Compaq Alphaserver DS20 - Revitalizing the Mammoth (Part 1)
Compaq Alphaserver DS20 is my first server in the collection. This server is not an x86-compatible computer. It is built on a separate alpha architecture. The server is old by modern standards, it was released in 1999, but at that time it was a very powerful computer. Its configuration is this: processor: Alpha EV6 - 500 MHz, 64 bit, 4 megabytes of cache! 7 hard drives, SCSI Ultra320, a total of 160 gigabytes! RAM maximum 4 gigabytes (I have one and a half gigabytes). Also, there are two power supplies on the server, if one fails, the server automatically starts working on the second. Each power supply is 675 watts! In 1999, this server cost $ 19,000! On the Internet, I found information that such servers were in the Pentagon, and 128 of these dual-processor servers, united in a cluster, were recognized as the fastest computer in 1999! At that time, I was just starting to work (play) on computers, and I would not even be able to imagine such a performance, especially the disk size and the amount of RAM.
The server came to me in a very terrible state, it was stored for some time in the barn (it’s good that without hard drives). Besides the fact that he was falling asleep with dust, the back of the case also began to rust. When I received it, it all started with a general cleanup.
I would also like to say that the server is very heavy, weighs about 50 kilograms, so I took it apart in the yard, cleaned everything and assembled it at home. This whole procedure took several days. When I removed the server covers, a spider sat inside:
I dismantled the entire server for spare parts, the case and side covers, as well as the plastic parts, I washed in the yard with water, and then it all dried out well in the sun. And if possible, wiped all the electronics with cotton wool and alcohol. It was worse with the case and power supplies, as the back of the server began to rust. I had to clean the back of the case with fine sandpaper, then cover everything with WD-40 liquid. And when everything dried up, I painted the back of the server with black spray paint. The power supplies in the back just cleaned up with sandpaper. This is what the server case looks like now:
And these are power supplies, they have 12-centimeter coolers in the back. Each block gives a line of +5 volts 85 amperes! The blocks are very heavy.
Motherboard and processor board. There could be two processor boards in the server, I have only one.
The motherboard has 6 64-bit PCI slots and one ISA bus. There are also SCSI, IDE and floppy interfaces. An interesting situation with USB. On the motherboard there are microcircuits that are signed as USB controllers, however, the USB connectors themselves on the back of the computer are not wired. It is only necessary after installing the system to see if the system itself finds a USB controller, then it makes sense to try to unsolder the USB.
Cleaning the server from dirt and dust was only one problem, another problem was that the server was not working. They warned me about this even when I took it. The previous owner did not understand because of a server failure. And the breakdown was this: if the server was turned on, it started, but after about half a minute it turned off. And it was not a simple button dropping or something else like that. I managed to repair this failure only after a month! First, I started reading manuals on the server. In the “problems” section, it was written that, indeed, the server can shut down 30 seconds after the start if you have problems with voltage, coolers or temperature. Everything was normal with voltage, the coolers were still spinning normally, and nothing seemed to overheat, at first glance everything was fine, but then what’s the matter? Dealing further with the server and with the documentation for it, my attention was drawn to one board, which was mounted separately inside the server case:
A server start button is connected to it, the wires from the power supply are also suitable here, and there are two loops connected to the motherboard. But the most interesting thing is that there are 4 LEDs on this board: “5v ok”, “cpu fan ok”, “sys fan ok” and “temp ok”. And after turning on the diode "sys fan ok" immediately turned off! Everything coincided, but in the manual they write that the server can turn off if something is wrong with the coolers, and here it is immediately clear that something is wrong with the system cooler. But the worse is that it actually spins! All coolers spin normally! Here the worst began, what is wrong? The board says that something is not good with the system cooler, but it works. The cooler, by the way, is connected directly to this board. The connector for its connection has only two contacts: +12 volts and ground. There is not even a speed sensor (which could be the cause of the breakdown). And then there are only two contacts, tension comes in, all is well. And over this problem I racked my brains for about a month. Thinking everything logically, I put together such a picture. When you click on the server’s power button, the electronics on the board include a power supply, which, in turn, turns on the entire computer, everything works, but the board itself at that moment starts to analyze the situation, and here she doesn’t like something with the cooler, it shows LED, that something is wrong, and after 30 seconds sends a power off signal to the power supply - and the server shuts down. Obviously, there are two ways to solve the problem: to figure out why the board does not like the cooler, and the second is to trick the board so that it cannot send a shutdown signal to the power supply. At first I tried to deal with the cooler, but it did not work, I put other coolers, more and less powerful, spread legs of food on different contacts - nothing helped. Therefore, I undertook to "fool" the board. But this is also not an easy task, nothing was clear here, so I started looking for manuals for all the microcircuits that were on it. Finding manuals on electronic components helps the site perfectlywww.alldatasheet.comIt turned out that on this board there is a microcontroller, a microcontroller - it's such a microcomputer, a processor, random access memory and read-only memory, I / O ports - and that's all in one chip! I thought that probably this particular microcontroller “counts” the time for 30 seconds, and then sends a shutdown signal to the power supply. A friend told me that you can disconnect the microcontroller from the power supply, then it will not work, and, obviously, no one will count the time, and the server will not shut down. So I tried to do it, found the power contact, measured it for verification, indeed, there was +5 volts, then I unsoldered the contact, turned on the server and ... It turned off after 30 seconds. :( It was sad. From despair, I began to turn off the power in turn for all the microcircuits, and when I turned off some kind of microcircuit, the server no longer turned off! It was a victory. :) Here it is,
Now the server turned on! Finally, you can play with him. :) But now the first minute of joy passed, and the server continued to resist further - it did not display anything on the screen! The blue screen just stood and the cursor blinked on top. But it was already relatively simple, I read on the Internet that it can display information not on the screen, but in the terminal, which must be connected to the COM port. On the second day, I soldered a null modem cable and connected the server to an old laptop with 2000 Windows, where I launched the terminal program. He took the settings from the Internet, “speed” - 9600 bits, “flow control” - no. I turned on the server and saw the inscriptions that the server started in the terminal program!
