andyc wrote on 2020-12-24, 13:55:

Paadam wrote on 2016-09-18, 08:24:

I have Dell Precision 620 board with PSU and accessories (pretty much everything except the case), I am planning to get the board to run 2MB CPU's on 133 MHz bus (with lower multiplier). On Dell you set the CPU speed with floppy disk (utrility called sprfrq2.exe). It basically sets multiplier and the resulted speed cannot exceed rated CPU speed, otherwise the board wont' boot iirc. I had a complete 933 MHz Precision 620 once and experimented with it, though did not have 100 Mhz CPU's to test. But it had CPU speed option in BIOS ("Compatibility") which seems to be 100 MHz speed setting and on the board there is unsoldered jumper stating "100 Mhz" so it is possible that it works.
Though not sure I can get it to work at 133 Mhz FSB. I can live with 700 or 900 MHz CPU but I want it to work at 133 MHz bus speed 😀 100 MHz seems such a waste for that reat CPU.

Hi! I also have a Precision 620 currently running 2 x 1Ghz 256K 133Mhz Xeons. Would 2 x 900Mhz 2MB work with it? I would like to see how they perform vs the 2 x 1Ghz that i have and don't want to buy them if they don't work since they can get quite expensive...
Thanks!

I'm pretty sure the 2MB Cascades would work even though they were not meant to be used with the i840 Carmel workstation chipset. I've successfully upgraded my Compaq 840 based workstation with a pair of 2MB 900MHz Xeons and did a comprehensive comparison with the old 1GHz 256K Xeons it used to have.
In every single test 900/100/2M was faster than 1000/133/256K averaging about 20%. In some games such as Quake III the 900 was faster by 60%!
Also what's pretty impressive, in several tests these 2MB monsters were faster than the Tualatin-S 1.4/512K and even Rambus-based 1.5GHz Willamette P4.

My machine also has a hidden 100MHz jumper that forces 1GHz to work at 750MHz but it won't overclock 100 to 133. I was hoping to get a 933/2M by forcing 700/2M to run on a 133MHz bus but unfortunately couldn't find a way to do so. Still, even though 2MB Xeons run at 100MHz bus they're fantastic chips that can put much higher clocked siblings to shame. You would never guess that a 900MHz P3 could beat 1.4GHz P3.

I don't suppose anyone knows where to get a CPU retention mech. compatible with a MS440GX? It's been 10 years now...

https://creality3d.shop/collections/ender-ser … -pro-3d-printer

its what i use

Mike V8 wrote on 2021-10-06, 21:30:

I'm pretty sure the 2MB Cascades would work even though they were not meant to be used with the i840 Carmel workstation chipset. […]
Show full quote

andyc wrote on 2020-12-24, 13:55:

Paadam wrote on 2016-09-18, 08:24:

I have Dell Precision 620 board with PSU and accessories (pretty much everything except the case), I am planning to get the board to run 2MB CPU's on 133 MHz bus (with lower multiplier). On Dell you set the CPU speed with floppy disk (utrility called sprfrq2.exe). It basically sets multiplier and the resulted speed cannot exceed rated CPU speed, otherwise the board wont' boot iirc. I had a complete 933 MHz Precision 620 once and experimented with it, though did not have 100 Mhz CPU's to test. But it had CPU speed option in BIOS ("Compatibility") which seems to be 100 MHz speed setting and on the board there is unsoldered jumper stating "100 Mhz" so it is possible that it works.
Though not sure I can get it to work at 133 Mhz FSB. I can live with 700 or 900 MHz CPU but I want it to work at 133 MHz bus speed 😀 100 MHz seems such a waste for that reat CPU.

Hi! I also have a Precision 620 currently running 2 x 1Ghz 256K 133Mhz Xeons. Would 2 x 900Mhz 2MB work with it? I would like to see how they perform vs the 2 x 1Ghz that i have and don't want to buy them if they don't work since they can get quite expensive...
Thanks!

