VOGONS


First post, by Paadam

User metadata
Rank Member
Rank
Member

As the subject says. I got XG-DLS board (rev. 1.03) from kixs. Apparently brand new in box. I have only 700Mhz/1MB/100 and 933/133/133, both 2.8v CPU's to test with. And a terminator. No POST sadly.

Is it possible that since BIOS is as per factory, it is just too old to POST with large cache Xeon? Or (most likely) something else?

VGA and memory are known good. It accepts regular unbuffered SDRAM right?

Last edited by Paadam on 2016-11-14, 08:10. Edited 2 times in total.

Many 3Dfx and Pentium III-S stuff.
My amibay FS thread: www.amibay.com/showthread.php?88030-Man ... -370-dual)

Reply 1 of 28, by kixs

User metadata
Rank l33t
Rank
l33t

I hope luckybob can help you more. My guess is that even 700Mhz might be too new for this revision.

My Amibay: SALE | BUY - Updated on 2021-07-06 | Requests also possible

Reply 2 of 28, by Paadam

User metadata
Rank Member
Rank
Member

I could hot-flash latest XG-DLS BIOS to some other chip and test (can't hotflash XG-DLS chip).

Many 3Dfx and Pentium III-S stuff.
My amibay FS thread: www.amibay.com/showthread.php?88030-Man ... -370-dual)

Reply 3 of 28, by luckybob

User metadata
Rank l33t
Rank
l33t

36339861.jpg

anyway...

it should still post. Do you have a bios code reader? they will shed some light on the situation. It could be wrong ram. I know I'm using ecc/reg in mine.

It is a mistake to think you can solve any major problems just with potatoes.

Reply 4 of 28, by Paadam

User metadata
Rank Member
Rank
Member

Thanks for chiming in luckybob. On the BIOS chip the sticker indicates that BIOS version is 1008.

I ordered POST card few weeks ago, it should arrive soon. I will try ECC memory even though the manual states that regular ones are also supported.

I did notice that on this particular board many jumper pads are unpopulated, like KBK EN/DIS (near coin cell battery), JP5 (JUMPLESS/JUMPER), JP1 etc. Which revision board(s) you have and are yours also unpopulated that way?

Many 3Dfx and Pentium III-S stuff.
My amibay FS thread: www.amibay.com/showthread.php?88030-Man ... -370-dual)

Reply 6 of 28, by Paadam

User metadata
Rank Member
Rank
Member
luckybob wrote:

Picture?

Here:

Attachments

  • XGDLS1.jpg
    Filename
    XGDLS1.jpg
    File size
    381.7 KiB
    Views
    764 views
    File license
    Fair use/fair dealing exception

Many 3Dfx and Pentium III-S stuff.
My amibay FS thread: www.amibay.com/showthread.php?88030-Man ... -370-dual)

Reply 7 of 28, by gdjacobs

User metadata
Rank l33t++
Rank
l33t++

Just an FYI, those are unpopulated because they're hard set. JP5, for instance, has a hard link soldered in.

All hail the Great Capacitor Brand Finder

Reply 8 of 28, by Paadam

User metadata
Rank Member
Rank
Member
gdjacobs wrote:

Just an FYI, those are unpopulated because they're hard set. JP5, for instance, has a hard link soldered in.

Yes, I see that. But none of the other ones have jumper bridged. I am just wondering if other XG-DLS boards have the same arrangement.

Many 3Dfx and Pentium III-S stuff.
My amibay FS thread: www.amibay.com/showthread.php?88030-Man ... -370-dual)

Reply 9 of 28, by luckybob

User metadata
Rank l33t
Rank
l33t

Nothing there is important. The chassis is for case intrusion detection. Basically if you open the box while its on it will power down.

If memory serves, the keyboard one lets the ps/2 keyboard wake up the unit. What key is selectable in bios.

Cant do much more without a 80h bios card.

It is a mistake to think you can solve any major problems just with potatoes.

