Notes and reviews about hardware

Toshiba U300-13U (080218)

In the post-festivities sales at the beginning of the year I found a good offer for a Toshiba U300-13U laptop, a very portable laptop with a 13.3" screen, 2×1.5GHz Core 2 chip, 2GiB of RAM, and 160GB of disk.

Overall I am quite happy with it. Some good points:

• Quite light and thin.
• Quite powerful configuration.
• The screen is quite legible, bright and has a good resolution of 110DPI, which means that antiliasing is even less necessary.
• Almost every component is support by GNU/Linux. For example there is a rather good Intel developed driver for the WiFi IPW3945 chipset, for the Ricoh flash memory interface and so on, and even for the Chicony webcam.
• The keyboard has clear keys with black lettering, which makes it rather more readable in low light, which often happens when travelling.
• The flash memory reader supports the xD format, which is what my Olympus camera uses.

• The most annoying is that the button that opens the CD/DVD drive protrudes from right hand side, and thus it is very easy to press it, in particular if one stores the laptop tightly. The Dell Latitude D620 has a recessed button for that reason, and it is so much better.
• The default battery is only a four cell 4000mAh one. This means a rather shorter than expected life at around 2-3 hours of light use. Obviously this laptop should have by default the 6 cells 5200mAh battery (which I have promptly bought).There is also a much bigger 9 cell 7800mAh battery that really qualifies as an accessory. I suspect that the default is the weaker battery for two reasons:
  • To keep the total weight of the laptop just under 2kg (it is declared as 1.99kg), which is the conventional limit of the ultraportable category.
  • To induce a significant portion of the customers to buy a higher margin accessory (as the base unit is competitively priced).
• The network interface only supports 100mb/s, instead of 1gb/s, which is disappointing even for home networks.
• The screen vertical viewing angle is rather small. This is because it looks like the backlight comes only from the bottom, so viewing the screen from above makes it washed out and from below rather dark. Fortunately it is not too difficultto tilt the screen until it is good.
• As some reviewers point out:
  • The keyboard flexes in the middle.
  • The top of the case flexes in the corners above the CD/DVD drive and the ExpressCard slot.
  • The screen is quite reflective.
• A relatively small number and type of ports, due to the back being devoid of them as the screen flips into it.
• MS-Windows Vista instead of XP. Or rather that I have to pay a few dozen extra pounds for something I don't really use.

Overall I like it more than the Dell D620, where some points of comparison are:

• The U300-13U is nearly half the price.
• The U300 is somewhat smaller and lighter.
• The build quality of the D620 is much better.
• The U300 has a much brighter screen with rather better colour rendition, but the D620 has a matte surface instead of a reflective one.
• The U300 has fewer but newer sockets, which may be an advantage or not.

Like with all laptops one should buy with it a spare battery, a spare charger, and at least one backup external disk, and also eventually an extended warranty.

I have already got the external backup disks; I shall buy an extended warranty near the end of the initial warranty (it would be very annoying to pay for an extra two years and warranty and have the laptop destroyed or stolen in the first year).

I have bought as spare the Toshiba 6 cells 5600mAh model, and for a charger a Targus compatible model. Considering how critical the battery is, and how dangerous a faulty battery can be, and how easy they are to design poorly or with bad materials1, I think that a third party battery can be an excessive risk. Buying a random power supply can also be dangerous, but I reckon that it is a lesser risk, and anyhow Targus are a fairly high end supplier with a reputation (and margins) to defend.

The list of GNU/Linux specific tweaks I have used is:

• Use of the ipw3945 driver and dæmon (superceded by the iwlwifi one).
• Applied the patch described in Method E to the ALSA ALSA 1.0.14 on my Linux 2.6.23 version.
• Used the intel driver of X.org 7.2, which works but only in XAA mode (EXA has an annoying rendering bug), and which crashes if using DRI.
• Used the Synpatics drive for the touchpad.
• Followed other advice in this article.

Teknik Office posture chair (061022)

I tend to find most comfortable when sitting somewhat high with respect to keyboard and mouse, so that my elbows are higher than either; this means that usually my legs tend to dangle down from the seat.

Having been a bit swamped with things to do recently, I have had to sit for longer times than usual, and this has been painful, because the edge of the seat than cuts into my thighs, even if I take frequent breaks.

