Inki

My thanks to Gurgelmeyer for this great contribution. I did notice that update KB832414 for MSXML 2.5 (msxml.dll 8.0.7002.0) is not included. I am not sure how useful or necessary this update is if later versions of MSXML are also installed. Nor am I certain how this update plays together with current updates of Internet Expolorer, or if it has somehow been made obsolete. KB832414 has been totally changed from its original appearance. It no longer has a download link to this update but points on to KB887606 which only covers later versions of MSXML. Anyway, the download link for this update can still be found at MDGx's site on the page for Internet Explorer. P.S. Edited addition (made after the following post by -I-): I just noticed that my laptop at work, which has XPSP2 installed by IT people, and where I do not have admin rights, also has the updated msxml.dll file among others. So I guess that at least it probably is not harmful.

Thank you for your replies. First of all, let me emphasize that the disk access delays that I am referring to are of a totally different magnitude than the mere slowing down due to DOS compatibility mode. Think in terms of having to wait ten minutes on an XP3200+ system before you can see the listing of your disk volumes after having double-clicked on "My Computer". These huge delays do not occur whenever the system is in safe mode and all drives are in DOS compatibility mode. Then let me describe a few cases: Case 1. If one does a clean install with SATA in IDE mode, or if one does a clean install with SATA disabled and then reboots with SATA in IDE mode, the situation becomes as is described in my previous posting: - The normal triplet of Hard Disk Controllers is present: -- Primary IDE Controller (dual fifo) -- Secondary IDE Controller (dual fifo) -- VIA Bus Master PCI IDE Controller - Three additional Hard Disk Controllers appear (the third one is the root for the other two): -- Primary IDE Controller (single fifo) -- Secondary IDE Controller (single fifo) -- Standard Dual PCI IDE Controller - Ludicrously large SATA access delays occur when the SATA disk is connected. - The SATA disk, when connected, is accessed in DOS compatibility mode. - The VIA SATA RAID Controller driver can not be installed. I have not done a systematic analysis of further options based on this case. However, the mere manual removal the Standard Dual PCI IDE Controller, for example, is not a solution. If that is done, then: - The SATA disk is accessed in DOS Compatibility mode - At each reboot the system asks for the driver to be reinstalled. - If it is reinstalled while the SATA disk is connected there is a risk of the same kind of halt on error that is described under Case 2. Case 2. If following Case 1 one reboots with SATA in RAID mode, the situation changes as follows: - The Primary IDE Controller (dual fifo) shows an error in normal mode. - If the SATA disk is connected at this point, then the systems halts at a big blue error screen: -- The Standard Dual PCI IDE Controller does not support a mixed 16-bit and 32-bit access configuration. -- After restarting, all disks are accessed in DOS compatibility mode and all four Primary and Secondary IDE controllers show an error in Device Manager in normal mode. With no SATA disk connected it is possible to proceed by installing the VIA SATA RAID Controller even though the related PCI IDE Controller device is not visible in Device Manager. The situation then becomes the following: - The normal triplet of Hard Disk Controllers works properly. - The three additional Hard Disk Controllers no longer appear in normal mode: -- The two (single fifo) -controllers are visible only in safe mode. -- The Standard Dual PCI IDE Controller has disappeared. - The PCI RAID Controller device is now listed under SCSI Drivers as VIA SATA RAID Controller. - The system functions normally. Case 3. If one does a clean install with SATA in RAID mode, then: - The normal triplet of Hard Disk Controllers is present. - The additional Hard Disk Controllers never appear. - The PCI RAID Controller device appears, and after installing its driver it is listed under SCSI Drivers as VIA SATA RAID Controller. - The system functions normally. Case 4/Case 5. If following Case 2/Case 3 one reboots with SATA in IDE mode, then: - No change is apparent among the devices listed in Device Manager either in normal mode or safe mode. - Ludicrously large SATA access delays occur when the SATA disk is connected. - The SATA disk, when connected, is accessed in DOS compatibility mode. Mysterious, is it not?

