speed of Ethernet

Ethernet speed test between 2 VGN-P21Z, 3 m CAT5 shielded cable, RAM-loaded Linux grml_2011.05.iso from USB Stick. IP 192.168.151.144 and .166, mask 255.255.255.0. Idea is to use netcat for TCP traffic, predefined amount of data supplied by dd.

Used after boot the offered "n" for "Configure network (grml-network)". All set ok, said "enable at boot", too, then Cancel/Cancel/Quit because this is only way to leave. ping says unrechable, ifconfig does not show eth0 but lo only. "ifup eth0" says "... already configured". "ifconfig eth0" does not report any IP settings! Manually "ifconfig eth0 192.168.151.166" and ifconfig reports it, as intended, OK.

ping to own IP ok, 0.09x ms usually, to partner IP ok, too, 0.7xx ms usually.

The netcat man page is "nc6", but switch "-4" shall do it anyway. "netcat --version" clearly shows it is with V6 support.

Gave cmd on .166 then .144, after report of dd hit Ctrl+C on .144.
ATTENTION: Give commands with very short time between only, otherwise report of .166 dd will include a significant wait time, and look slower than real.

Single direction only test (.166 receives as long as you do not type anything there):
.166: netcat -4 -l -n -p 7788 -s 192.168.151.166 >/dev/null
.144: dd if=/dev/zero bs=1M count=1000 | netcat -4 -n 192.168.151.166 7788
dd reports approx 55 MB/s. Before Ctrl+C can send text lines into opposite direction. Repeat using "count=10000" got 65 MB/s (yes, 10 more), repeated got 57 MB/s (the number of elapsed seconds matches the speed), "bs=10M count=1000" got 57 MB/s

Does CPU clock change in-between? May be. Console 2 and higher have a bottom status line, showing on .166: 1066 / 1333 and on .144: 800 / 1333. Seen .144: 1333 / 1333 for a short time _after_ transfer finished already.
ATTENTION: Atom Z520 is 1 core 2 threads run at same frequency. Guess the correct frequency value is 1st one.

The shell reported CPU usages well above 50, more close to 100 % on both sides, probably due to small default buffers (8 KByte for TCP) of netcat.

Trying "netcat --buffer-size=1048576 ..." on both sides, .144: "dd bs=1M count=10000 ...". Got 73.0 MB/s, repeated 72.5 MB/s, 73.8 MB/s.

Bidirectional test with defined volume in both directions, large netcat buffers:
.166: dd if=/dev/zero bs=1M count=10000 | netcat -4 -l -n --buffer-size=1048576 -p 7788 -s 192.168.151.166 >/dev/null
.144: dd if=/dev/zero bs=1M count=10000 | netcat -4 -n --buffer-size=1048576 192.168.151.166 7788 >/dev/null
Got 41.3 + 42.6 MB/s, repeated 41.4 + 41.3 MB/s. For "bs=1000" 51.3 + 39.8 MB/s, indicating there is significant buffering, or just skew of CPU time?

During all these tests, console switching by Alt+<num> showed quite slow redraw. .166 is my 1st VGN-P21Z with Accu installed, shows 800 MHz CPU most of the time, although has better cooling than the other. .144 is my 2nd VGN-P21Z, shows most 1333 MHz, later 800, too.

Ethernet speed summary

The GBit Ethernet is indeed the fastest digital interface VGN-P21Z has. Using ftp to transfer files should achieve same 72 MB/s as wget, if ftp client and server are well implemented, and the mass storage is not the limit.

Data rates of USB mass storage and internal HDD

IMPORTANT: Switch WIRELESS=OFF for WinXP. Does not matter for Linux.

Done 2011-06-09. Compared with a Samsung Q45, it seems to be slower. Q45 is the more powerful hardware, but I see no reason VGN-P21Z cannot fully use USB 2.0 speed. My main data collection is in file USB-Mem-Sticks_uam.rtf.

Always consider where you write, and what. Writing 0-bytes or any regular pattern may lead to biased results, if the storage device (or file system on top of it) does some compression or optimisation, which I would expect on Flash Memory (like USB-Sticks or SSD), but have not seen so far.