Reply 10 of 28, by Paadam

User metadata
Rank Member
Rank
Member

Yeah, I know the function, just was wondering why these were omitted from mine. But Jumperless? any ideas?

I really want to get this thing alive, have some plans with it, like adding a 4th FSB jumper for 133/33 MHz 😉 It already has correct clockgen.

Many 3Dfx and Pentium III-S stuff.
My amibay FS thread: www.amibay.com/showthread.php?88030-Man ... -370-dual)

Reply 11 of 28, by luckybob

User metadata
Rank l33t
Rank
l33t

it just means it will ignore the jumpers and go with what is best. So you cant set multi and fsb. you can disable this in bios iirc. (not the multiplier though)

It is a mistake to think you can solve any major problems just with potatoes.

Reply 12 of 28, by Paadam

User metadata
Rank Member
Rank
Member

WOOOOOOOOOT!!!!!

Went to my workshop, dug out some 2x256MB ECC REG modules and decided to try, board POSTED!!! Displays as Unknown CPU at 689 MHz but none the less.

Next: rig the board into one of the cases and start testing various Slot2 CPU's. I have 900MHz pair (5/12v), I will probably try to replace the VRM's on the board with those of 2.8v units and see how it goes. Worst case I will need to buy one 700/1M 2.8v CPU (I have another one) and just resolder 900Mhz CPU's to those cartridges. Have been messing with BGA stuff before so that will be a nice task.

Many 3Dfx and Pentium III-S stuff.
My amibay FS thread: www.amibay.com/showthread.php?88030-Man ... -370-dual)

Reply 13 of 28, by Paadam

User metadata
Rank Member
Rank
Member

Did the 4th FSB jumper mod to XG-DLS. It's PLL fully supports 133/33 operation but Asus omitted FS3 jumper for some reason which means 133MHz was only possible with PCI at 44 MHz (totally unacceptable). Soldered a jumper header with center pin going to PLL pin 9 through 10k resistor, now it works perfectly. And since Xeons have multipliers unlocked to lower ones (like some Pentium MMX), I just set multi to 5.5 and set FSB to 133 MHz, now my 700/100/1M works at 733/133/1M 😀 Next is to get the 5/12v Xeons working at 2.8v board, already dug through datasheets and OCVR's (on-board voltage regulators) seem to be different between the types. At first I though it was usual Intel stuff (playing with RESET# pins, HV_EN etc) but seems that it requires OCVR swap (there's two of them), CPU change itself (BGA rework) or just separate VCC_CORE pins from CPU and feed it with 5v supply (5/12v works with either). Will test the CPU's with my Dell Precision 620 board, don't want to risk at all with Asus one 😀

Attachments

  • XG-DLS_133_1.jpg
    Filename
    XG-DLS_133_1.jpg
    File size
    166.91 KiB
    Views
    660 views
    File license
    Fair use/fair dealing exception
  • XG-DLS_133_2.jpg
    Filename
    XG-DLS_133_2.jpg
    File size
    276.66 KiB
    Views
    660 views
    File license
    Fair use/fair dealing exception

Many 3Dfx and Pentium III-S stuff.
My amibay FS thread: www.amibay.com/showthread.php?88030-Man ... -370-dual)

Reply 15 of 28, by Paadam

User metadata
Rank Member
Rank
Member

Yes, me too!

But just getting it running was not my overall goal, I also wanted to get it running at 133 MHz FSB and that is also done. Next move is to get those 900mhz Xeons running at 133MHz FSB at max multiplier it runs stable. Hope that they work at 1 GHz (7.5x133) and do some testing how FSB affects pereformance of this uber cool CPU.

Tried different RAM sticks and to my surprise the board happily POSTed with 2x 1GB sticks, though adding more sticks caused the memory count to register just 1 GB total so I left it at 2x 1GB now.

Oh, and I need to source CPU retention mechanisms, these were sadly not included with the board (just the brackets and holders).

