First post, by shamino
- Rank
- l33t
Earlier I was poking around on a Tyan S1590 rev C motherboard, trying to figure out it's 3.3v supply limits. This is a notorious concern on some early AGP motherboards, where higher powered AGP cards might overload and damage them. I got carried away researching this and so I figured some of this info might be useful if anybody else uses this board.
If you use an ATX power supply, then it's 3.3v output will help feed the onboard 3.3v rail. Therefore, I don't anticipate that there would be any problem with using high powered AGP video cards on this board *IF* you use an ATX PSU.
If you use an ATX PSU, then the rest of this post should not apply.
If you are using an AT PSU, then the 3.3v rail has limited output and could be overloaded if you use a heavy video card.
TLDR: When using an AT PSU, the Tyan S1590 supports up to 7A of 3.3v current for the whole board. This includes your RAM, AGP card, up to 0.75A of the CPU, L2 cache chip, and possibly some amount of power on some PCI cards.
The Tyan S1590 supports both AT and ATX power supplies. AT power supplies do not have a 3.3v rail of their own, so boards with an AT connector will always have an onboard regulator to provide the onboard 3.3v.
This board has some jumpers to select between the onboard regulator or an ATX 3.3v supply, but on my board rev C, they are soldered so that both sources are always connected. I have measured and confirmed that the 3.3v ATX pins are connected with the output of the onboard 3.3v regulator. If you use an ATX PSU then they will work in parallel. If you use AT, then you only have the onboard regulator.
The S1590 provides onboard 3.3v by feeding the 5V power supply input into a linear regulator. It is the heatsinked component Q2 located near the corner of the board, just under the DIMM slots. It's a CS5207-1. There's a datasheet from ON Semiconductor here:
http://www.onsemi.com/pub_link/Collateral/CS5207-1-D.PDF
Mine has a different logo on it, so maybe their logo changed or it's a competing brand. Either way, I assume the basic specs should still be the same.
The 3.3v regulator can handle up to 7 Amps total output.
This assumes the heatsink can actually dissipate enough heat at 7A without letting things get too hot. I have no idea how good the heatsink is with respect to the 7A limit.
This 3.3v regulator feeds the AGP slot, the RAM, the PCI slots, the L2 cache, and part of the CPU socket.
AGP: According to wikipedia (unknown if accurate), the AGP spec allows up to 6A to be drawn from the 3.3v pins on the AGP slot. So in the worst case, that leaves only 1A for everything else. I think it would be pretty easy to get overbudget at that point and blow the regulator.
I don't know how much 3.3v current is drawn by well known cards.
PCI: Even though these are generally referred to as 5V PCI slots, they do have some 3.3v power supply pins. I don't know how high their power draw can be. I found a PCI 2.1 spec which seems to say that no 3.3v power is required to be provided at all, but this board is apparently providing it.
RAM: The same 3.3v supply also feeds into the DIMM slots. If you've jumpered the SIMM slots to use 3.3v, then it feeds those also.
I looked up a couple datasheets for 128MB 8-chip PC100/133 SDRAM modules. They both implied a max operating current of 1.2A for an 8-chip 128MB module. However, I think only one module would be active at a time. In standby power on mode, the higher rating I saw was 320mA. There are 3 DIMM slots, so assuming you have all 3 filled with 8-chip 128MB at PC100 speed, I guess ~2A is a safe worst case estimate. I could be way off about this.
L2 Cache: One chip: EliteMT LP61"L"64128F-4 The "L" is stylized, not sure if it's part of part number. Don't know how much power it uses.
CPU: the portion that's powered by this regulator can be a max of 0.75A from what I found. More detail below.
So if you have 0.75A max to the CPU, maybe 2A to the RAM, unknown to the cache chip and possibly nothing to the PCI cards. Let's say 3A is realistic, that would leave 4A for your AGP card. Staying below that might be safer, but maybe my figures are too paranoid.
I'm really unsure what to expect from the PCI cards, but the spec makes it sound like most contemporary cards might not draw any 3.3v at all.
=============================
CPU impact on the 7A regulator
=============================
Split voltage CPUs have 2 power inputs, Vcc2 and Vcc3. Vcc3 is always 3.3v, Vcc2 is whatever core voltage you set with jumpers.
The Vcc2 pins are connected to the CPU Vcore regulator.
The Vcc3 pins are connected to the 7A onboard 3.3v regulator. It eats into the 7A budget.
Vcc2 and Vcc3 each occupy one pin at jumper JP11. Closing that jumper ties these circuits together. It is left open for split voltage CPUs.
Split-voltage CPUs draw most of their power from the Vcc2 pins, not the 3.3v.
For an Intel Pentium MMX 233MHz - according to Intel's Pentium MMX datasheet (24318504), the worst case max current from the 3.3v pins (Icc3) is 0.75A.
For an AMD K6-3 450MHz - according to the "AMD K6-III Processor Data Sheet" pg261, max Icc3 current is 0.66A.
For Cyrix 6x86MX and M2 - all models shown in the datasheets are only 0.1A Icc3.
Classic Pentium P54C - these are single voltage 3.3v CPUs. The Vcc2 and Vcc3 pins described above are all the same on this CPU.
According to Intel's Pentium datasheet (24199710) pg29 - For a Pentium 200MHz, the max 3.3v current is 4.6A.
For a Cyrix 6x86 133MHz (P166+) it's 6.6A. The 150Mhz P200+ model is unlisted, but would be even higher.
These CPUs could be scary if the board isn't jumpered correctly.
When the board is jumpered for a 3.3v-only CPU, you are supposed to close jumper JP11. When you do this, current flows between the 7A regulator described above and the CPU's Vcore regulator circuit. This bolsters how much 3.3v current can be provided. It could even have the side effect of increasing the 3.3v budget for the rest of the board.
The jumper itself is providing the current flow, so you don't want a loose/flimsy jumper at JP11.
If JP11 is closed but you don't have the Vcore jumpers set for 3.3v, it appears it could be catastrophic.
Failing to close JP11 with a 3.3v CPU (like the P54C) might add substantial stress to the onboard 3.3v regulator. - I'm not sure about this, because it depends how much current is drawn at those particular pins on the P54C. Perhaps even on the P54C these pins don't draw any more current than they do on later CPUs, but Intel doesn't say this. On the P54C all the "Vcc2" and "Vcc3" pins are shown as a unified "Vcc", implying they are one unified power supply for the whole chip. As such, the pins formerly known as "Vcc3" might draw a lot more current on that type of CPU.
