[dencorso]

GreyPhound:

Awesome! You rock, you really do!

As for your questions:

1) For some reason I don't understand, the 'attach file' option has been disabled in the "Windows 95/98/98SE/ME" main forum, for as long as I can remember, although it is enabled in the "Unofficial Win98 Se Service Pack" subforum...

2) IMHO, it'll be much better if you can add part 1 to your original post, regardless of the final size of the post. After all, if you do that, every one who reads your translation, from now on, will be able to read it in order.

[GreyPhound]

Done.

see Post #23:
Windows 9x and lots of memory: Dotting the i's
Igor Leyko

This post has been edited by GreyPhound: 05 January 2008 - 02:19 PM

[Max Monroe]

Wonderful job translating the article, things are nicely explained in it. However, the original question still stands, and I seem to have the same problem, namely:
- I have Win98SE (no patches as far as I know, other than the one for large disk access) on an EPOX EP-8KRAI with an Athlon XP, with 2GB RAM in two sticks (DDR400, PC3200). Used to have 512MB, without problems up to the upgrade;
- After the customary MaxPhysPage and MaxFileCache restriction to 40000 (1GB) and 524288, windows boots nicely, no immediate apparent problems - but MS-DOS produces the "not enough memory error", no matter what I do.
- I've tried EVERYHING, nothing makes DOS work.

- AGP aperture is 64MB; tried to reduce to 32MB: no luck
- VCache down to 32768 changes nothing, below that windows starts to refuse to boot (not enough memory). The setting is applied correctly, verifiable with System Monitor.
- PhysPage as 60000 (1.5GB) does not boot at all; as 3C000, 30000 or even 20000 (back to 512MB!!!) does not help. windows boots, DOS does not work. The setting is applied correctly, verifiable on every about page and with System Monitor.

...now what? I'd rather not physically remove RAM, because I'm dual booting with XP which is fine thank you and makes good use of it. Any ideas, please...?

- Max

[dencorso]

GreyPhound, on Jan 5 2008, 11:36 AM, said:

see Post #23:
Windows 9x and lots of memory: Dotting the i's
Igor Leyko

Awesome, GreyPhound! Thanks so much!

=======

@Max Monroe:

Pleases try this extreme configuration:

1) in Config.SYS:
(i) NO EMM386.EXE
(ii) NO XMSDSK.EXE
(iii) HIMEM.SYS /NUMHANDLES=96 /TESTMEM:ON /V

2) NO Autoexec.bat at all

3) in System.ini
(i) under [386Enh], MaxPhysPage=38000
(ii) under [VCache],
MinFileCache=0
MaxFileCache=29696 ; 29M
ChunkSize=512

>>> Note that MaxFileCache=29696 really, this is not a typo. <<<

4) AGP aperture 64MB

...and let me know what happens. Good luck!

This post has been edited by dencorso: 05 January 2008 - 07:13 PM

[Max Monroe]

Did you mean "DEVICE=HIMEM.SYS /NUMHANDLES=96 /TESTMEM:ON /V" ? Well, it didn't work, anyway. Win98 didn't boot, with "no memory". Had to return to 40000 / 65536. Aaaaaanyway, thanks for trying, and here's the RAM dump, perhaps it helps....

