VOGONS


Outrigger: an ongoing adventure with the OR840

Topic actions

Reply 220 of 243, by slivercr

User metadata
Rank Member
Rank
Member

Welcome to vogons, d4m3d!

d4m3d wrote on 2021-07-01, 13:18:
... so, i've got the workstation in this config: single 933/133/256 slot1 cpu, dummy in the other slot, dual 128mb 40ns rimms wi […]
Show full quote

...
so, i've got the workstation in this config: single 933/133/256 slot1 cpu, dummy in the other slot, dual 128mb 40ns rimms with no crimms in the other 2 slots.
it was sold as 'broken (not starting)', but i have figured out it was just the battery so it's now starting, giving the error beeps about the absent memory modules.
i am about to buy some memory now (serial number is 750711-702, so i guess 512mb rims is the way to go), and need a piece of advice - i have read the Hannibal Ram Guide about rimms and basically it says that more rimm modules = more latency (cause rimm uses serial principle of data transfer).
my question is - will using crims in slots 2 and 4 improve latency? i know its a silly one but i dont want to spend extra money on another 2x512 set if dual channel mode does not require you to populate all the slots and there may be even performance hit. 1gb is plenty for win2k.
there is also 40ns vs 45ns performance difference, i could not find any benchmarks anywhere but im just curious if there is any (if or840 even supports 40n timings)

I have not measured the difference in using CRIMMs instead of RIMMs or if it would improve latency in any significant way, so I can't really answer your question. The board does accept 40 ns RIMMs, I use modules with that timing.

I can confirm there are measurable differences between 40 ns vs 45 ns, and also between non-ECC and ECC modules. This is in both synthetics (SiSoft Sandra, AIDA64, 7zip benchmark) and real applications (Quake3).

Keep in mind that, as it was said before, the board may not boot with 512 MB modules right away, so you need to get CRIMMs or another set of small RIMMs (< 256 MB) to boot it and flash the newest BIOS. Supposing you buy a 512 MB set, and a CRIMMs set, you could test the scenario you propose and tell us)

Have fun with the board!

Outrigger: an ongoing adventure with the OR840
QuForce FX 5800: turn your Quadro into a GeForce

Reply 221 of 243, by freekymachine

User metadata
Rank Newbie
Rank
Newbie

Ah funny, still alive the or840

I refound it due surfing , 2 people sell currently 2 Machines on Ebay ;
https://www.ebay.de/itm/115065487275?hash=ite … sK8AAOSw5XVfzSI~
https://www.ebay.de/itm/324858967959?hash=ite … 7QAAOSwLpFhfu-o

bit expansive, but mabye pressure them a bit hard 😀 "nobody" is buying RD-Ram - i bought modules in germany, where people had them listed for up to 3 years and i was the first to ask..some i even got for free, just needed to pay shipping, some 4x 512MB i got for 15 € incl shipping. If they are on Ebay for 10 €, nobody buys them, Rambus is such a freak-corner, barely anyone had it, used it, knows about it.

going and going the OR840^^- i remember i posted years ago, but still didnt do anything with my OR840 - i mean, i would just need to put it back together, powerleap adapters i have, the p3 1266 too, didnt had much rd-ram left, sold em all but..i went mad, due...nobody wants anymore rd-ram modules, you can get an infinite ammount for like 1,99 € or 3,52 Dollars - so i did pile up some again, just as good memory^^
and built fun-shelf^^ Even found lots of 1GB PC1066er Modules. https://imgur.com/a/5MHZ1l1

Impressive to see what youre all still doing 😀 especially the rd-ram overclock.

Not to strive off, but i still like the Compaq SP750 Board, max with 2x P3 1000 Xeon, but more space for Rambus, 4Gb, SCSI Onboard and even PCI-X. In my opinion the best 2000er machine..
On the other Hand..the OR840 ran with Tualatins - i remember when i had both running.."where do i play"..man what 1st World Problems 😀

https://abload.de/image.php?img=img_09646zk22.jpg
https://abload.de/image.php?img=img_09650cj7n.jpg

Well enjoy - should keep looking again and maybe fix up both machines again

Reply 222 of 243, by astigmatism

User metadata
Rank Newbie
Rank
Newbie

An amazing resource for the OR840 slivercr! As an authority on this board, perhaps you don't mind a necro-post? 😉 I imagine as more of these boards come into the hands of collectors, you'll get more.