Thank you for your suggestion, LLXX. In fact, the SATA drive is accessed through DOS-compatibility mode when set to IDE mode. However, I suspect that something else is also going on. As an example of what kind of delays are involved: after right-clicking on "My Computer", it takes a full nine and a half minutes for the dialogue box to appear where one can then select "Properties". Furthermore, when SATA IDE-mode is activated, the Device Manager displays all of the following Hard Disk Controllers: - Primary IDE Controller (dual fifo) - Primary IDE Controller (single fifo) - Secondary IDE Controller (dual fifo) - Secondary IDE Controller (single fifo) - Standard Dual PCI IDE Controller - VIA Bus Master PCI IDE Controller When one then includes the actual SATA disk, the behaviour with long delays results. All the Disk Controllers and Disks are reported to be "working properly". I am guessing wildly here, but perhaps with that whole bunch of disk controllers present 98SE might be attempting to simultaneously perform incompatible actions but somehow manages not to crash, and perhaps the hardware RAID management layer is required for 98SE to make sense of the situation. Maybe it is just not meant to be that this system would work with 98SE if SATA is set to IDE mode. I really have no idea. This is something that I have been wondering about for at least a year or so, and I have usually tested some aspect of it each time that I have re-installed 98SE. Currently I am resigned to thinking that this is a feature and that the reasonable thing to do is just to be happy and stick with the RAID mode.

Out of pure curiosity, I wonder if any of the experts on this forum might be able to shed some light on what could be going on in the following situation. Still, this is not an issue worth making a big a big fuss about, and I am sure that there are better ways of spending time than digging into it if there is no immediately obvious explanation. Anyway, here goes: The Gigabyte GA-7VT880-L (VIA KT880/VT8237) mainboard allows its SATA ports to be set either into IDE mode or RAID mode through the BIOS. If the system is used in DOS, everything appears to work fine with both PATA and SATA disks irrespective of which mode SATA is set in – no issue here. If SATA is in RAID mode and 98SE is installed (on one of two installed PATA disk with an additional SATA disk installed) everything appears to work fine as one might expect – no issue here either. However, if SATA is in IDE mode and 98SE is installed in the same disk configuration as above, then any access to the SATA disk takes up to five minutes or more to complete (but is eventually completed). Installing 98SE, for example, proceeds quite normally (including the initial disk checks) until 98SE is activated and begins to do the disk accesses, at which point the waiting game begins. This behaviour does not appear to depend at all on whether VIA drivers are installed or not, what version they are, or whether their SATA/RAID-component is installed or not. It does not seem to depend on the BIOS version either, and I remember having previously observed something like it also on a VIA KT600/VT8237-based mainboard. The behaviour, however, does not occur with Windows 2000 on the same platform, and I assume that it is linked to 98SE or its disk drivers and possibly some incompatibility with VT8237 hardware in this particular mode. My best guess is that somehow 98SE gets confused by the large number of available IDE ports (I have not yet tested the configuration where I would disable all PATA ports and only have SATA in IDE mode, nor will I shortly due to having other things to do). Still, no confusion is apparent under DOS, and 98SE appears to run, although with the ridiculously long disk access delays it is difficult to be certain of this. Nor have I (yet) observed any indication of disk corruption. This is not really a problem, because one can always just have the SATA ports set to RAID mode without actually setting up a RAID array, and everything is fine. However, I am just simply amazed by how much resultless activity the system seems to be capable of generating for several minutes over nothing but the simplest SATA disk access in IDE mode.

THE SEMPRON EFFECT? I guess I must answer this as I just made it up. The Sempron effect means the replacement of a previous high-end part by another part that may not be not quite as good but is more cost effective to manufacture. This happens when the performance of the previous high-end part begins to lag behind newer parts, and its selling price goes down so as to seriously bite into the manufacturer's margins. When socket 939 was launched, this happened to the Athlon XP product line which was replaced by Semprons that were otherwise fairly similar but had a smaller cache and a lower front-side-bus speed. It may begin to happen to the Athlon 64 product line as a result of the upcoming AM2 launch. This has also happened or is still happening to the GeForce 6800 graphics card series. It will be interesting to observe whether or not the Radeon X800/X850 graphics cards manage to receive non-"beta" 98SE drivers before something nasty possibly begins to happen to them. Still, they may get those drivers even posthumously. Partly because of this effect there exists only a relatively short window during which it is possible to make good deals on high-end parts with legacy support before they are completely wiped off the market and replaced by newer parts without this support. I guess the new 6800 GS may be an example of a newer replacement part, which possibly may end up never being supported for 98SE, but I really cannot know whether or not this is the case. (For your information: I have no involvement whatsoever in the electronics or software industries) Inki And, by the way, Molecule, my initial response to your other question, that I can comfortably relate with, is that, as a graduate-level physicist I do not immediately see any obvious way to link platonic solids with gaussian modulation of rotation.