I have tried two solutions. The first is standing up while using screen and keyboard; this work very well and results in avoiding the problem completely, but it rather tires my feet, which then ache; I wonder how shop assistants cope.

The other solution has been to try one of those kneeling chairs, specifically a Teknik Office posture chair (SKU 309854) which is one of those who try to improve posture by having the sitting person rest their weight in part on a kneeling pad.

I usually have a good sitting posture in part because I am careful to have my monitor at eye level, and not much below, but the idea attracted me because it means that the weight of the person rests not so much on buttocks and thighs but also on the knees, and also the legs are not horizontal and point down.

Well, I was a bit disappointed because of two issues:

• The higher seat tend to remain too near to level, thus cutting into thighs too. I found that this can be fixed by raising it, which also inclines it forwards, but raising it of course means that one ends up sitting too high relative to the desk. The problem is that the higher seat is not tiltable, but is fixed to the beam on which it is connected. Well, I avoided the issue by removing the front castor wheels, which both lowers the chair and tilts it forward.
• The pad where the knees or the forelegs rest has an excessively soft cushion which means it compresses against the wood base too easily. This results in aching knees or forelegs fairly quickly, I guess until one gets used to it. This seems to be a common problem, and a friend solved it by removing the padding and replacing it with a cutout from the kind of dense foam mattress used by backpackers.

Also one of the seams of the upper seat ripped soon (probably all too weak sewing).

Well, neither solution is perfect, but both are better than an ordinary chair. The Teknik product has some issues, but it avoids my main problem, and is fairly cheap for a chair, so I think that overall it is fine, but I say this a bit begrudgingly.

Zalman Fanmate2 (060505)

The Zalman Fanmate2 is a little device that allows controlling the speed of fans that use the usual 3 pin berg connector style.

Like all the devices of its type it does so by changing the voltage supplied to the fan, and it changes the voltage by applying variable resistance, which means it reduces the power supplied to the fan by dissipating it into heat.

Thus the major issue with such devices is how hot they become. Well, the Fanmate2 behaves very well here: it can dissipate enough to support fans drawing up to 6W, and most fans draw at most around 2W. Indeed with two such fans in my PC (CPU and case fans) the Fanmate does not even feel warm.

It is possible to see a pretty huge heatsink inside the little perforated box that is the body containing the electrical bits, which surely helps a lot. I had tried some other mainfestly poorly built fan voltage control device and it would get very hot, by contrast.

The Fanmate2 also has the considerable advantage of having a quite long cable, which allows great flexibility in positioning the box with the control dial; I have for example put them in an otherwise unused 5.25" slot on the front of my case, for easy access when tuning the speed of the two fans.

Overall so far the Zalman Fanmate2 has been an excellent choice, way better (which mostly means cooler) than others, and still fairly cheap at around £4 each. The only limitation that I can see is that it does not support directly fans that have a 4 pin Molex connector, but it is fairly easy to get 4 pin to 3 pin cables.

No Name ULT31311 (060418)

The No Name ULT31311 is an external enclosure for 3.5" disc drives with an external brick style power supply.

It uses a Prolific PL3057 chip for both USB2 and IEEE1394 bridging to ATA.

It is fairly cheap and convenient, with a clamshell style case that is easy to open and close, but which also seems to be very ugly.

The power supplys is one of those cheap 12V ones that tend to burn out, and is rated at 2.0A, which is not too bad but would not support many 3.5" hard discs that require more power.

The PL3057 chip is a bigger issue, as its IEEE1394 side is, according to my experience and many others reported on the web, practically unusable because of poor reliability. Some people report that updating the firmware in the chip may improve it, but that are contradictory reports even as to that.

The USB2 side of the PL3057 chip actually works fairly decently and with good speed.

Overall I was not that happy with this enclosure, but then the same power limits and reliability problems are pretty common.

No Name TT-346U2F (060418)

The No Name (probably Triumph) TT-346U2F is an external USB2/FW enclosure for 5.25" and 3.5" disc and CD/DVD drives; it has an USB2 to ATA chip from ALi, the M5621, and a FW to ATA chip from Oxford Semiconductor, the OXFW911.

It also has an internal power supply rated at 1.8A for 12V.

The bad points are:

• The M5621 chipset seems incompatible with several USB2 host chipsets, and/or some ATA hard drives. It seems to work better with CD/DVD drives though.
• The power supply's rating of only 1.8A at 12V may cause issues with many higher capacity or performance hard discs.
• The tiny fan is rather loud.