00000000-0009FFFF : System board extension for ACPI BIOS
000A0000-000AFFFF : RADEON 9600 SERIES (Omega 2.6.37)
000B0000-000BFFFF : RADEON 9600 SERIES (Omega 2.6.37)
000C0000-000CCFFF : RADEON 9600 SERIES (Omega 2.6.37)
000CD000-000CFFFF : System board extension for ACPI BIOS
000F0000-000F7FFF : System board extension for ACPI BIOS
000F8000-000FBFFF : System board extension for ACPI BIOS
000FC000-000FFFFF : System board extension for ACPI BIOS
00100000-7FFEFFFF : System board extension for ACPI BIOS
7FFF0000-7FFFFFFF : System board extension for ACPI BIOS
C0000000-CFFFFFFF : RADEON 9600 SERIES (Omega 2.6.37)
C0000000-DFFFFFFF : VIA CPU to AGP Controller
D0000000-DFFFFFFF : RADEON 9600 SERIES - Secondary (Omega 2.6.37)
E0000000-E3FFFFFF : VIA Standard Host Bridge
E4000000-E43FFFFF : Intel® 536EP Modem
E4400000-E441FFFF : RADEON 9600 SERIES (Omega 2.6.37)
E4400000-E44FFFFF : VIA CPU to AGP Controller
E4420000-E442FFFF : RADEON 9600 SERIES (Omega 2.6.37)
E4430000-E443FFFF : RADEON 9600 SERIES - Secondary (Omega 2.6.37)
E4500000-E45007FF : PCI OHCI Compliant IEEE 1394 Host Controller
E4501000-E45010FF : VIA PCI to USB Enhanced Host Controller
E4502000-E45020FF : VIA Rhine II Fast Ethernet Adapter
FEC00000-FECFFFFF : System board extension for ACPI BIOS
FEE00000-FEEFFFFF : System board extension for ACPI BIOS
FFF80000-FFFEFFFF : System board extension for ACPI BIOS
FFFF0000-FFFFFFFF : System board extension for ACPI BIOS

[dencorso]

Max Monroe, on Jan 5 2008, 11:28 PM, said:

[...]
C0000000-CFFFFFFF : RADEON 9600 SERIES (Omega 2.6.37)
C0000000-DFFFFFFF : VIA CPU to AGP Controller
D0000000-DFFFFFFF : RADEON 9600 SERIES - Secondary (Omega 2.6.37)

1) This looks as a memory conflict to me. Isn't there any typo?

Max Monroe, on Jan 5 2008, 11:28 PM, said:

E0000000-E3FFFFFF : VIA Standard Host Bridge
E4000000-E43FFFFF : Intel® 536EP Modem
E4400000-E441FFFF : RADEON 9600 SERIES (Omega 2.6.37))
E4400000-E44FFFFF : VIA CPU to AGP Controller
E4420000-E442FFFF : RADEON 9600 SERIES (Omega 2.6.37)
E4430000-E443FFFF : RADEON 9600 SERIES - Secondary (Omega 2.6.37)

2) Same here. Isn't there any typo?

Max Monroe, on Jan 5 2008, 11:28 PM, said:

E4500000-E45007FF : PCI OHCI Compliant IEEE 1394 Host Controller
E4501000-E45010FF : VIA PCI to USB Enhanced Host Controller
E4502000-E45020FF : VIA Rhine II Fast Ethernet Adapter
FEC00000-FECFFFFF : System board extension for ACPI BIOS
FEE00000-FEEFFFFF : System board extension for ACPI BIOS
FFF80000-FFFEFFFF : System board extension for ACPI BIOS
FFFF0000-FFFFFFFF : System board extension for ACPI BIOS

Now, I see the following free memory spaces:
E4440000-E44FFFFF = C0000
E4500800-E4500FFF = 00800
E4501100-E4501FFF = 00F00
E4502100-FEBFFFFF = 1A6FDF00
FED00000-FEDFFFFF = 100000
FEF00000-FFF7FFFF = 1080000

Which totals to 1B93FE00 => 451839 kB on doing Igor Leyko's calculation.
Subtracting a VCache of 65536 KB, the result is 386303 kB... But note that this is fragmented real memory.
And it seems not to be enough for the Ring 0 VM (all Windows VxDs) and additional DOS Boxes (DOS VMs).

I'd do a backup first and then...
I'd try to remove ACPI (and APM) from Win 98, to liberate some more memory on the 4th GB...
I've posted a how-to here (link) some time ago, for other reasons, but it might solve your problem.
Then again, this is just a wild guess...
Anyway, good luck! And please do keep me posted on it.

[dencorso]

Hi again, GreyPhound!