512 Meg of system memory
probing hose 1, PCI
probing hose 0, PCI
probing PCI-to-ISA bridge, bus 1
bus 0, slot 5, function 1 - dqa - Cypress 82C693 IDE
bus 0, slot 5, function 2 - dqb - Cypress 82C693 IDE
Testing the System
Testing the Memory
System Temperature is 21 degrees C
initializing GCT / FRU at offset 1b8000
AlphaServer DS20 500 MHz Console V5.4-23, May 19 1999 14:48:15
CPU 0 booting
(boot dk * 100.1.0.2001.1 -flags A)
failed to open dk * 100.1.0.2001.1
What to do next I was told by a friend from the Internet . You need to press "Ctrl + C" to abort the download attempt and go to the SRM console, and there run the command "set console graphics", which will switch the output from the terminal to the monitor. Here's what the first launch looked like. On the laptop in the terminal, old data still hangs, and new ones are already displayed on the monitor.
Separately, I want to say about the SRM console, this is something like an operating system and BIOS combined. Here you can change the configuration and parameters, as well as revise the configuration of the server and devices. You can test the server and do many different things. I know only a small fraction of what you can do there, because I have not yet found the manual on the SRM console. And if you type alphabios in the console, the server BIOS will start. Further, in a few days I already got used to it; the FreeBSD and OpenVMS operating systems for alpha systems were found, however, none of them were installed. :( FreeBSD reached the breakdown of hard disks, and for some reason did not see any, and when I called any installation function, it crashed into the console. And OpenVMS wanted a BIOS update, which I could not do, because the update program freezes.
But recently I was given two DVD discs packed with the images of operating systems for alpha servers. So we are waiting for new adventures. In the next part I will write how I installed the operating systems on this server (when I can do this, since the situation is not good at all)
The server came to me in a very terrible state, it was stored for some time in the barn (it’s good that without hard drives). Besides the fact that he was falling asleep with dust, the back of the case also began to rust. When I received it, it all started with a general cleanup.
I would also like to say that the server is very heavy, weighs about 50 kilograms, so I took it apart in the yard, cleaned everything and assembled it at home. This whole procedure took several days. When I removed the server covers, a spider sat inside:
I dismantled the entire server for spare parts, the case and side covers, as well as the plastic parts, I washed in the yard with water, and then it all dried out well in the sun. And if possible, wiped all the electronics with cotton wool and alcohol. It was worse with the case and power supplies, as the back of the server began to rust. I had to clean the back of the case with fine sandpaper, then cover everything with WD-40 liquid. And when everything dried up, I painted the back of the server with black spray paint. The power supplies in the back just cleaned up with sandpaper. This is what the server case looks like now:
And these are power supplies, they have 12-centimeter coolers in the back. Each block gives a line of +5 volts 85 amperes! The blocks are very heavy.
Motherboard and processor board. There could be two processor boards in the server, I have only one.
The motherboard has 6 64-bit PCI slots and one ISA bus. There are also SCSI, IDE and floppy interfaces. An interesting situation with USB. On the motherboard there are microcircuits that are signed as USB controllers, however, the USB connectors themselves on the back of the computer are not wired. It is only necessary after installing the system to see if the system itself finds a USB controller, then it makes sense to try to unsolder the USB.