Hi there, once again, I noticed that quite a few people have had a look at this topic, so maybe there is some general interest towards it. Although this posting does not contain much that is new, I decided to satisfy my own striving for completeness by coming back just to post an overall summary of my recipe to build a high-performance 98SE system. Please note that some areas are covered incompletely and only to my own satisfaction, but maybe this will at least be of some guidance if you wish to do further research. How to assemble a relatively up-to-date and 98SE-compatible system: 0. MAIN POINTS Choosing to have 98SE compatibility will limit your system performance below what would otherwise be possible. Build a socket 939 based system with a single graphics card. The performance of your system will be limited mainly by the graphics card so you may want to consider how costly a CPU to buy. The first thing to do is to decide upon and/or secure a suitable graphics card (see below). You are most likely better off with an AGP system rather than a PCI-E system, because: - Nearly all graphics cards with current 98SE support have AGP versions, and AGP therefore provides most flexibility. - In their most recent 98SE drivers, ATI offers only "beta" support for PCI-E graphics cards, and nVidia none at all. - There is no real performance benefit from PCI-E in itself. - The benefit from PCI-E is restricted to making it possible to install a faster but non-98SE supported graphics card. 1. The GRAPHICS CARD is the most critical and problematic component. I would suggest three main categories to choose from: i) Fastest AGP/PCI-E cards with current "beta" driver support: ATI Radeon X800 and X850 series probably offer the fastest cards with some level of 98SE driver support: - Radeon X850 XT Platinum Edition would be the fastest - Radeon X850 XT would be the second fastest. The latest 98SE drivers (Catalyst 5.9) for these cards are "beta". This status has actually worsened as the previous 98SE drivers for the same cards (Catalyst 5.2) are not labeled as "beta". (you may want to check the most recent driver status) The Catalyst 5.9 drivers for 98SE also claim to support PCI-E, but I have no direct experience of this. The happiest situation would occur if the "beta" status refers mainly to PCI-E support, in which case you might be roughly OK with AGP versions of the cards, but again, I have no experience on this. ii) Slightly slower AGP cards with proper driver support but poor availability: NVidia GeForce 6800 series AGP cards have proper 98SE driver support: - GeForce 6800 Ultra would be the fastest - Geforce 6800 GT would be the second fastest (I have them and they are OK) These cards could roughly be compared to the X800 series in performance, but they are poorly available and may be impossible to find. There are other members in the 98SE-supported 6800 family, that generally have lower performance but may have some availability. The latest nVidia driver (81.98) only supports AGP with 98SE. (Whether previous drivers do support PCI-E, or their release notes are incomplete, is unclear to me.) The one exception within the 6800 series is the new 6800 GS, which has performance similar to the 6800 GT, but is so far only available as a PCI-E version and without 98SE support (you may want to check the most recent status). iii) A safer but even slower AGP card: NVidia GeForce 6600 GT is a safer but slower solution for 98SE: - Performance is roughly comparable to the slowest 6800 cards, but perhaps with better working hardware video decoding support - Proper 98SE driver support - Good availability - Possible issues concerning rigidity of cooling assembly As above, the latest drivers only support AGP with 98SE. 2. The choice of MAINBOARD is less difficult but may still be tricky I would recommend a mainboard with the VIA VT8237 southbridge because it is fairly mature and works nicely with 98SE. It supports SATA/RAID as well as USB 2.0, and all drivers are easily downloadable. See previous postings for more thoughts and info on this. It also appears to have the benefit of being able to recognise SATA hard drives as IDE drives before installing any drivers, so you just may be able to boot 98SE from a SATA drive. However, I have not had any need to test this in practice so you should not take my word on it working in reality. I would recommend a mainboard with an AGP graphics port to be on the safe side, in which case you would most probably have the VIA KT800 Pro northbridge. If you decide to go for a VIA-based PCI-E mainboard, you would probably have the VIA K8T890 northbridge. Note, however, that there are two versions of this: the later one supports dual-core processors and the earlier one does not. The concern here is that as the world turns towards PCI-E, the availability of suitable AGP mainboards is reduced. Also, VIA chipsets in general do not appear to be very popular these days and may therefore have limited availability. Remember to make sure that your graphics card physically fits on your mainboard! 3. The CPU is no major issue as long as it is AMD socket 939 Just go for any socket 939 processor that pleases you (64, X2, FX). For 98SE alone you would get the most bang for the buck from a single core CPU as one core is all that 98SE can utilize. To benefit from two cores you would have to run some other dual-core capable operating system on the same system. Personally I am still holding back on this to see how the situation develops: - Maybe FX-55 will be replaced by a cheaper non-FX part (who knows?) - Maybe the 64 4000+ will become still cheaper as X2 matches it in speed. - Maybe the whole single core line will come down in price - Beware, however, a possible "Sempron effect" in the future and pounce before that 4. RAM - yes, please. I would go for 2x512MB, but this is discussed in more detail elsewhere in this forum. Just remember to do the necessary tweaks to the 98SE VCache settings. 5. HARD DRIVES Leaving out any SCSI-stuff or other special arrangements, three things are to consider: - Have two hard drives and locate the Windows Swap file on the second one. - 98SE typically needs a Parallel ATA hard drive to boot from - Keep in mind the max. 137 GB drive size limitation for 98SE however you do it. I actually use in my pure 98SE system two Maxtor (supports UDMA133) 9 or 10 series 160 GB drives, which I have configured with the help of a Maxtor utility program to report their sizes to the BIOS as 137 GB exactly. 6. RESPONSIBILITY For the purpose of assigning responsibility over making hardware investment decisions, please consider these guidelines as the mindless ravings of an over-aged lunatic, who wildly overestimates his understanding of modern computers, and only blame yourself if you are reckless and gullible enough to base any actions on them. Bye again, Inki