Overall I like this box, even if it is a bit annoying to open and close, but both slightly differentw variants that I have seen (I bought two of these boxes) look faily good and robust.

The internal power supply is a lot more convenient than an external one, and the 5.25" form factor means that it can house a CD/DVD drive or a 5.25"-to-3.5" disc drawer., and the OXFW911 IEEE1394 chipset is fast and reliable.

Belkin F5D7630 ADSL router (060929)

This is a combined ADSL modem, wireless AP, and 100MHz Ethernet switch, with firewall and routing. It is a rebadged product from the Far East, and very similar variants are also sold by SMC and 3com. Indeed many models of these variants are firmware compatible, which is good because the Belkin firmware seems rather buggy and incomplete to me.

The good thing is that it is fairly cheap (I paid £80 at the local Comet shop in Oct. 2004, and it costs only a little less from mail order companies). It also works fairly well and it is easy to set up, with fairly sensible defaults customised for the UK.

Apparently it is also quite reliable and not crash prone (except for cooling problems, which are easily fixed, see below) like other similar products.

It has some bad points though:

• Fast repeated DNS queries by a DNS server running on a host behind the firewall trigger flood protection, which is very annoying. Even more annoying is that quickly repeated HTTP requests also trigger SYN flood protection, again from inside the firewall (which runs ar 100Mb/s with <1ms latency, that is very fast). This to me seems eminently moronic.
  I think that the only obvious way to avoid this is to disable what is called the firewall, provide some protection, and it is a quite safer to rely on per-host firewall software than one just at the gateway. Unfortunately if that is disable the DMZ feature is disabled too.
  However see below for access to hidden menus that might be contain configuration parameters that can be tweaked to avoid this.
• The port forwading menus are clumsy, and it might be in any case better to simply enable the DMZ feature to designate a node as an externally visible server, which is practically necessary if one uses one of several and important protocols that do not traverse well NAT gateways. Unfortunately to enable the DMZ one has to enable the firewall and this causes trouble with overzealous flooding detection. Anyhow even so the DMZ feature simply seemed to be permanently disabled.
• The device is silent because it has no fan, but the heatsink is the metal base plate on the bottom, and since heat rises, the device is prone to overheating. The solution is very simple, and it is to keep it vertical and not horizontal, and this requires a simple holder, which is not supplied. I use a letter holder, and that is fine.
• There is no display of the statistics, either traffic or status, of the ADSL line, like error rate, up and down speed, SNR and attenuation, which are very useful to diagnose line problems.
  However see below for access to hidden web pages that do contain ADSL statistics.
• There is no display of the statistics, either traffic or status, for the wireless AP.
• By default it does not keep the line up if there is no traffic for a little while. This provokes a reconnection delay when traffic starts again, and while the delay is usually one or two seconds, on some ISP it can rather longer as their authentication servers are often undermaintained and overloaded.
  However, with the latest firmware release, this rather annoying thing can be configured away, for PPPoA in my case, in the WAN>Connection Type>PPPoA(Routing Mode, for multiple PCs) for the UK form.
  Unfortunately this form is partially broken for PPPoA in that the dial-on-demand and idle-timeout values are ignored, as checking with this undocumented status page shows, but there is a workaround: the otherwise almost identical WAN>Connection Type>PPPoE(Routing Mode, for multiple PCs) form sets the dial-on-demand and idle-timeout values correctly, and the values it sets apply also to the PPPoA case, so one can first set them in the PPPoE form and then submit the PPPoA form without changes.
  While there it may also be useful to further reduce the MTU to 1400, but the default of 1454 seems already good. That the default is less than 1500, to account for PPP and L2TP header overheads, means that the people who do the firmware have more than usual clues.
• The led that indicates whether the ADSL line is connected or not does not flicker to indicate traffic on that line, while the similar LED for wireless and Ethernet does.
• Another review points out that the wireless AP is somewhat weak but not that bad. I'd rather have a lower power transmitter that lasts longer. Also, in my experience it is better than what the review says, but perhaps I have thinner walls.
• Another review mentions that the DNS resolver in the firmware has a hardcoded limit on the length of domain name components that is too low for some sites:

  Hi, the German C'T magazine nr. 23 of 2004 gives the cause of these problems in a exhaustive comparative test of 19 wireless 54G modem/routers. I quote translated:

  Errors in DNS relays lead to peculiar problems. Some routers crash when opening specific Ebay pages. This is related to the customer monitoring system used by Ebay. This system does not use cookies, but customer data encrypted inside DNS queries instead.