Cleaning the server from dirt and dust was only one problem, another problem was that the server was not working. They warned me about this even when I took it. The previous owner did not understand because of a server failure. And the breakdown was this: if the server was turned on, it started, but after about half a minute it turned off. And it was not a simple button dropping or something else like that. I managed to repair this failure only after a month! First, I started reading manuals on the server. In the “problems” section, it was written that, indeed, the server can shut down 30 seconds after the start if you have problems with voltage, coolers or temperature. Everything was normal with voltage, the coolers were still spinning normally, and nothing seemed to overheat, at first glance everything was fine, but then what’s the matter? Dealing further with the server and with the documentation for it, my attention was drawn to one board, which was mounted separately inside the server case:
A server start button is connected to it, the wires from the power supply are also suitable here, and there are two loops connected to the motherboard. But the most interesting thing is that there are 4 LEDs on this board: “5v ok”, “cpu fan ok”, “sys fan ok” and “temp ok”. And after turning on the diode "sys fan ok" immediately turned off! Everything coincided, but in the manual they write that the server can turn off if something is wrong with the coolers, and here it is immediately clear that something is wrong with the system cooler. But the worse is that it actually spins! All coolers spin normally! Here the worst began, what is wrong? The board says that something is not good with the system cooler, but it works. The cooler, by the way, is connected directly to this board. The connector for its connection has only two contacts: +12 volts and ground. There is not even a speed sensor (which could be the cause of the breakdown). And then there are only two contacts, tension comes in, all is well. And over this problem I racked my brains for about a month. Thinking everything logically, I put together such a picture. When you click on the server’s power button, the electronics on the board include a power supply, which, in turn, turns on the entire computer, everything works, but the board itself at that moment starts to analyze the situation, and here she doesn’t like something with the cooler, it shows LED, that something is wrong, and after 30 seconds sends a power off signal to the power supply - and the server shuts down. Obviously, there are two ways to solve the problem: to figure out why the board does not like the cooler, and the second is to trick the board so that it cannot send a shutdown signal to the power supply. At first I tried to deal with the cooler, but it did not work, I put other coolers, more and less powerful, spread legs of food on different contacts - nothing helped. Therefore, I undertook to "fool" the board. But this is also not an easy task, nothing was clear here, so I started looking for manuals for all the microcircuits that were on it. Finding manuals on electronic components helps the site perfectlywww.alldatasheet.comIt turned out that on this board there is a microcontroller, a microcontroller - it's such a microcomputer, a processor, random access memory and read-only memory, I / O ports - and that's all in one chip! I thought that probably this particular microcontroller “counts” the time for 30 seconds, and then sends a shutdown signal to the power supply. A friend told me that you can disconnect the microcontroller from the power supply, then it will not work, and, obviously, no one will count the time, and the server will not shut down. So I tried to do it, found the power contact, measured it for verification, indeed, there was +5 volts, then I unsoldered the contact, turned on the server and ... It turned off after 30 seconds. :( It was sad. From despair, I began to turn off the power in turn for all the microcircuits, and when I turned off some kind of microcircuit, the server no longer turned off! It was a victory. :) Here it is,
Now the server turned on! Finally, you can play with him. :) But now the first minute of joy passed, and the server continued to resist further - it did not display anything on the screen! The blue screen just stood and the cursor blinked on top. But it was already relatively simple, I read on the Internet that it can display information not on the screen, but in the terminal, which must be connected to the COM port. On the second day, I soldered a null modem cable and connected the server to an old laptop with 2000 Windows, where I launched the terminal program. He took the settings from the Internet, “speed” - 9600 bits, “flow control” - no. I turned on the server and saw the inscriptions that the server started in the terminal program!
512 Meg of system memory
probing hose 1, PCI
probing hose 0, PCI
probing PCI-to-ISA bridge, bus 1
bus 0, slot 5, function 1 - dqa - Cypress 82C693 IDE
bus 0, slot 5, function 2 - dqb - Cypress 82C693 IDE
Testing the System
Testing the Memory
System Temperature is 21 degrees C
initializing GCT / FRU at offset 1b8000
AlphaServer DS20 500 MHz Console V5.4-23, May 19 1999 14:48:15
CPU 0 booting
(boot dk * 100.1.0.2001.1 -flags A)
failed to open dk * 100.1.0.2001.1
What to do next I was told by a friend from the Internet . You need to press "Ctrl + C" to abort the download attempt and go to the SRM console, and there run the command "set console graphics", which will switch the output from the terminal to the monitor. Here's what the first launch looked like. On the laptop in the terminal, old data still hangs, and new ones are already displayed on the monitor.
Separately, I want to say about the SRM console, this is something like an operating system and BIOS combined. Here you can change the configuration and parameters, as well as revise the configuration of the server and devices. You can test the server and do many different things. I know only a small fraction of what you can do there, because I have not yet found the manual on the SRM console. And if you type alphabios in the console, the server BIOS will start. Further, in a few days I already got used to it; the FreeBSD and OpenVMS operating systems for alpha systems were found, however, none of them were installed. :( FreeBSD reached the breakdown of hard disks, and for some reason did not see any, and when I called any installation function, it crashed into the console. And OpenVMS wanted a BIOS update, which I could not do, because the update program freezes.
But recently I was given two DVD discs packed with the images of operating systems for alpha servers. So we are waiting for new adventures. In the next part I will write how I installed the operating systems on this server (when I can do this, since the situation is not good at all)