Many 3Dfx and Pentium III-S stuff.
My amibay FS thread: www.amibay.com/showthread.php?88030-Man ... -370-dual)

Reply 16 of 28, by Paadam

User metadata
Rank Member
Rank
Member

While I get the 900MHz CPU's to run I was playing a bit with the 700 MHz one. Since they are known to have dried out thermal paste I removed the heat spreader (not the big one that covers whole PCB but the one on the CPU core itself) with a sharp knife. Cleaned it, applied new paste and assembled the cartridge.
It worked on all multis up to 7 (factory max) at 133 MHz FSB which means it posted nicely at 933 MHz but failed to POST at 1 GHz. Maybe I will install OS with the single 700 MHz one to check the real world stability but so far results are promising. Real vCore (the one that goes to core not the one that goes to cartridge that is 2.8v) is 1.67volts and 900MHz ones also have it at 1.7v range, so quite typical with 0.18um manufacturing process like Coppermines.

It seems even more that 100 MHz bus speed is such a waste on those insane CPU's, beginning to think Intel deliberately left them 100 MHz so it does not compete with Foster Xeon and also probably because they did not have ecent chipsets themselves for the 2M Xeons. Serverworks had but most likely Intels greed as one factor.

Attachments

  • 700xeon.jpg
    Filename
    700xeon.jpg
    File size
    187.53 KiB
    Views
    601 views
    File license
    Fair use/fair dealing exception

Many 3Dfx and Pentium III-S stuff.
My amibay FS thread: www.amibay.com/showthread.php?88030-Man ... -370-dual)

Reply 17 of 28, by Paadam

User metadata
Rank Member
Rank
Member

Played a bit with the XG-DLS. Dragged an industrial PC desktop case (this thing is heavy!) from storage, put that board into it, installed XP on IDE drive (later will use SCSI RAID 0) and test it.

As I feared in the beginning, XG-DLS is quite sensitive to PSU's. It would work probably ok at 100MHz bus speed but at 133 MHz VIO was at lousy 3.2volts and that was with Fortron 400w unit. Examining the XG-DLS I found out that it does not have on board regulated 3.3v supply (unlike other boards, like P2B-DS etc) where you can adjust VIO with pulldown resistors and on P3B-F there is even a jumper to adjust VIO (3.3v).
XG-DLS depends entirely on PSU in that regards so I needed to raise IO voltage.
On the Fortron unit I have rails are not adjustable separately so I used Chieftec 550W unit and set VIO to 3.5v with internal pot. That is helpful btw with BX chipset boards too, especially when all 4 DIMMs are in use.

Earlier the board was not stable at high loads (would lock up in the middle of 3DMark2001SE), now it passed it with flying colors. My 700MHz CPU runs perfectly stable at 866 MHz (6.5x133), had it running with Sandra burn.in test for hour or som, maybe later will test it on 933 MHz too unless I finish the 900 MHz CPU project sooner.
Tested different graphics cards, had Ti4600, FX5900 XT and Quadro FX 1100 (which is basically 5800). 5900 XT got 7300 points and Quadro FX 1100 got 7080 points so I left Quadro FX in as it smeed like more approriate for dual Xeon machine 😀 And it has green PCB which fits better with XG-DLS green PCB (5900 has blue PCB).

Many 3Dfx and Pentium III-S stuff.
My amibay FS thread: www.amibay.com/showthread.php?88030-Man ... -370-dual)

Reply 18 of 28, by kixs

User metadata
Rank l33t
Rank
l33t

It's good to see the board got into the right hands... You certainly know your way with hardware hacks 😀

Meanwhile 486 boards are still waiting for some spare time 🙁

My Amibay: SALE | BUY - Updated on 2021-07-06 | Requests also possible

Reply 19 of 28, by luckybob

User metadata
Rank l33t
Rank
l33t

question: are you using the aux 3.3v supply connector?

Also, i ran my board on a new corsair 400w supply, and never once had any issues. Then again, it was working as a router, and not really using any video, but still it ran for years without a restart.

It is a mistake to think you can solve any major problems just with potatoes.