I just finished reading it and found a small typo, that needs correction:

GreyPhound, on Jan 2 2008, 06:58 PM, said:

As a rule, AGP devices use addresses starting from E0000000, that is 3,5 Gb. Disk cache at that can use a half gigabyte. This is exactly what determines a recommendation to limit cache to the value of 524288 Kb (512 Mb).

This notwithstanding, you did a fantastic job! Thanks again!

[dencorso]

vick1111, on Dec 30 2007, 02:25 PM, said:

Well I have tried everything that was possible (given my knowledge of the problem) and I have not been able to make XMSDSK work without problems.
[...]
-----------------------------------
XMSDSK Failure
-------------------------------------
Whenever XMSDSK succesfully creates a ram drive then my dos windows become unavailable for the "out of memory error"

MOREOVER XMSDSK does not work properly: when I start the scandisk, to check the sectors into the ramdrive Z, the write attempt fails after a while.
A more simple copy write attempts fails too.

The failure happens not at a specific point.

what happens is that I got a blue screen with not recoverable error sayng things like:

"An exception 0E has occurred at 0028:C0004D6D in VxD Vmm(01) + 00003D6D. This was called from 0028:C19359C4 in Vxd Rmm(01) + 00000254"

I have tried many different settings for XMSDSK.
[...]

Hi, vick1111!

Please try:
c:\ramdisk\xmsdsk.exe 262272 Z: /c1 /y

**yes, without the "/t", despite all you can read about it: there are reports this may solve your problem.**
It's said it works for 2GB or more RAM. I never tried it, because I've not encountered such a problem, probably because I've got only 1.5GB, I guess. Good luck!

This post has been edited by dencorso: 06 January 2008 - 01:17 AM

[GreyPhound]

dencorso, on Jan 6 2008, 06:57 AM, said:

Hi again, GreyPhound!

I just finished reading it and found a small typo, that needs correction:

GreyPhound, on Jan 2 2008, 06:58 PM, said:

As a rule, AGP devices use addresses starting from E0000000, that is 3,5 Gb. Disk cache at that can use a half gigabyte. This is exactly what determines a recommendation to limit cache to the value of 524288 Kb (512 Mb).

Corrected.

Sorry for that one, it must have been a browser encoding thing, as that misrepresented value is correct in my original text file.

[Max Monroe]

dencorso, on Jan 6 2008, 05:38 AM, said:

Max Monroe, on Jan 5 2008, 11:28 PM, said:

[...]
C0000000-CFFFFFFF : RADEON 9600 SERIES (Omega 2.6.37)
C0000000-DFFFFFFF : VIA CPU to AGP Controller
D0000000-DFFFFFFF : RADEON 9600 SERIES - Secondary (Omega 2.6.37)

1) This looks as a memory conflict to me. Isn't there any typo?

Nope, that's copy-pasted as it came. Anyway, both are kinda "video card" stuff, so I assume it might not be entirely wrong to overlap / be dual allocated / accessed. Actually, I do kinda wonder WHAT THE HECK does the video adapter with THAT kind of memory allocated to it... It's a 128MB ATI Radeon 9600, for Pete's sake...

dencorso, on Jan 6 2008, 05:38 AM, said:

I'd try to remove ACPI (and APM) from Win 98, to liberate some more memory on the 4th GB...
I've posted a how-to here (link) some time ago, for other reasons, but it might solve your problem.
Then again, this is just a wild guess...
Anyway, good luck! And please do keep me posted on it.

Thanks, but as one of the posters in the other thread noted, I'm not in a hurry to get rid of ACPI (if I understand correctly, the 3 to 4 GB area really matters, and ACPI seems to hold a lousy 20MB there), especially since disabling it in the BIOS would shoot it under my other OS (XP) too, I suppose.

As much as I hate that, I might just finally dump '98 alltogether, because I'm stuck with disturbingly-not-current versions of much everything lately, things crash a bit too often for lack of resources and so forth. It has recently come to the point of nobody giving a crap whether their app runs on 98 or not (mostly not), and I'm NOT willing to get stuck in the stone age of software if only that keeps working. Not being able to watch a current MOV clip on the net just about does it...