  It requests names, comprising name-parts separated by periods within some labels, being up to 63 characters long. This complies with the DNS standard of 1987 (RFC 1034), however not for some router programmers in 2004 who designed routers with less memory space for using labels. Direct result: Buffer Overflow and Router crash. In the test this applied to the 3COM, Belkin and SMC wireless routers.

By looking at the source of the web pages of the user interface I have formed the impression that it is quite badly written by rather unskilled people, and this might explain why it seems to me rather awkward and buggy. Reading the impression of the various people who have looked into the firmware update files also gives me an overall sense of shoddy work.

The final impression I have got is that I should have gotten a Netgear or Linksys instead, as they are far more configurable and informative (Linux based...), but then I read bad things about the reliability of many Linksys and some Netgear models.

All in all I am going to keep it. It works most of the time despite the bugs, and its reported hardware reliability is somewhat more important to me than customizability.

I found important to at least upgrade the firmware to the latest version, which as I write seems to be V.A.1.08.03UK (Jul 8 2004 15:41:37) because this allows configuring the disconnect-on-idle timeout and the MTU for PPPoE, PPPoA and similar connection types.

But after some long frustration I decided to try and load as some other users have done the firmware for the equivalent 7804WBRA router from SMC, and what a change. Not only the user interface is much better organized, it is far more complete, and less buggy (it is still written in a way that to me seems very poor). Also virtually all the most annoying bugs with the Belkin firmware seem fixed in the SMC one.

Note: one annoying bug remains: while the DMZ option works, the flood protection continues to be applied to the internal interface, and even if it is disabled for the external interface. The protection gets triggered by TCP connection rates higher than 300 connections per minute, or 5 per second, which is absurdly low, as just opening a page with half a dozen images will trigger it. As with the Belkin version of the firmware this can only be avoided by disabling the firewall entirely, and then one loses the DMZ. But there is a workaround: in the NAT section, under Virtual Servers it is possible to forward all ports (1-65535) one-to-one under both TCP and UDP to a given internal host, which is the same as DMZ for a single external address. Since I have only one external address that is fine for me. Perhaps full DMZ functionality can be obtained by using the Address Mapping feature in the same NAT section.

On top of all this the 7804WBRA manual is well written, complete, documenting all the options and expalining them tersely but correctly. I am going to buy SMC in the future rather than Belkin.

Note: a number of people have also discovered (1, 2) that Belkin have licensed a generic and fully featured ADSL firmware used also by SMC, and that Belking have removed from the main web configuration page the links to a number of really useful control pages, and here is a list of these:

Status and statistics

• Configuration summary.
• Configuration and status details.
• Options about mass deployement (probably).
• Web based GUI version information.
• Downloadable security log.

ADSL

• ADSL advanced configuration in particular the ability to select the handshake mode.
• ADSL modem statistics including SNR margins and dB attentuation.

WAN

• Form to register multiple ISP details.
• ATM WAN configuration.
• ATM WAN PPPoA configuration.
• WAN DHCP configuration.
• WAN DHCP MAC configuration.
• WAN configuration for Telstra Bigpond.

Wireless

• Enable or disable the wireless radio.
• Wireless security configuration.
• Wireless security WEP configuration.
• Wireless security WAP configuration.
• Wireless RADIUS authentication server configuration.

Firewall

• Firewall NAT configuration.
• Firewall extended configuration (#1).
• Firewall extended configuration (#2).
• Fireall configuration by named protocol or ports.

Other services

• Enable or disable UPnP.
• Registration with Dynamic DNS services.

Internet Gateway Device XML status pages

• http://192.168.2.1/igd.xml.
• http://192.168.2.1/igd_osf.xml.
• http://192.168.2.1/igd_wcic.xml.
• http://192.168.2.1/igd_wdsl.xml.
• http://192.168.2.1/igd_wec.xml.
• http://192.168.2.1/igd_wic.xml.
• http://192.168.2.1/igd_wpc.xml.