I'm pretty sure the 2MB Cascades would work even though they were not meant to be used with the i840 Carmel workstation chipset. I've successfully upgraded my Compaq 840 based workstation with a pair of 2MB 900MHz Xeons and did a comprehensive comparison with the old 1GHz 256K Xeons it used to have.
In every single test 900/100/2M was faster than 1000/133/256K averaging about 20%. In some games such as Quake III the 900 was faster by 60%!
Also what's pretty impressive, in several tests these 2MB monsters were faster than the Tualatin-S 1.4/512K and even Rambus-based 1.5GHz Willamette P4.

My machine also has a hidden 100MHz jumper that forces 1GHz to work at 750MHz but it won't overclock 100 to 133. I was hoping to get a 933/2M by forcing 700/2M to run on a 133MHz bus but unfortunately couldn't find a way to do so. Still, even though 2MB Xeons run at 100MHz bus they're fantastic chips that can put much higher clocked siblings to shame. You would never guess that a 900MHz P3 could beat 1.4GHz P3.

Cool! Thanks for the info. I already have a Tualatin-S 1.4 in a D815EEA2 mobo and a RAMBUS Willamette is on my list. I'll start searching for a pair of 2MB Xeons for now.

luckybob wrote on 2021-10-07, 00:19:

https://creality3d.shop/collections/ender-ser … -pro-3d-printer

its what i use

thanks. Do you have the files to make the brackets? Or do I have to create them?

oh, I haven't made one, all my boards have their brackets.

however, if I was going to do it, it's a relatively simple process. just take a standard pentium 2 bracket and stretch it to fit the slot 2. I don't think I own a MS440GX, so I can't do it here.

My compaq sp750 works with dual SL4XZ 900/2 @ 5/12V with i840/rambus and it is sometimes faster than dual Tual 1,4/512@sdram reg @via694

And 900s are much faster than Cascades 1GHz/256/133 which were original cpus in SP750

Also i have working s2dg2 with V5 5500 with dual tanner 550/1MB
which they wait for todays swap to 550/2MB.
It for sure supports only 2,8V Xeons. I’ve tested it with 700/1MB. It boots @700 without jumpers touching with microcode error but works ok.
unfortunately I don't even have one 2,8V 900 to test it with it.
I’ve got two MS440GX (one made for ibm Intellistation Z Pro) but again i didn’t tested it even once…

Callahan wrote on 2021-10-30, 10:15:

My compaq sp750 works with dual SL4XZ 900/2 @ 5/12V with i840/rambus

I have never tested my SIV utility on a compaq sp750 and would be interested to see how well it does. It works well on my compaq ap550 so suspect it will do OK.

Please will you try SIV and post the two Menu->File->Save Local files so I can check all is as it should be.

It took me a half an hour to dig out that heavy compaq. But it still fully operational. So here it ts an almost full report.
RIMM spd section is empty...

Here is a short comparison between few my rambus workstations, all with NV10GL/NV10 DDR/NV10SDR
http://3dfx.pl/cgi-bin/yabb2/YaBB.pl?num=1589569986
All with dual cpu, same ram, same OS, same DX: (P3: Cascades/Coppermine), xp sp1,dx 8.1, 1GB ram. VGA driver NV: 45.23

Compaq SP750 i840 NV10GL 64MB DDR 135/166/333DDR
@1000/256/133 s2:
3DM2001se: 2956
3DM2000: 5140
Everest 5.01.1700 MemCopy: 834MB/s MemRead: 1029 MB/s MemWrite: 795 MB/s
7zip 4.57 Test 2cpu: 694MIPS/1265MIPS

@900/2MB/100 s2:
3DM2001se: 3167
3DM2000: 5501
Everest 5.01.1700 MemCopy: 698MB/s MemRead: 783 MB/s MemWrite: 707 MB/s (i think in that case RIMMs makes only 600MHz: FSB:100MHz 3x100*2=600 vs 3x133*2=800)
7zip 4.57 Test 2cpu: 712MIPS/1272MIPS