My board is missing the DC-to-DC adapter. This adapter sits just above the ATX power connector and presumably came installed on the board. I can't find anything about this part other than the IC used on the adapter is a "12S3312A". Do you have any additional information on this? A part number would be perfect, perhaps its given on the underside of the part (I haven't seen pictures of that). Also - is it even necessary at all? I've used dual-proc boards with VRM's which were only required when the 2nd CPU was present. I suspect it is required but would love to hear what you know about it.

Anyway, crossing my fingers for a working board. Like you I intend to build out a rig with Coppermine's (P3 1000/133) for period's sake

Reply 223 of 243, by pshipkov

User metadata
Rank Oldbie
Rank
Oldbie

Got rid of a Type 6868 motherboard last week.
It had one of these adapters on it, but i think i have spare. Will check and let you know.
Bought it (the adapters) at end of 2020 beginning of 2021 from some online store selling old server hardware. High chance they still have some.
Remember that locating it was not easy - some obscure naming at play.

retro bits and bytes

Reply 226 of 243, by astigmatism

User metadata
Rank Newbie
Rank
Newbie

Thanks for the quick help everyone 😀 I've found some Artesyn parts catalogs (without pictures) so I'll get started there.

Bought it (the adapters) at end of 2020 beginning of 2021 from some online store selling old server hardware. High chance they still have some.

If you have the vendor and item/part in your internet search history this would be incredibly helpful!!

Reply 227 of 243, by PentAmd

User metadata
Rank Newbie
Rank
Newbie

Yesterday I Memtest-ed my intel OR840 motherboard with 4x512MB rambus (PC-800) memory. I received "new" memory moduls from EBay.
The mobo is dual channel, so the max RAM speed must be about 3200 Megabytes/s
According to the math: 400Mhz (memory base freq) * 2 (both edge of the clock) * 16bit (rambus memory module bit-width) * 2 (interleves - dual channel) = 25600 Megabit/s = 3200MByte/s
I don't know how Memtest86 calculates the memory speed, but according to the picture it is far away from what I expected. It is 13% of the theoretical maximum.

I think the theoretical memory access speed is limited between the i840 chipset and the CPU.
The FSB is 133MHz, and the P3 CPU has 64 Data lines.
Math: 133MHz * 1 (only falling edge of the clock) * 64bit = 8512 Megabit/s = 1064 MB/s
(for a P4 1600MHz it looks like so: 100Mhz (base frequency) * 4 (quad data rate, 4 sampling during one clock cycle) * 64 bit = 25.600 Megabit/s = 3200MByte/s . So the RAMBUS PC-800 is constructed rather for this CPU)

That can be valid only, if the CPU communicates ONLY with the Memory. In the real world the CPU in between is polling the I/O channel as well, which uses the same 64bit databus of the CPU.

I had the question yesterday, why using Rambus with P3 if the 133 MHZ FSB has so slow memory access.
I think it is because the OR840 MOBO has 2 CPU capability, so both CPU can transfer from/to memory without speed drawback.

I have also a dual P3 MOBO with 4x512 Kingston SDRAM (133Mhz) where the max speed is around
Math: 133MHz * 1 (only falling edge of the clock) * 64bit (memory module) * 1 (interleaves - single channel)= 8512 Megabit/s = 1064MByte/s

Important to mention, that the CPU is not reading memory itself on demand. The CPU is filling it's cache with block of data from the i840 chip. The i840 has a very fast (3 times faster than the CPU) memory access, so I can imagine that after i840 gets the memory content very fast and buffers it, the i840 is feeding the P3 slowly 😁

I am curious if the i840 chip can feed the two P3 concurrently, I mean whether the data lines are shared between the 2 cpu, or separate.

Attachments

  • IMG_7540.JPG
    Filename
    IMG_7540.JPG
    File size
    117.83 KiB
    Views
    399 views
    File license
    Public domain

Reply 228 of 243, by PentAmd

User metadata
Rank Newbie
Rank
Newbie

I investigated my last question by reading the manual of the i840 chipset from this PDF:
http://ww.ic72.com/pdf_file/8/81828.pdf

According to the picture below, there is only one shared address/data bus between the 2 processors.
That means, no matter how fast our RAMBUS ram is, every time when CPU1 and CPU2 wants to access to the memory in the same time the one or the other has to wait until the i840 feeds the one.

The 2 CPU does not have shared "L3" cache so if the CPU2 wants to access the same memory where CPU1 already wrote some data, CPU2 has to wait, until CPU1 is writing back the content of the cache into the RAM, by using the slow FSB speed (ca. 1GB/s) and waiting again until the i840 feeds the CPU2.
Here comes the interesting part: If Windows NT (or 2000, XP) is smart enough, it does not let CPU2 to work on the same thread with the same memory content, to avoid the cache write back thing, which steals a lot of execution time.