Not very useful or broken

• http://192.168.2.1/login.htm.
• http://192.168.2.1/loginpserr.htm.
• http://192.168.2.1/data.js.
• http://192.168.2.1/function.stm.
• http://192.168.2.1/main_router.css.
• http://192.168.2.1/menu_style.txt.
• http://192.168.2.1/my_function.txt.
• http://192.168.2.1/nat_main.stm.
• http://192.168.2.1/advance_main_0.stm.
• http://192.168.2.1/c_adsl_status.stm.
• http://192.168.2.1/routine.txt.
• http://192.168.2.1/routine2.txt.
• http://192.168.2.1/setupw_main2.stm.
• http://192.168.2.1/system_syslog_svr.stm.

LG L1710B 17" LCD monitor

Pretty good display quality, with many nice details: DVI input with DVI cable included, the power supply is builtin, the base is telescopic and contains an USB hub, narrow bezel, fairly good analog autosync, the base can be removed leaving a standard 100mm VESA mounting plate, for example for a telescopic arm, and it has very little to no trails/ghosting when playing games or movies.

Almost the only defect is that backlighting is slightly uneven, with the edges of the screen being brighter, but this is almost unnoticeable. A secondary issue is that the display is somewhat fuzzier on the VGA input; but it is supposed to be used in DVI mode. Those monitors that only have a VGA input tend to have better VGA input quality.

ASUS DRW-0402P DVD-R/DVD-RW

Good quality drive, a clone of the Pioneer DVR-105 drive. Speed is not bad, random access is a bit slow relative to some others.

The best news is that there are ways (1, 2) to turn it into a Pioneer DVR-105 with firmware that contains some useful hacks.

Adaptec DuoConnect AUA 3211

This card is a quality USB2 plus IEEE1394 host adapter for the PCI bus.

Major advantages:

• Three external USB2 sockets and two external IEEE1394 ones one card.
• Both USB2 and IEEE1394 internal sockets, which makes it easy to have a front panel with either or both.
• Power connector so it can be a powered hub for fairly power hungry peripherals that draw more than what can be supplied via the PCI bus.
• Uses a NEC EHCI chipset for USB2 (uPD720100A) and a Texas OHCI chipset for IEEE1394 (TSB12LV26) that are well supported under recent versions of Linux, and are quite reliable.
  Unfortunately the NEC EHCI chip has a well documented defect (apparently it can only support half duplex on the PCI bus) that makes it one of the slowest USB2 chipsets.
• It comes with a good quality bundled 6-to-4 pin IEEE1394 cable which can be pretty expensive if purchased separately.

The only big disadvantage is that it is somewhat expensive; but compared to two separate cards it is rather more convenient, and the features are rather better thought ought than other combo cards.

Epson 1660 (GT-8300) scanner

This is probably the best value midrange scanner around, and it has four large advantages:

• USB2 output, which is much much faster than USB or parallel output for non-small color images.
• A built in backlight for scanning 35mm negatives and slides.
• Very good speed, resolution and color quality; at 1600 DPI one can see the grain of 35mm slides.
• It is fully supported under recent versions of Linux by SANE and xsane (which are rather better than Epson's own Linux frontend, iScan).

I am not aware of any disadvantages.

Epson 680 (or 777) color inkjet printer

Major advantages:

• The speed and output quality are quite good both in monochrome and color mode, and on plain paper too. B&W output quality on good plain paper is amazingly good.
• It is well supported under Linux, and supply their own binary driver that is really quite good, even if not very Unix-like.

There are three disadvantages, two large (one of them is not confirmed) and the other small:

• According to a reviewer some Epson Stylus Color printers (he mentions the 580) have a built in logic that disables the printer after a certain amount of use (the metric is the number of head cleaning operations):

  After 1 year and 1 month the printer broke down: "unknown error has occurred bring the computer to a service center".

  Of course this was too expensive. I looked on the internet and found out that at least hundreds had the same error. The official epson site personel did not know anything, they say. i found out that the epson 580 and other epson printers are more or less programmed (EEPROM) to stop after a certain time.

  This time depends on the number of times it has used ink to clean the head. In short, it probably wants to prevent that ink is leaking onto the desk, and therefore - btw thank you Epson - tells the printer to stop printing and so buy a new printer. It displays a message that nobody understands and at Epson they of course say they dont know what is wrong.