Quake III 1.32c Demo Four:
r_smp "0"
Fastest: 141 fps
High Quality:113 fps

r_smp "1"
Fastest: 142 fps
High Quality: 95 fps

Compaq AP550 i840 NV10GL 64MB SDR 135/166
@1000/256/133 s1:
3DM2001se 2468
3DM2000: 4507
Everest 5.01.1700 MemCopy: 842 MB/s MemRead: 1040 MB/s MemWrite: 795 MB/s
7zip 4.57 Test 2cpu: 716MIPS/1287MIPS

Quake III 1.32c Demo Four:
r_smp "0"
Fastest: 108 fps
High Quality: 79 fps
r_smp "1"
Fastest: app crash
High Quality: 65fps

HP Kayak XM600 i820 NV10GL 64MB SDR 135/166
@1000/256/133 s370:
3DM2001se: 2449
3DM2000: 4482
Everest 5.01.1700 MemCopy: 784 MB/s MemRead: 1032 MB/s MemWrite: 1032 MB/s
7zip 4.57 Test 2cpu: 686MIPS/1256MIPS

Mam dwie Quadro SDR i jedną Quadro DDR, a więc jednej konfiguracji tu jeszcze nie testowałem...

cdn.

IBM Intellistation M Pro type 6868 i840 2048MB RAM
@2x800/256/133

Geforce 256 32MB (Creative Annihillator) SDR (CT6940) @120/166
3DM2001se: 2269
3DM2000: 4126

Quake III 1.32c Demo Four:
r_smp "0"
Fastest: 92 fps
High Quality: 78 fps

Geforce 256 32MB (Creative Annihillator) DDR (CT6970) @120/143/286DDR
3DM2001se: 2611
3DM2000: 4531

Quake III 1.32c Demo Four:
r_smp "0"
Fastest: 93 fps
High Quality: 83 fps

Everest 5.01.1700 MemCopy: 842 MB/s MemRead: 1031 MB/s MemWrite: 796 MB/s
7zip 4.57 Test 2cpu:602MiPS/1083MiPS

Callahan wrote on 2021-10-31, 10:00:

RIMM SPD section is empty...

Thank you, please may I also have D:\siv_v5.60\SIV_SP750.dmi which is the raw data that the [Machine] panel displays as SIV runs better in test mode when I have this?

I suspect the RIMM SPD is missing as the system has an SMBus multiplexor and I will need to add support for it, but I don't know how to drive it. Do any programs report the RIMM SPD and if so which?

SIV should report the RIMM SPD for your W8000 and if you look at [SMB Bus] you should notice the Xeon PIRs on on SMBus 0 and the RIMM SPD on SMBus 1, SIV supports the W8000 + W6000 muxes.

I noticed that SIV failed to report the CPU temperatures which can be read via the Analog Devices ADM1021 so I will fix this for SIV 5.61 Beta-11.