Well, Krick, now for the final run: I glanced at your list of VT8237 mainboards and recognized a few: I briefly considered the ABIT AX8 before I decided that 98SE and PCI-E probably do not mix well. The ASUS A8V is very similar to the ABIT AV8 that I bought but the ASUS board did not have integrated Firewire. Some of the others were a bit more unfamiliar. I am not sure what the R stands for in VT8237R. It seems to be used very inconsistently and I imagine that it cannot stand for anything of critical importance. Keep in mind that the Asrock may be an excellent board, although everything nowadays is "Designed for Windows XP", which means that complete 98SE compatibility may be a question of chance. With a full set of 98SE drivers there is no reason, however, to doubt that it would run 98SE, and it may even be completely compatible. Or then incompatibilities may present themselves as minor irritants such as the game port on a soundblaster sound card not working. Without trying it is impossible to say. If I were in your shoes I might just buy the Asrock for the heck of it, and then make something roughly useful out of it for a grateful relative if it didn't work out. At least it would seem to offer roomy accommodations for your slightly bulky graphics card. Still, you might not be a well-paid 40+ bachelor executive, and you may want to consider your options more carefully. To test the card before all the other options vanish over time you would also have to buy immediately the CPU which is still probably on a declining price slope. Still, this is the name of the game if you want a very up-to-date and fully compatible 98SE system. You have to be opportunistic, fast and possibly take some risks unless you can find somebody with even more experience in the matter. I myself had to make a judgement call when I bought the last of my GF6800GT cards. I had a gut feeling that they may be the last (and best apart from the Ultra) fully 98SE supported cards and that it might just be the right moment to make good deals on them before they are pulled from the market. I was a bit concerned that the Radeon X800 and X850 series may yet turn out to be better choices, but as the driver support for them remains a bit shaky I an still happy with my choice. Now I am afraid that I have exhausted my thoughts and experience in this matter, and I must leave you to make your own judgement calls. Therefore I thank you for the discussion, wish you happy hunting, and it is time for me to turn back to my normal business. Bye, Inki