For mechanical drives, speed may differ depending on cylinders you hit.

If unsure whether any cache for mass storage is active, make test size large so it exceeds any possible cache (RAM size) by at least factor 4.

WinXP, h2testw.exe

h2testw.exe published by magazine c't is designed as integrity test, but serves well as linear large block write and read speed test, too. It works on the file system. It writes and reads in blocks of 2^20 Byte. Maximum size per file is 2^30 Byte (1 GiB). It writes non-0-data.

Accessing a fast USB stick FAT32 formatted. Read 20.x MByte/s (Q45 29.8). Write 11.8 MByte/s (Q45 20.x). Accessing USB connected HDD WesternDigital myBook 500 GB. Write 12.3 (HDD rattles intensively, is FAT frequently re-written?). Read 20.5 not better than USB-Stick. NTFS much less noise (still "Fuer schnelles Entfernen optimieren" Write 13.6 (little more). Read 18.9 (little less).

On internal HDD (cache on) Write 19.4 Read 20.4, testing 2048 MByte on NTFS.

Linux grml, dd

"dd" is a nice tool to transfer data in blocks of defined size. It can easily be used for mass storage linear access tests, either going via any file system mounted, or directly to the device. To test reading, "of=/dev/null" is a safe and fast destination on Linux. To test writing, the only reliably fast source is "if=/dev/zero", but it provides 0-bytes only. "if=/dev/random" is a bad idea, because used to be extremly slow. Either double-check with another, slower, source that there is no optimisation for 0-byte writes, or prepare a RAM disk as a fast source with e.g. a copy of sectors from your computers HDD, or part of an *.iso image. Do not forget to speed-test the RAM disk!

The damn 10^3 versus 2^10 issue. dd bs=1M means 2^20, MB would mean 10^6. The reports of dd are in MB, means 10^6 byte/s.

grml_2011.05.iso "dd bs=1M ..." is significantly faster than grml_2009.10.iso. Accessing USB connected HDD WesternDigital myBook 500 GB raw data (/dev/sdX with seek or skip, to keep existing FAT file system intact without using it). Read 29.x MB/s (Q45 34.x MB/s). Write 16.5 MB/s. (Q45 writing Hama 32 GB USB-Stick already 28 MB/s). These are the fastest USB rates I could get. All USB-Sticks were slower on VGN-P21Z.

Internal HDD Read 25.9 MB/s. Write not tested due to valuable data on it.

Summary

Results mean the VGN-P21Z cannot serve for USB mass storage speed measurements, because it is the limit itself.

Researching slowness

The RAM bandwidth available cannot be the cause: GBit Ethernet via PCI Express does >= 70 MByte/s in both directions. Dramatically reducing bandwidth demand of display refresh had no effect. Cause may be FIFO sizes, scheduling, setup, software, using only small part of expected EHCI capabilities due to chip errors, hardware limits. Linux and WinXP show same behavior. Strictly, this means nothing.

2011-06-27 My good old HP16500B + 16557 module directly attached to USB lines gives insight into USB communication. The 64 KByte requests made by higher software layers appear as contiguous blocks of traffic here in best cases - no other traffic inserted in-between. Not waiting whole frames more often than other (faster) computers. Slowness caused by long silent periods between 512-byte-payload USB packets. Such a packet with surrounding protocol takes approx 10 us.

Data transfer to Host (read from USB drive) shows different silent periods, often 2.5 us and 8 us, but larger and intermediate values, too. HOPE - may be there is a way to get always 2.5 or less. Data transfer from Host to Device (write to USB drive) mostly has silent period of 19 or 20 us.

The HP attached to USB cannot tell me why. Attaching it to RAM is much to large effort.

I try to ask Intel, because US15W data book is incomplete.

Data rate comparision microSD native and in M.S. slot

What can be expected at best?

