XPC Review: Tyan Tiger Mpx (s2466n-4m)

[NDC]

Click here (http://www.xtremepccentral.com/articles/archive/05_tyan-s2466n-4m/tyan.html)

[Mntsnow]

Done a great job bud! Hopefully next week I will have a chance to get the review done on the Gigabyte board!

[NDC]

Thanks! :)

I will be looking forward to the review you're going to write up... I haven't seen an English version for your mainboard. Actually, I don't think there is one around yet! LOL

[otheos]

Great stuff NDC.

You wouldn't mind a few additions, comments? :)

The 64bit PCI operates at 3.3V at 66Mhz and most 32bit PCI devices operate at 5.0V at 33Mhz. Now here is where careful decisions need to be made. Although most mainboard manufacturers claim backward compatibility using 32bit PCI devices in the 64bit PCI slot, this isn't the case. The reason for this is because most 32bit PCI devices operate at 5.0V, not 3.3V like 64bit PCI devices do. However, there are some 32bit PCI devices that support both 3.3V and 5.0V such as storage controllers, RAID controllers, and a few sound cards. You can easily distinguish if your 32bit PCI device is capable of running on 64bit PCI slots simply by looking the way the gold contacts are keyed.

I'm glad you brought this forware as most reviewers seem to forget. The difference here is the PCI standard. 5.0V only cards (no notch near the bracket) are NOT 3.3V tolerant. However, PCI2.2 brought the second notch and 3.3V tolerance. Cards the explicitly state compliance to PCI2.2 can be used to the 64/66 slots without any electrical issues.

But before you get too excited, keep in mind that connecting 32bit 33Mhz PCI devices along with 64bit 66Mhz devices in the 64bit PCI bus will cripple the whole PCI bus down to 32bit 33Mhz.

Now this is the point of debate that brough about that article I posted about earlier this week. One small correction here: 32bit cards will NOT bring the 64bit addressing to 32bit. Addressing is done per cycle by each device and 32bit devices will only use 32bit, while 64bit devices can use the full 64bit. However a 33Mhz device will bring the bus down to 33Mhz from 66Mhz. Absolutely!. Now if you don't have any 64/66 cards there you should really care, but unfortunatelly this also affects the AMD768 southbridge connectrion. The AMD768 is a 32bit/66Mhz device and out of the 266MB/s available to it, a reduction to 33Mhz brings it down to 133MB/s (just enough for the secondary PCI bus). This makes the whole MPX deal indifferent, as it matches exactly the specification of the MP (762/766) chipset.

Another difference on the latest version of this mainboard is that you don't have to use registered memory!

True, but the TigerMP was the same. As long as you only use two dimms you can use unregistered non-ECC RAM. It is not recommended though. What I am not sure off, is the number of banks that can be accessed. The TigerMP had a limit of 6 banks. If you used standard Crucial ECC dimms, they are typically two sided/bank per module and 4 of them would cause instabilities. This was unfortunately undocumented by Tyan, and I surely hope the MPX has the full 8 banks available for use.

A few additions that are not imporant but I would waste your time reading :)

The -4M distinguishes this board from it's twin TigerMPX. The difference is the BIOS chip. The -4M uses a 4MBit part while the standard TigerMPX comes with a standard 2Mbit. Why? It beats me but I can find out if you want :) Also the -4M was the first out of the two boards to use the revised AMD768 southgate with the fixed USB bug.

A couple of unique features to the MPX line of motherboards's BIOS (all Award and Phoenix and AMI I've seen have it) is the ability to run a terminal off the COM port for verbose. A standard vt100 or ansi terminal can be used for headless operation (no monitor) as these boards are typically used in U1/2 enclosures as servers (and this is why the IDE controllers are so low and not well placed for a tower).

I'm glad you pointed out the ESC at POST that birngs out the boot selection menu, nice touch, and btw many motherboards use it, only it's undocumented.

[NDC]

Thanks for your comments on the review. Your comment are welcome as always since you use practically the same board... :)

I'm rather confused now about the 64bit PCI ports, whether if does drop the bits and speed down to a standard PCI spec. To be certain, I have mailed Tyan for an answer.

And another thing is I was told by Tyan that you CAN'T use non-registered at all with Tyan MP. I've also read this in many articles. Have you actually tried non-registered DDR memory in your TYAN MP mainboard?

EDIT - I just called Tyan right now and from what they say, the bit and mhz will follow the specs of the slower device... So in other words, when mixing 32bit/33Mhz and 64bit/66Mhz in the 64bit/66Mhz slots, the bit rate will drop down to 32bit and the Mhz down to 33Mhz...But just to be certain, I'm going to mail TYAN USA and TYAN TAIWAN as well..

[NDC]

Looks like Tyan in Europe had different opinions... take a lok at the first review page if you'd like to know what they said... :)

[otheos]

That's quite embarassing given they are contradicting each other and that both are wrong :)

Straight from AMD's white paper, on the 64bit issue:

The AMD-762 system controller supports both 32-bit and 64-bit PCI agents on a cycle-by-cycle basis as defined by the PCI bus specification. The AMD-762 system controller asserts the REQ64# pin during reset to allow 64-bit devices to detect that the host bridge supports the full 64-bit data width. All 64-bit transactions from these bus masters are then negotiated with the PCI bus REQ#64/ACK54# protocol.

So both are supported at a cycle by cycle basis, hence can be both done (32 and 64).

And for the 66/33 issue:


If any of these devices are 33MHz only (the M66EN pin is low), then the AMD-762 system controller automatically drives 33MHz on the PCI_66CLK[2:0] pins during reset.

So this means the bus will explicitly set to 33Mhz if a 33Mhz device is used.

Thanks for taking the time to take it up with Tyan, I hope this resolves the issue :)

[NDC]

Yeah, that is pretty embarrassing! LOL :) Thank you for taking the time to look into the matter further. The page has been updated. :)

[otheos]

I hope this will solve the issue :)

On the RAM front, as you said th MPX supports non buffered ram on the first two dimms.

This is what Crucial says about it:

The Tiger MPX supports non-registered DDR SDRAM in the first 2 memory sockets only (DIMM1 and DIMM2, as labeled on the motherboard). Registered DIMMs are supported in all sockets.

It does'n mention any bank issues, which hopefully means you can fill all 4 slots with double sided ecc/reg ram.

This is what Crucial says about the MP board:

This motherboard only supports 6 banks of memory.

This is a great improvement if it's true as people were having problems with all 4 slots filled with 2 sided dimms and they couldn't say what the problem was and once they found out they started looking for single sided dimms (not many around). So this is a great improvement!!

[jadison]

Well-written! :)
Love the pics too...that's what grabs most readers eyes, are lots of pics. I have a few suggestions as well, will relay them to you via ICQ.

[fshanda]

Congrats NDC and XtremePC.

You got front page pimpage at 2CPU.com.

[jadison]

I sent the review out to a list of sites I had in my contact list...
2CPU.com just happened to be one of them, after all it is a dually mobo review ;)

We've also made frontpage at several other tech sites such as Legion Hardware (http://www.legionhardware.com) and Tweakers Australia (http://www.tweakers.com.au), etc. :)