[dencorso]

Max Monroe, on Jan 6 2008, 01:15 PM, said:

dencorso, on Jan 6 2008, 05:38 AM, said:

I'd try to remove ACPI (and APM) from Win 98, to liberate some more memory on the 4th GB...
I've posted a how-to here (link) some time ago, for other reasons, but it might solve your problem.
Then again, this is just a wild guess...
Anyway, good luck! And please do keep me posted on it.

Thanks, but as one of the posters in the other thread noted, I'm not in a hurry to get rid of ACPI (if I understand correctly, the 3 to 4 GB area really matters, and ACPI seems to hold a lousy 20MB there), especially since disabling it in the BIOS would shoot it under my other OS (XP) too, I suppose.

Note that my how-to was written with other purposes in mind. For a double boot machine, I'd recommend to skip the BIOS part, so as to leave ACPI for the other OSes. And, then, remove ACPI from Win 98 only (this causes Win 98 to switch to APM). And the remove APM also from Win 98 only. As well as disable NVRAM/ESCD updates and IRQ Steering in Win 98 only. It can be done and works OK. What remains to be seen is whether this also can solve your memory problem.

This post has been edited by dencorso: 06 January 2008 - 10:54 AM

[diskless]

Does the WinME virtual memory manager work with Win98SE?

If so, I think this would be the only way to get 2 GB working, otherwise the most is a little more than 1.5 GB.

[BenoitRen]

Quote

I'm NOT willing to get stuck in the stone age of software if only that keeps working

You want stone age? Get a computer with an early version of DOS. That's stone age.

[j7n]

GreyPhound, on Jan 6 2008, 09:29 AM, said:

Corrected.

Sorry for that one, it must have been a browser encoding thing, as that misrepresented value is correct in my original text file.

Don't the last 512 Megs start from A0000000? Originally you had Cyrillic "A" there.

[GreyPhound]

j7n, on Jan 7 2008, 11:47 PM, said:

GreyPhound, on Jan 6 2008, 09:29 AM, said:

Corrected.

Sorry for that one, it must have been a browser encoding thing, as that misrepresented value is correct in my original text file.

Don't the last 512 Megs start from A0000000? Originally you had Cyrillic "A" there.

Here's the original text in Russian:

Quote

Как правило, устройства AGP используют для своих нужд адреса, начиная с е0000000, то есть 3,5 Гб. Дисковому кэшу при этом можно использовать полгигабайта.

Same text in English:

Quote

As a rule, AGP devices use addresses starting from E0000000, that is 3,5 Gb. Disk cache at that can use a half gigabyte.

[Max Monroe]

BenoitRen, on Jan 6 2008, 10:19 PM, said:

Quote

I'm NOT willing to get stuck in the stone age of software if only that keeps working

You want stone age? Get a computer with an early version of DOS. That's stone age.

My oh my, someone is in a cranky mood today... We certainly can go back to the Spectrum, the punched tape, or Babbage if you wish, but let's not, shall we? I arbitrarily define as stone age version any software that does not enable me to use features considered standard these days, features that my hardware would otherwise be quite capable to perform. And most software today is a GENERATION beyond their last version that supported win98, usually more than fulfilling the criteria above. I could even live with that, I suppose. But when a properly set-up memory upgrade wrecks the runnability of a quarter of my programs, plays tricks with the audio with the rest, and the best advice I get is to shoot the ACPI too to solve all this, I start having second thoughts, pardon me.

So.

You say all this isn't very constructive? Excellent observation. Perhabs you have an idea then why is a 128MB Radeon allocating 512MB of virtual address space, given that most people's video adapters take a QUARTER of that...?

...Thought so. Thank you.

[erpdude8]

Max Monroe, on Jan 7 2008, 06:47 PM, said:

BenoitRen, on Jan 6 2008, 10:19 PM, said:

Quote

I'm NOT willing to get stuck in the stone age of software if only that keeps working

You want stone age? Get a computer with an early version of DOS. That's stone age.