SD-Card (and microSD) spec version 2.00 as of 2006 have at best a 4 bit data bus operating at 50 M transfers/s (50 MHz), means 25 MByte/s. The speed class gives only the minimum sustained write data rate in MByte/s, usually printed on SDHC card in a (partial) circle, values e.g. 2, 4, 6, 10. It says nothing about read.

SD-Card V3.01 spec 2010 extends transfer rates up to 104 MByte/s. Of course, Host needs to be prepared, too. The VGN-P is not.

Memory Stick PRO spec by year 2004 (downloaded 2011) uses 4 bit data bus at max 40 M transfers/s (40 MHz clock), means 20 MByte/s, or as listed in spec more accurate 19.7 MByte/s in burst considering overhead.

Test conditions, setup

Done 2011-07-13..14, VGN-P21Z, WinXP SP3 current updates, h2testw.exe V1.4, WIRELESS=OFF, no special things at USB, only 1 of internal slots SD-Card or Memory Stick occupied at a time. That means file system performance is tested, which is less than raw sector access.

A fast microSD card is needed to see differences. Found 2 pieces Kingston 2 GByte microSD "SDC/2GB 35" (printed on card, datecodes look like made 2008) and package label "SDC/2GB-2P1A" (this means 2 pieces in package, and 1 adapter to SD-Card form factor) in my storage box, purchased for little money probably 2009. Turned out they are quite fast.

Results

Using "Kingston microSD to MS PRO Duo ADAPTER", which contains active electronic circuitry (of course), got Write 7.28 MByte/s, Read 9.02 MByte/s. Using wire-only Adapter to SD-Card form factor got Write 9.19 MByte/s, Read 16.5 MByte/s.

Summary

This is a considerable performance loss for Adapter to Memory Stick host slot, especially for Read. Unsure whether an other Adapter could be faster. WILL TRY.

This microSD to Memory Stick adapter

"Kingston(R)" adapter contains a tiny PCB with 2 integrated circuits (chips) and some passive components. One 8 pin chip is labeled with two lines "416R" and "8807", looks quite ordinary, pin 1 to 4 inter-connected and to WHAT?, Other chip looks odd, is rectangular, has some pins cut off to allow traces on PCB, has only 4 (3 missing) + 11 + 6 (1 missing in center) + 11 pins, labeled 3 lines "SM281CF AC", "H31982", "0827 97C". Internet search found nothing useful.

Startup time of BIOS

Done 2011-06-15

BIOS R2073U3 startup time: From power switch to large text "VAIO" on display 7 s, to first blink of HDD activity LED 8 s, to WinXP boot menu 11.5 s (multiple HDD accesses). It is not faster than older version R2060U3. Only USB-mouse connected, no USB mass storage. Boot from USB and Network disabled, Boot priority USB / HDD / Network. Enabling USB boot did not change time.

Mains adapter

(At 230 V AC input) It is completely cool while VGN-P21Z is OFF and Battery is fully charged. It is warm when heavily loaded.

It consumes 2.45 mA when Mains disconnected, but 10.5 V driven back into output (measured 2011-06-26). This current is consumed by the green LED and may be some control circuitry. It sets a lower limit for standby power consumption of 26 mW.

It has most likely no power factor correction, means it eats input current only close to the tops of input voltage wave.

It has no screws, seems to be glue-closed. At least not epoxy-filled. Electrical repair seems possible with little mechanical destruction.

External display detection, DDC

(Win XP SP3 current updates, Intel graphics driver 6.14.10.1097 date 26.02.2009, done before and at 2011-06-24, 2011-07-05)

The external display (monitor) is detected by its resistance of 75 Ohm from R,G,B lines to GND, even if it has no DDC. There needs to be load on all 3 lines. An old monochrome monitor loading G line only will not be detected (tried using resistors!). DDC2b (I2C protocol) is queried, too, to get timing and mode information. NOT TESTED: DDC2b present without R,G,B load.

The Intel driver allows the user to activate external display only if presence has been detected.

Without DDC2b providing more mode info, the Intel driver allows only modes 640*480, 800*600, 1024*768 all 60 Hz. Bad with Intel driver is: It reports only the modes it believes to be good for display to Windows. Other drivers report all modes and allow (after special selection) the user to choose and try modes which may be unsupported by display.