I was reading somewhere else, that the P6 architecture (P Pro, P2, P3) has the so called "out of order execution" which means, if the cpu wants to access a memory, which is not available in the cache, then during the waiting, the CPU executes the codes, after this memory request, and temporarly stores the values somewhere, and then stops, until the memory content is available in the cache. Then executes the instruction with the memory (which is available now), and then uses the prevoius stored result immediately. Unfortunately it is not much help, if the following instructions are working only with the memory, which is missing from cache.
More info about out-of-order https://slideplayer.com/slide/9408677/

I was also thinking, why the OR840 has so fast memory access (Rambus PC-800), which is 3 times faster than the FSB. The answer probably they wanted to make a 2 times faster access( because we have 2 CPU), with the PC-600 memory, but they probably decide to let PC-800 also working with the board.

The only way to have the advantage of that fast dual channel rambus memory, is to use a P4 processor with the quad data rate FSB, which datarate I calculated in my last post.

I am going to do some benchmarks soon, where I test the single/dual processor performance, and also the SDRAM/RAMBUS differences. Probably there will be not much difference.
I am also curious how GTA3 runs with the single/dual processor set on different Geforce and Ati cards from that time.

Attachments

  • Block_DGR.png
    Filename
    Block_DGR.png
    File size
    85.91 KiB
    Views
    371 views
    File license
    Public domain

Reply 229 of 243, by Xray1281

User metadata
Rank Newbie
Rank
Newbie

Idk if anyone here can help but I’ve been having an ongoing battle with this motherboard. I simply can’t get it to read anything. I’ve been trying to install a bios update, windows 98 se, and I’ve tried everything I can think of. I have no idea what’s going wrong but I really want to get this computer running. I’ll take any tips on how to get this motherboard or any disk readers working…

Reply 230 of 243, by chrismeyer6

User metadata
Rank l33t
Rank
l33t
Xray1281 wrote on 2022-06-15, 21:16:

Idk if anyone here can help but I’ve been having an ongoing battle with this motherboard. I simply can’t get it to read anything. I’ve been trying to install a bios update, windows 98 se, and I’ve tried everything I can think of. I have no idea what’s going wrong but I really want to get this computer running. I’ll take any tips on how to get this motherboard or any disk readers working…

I'd first make sure the bios battery is good And then do a clear cmos just to make sure there aren't any weird settings. I've had quite a few motherboards act really weird with a dead/missing or low bios battery.

Reply 231 of 243, by Xray1281

User metadata
Rank Newbie
Rank
Newbie
chrismeyer6 wrote on 2022-06-15, 21:28:
Xray1281 wrote on 2022-06-15, 21:16:

Idk if anyone here can help but I’ve been having an ongoing battle with this motherboard. I simply can’t get it to read anything. I’ve been trying to install a bios update, windows 98 se, and I’ve tried everything I can think of. I have no idea what’s going wrong but I really want to get this computer running. I’ll take any tips on how to get this motherboard or any disk readers working…

I'd first make sure the bios battery is good And then do a clear cmos just to make sure there aren't any weird settings. I've had quite a few motherboards act really weird with a dead/missing or low bios battery.

I will do that. Here’s a previous conversation I had about this problem if you think it’ll shed any light. Re: Configuration questions

Reply 232 of 243, by chrismeyer6

User metadata
Rank l33t
Rank
l33t
Xray1281 wrote on 2022-06-15, 23:27:
chrismeyer6 wrote on 2022-06-15, 21:28:
Xray1281 wrote on 2022-06-15, 21:16:

Idk if anyone here can help but I’ve been having an ongoing battle with this motherboard. I simply can’t get it to read anything. I’ve been trying to install a bios update, windows 98 se, and I’ve tried everything I can think of. I have no idea what’s going wrong but I really want to get this computer running. I’ll take any tips on how to get this motherboard or any disk readers working…

I'd first make sure the bios battery is good And then do a clear cmos just to make sure there aren't any weird settings. I've had quite a few motherboards act really weird with a dead/missing or low bios battery.

I will do that. Here’s a previous conversation I had about this problem if you think it’ll shed any light. Re: Configuration questions

I did read through that thread. I'd still start with making sure your bios battery isn't on the low side as it can still cause all kinds of wonkyness even if it's saving settings and keeping time. You can also try pulling the battery over night then the next day reinstall the battery and still do a cmos clear with the jumper and try again. Do that has made boards that we're working properly be just fine afterwards. I've seen some crazy things when systems have low, dead, or missing batteries.