In answer to Krick above: Unfortunately I do not have any opinions on the ULi chipsets. The Asrock board looks very impressive indeed, although it would not be optimal for my needs: In my case I have replaced the original coolers on my GF6800GT cards with Arctic Cooling NV Silencer 5 coolers which cover an additional PCI slot. This would leave me with only two free PCI slots on the Asrock board, which would not be enough for all my needs since I would definately need to insert also a card to provide Firewire that is not integrated into the board. My thinking in going with VIA is based on the fact that they have been around for some time and are therefore more likely to have managed to carry over support for 98SE from one design generation to another, for good or for bad. Also I know from experience that the VT8237 southbridge works for 98SE (whereas I have understood from their driver release notes that the VT8251 southbridge no longer has full 98SE support, at least for SATA/RAID). I believe that it is less likely for any "new" chipset to have complete 98SE compatibility, because it is unclear why the designer should be motivated to take this compatibility into account from scratch. I suspect that this happened even to the NForce2 chipset, which basicly had a full set of 98SE drivers but still would not support a soundcard game port under 98SE. Based on what I have found on the net, this appears to be an internal architectural issue concerning addresses within the chipset. Therefore my recipe is to take the "most mature" chipset that has been stretched to work with socket 939. However, even if I indulge myself in a little PC building, I am by no means an expert and may be mistaken in this. I guess the only way to know for sure is to try it out. Concerning the ABIT AV8, I looked up the picture of your graphics card, and my impression is that you would probably not be able to fit it on the board due to the blocking capacitors that I mentioned in my previous post. Still, it might be a close call and depend on the precise length of you card. Best regards, Inki

Hi, This is an interesting question that I have given some thought to, because I like my machines to be backwards compatible as far as possible. I also require full DOS support, meaning that I wish to play legacy games in a DOS box in 98SE with full sound and game port support. As I don't have much to contribute to the excellent 98SE Service Pack, I will be happy to share with you my thoughts and experience in this area. I currently have a fairly good 98SE system and I have another one planned and waiting to receive a CPU. My current system is the following: CPU: AMD Athlon XP 3200+ (Barton) - 32 bit CPU for a 32 bit system (I bought almost the last one that was locally available in stores) Mainboard: Gigabyte GA-7VT880 Pro - Nortbridge VIA KT880 allows use of fast dual channel memory, - Southbridge VIA 8237 provides SATA, USB2. Also Firewire - AGP graphics port, as I am not sure about 98SE support for PCI-E. - Full driver support for 98SE - I had also tried out an NForce-based ABIT AN7, but I donated it away for the following reasons: - The NForce2 mainboard seemed to have an address-overlap issue with the game port on a soundcard, and the game port of the sound card could not be used on such a mainboard - Additionally, if not using 98SE, there seemed to be some kind of timing issue in NForce2 that creates a problem for some USB 2.0 devices in full APIC mode (e.g. Haupauge USB WinTV in W2k) RAM: 2x256MB Kingston HyperX DDR400 Graphics card: MSI Geforce 6800 GT (AGP) (again hoarded nearly the last one from the store) - Although a bit lax in this respect, I still wouldn't dish out the cash for an "Ultra", apart from which this is the fastest Nvidia card with 98SE support. Drivers are downloadable and functioning. - ATI Radeon X850 might be even faster, but according to release notes, the 98SE drivers for it are "beta", and I sometimes like to toy around with the stereo viewing functionality that Nvidia supports Soundcard: Soundblaster 16 PCI (jumped one day on the occasion and bought several when available) - Full SB16 functionality in DOS and emulation in 98SE TV-card: TechniSat SkyStar 2 TV PCI - Supports DVB subtitles with 98SE drivers My ultimate 98SE-system will be the following: - Graphics card: XFX Geforce 6800 GT (AGP) (from my hoard) - For an AGP mainboard with the VIA 8237 southbridge (which I believe may be the last one fully supported with 98SE drivers) I have the ABIT AV8 (Because of my previous issues with NForce2, I feel that I would not trust even newer NForce chipsets to offer full support for everything under 98SE, although I have not tested this. In this respect VIA might be more conservative, perhaps.) - Other components apart from the CPU identical to what is above This systems now sits in my closet waiting for the price to drop on the Athlon 64 4000+. Quite frankly, I believe that building the "ultimate" 98SE machine may be an interesting side project, even if maybe not so very useful. The sad thing is that many of the best components for this kind of project are very rapidly disappearing from view and the marketplace. Therefore you have to act fast and be opportunistic. Inki P.S. Just in case somebody might be tempted to use the ABIT AV8, it is good to note that at least the one I have has some capacitors close enough to the AGP slot to potentially block inserting the graphics card. I actually have one MSI GF6800GT and two XFX GF6800GT cards. Of these three only the MSI card and one XFX card would fit. Although the XFX's are basicly the same product, one of them has capacitors low on the card, close to where the second power dongle for an Ultra card would be, and these get in the way. So beware.