Without DDC, Intel driver seems (at least sometimes) to report some kind of NEC monitor to WinXP, plug and play ID NEC61BE, but without any useful timings. (2011-07-05 while driving different content to both, it reported for external display without DDC the Microsoft standard thing MS_0A18 and for internal display SNY09FA.) By its basic parameters

it shall allow selection of nearly every mode, because it is considered safe, but something else must be going wrong. Either Intel driver does not build the usual modes at all, or ???, and I believe Intel driver may hack around in this registry entry while running. FACT IS: The bit fields for "Established timings" in EDID are all cleared (Byte at offset decimal 35 and 36). ##### Patching registry is futile, seems to be overwritten.

This is very misleading if you have indeed some NEC connected, but no DDC e.g. because of a broken or incomplete VGA cable (Pin12 and Pin15 need to be wired-thru).

DDC is queried when entering Display dialog, may be queried on operation within it, not further checked. DDC is queried on right-click on Desktop to invoke Context menu. It is not queried on idle Desktop, means no time driven queries.

DDC query means reading the 128 Byte EDID the usual way. SCL period 10.4 us, frequency little below 100 kHz. The US15W GMA500 has a hardware engine for DDC/I2C transfers - REALLY ???.

There are no traces of any other I2C activity on DDC lines, neither during screen X*Y size change nor during bpp change, and even not during VGN-P startup, or WinXP startup. DDC has been queried 2 times during WinXP startup while I watched. Such traces are common for specific SDVO chips, which attach the DDC to the I2C Bus controlling them.

OPEN QUESTION: Will Intel driver allow and obey DDC override by *.inf file supplied to WinXP, as documented by Microsoft? This would allow to enable more resolutions and timings, e.g. flicker free 75 Hz drive for old CRT monitors.

In WinXP, a lot of things are in registry. To see directly what is in there regarding displays and EDID, use regedt32.exe and look inside (open the keys until deepest level): HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Enum\DISPLAY\

Intel graphics driver for WinXP

This thing for GMA 500 seems to be much worse and more limited than other Intel graphics drivers. As of 2011-06-25, the latest version has been published 2009-12.

It supports only 32 (x-8-8-8) and 16 (5-6-5) bit per pixel (bpp), but not 8 bpp (indexed). The GMA 500 supports 8 bpp according to its bad (much infos missing) data sheet sections in US15W data sheet.

Although WinXP does not "like" 8 bpp, means it makes it a bit more difficult to select such a mode (usually via graphics card panel, "List all modes"), it works flawlessly in 8 bpp, as far as I have seen. 8 bpp is only half of data used in 16 bpp. Some older applications prefer it for good reasons.

Why 5-6-5 is bad

The 16 bpp 5-6-5 mode (5 bit red, 6 bit green, 5 bit blue) is despite its widespread appearance a bad idea on its own, because you see a little bit wrong colors in all non-optimized brightness variation flows, simply because of the different number of bits, because of this extra green bit. Basically, Windows supports every combination of bit width (according to documentation, I have not tried so far), and apps shall be prepared, so some small patch to either Intel driver, or a small piece of software inserted between Intel driver and Windows, should allow to change pixel format to 5-5-X-5 and all brightness variations should look like in 32 bpp, a little more coarse, but just brightness changes, not color. Or load a modified CLUT (gamma table) masking the lowest Green bit.

I am not sure what kind of video overlay is supported by driver, if any. But looks somehow limited.

How to get 8 bpp

Observed 2011-07-01 on a VGN-P530H with old BIOS R1250U3, WinXP Media Center SP2 US english, still no Intel driver installed: All 3 bpp modes (8, 16, 32) are listed in the "Display Properties / Settings / Advanced / Adapter / List All Modes..." list. The driver is VgaSafe as can be seen under Properties.

I have still not checked on BIOS R2073U3, but probably it works there, too. That means to get 8 bpp (256 colors), you have to disable or uninstall the Intel driver. Yes, you get it in 1600*768, too.