1[SMB Bus] <- SIV32X - System Information Viewer V5.61 Beta-11 RXP::ray
2
3#  Device Name   |Bus-Numb-Fun|  Port     Multiplexing  RM  GB  SO  CO  Vendor-Dev-Sub_OEM-Rev  [Y]INUSE Wait 0 of 812 ms Compaq W8000
40  W8000 SMBus   [ 0 - 31 - 3 ]  FC00 10  [Y][_]#0      BF              8086-2443-00780E11-04   Intel 82801BA SMBus Controller (ICH2 B4 step)
51  W8000 MUX-A   [ 0 - 31 - 0 ]  FA00     [Y][_]#1      01  18          8086-2440-00000000-04   Intel 82801BA LPC Interface (ICH2 B4 step)
62  MRHR 82803 A  [ 0 - 00 - 0 ]           @ 50->54      20  16  90  94  8086-2531-00000000-04   Intel 82860 Chipset CPU to I/O Hub Bridge
73  MRHR 82803 B  [ 0 - 00 - 0 ]           @ 50->54      40      90  94  8086-2531-00000000-04   Intel 82860 Chipset CPU to I/O Hub Bridge
8
9Slave   SMBus Device   Host     Peak  Data  Hex     Data
10Device  Description    Status   Time  Size  Offset  00  01  02  03  04  05  06  07  08  09  0A  0B  0C  0D  0E  0F
11
12|0_10|                 44         52  0000
13...
14[0_18]  NE1617A (NXP)  42        185  0010  00      35  3F  ..  ..  02  7F  C9  48  C9  __  __  __  __  __  __  __
15...
16[0_4C]  NE1617A (NXP)  42        185  0010  00      3E  49  10  80  02  7F  C9  4B  C9  __  __  __  __  __  __  __
17...
18[0_50]  Xeon PIR Data  42        186  0010  00      02  ..  80  0E  16  25  32  38  70  74  7E  ..  ..  69  20  53
19[0_51]  Xeon PIR Data  42        186  0010  00      02  ..  80  0E  16  25  32  38  70  74  7E  ..  ..  69  20  53
20...
21|0_7F|                 44         52  0000
22|1_10|                 44         52  0000
23...
24[1_37]                 42        185  0010  00      __  __  __  __  __  __  __  __  __  __  __  __  __  __  __  __
25...
26[1_50]  RIMM SPD Data  42        185  0010  00      02  08  01  01  97  C5  05  20  02  05  08  14  0A  08  08  13
27[1_51]  RIMM SPD Data  42        185  0010  00      02  08  01  01  97  C5  05  20  02  05  08  14  0A  08  08  13
28[1_52]  RIMM SPD Data  42        186  0010  00      02  08  01  01  97  C5  05  20  02  05  08  14  0A  08  08  13
29[1_53]  RIMM SPD Data  42        185  0010  00      02  08  01  01  97  C5  05  20  02  05  08  14  0A  08  08  13
30[1_54]  RIMM SPD Data  42        185  0010  00      02  08  01  01  96  C5  05  20  02  01  08  14  0A  08  08  13
31[1_55]  RIMM SPD Data  42        185  0010  00      02  08  01  01  96  C5  05  20  02  01  08  14  0A  08  08  13
32[1_56]  RIMM SPD Data  42        186  0010  00      02  08  01  01  97  C5  05  20  02  05  08  14  0A  08  08  13
33[1_57]  RIMM SPD Data  42        185  0010  00      02  08  01  01  97  C5  05  20  02  05  08  14  0A  08  08  13
34...
35|1_7F|                 44         51  0000
36|2_50|                 20        334  0000
37...
38|2_53|                 20        337  0000
39|3_50|                 20        337  0000
40...
41|3_53|                 20        337  0000
42
43Total of 4 SMBuses and 13 SMBus Devices in 14.993 seconds (0.867 KBps). Slaves Probed 232. [Scan Range]0_10 to 3_80. [.. -> 00 : __ -> FF]
44
45[  OK ]  [ Copy]   [Windows]  [Machine]  [ DIMMs]  [Sensors]  [USB Bus]  [ SPD ]   [PMB Bus]  [Volumes]  [SMB Bus]  [PCI Bus]   [ALL Dev]

Here it is. Just delete .pif extension

Callahan wrote on 2021-10-31, 13:01:

Here it is. Just delete .pif extension

Thank you, I downloaded it eventually, but W10 is none too keen on .pif files and .dmi.txt would be less of a battle.

The attached SIV32X 5.61 Beta-11 should report the CPU temperatures, does it please?

After pondering the RIMM SPD I feel it's worth trying GPIO 18 to see if this controls the SMBus MUX, but it may cause the system to reboot when SIV toggles it. If you would like to try it do SIV32X -SMBMUX=SP750 then check/post what Menu->System->SMB Bus reports. If SMBus 0 has different devices to SMBus 1 then we have found which GPIO to toggle.

Back in around 2010 when I made the W8000 MUX work I did this by trial and error so if it's not GPIO 18 you could do the same using commands such as SIV32X -SMBMUX=GPIO:21 and working up.