  But luckily on internet they found out: http://www.fixyourowncomputer.com/

  This sounds rather appalling especially as it is not documented, except in this rather unofficial web site.
• The ink cartridges are chipped, which makes it more expensive or annoying to use compatible ink. The cartridge design also makes it a bit more annoying to do refills. Fortunately there are now pre-chipped compatible ink cartridges, but this comes at a (small) price.
• The output resolution is based on 360DPI (actually 72DPI originally) and even multiples or submultiples. Most scanners and other digital devices are based on 300DPI and even multiples or submultiples, which may result in some non even scaling of the image.

All in all I am fairly happy with it, but I think I should have gone for a Canon inkjet instead, as they have much lower ink costs even if they cost about 50% more to purchase. Another alternative would have been a slightly more expensive HP inkjet, like a 940c, which does not use chipped cartridges and they are easy to refill too.

ABIT KG7 motherboard

The KG7 is one of the better Athlon motherboards, if what one is looking for is solidity rather than flashy features. Its main advantages are:

• It supports ECC memory. This is very important; unfortunately there are very few Socket A motherboards support it, and they tend to be all AMD 761 chipset based. Most VIA chipsets also support it, but most VIA motherboards don't.
• Six PCI slots and four DDR-DRAM slots. This gives more options for expansion and customization.
• All power regulators are switching.
• According to a reviewer the presence of six power transistors means that the onboard power supply is triphase, which helps stability. Indeed this board is rather more stable than my previous Soltek, which was fairly new anyhow.
• The usual excellent jumperless CPU setup menu of ABIT motherboards.
• It can be found for pretty low prices. The Lite version, which has only two DDR-DRAM slows, can be bought for 41 pounds including taxes.
• Pretty good layout, with a number of case fan sockets and good positioning of the CPU and most sockets. I think the better proximity of the CPU to the back and top of the case (case fan and PSU fan) is the reason why my CPU runs now about 5 centigrade cooler than with my previous motherboard. This may also be because the power regulators being switching no longer become hot.

The main disadvantage is that it uses DDR-DRAM, which is twice as expensive as SDRAM. If you want to use SDRAM, the best bet may be one of the few Socket A motherboards that support both SDRAM and DDR-RAM.

Another advantage is that the bracket with the sockets for the second USB host adapter is bundled. Unfortunately the IR bracket is not included, and is hard to find.

The few recent Socket A motherboards that support ECC seem to be all based on AMD chipsets, and are mostly expensive dual processor ones. The ABIT KG7 is together with the EPOX 8K7A the only cheap uniprocessor socket A motherboard that supports ECC and has six PCI slots. The EPOX seems nice, but some power regulators are non switching, so I went for the ABIT. Another two uniprocessor socket A motherboards support ECC, the Gigabyte XXX and the DFI YYY, and the both trade stupidly a PCI slot for an AMR or CNR slot.

Maxtor 133 IDE/ATA PCI card (Promise ultraATA TX2)

This is based on a Promise PDC20269 chipset. It is OK and well supported by recent Linux kernels. The ATA133 support is largely pointless, so one could buy the similar ATA100 board by Promise, which can be found for somewhat cheaper.

3ware 7410 IDE/ATA RAID PCI card

Very very good, amazing tagged queueing performance.

Lian Li PC-7 and PC-60 cases

In brief: amazingly good at a relatively low price. But on reflection the PC-60 case is only a little more expensive and has a few nicer details, and is a slightly better value.

Samsung 570S 15" TFT LCD

In brief: amazingly good even if a bit old styled. Fast enough for DVD and games.

Hansol H530 15" TFT LCD

In brief: amazingly good even if a bit old styled, and despite some slightly uneven backlight. Fast enough for DVD and games.

CMI 8378 sound cards

In brief: quite good, well supported under Linux ALSA (but they need software sharing). The mixer structure is simple enough, which of course help a bit with ALSA that by default gives a raw view of any card's mixer.

For some people they are the cheapest (and really cheap usually) way to get full optical and coaxial digital SPDIF/IEC958. The CMI 8738 based Trust 514DX has a full set of digital inputs and outputs, for example. But the gameport does not work (known defect, never fixed).

The sound quality is good for the price, even if it has limits at high frequencies. The CMI8738 chip is really 48kHz only, even if it can accept and produce 44.1kHz, but using 44.1kHz should not be done as it involves low quality resampling, sometimes twice.

The MS Windows drivers seem to me to be quite slow though; seems better under MS Windows 2000 than 98.