CPU braking observed

Some more investigation into CPU braking from 1.33 GHz to 800 MHz, done 2011-06-25. Mains powered, external display only, 1920*1080*16bit*60Hz, Ethernet deactivated, WIRELESS=OFF. The latter 2 to keep the Comm board cool, which was successful. My usual setup on table with 1 cm air below, room temperature 24 celsius. Using "Core Temp 0.99.8" to read temperature.

It was not easy to get CPU into braking in this setup. Two Calculator ("Taschenrechner") computing faculty were insufficient, CPU core only up to 54 celsius, full 1.33 GHz. My improved gemul8r.exe (Atari ST emulator) did a better job, in parallel with 1 Calculator. Within a few seconds after core temp reaching 57 celsius, clock was reduced to 800 MHz. After some time, core temp below 50 (exact value not tried to observe), clock was set to 1.33 GHz again. The case in CPU board area got warm to hot, but the Comm board area stayed cool.

My best guess from all observations is: There is at least 1 more temperature sensor. 2011-07-02 revisited: Yes, and guess it is near Ethernet chip, at Comm board. (CPU had below 45 celsius, but was still 800 MHz. Soon after Deactivate of Ethernet chip clock raised to 1.33 GHz.

So far I cannot derive exactly the condition for braking CPU.

CPU load displayed

WinXP (SP3, updates as of 2011-06) has a very inaccurate accounting for CPU load. It counts towards load only when scheduler preempts a task at the end of a time slice. It does not count as load when that task returns the CPU before the time slice ends, e.g. by calling Sleep().

This can bias the graphs of "Task Manager" significantly, and all other load accounting, too, by hiding nearly all load. My practical experience is 1/3 of load hidden, without any bad intent, but fine tuning that application could have hidden 90% or more.

A normal WinXP time slice is approx 15.6 ms. This can be adjusted by special software, but not by WinXP supplied control panels. Shorter values will reveal more and more hidden load, but eat some more CPU power just for switching between tasks.

Other stuff, e.g. found on web

Wrong ebay advertisements / listings

There is a lot advertised / listed at ebay under VGN-P device identifiers which does not fit, e.g.:

Sending a message to seller seems to be of very little use, at least does not cause correction or withdrawal of listing.

RAM upgrades?

I do not know any VGN-P which has a socket for RAM. I found no hints on SONY websites pointing into that direction. All statements are the opposite, means no slot, no upgrade, all fixed.

Nevertheless some guys are marketing RAM upgrades, by SO-DIMM module 2 GB to 2 GB (SO-DIMM is large, only Micro-DIMM had a chance to fit in!), for e.g. VGN-P530H, tell you according to SONY it has slot and can be upgraded to 2 GB, do not understand if you refer them to SONY website stating the opposite - and it has 2 GB already. Do not let you fool by high ebay scores!

The US15W chip data sheet rev 003 May 2010 states "Support for a maximum of 2GB of DRAM". This makes it unlikely to get 4 GB working (except may be with super tricks).

US15W data book?

Unfortunately, the US15W as a basic building block is another badly documented piece. Some parts are looking very well, but I did not verify whether the register descriptions are really matching silicon.

All memory setup, all information regarding buffers and FIFOs for read and write is missing. Nearly all graphics registers are missing.

The VGA Port pins DDC SCL and SDA: They are probably controlled by an I2C protocol engine, and all control is not described in US15W data sheet, not even the registers are mentioned. All other chips with I2C protocol engines I have looked at so far, offer a way to bypass that engine and control the pins directly.

Internal Card Reader

The USB ID information comes from firmware: idVendor==0x054C Sony, PID==038B, Name UMH-U10. Chip is SMSC USB2256AHZS. Firmware is in FlashROM. See hardware file for more details.

Location of Antennae

Applies at least to VGN-P21Z: The Bluetooth antenna is at right edge, right from right [Fn] key. WLAN and WWAN (UMTS) antennae are in Display frame. (still not checked where)

DC IN socket = DC jack, and compatible plug

I do not have the necessary precision mechanical measurement equipment at hand at this time, but want to note nevertheless ... UPDATE LATER!

