@all

I want to ask a quick question to get your oppinion on a possible feature change from the eval board.
Do you think we need Gigabit Ethernet? How many home or busniness environments support it?
There are 4xRJ45 Ethernet ports on the reference board:
2x10/100/1000 coming off an Intel combined MAC and PHY on one of the two 64bit PCI-X 133mhz interfaces
1x10/100 coming off the 405EP service processor
1x10/100 coming off the AMD 8111 southbridge
My idea is that by removing the Intel Gigabit MAC+PHY we could reclaim the second PCI-X channle and ahve two independat sets of PCI-X slots, bearing in mind that with 1 item on PCI-X it runs at 133mhz but with 2 it slows to 100mhz IIRC, combine the speed advantage gained by a higher clockspeed with the fact that the buses would each be able to handle data trsfer simultaniously and this could be one riproaring system. The reference board uses 2xPCI-X slots off one PCI-X bus and 4x32bit 33mhz PCI sltos off the AMD8111 southbridge.
My idea is to sacrafice the Gigabit LAN to get a second sed of 133mhz 64bit PCI-X connectors and use only 2 32bit 33mhz PCI connectors. The chipset implementation of PCI-X will determin whether it is possible or not but I can see no reason why it wouldn't be. I'm not sure what effect this would have on teh cost of teh moptherboard, almost certainly decrease it thought how noticable this would be I don't know. You replace 2 PCI slots with 2 PCI-X slots so a few pence their but you save yourself several dolars on the Intel Gigabit Mac+PHY combo as well as teh 2xRJ45 connectors that go with it. If people want Gigabit ethernet they can get a cheap PCI card (for example £ from ebuyer.co.uk) and stick it in one of the PCI slots (They can take the bandwith with 16Megabit per second to spare if both run at full speed, which neither ever do!) thus saving yourself 2 nice fast PCI-X slots

@Hammer

Quote:

Hammer wrote: @BrianK Quote: Another criticizm I've seen leveled here is, if you don't have enough to spend on a test motherboard you don't have enough money to spend on manufacturing the board itself That particular purpose could be use for gauging real interest and spreading the risk across the community. Atm, a competitive PPC clone hardware is at its infancy compared to X86 laptop/desktop/server world.

Indeed, look at it this way. He could spend $6000 and then it totally collapse at negotiations etc., or he could spred the risk ($10 or $20) across a few hundred people. Once the initial hurdles are sumounted then it becomes easier for him to justify a large personal financial investment. As he said he'd also donate that board to Hyperion for the porting, so it's kinda a double investment.

$10 isn't a huge risk compared to say $x00 for a G4 PPC accelerator card or $1x0 for a Mac Emulator for Blizzard PPC is it?

@Uncharted

Thank you, I was begining to wonder if I was talking in a different langauge.

@digitaldisaster

Quote:

My idea is that by removing the Intel Gigabit MAC+PHY we could reclaim the second PCI-X channle and ahve two independat sets of PCI-X slots

Why not try to hook it up to a SATA controller instead, to much work/cost?

@Eric_S

Making one yes, sticking a pre-done one on the bus, no. Ceratinly possible, I was wondinerg about SATA but figured it would turn up on a later AMD southbridge, along w/Giagbit Ethernet. until then I will investaigte who makes SATA controllers (I know Marvell do - SATA II infact - and I'm under NDA with them so that may be our best bet)

@Null

Quote:

A much lower priced G4+CPU combination would be much better than a G5 mobo.

In my POV, G5 is would be a better choice since any "32bit to 64bit? transition phase will involve some migration activities in terms of operation system and applications. The degree of ?pain? in transition phase is dependant on a solution involved.

In another words, if a biz needs to migrate to another platform (from a 32bit solution) why not a 64bit based solution? A G5 can surf the mainstream 32bit-to-64bit transition path.