Reply 233 of 243, by Xray1281

User metadata
Rank Newbie
Rank
Newbie
chrismeyer6 wrote on 2022-06-15, 23:42:
Xray1281 wrote on 2022-06-15, 23:27:
chrismeyer6 wrote on 2022-06-15, 21:28:

I'd first make sure the bios battery is good And then do a clear cmos just to make sure there aren't any weird settings. I've had quite a few motherboards act really weird with a dead/missing or low bios battery.

I will do that. Here’s a previous conversation I had about this problem if you think it’ll shed any light. Re: Configuration questions

I did read through that thread. I'd still start with making sure your bios battery isn't on the low side as it can still cause all kinds of wonkyness even if it's saving settings and keeping time. You can also try pulling the battery over night then the next day reinstall the battery and still do a cmos clear with the jumper and try again. Do that has made boards that we're working properly be just fine afterwards. I've seen some crazy things when systems have low, dead, or missing batteries.

Thank you for the help. The batter was recently replaced. I’m not sure if I reset the cmos correctly as it did not reset the bios to the factory original. I’m going to get a new cmos battery anyways to see if that does anything. I hope it does because I’ve been trying to get this computer up and running for half a year now. It’s most likely something stupid that I’m doing but who knows.

Reply 234 of 243, by PentAmd

User metadata
Rank Newbie
Rank
Newbie
PentAmd wrote on 2022-06-08, 11:41:

Yesterday I Memtest-ed my intel OR840 motherboard with 4x512MB rambus (PC-800) memory. I received "new" memory moduls from EBay.

I got now time to install a Windows XP SP3 on my OR840 with a dual 933MHz Pentium processor.
I used a Matrox Millenium PCI VGA card for the first trial.

The XP installer was going without any problem.
The LAN and USB was immediately working, there were no unknow device in the Device Manager.

I decided to put an Abit Siluro Geforce 4 Ti 4200 AGP card in it.
After the first boot windows was recognised the Nvidia card, and activated the WDM driver.
After a restart the WinXP load picture shows up, and then as the 3 LEDs on the keyboard lit on, the system restarts. And it hapens over and over again in endless loop.
In the WinXP event log i did not find anything. And this is not a blue-screen-of-death, because there is no minidump log in the proper folder.
I don't know what to do.

I could start windows in safe mode.
I could start windows in normal mode with the PCI Matrox card.
Just for test, I tried a Vanta TNT 16MB card but the same happend as with the Ti 4200.

Reply 235 of 243, by PentAmd

User metadata
Rank Newbie
Rank
Newbie
PentAmd wrote on 2022-06-22, 20:29:

After a restart the WinXP load picture shows up, and then as the 3 LEDs on the keyboard lit on, the system restarts. And it hapens over and over again in endless loop.

I tried to instal WinXP SP2, but the same happenes.
I selected "SOS" boot without GUI (Graphich User Interface) in the "msconfig", and I was able to see, that the boot was reached the AGP440.sys was loaded, which is probably the last driver, before the session manager starts. The problem is, that the bootlog is not saving after every event, probably Win is going to save the log later, where the restart already happened. So I cannot see what happens between the session manager start and the reboot.

I tried the "VGA" mode, and the system was able to boot, but that also means, that there is a problem as the card starts with 100% performance.

The card itself is working in another config.
I need a workaround to solve the problem, or at least find the root cause.

Reply 236 of 243, by PentAmd

User metadata
Rank Newbie
Rank
Newbie
PentAmd wrote on 2022-06-22, 22:03:

I need a workaround to solve the problem, or at least find the root cause.

Update:
I booted in VGA mode. And installed the Detonator W2K_XP_V29.90 Driver.
After reboot everything was fine. I could see the spinning cube in the dxdiag 😁

During the driver install the SW asked me to replace some "newer" files in the win folder. I said "sure!". That was the magic, I think.

Reply 237 of 243, by PentAmd

User metadata
Rank Newbie
Rank
Newbie
PentAmd wrote on 2022-06-22, 22:56:

After reboot everything was fine. I could see the spinning cube in the dxdiag 😁

My happyness was to early. After the next boot it happened again.
For testing purpose I disabled the 2nd processor, and the Windows boot was successful, many times.

I thought first that my 450W PSU was not good enough, so I changed to a 500W one. But it helped nothing.

I have the feeleing, that this Nvidia driver for the Ti 4200 maybe is not written for multiprocessing (multicore) computers. I mean in 2002 May (the Detonator version came out) there were not much multiprocessor computers on the market for gamers.
I am going to check the latest driver.