The socket has a central pin of 1.65 mm diameter (may be 1.7 mm). The plug has 2 contact fingers inside grapping this pin from opposite sides. The plug outer diameter is 4.8 mm (definitely not 5.5 mm), the length approx 9.5 mm from plastic body to tip. Socket is yellow plastics, plug tip the same yellow color. Plug has small impression ring 2 mm from tip, but this is not important for function.

There is an other connector system called in German "Hohlstecker" which looks very similar, and may somehow fit, but is not the same, because the central pin is expected to have some elastic properties, (or not?) but the plug is only a rigid pipe, inside and outside.

To the best of my knowledge, the "DT001 1.65 mm" offered by www.laptoppartstrading.com as of 2011-07 should be a compatible plug.

Considerations

Unfair critics? (lifestyle)

SONY advertises "VAIO(R) P Series Lifestyle PC". What does it mean? Simply and badly a stylish thing without any utility, or impacting style of life, as SONY asked in some contest?

It cannot change style of my life, because I use such devices, even smaller ones, since 1990. In 1991, I have been sitting under a tree after finishing my social duty for that day, programming in x86 asm on an Atari Portfolio.

I treat these computers as tools designed for real work, not as geek toys. I may be completely wrong.

HDD or SSD

Why mistrust SSD? Because it is strongly pushed into market with a lot of advertising. Some HDD makers do a bad job, too, but risk at SSD is much higher - my feeling.

Please see this page for details, because it is a general issue.

Power budget (electrical)

Not considering losses of a few % for voltage conversions and cables. MAKE MORE DETAILED LATER.

Source is Mains adapter 1.9 A at 10.5 V, means little below 20 W. Or Battery.

Consuments are given in list, with max power and a typical value, which may seem arbitrarily chosen.

Looks like the Mains adapter is driven to its limits easily. This is common to many notebook designs. Some even allow the user to connect an under-powered supply and limit the computational power to operate at all, when original adapter not available.

Digital I/O lines for experiments

One of my major complaints with most current computers is their lack of some simple to use digital I/O lines for e.g. controlling a LED or reading a mechanical switch. The old machines had a parallel printer port, serial ports, for quasi-static analog input (resistive) joystick ports.

The VGN-P has none of these ports, either. But remember what I did 1995 to 2000 for earning money: Programming PC Graphics chips. Assume the VGA for external display is not in use. There are two 3.3 V CMOS output lines available: HSYNC and VSYNC. Provided the GMA500 is sufficiently VGA compatible, both can individually be set to static levels, via old-style VGA I/O ports, in a known and documented way. CHECK whether this may interfere with SDVO!

The DDC consists of lines SDA and SCL. SDA is bi-directional open drain. SCL should be the same, but worst case it is output only. Best case both can actively drive H, too, means are push/pull/tri-state. But needs documentation research because d.s. incomplete, or use Linux drivers as info-source. It cannot be used easily if DDC is implemented via the I2C used to control SDVO.

3x analog out

There are 3 powerful (driving current into e.g. 75 ohm resistive load) analog outputs, too, if GMA500 is sufficiently VGA compatible, or allows setting this an other way (missing register docs). I mean R,G,B. Basically, enable a color for frame (around active image area), turn image display off, and you get already a value for nearly all the time. By appropriate setting of timing, or stopping of display output, it was possible to contiguously (statically) output the same R,G,B in all VGA and VGA-like chips. Just need to refresh that knowledge, and may be need a little driver for WinXP, or just use the general hack-driver "give-io".

Use USB a different way

It may be possible to switch the USB port power supply programmatically on/off, as a very powerful output, driving up to 500 mA at close to 5 V. It may be possible to read the USB port presence and speed detection, which is done by a resistor of 1.5 kOhm from either D- (low speed present) or D+ (full speed present) to 3.0 to 3.6 V, giving a digital input with 3 states. Check whether this can work at down-stream ports of external Hub, too, may be even better documented.