For -SMBMUX to work on the SP750 you need to use SIV32X 5.61 Test-11.

Once/if we figure out which GPIO to use I will change SIV to automatically use the MUX.

Callahan wrote on 2021-10-31, 10:00:

RIMM spd section is empty...

After looking at the ICH1 datasheet it can't be GPIO18 as it's not implemented!

Looking at Menu->Hardware->H/W Detail->I/O GPIO then GPIO21 + GPIO22 + GPIO23 + GPIO25 are set as GP Output so I suspect it's one of these we need to toddle.

red-ray wrote on 2021-10-31, 13:31:

T..... For -SMBMUX to work on the SP750 you need to use SIV32X 5.61 Test-11. ..... […]
Show full quote

Callahan wrote on 2021-10-31, 13:01:

Here it is. Just delete .pif extension

T.....
For -SMBMUX to work on the SP750 you need to use SIV32X 5.61 Test-11.
.....

I'm sorry, I do not have enough time to test it now. I will check when the weekend comes

Callahan wrote on 2021-11-03, 18:26:

I'm sorry, I do not have enough time to test it now. I will check when the weekend comes

OK and no problem, I have been wondering about SP750 for a few years, a few days is fine.

I have now released SIV 5.61 Beta-11 so please use that (or later) for the test.

definitely curious whether I could get some better use out of my 800-45 and 800-40 RIMMs in my SP750, bought it a while ago cheap on ebay and it's an absolute dream

being able to get 133MHz FSB on the 2MB xeons would be amazing, but I haven't yet figured out how it could be accomplished other than a heavily modified BIOS

libby wrote on 2021-11-04, 05:08:

being able to get 133MHz FSB on the 2MB Xeons would be amazing, but I haven't yet figured out how it could be accomplished other than a heavily modified BIOS

The 700 + 900 MHz 1 + 2MB Xeons are multiplier locked so I suspect if you installed a 700 MHz one it would try and run at 933 MHz and for a 900MHz one it would be 1200MHz, but I doubt that would work. I can't see why a modified BIOS would be needed.

They have only higher multis locked, multis lower than original (7x or 9x) are unlocked.
I ran 700 MHz 2MB Xeon on my XG-DLS at 133 MHz and it ran fine (at 933 MHz), completely stable. But could not get the system to POST with both of them installed and FSB set to 133 MHz. Using SetFSB resulted in hang even though both CPU's worked just fine at 133 MHz bus. And that did not matter on multi, it would not POST with both CPU's even at 533 MHz.
Not sure whether it was XG-DLS or 440GX specific issue.

Asus XG-DLS POST issues [solved, now working at 133MHz FSB!]

red-ray wrote on 2021-11-04, 07:31:

libby wrote on 2021-11-04, 05:08:

being able to get 133MHz FSB on the 2MB Xeons would be amazing, but I haven't yet figured out how it could be accomplished other than a heavily modified BIOS

The 700 + 900 MHz 1 + 2MB Xeons are multiplier locked so I suspect if you installed a 700 MHz one it would try and run at 933 MHz and for a 900MHz one it would be 1200MHz, but I doubt that would work. I can't see why a modified BIOS would be needed.

it would be in the compaq SP750, which appears to auto-detect the CPU and doesn't permit manual FSB setting from what I can tell.

(if modifying the BIOS, ideally the mod would allow other FSB settings like 112/124 in order to be able to boot without overclocking too far for POST to be achieved)

libby wrote on 2021-11-04, 16:49:

the compaq SP750 appears to auto-detect the CPU and doesn't permit manual FSB setting from what I can tell.

OK, all I have is the June 2000 Part Number 190709-001 Compaq SP750 Hardware Reference Guide which is next to useless, is there a Maintenance Manual or similar please?

BTW would you like to check if SIV32X 5.61 Beta-11 reports the SP750 CPU temperatures and/or help to try to fix SIV to report the RIMM SPD data?

When I tested the changes using the SIV save file they seem to work and now it needs testing on real hardware.