My oh my, someone is in a cranky mood today... We certainly can go back to the Spectrum, the punched tape, or Babbage if you wish, but let's not, shall we? I arbitrarily define as stone age version any software that does not enable me to use features considered standard these days, features that my hardware would otherwise be quite capable to perform. And most software today is a GENERATION beyond their last version that supported win98, usually more than fulfilling the criteria above. I could even live with that, I suppose. But when a properly set-up memory upgrade wrecks the runnability of a quarter of my programs, plays tricks with the audio with the rest, and the best advice I get is to shoot the ACPI too to solve all this, I start having second thoughts, pardon me.

So.

You say all this isn't very constructive? Excellent observation. Perhabs you have an idea then why is a 128MB Radeon allocating 512MB of virtual address space, given that most people's video adapters take a QUARTER of that...?

...Thought so. Thank you.

eh don't mind Benoitren. he can sometimes act like a punk!
some of us are really fed up with his immature childish behavior; one of these days, if he "crosses the line" he might get banned from this 9x forum

This post has been edited by erpdude8: 07 January 2008 - 06:14 PM

[erpdude8]

dencorso, on Jan 2 2008, 10:46 PM, said:

Ninho, on Jan 2 2008, 11:53 AM, said:

[...]

With this in mind, let's see if I understood well : dare I ask, what you are saying amounts to Windows/VMM at startup remapping to system memory space (the 3rd virtual gigabyte) any XMS memory blocks that were allocated (in DOS presumably) before Win started ? Makes sense !

--
Ninho

Ninho:
Thanks for the heads up. I've, now, corrected my post.
But I forgot to answer your question, so I'm doing it now:
Yes, you've got it right. That's exactly what I meant.
And XMSDSK is just a good example of a program that allocates XMS from DOS (run from autoexec.bat or from config.sys), before Windows even begins to load. There are other examples, but I think ramdisks are the most useful nowadays. I think Igor Leyko didn't cover this subject in his article because it would render even more complicated a subject that, as you yourself have said, is not easy at all. But in the present thread it is unavoidable to consider the case of ramdisks, because, IMHO, they are a must for systems having more than 1GB of RAM installed.

Speaking of RAMDisks, the max size you can create is 2Gb. see this MDGx tip on how to use RAMDisks:
http://www.mdgx.com/newtip12.htm#RAM

thanks to GreyPhound and dencorso for the english translation.

This post has been edited by erpdude8: 07 January 2008 - 06:32 PM

[BenoitRen]

Quote

eh don't mind Benoitren. he can sometimes act like a punk! mad.gif
some of us are really fed up with his immature childish behavior; one of these days, if he "crosses the line" he might get banned from this 9x forum

What the hell are you talking about?

I'm so sorry I was defending Win98, which is what this entire forum is mostly about. Sheesh.

Features considered standard these days? Eh, people take anything for granted these days. I wouldn't follow the sheep. Those same sheep that fall for M$' lies every time and continue to buy software that is more than a generation behind when it comes to actual programming techniques and optimisation. People will accept anything, apparantly.

[dencorso]

erpdude8, on Jan 7 2008, 09:13 PM, said:

Speaking of RAMDisks, the max size you can create is 2Gb. see this MDGx tip on how to use RAMDisks:
http://www.mdgx.com/newtip12.htm#RAM

Hi, erpdude8!
You're right: the max size for XMSDSK is 2Gb. That's because it is FAT16, which has an upper upper limit is 2Gb. It also causes XMSDSK to behave strangely and throw errors, refusing to load, with more than 2GB of RAM installed. Although I never faced such a problem myself, I've read that, strictly for these cases, the solution is to let it load below the top of the memory, by ommiting the "/T" switch, which is otherwise mandatory for use with Windows. See the last lines of this TechArena post (#2), by Mark-Allen Perry, about it. BTW: no, it cannot be made FAT32, because its format is part of how it's coded...

This post has been edited by dencorso: 07 January 2008 - 10:57 PM