Cheap bus-powered USB hubs do not have hardware for power switching, although they shall as far as I understand spec.

US15W data sheet Rev003 gives a strong hint: D+ and D- lines can be used as static digital I/O by activating test mode. (The small voltage levels make output less desirable, but input is ok.) See EHCI section, registers "PORT[7:0]SC-Port N Status and Control Register".

CPU Control (clock frequency)

What is "Power Options / VAIO Power Management / CPU Control" setting good for? (German: Energieoptionen / VAIO Power Management / CPU-Steuerung) Without any work to do, the CPU is in a sleep mode anyway in WinXP, consuming very little electrical power, cooling down. This setting controls clock frequency of the running CPU. As far as I have seen, only 2 different clocks are used, maximum (1.33 GHz for Z520) and low (800 MHz for Z520).

See CPU load displayed for important info. Energy or Battery capacity consumption is electrical power multiplied with time.

"Performance" (Leistung)

means maximum possible clock. If there is only a limited amount of work to do, CPU finishes it as fast as possible and returns to sleep. According to CPU Maker Intel, this is the most efficient way to run a system. The arguments given are reasonable.

"Adaptive" (Anpassend)

selects and switches the clock according to the CPU load. Higher load means max clock, lower means low clock. There is no means to know load in advance in the general case, so decision bases on history, on load accounting, on running faster or slower than needed. Because load accounting can be very inaccurate, and because load is counted simply for the sum of the 2 threads of the CPU, I have observed already that launching a Calculator for faculty calculation speeded up another application, which used only 1 CPU thread.

"Battery Life" (Akku-Lebensdauer)

runs CPU at low clock. The CPU consumes less power than during max clock, but needs significantly longer to return to sleep. Assuming "Performance" is the most efficient way to run a system under partial load, can this here save energy? Yes, if the software does not put the CPU to sleep, but runs it in busy wait loops until something else happens.

Instand mode replacement

Standby mode is a poor but usable replacement, if you need near Instant On behavior, and can afford to have 20 to 25 % less Standard Battery capacity available for a working day, connect the VGN-P to Mains anyway every night. You must not forget to feed your VGN-P.

RAM bandwidth

This computation shows the upper limits and estimates a little. You cannot get more out of or into RAM at the given clock freq.

266 MHz DDR2 64 Bit means 533 M transfers/s of 8 Byte peak data rate, product is 4264 MB/s or approx 4.3 GB/s. Real rates can reach 95 to 99 % of this for large linear blocks (and even some short distance random accesses, if chip is perfect, by just changing column address).

Really available bandwidth is less due to necessary setup and tear down (open and close of banks, refresh), which can only partially be hidden by accessing alternate banks.

Worst case bandwidth for random access is dramatically less.

Maximum internal display 1600*768*32bpp*60Hz refresh is approx 300 MB/s (294912000 B/s), less because blank and sync times, more because setup and tear down. And some accesses for cursor.

Maximum external display 1920*1080*32bpp*60Hz refresh is approx 500 MB/s (497664000 B/s), see previous chapter.

UNSURE: A 3rd display plane might be open for video overlay / playback.

Display refresh will have a limit on block size if setup is optimized for drawing performance by using memory tiles.

Some remarks

The MHz

I try to use "800 MHz" here when referring to display of "799.99 MHz" by some tools. I do not always write all after-dot digits of the 1.333 GHz.

Accu or Battery?

The English words seem to applicable to even more areas, but in Electro area their strict meaning seems to match their similar-sounding German translations:

Considering that, both terms can be used on VGN-P's mobile power supply. In English GUI term "Battery" is used, while in German it is "Akku".

PATA / atapi listed under SCSI

Just seen in WinXP registry 2011-07-05: "HKEY_LOCAL_MACHINE / HARDWARE / DEVICEMAP / Scsi / Scsi Port 0" and the "Scsi Port 1" both contain key Driver value atapi. The first lists in a deeper entry the HDD.

--- Author: Harun Scheutzow ------ Last change: 2012-02-20 ---