_________________

@Uncharted

Quote:

Indeed, look at it this way. He could spend $6000 and then it totally collapse at negotiations etc., or he could spred the risk ($10 or $20) across a few hundred people. Once the initial hurdles are sumounted then it becomes easier for him to justify a large personal financial investment. As he said he'd also donate that board to Hyperion for the porting, so it's kinda a double investment.

Reducing the risk by spreading it across a community may give an intangible boost to the project i.e. less worries on the personal side for project initiator and greater focus on the technical aspects.

_________________

@digitaldisaster

Quote:

Do you think we need Gigabit Ethernet? How many home or busniness environments support it?

AmigaOS is not a server OS and clients rarely need 100Mbps. Also, I don't know if any Amiga TCP-IP ports will fully utilize a Gig-E. In my opinion it'd be safe to say that Gig-E isn't a requirement. There are some users and critics who will say this is necessary for their application but my opinion is they are a minority of the Amiga users.

Everyone likes low cost/ high performance in their computers so a good balance of price/ usable features needs to be retained. It'd be nice to leave two 100Mbps ethernets available.

@digitaldisaster

Quote:

Do you think we need Gigabit Ethernet?

Not at the moment, due to the lack of gigabit wires for homes, and the cost of installing it.

But a 10/100/1000 Gb built-in lan-card would secure future compability, when the Gb-periferals (routers, wires etc) gets a bit cheaper.

_________________
Site admins are people too..pooff!

@digitaldisaster

I was just about to chime in with "why not add SATA controllers" but someone beat me to it already. The Gigabit Ethernet could always go into an add on controller if needed (usually only in a server role), although it may be useful if you had a render farm of G5 Amigas.

So my vote for deviation from the eval board would be to add SATA. I think UDMA IDE is like AGP... it's going to go the way of the dinasaur soon, so better to look at forward technologies as much as possible. SATA drives are also hot swappable so that would be nice too.

BTW, if you are just talking about a donation of $10 - $ 20 bucks to help out the project, I could easily do that. It's not a large price to help start a project that I've been hoping someone would start on, and not much more than the cost of a couple of Pizza's, so count me in.

If as many people invested $10 into this project as purchased a Club Amiga membership, you should be well on the way to getting that first board designed and have something concrete to start out with.


@ Tomazkid

Quote:

Not at the moment, due to the lack of gigabit wires for homes, and the cost of installing it.

Gigabit will run across the same CAT5 cable that is used for 10/100 connections.

@digitaldisaster

GigE is essential if you wanna look professional nowadays.

_________________
This weeks pet peeve:
Using "voltage" instead of "potential", which leads to inventing new words like "amperage" instead of "current" (I, measured in A) or possible "charge" (amperehours, Ah or Coulomb, C). Sometimes I don't even know what people mean.

@Herewegoagain

Thanks, it is my hope that the small donation price will attract as many peopl as possible. Until the website is live (I need DaveyD to set the FTP, subdomain, email etc. up for me so any proding in his direction would be apprecaited but I understand that he may be busy) donations can be made by transfering the money to the amiga@bezilla.org paypal account.
If we are looking at adding SATA which I think would be usefull, especially for a high end board (No P.C's feature SATA II to my knowledge so lets add that - twice the speed of SATA at 3gbps) then we should keep the Gigabit ethernet as long as the extra chips+connectors don't push the costs up too much. Having one device allready on the bus would start to bring speeds down and if you are trying to go high end or look profesional then you need both SATA and Gigabit Ethernet, esspecially in the low to mid end server markets it seems this system coulbe be targeted at with Linux.
Also Gigabit Ethernet requires CAT5e cables so only newer CAT5 stuff works. Just me being picky
IDE may be going the way of AGP for hard drives but it is still needed for CD/DVD drives unless you use SATA->IDE converters. Like AGP it is still needed for legacy compatability with hard drives.

@digitaldisaster

Here's an article on PPC on Linux which could be relevant to the discussion. It focuses mostly on how to run Linux on PPC, but also indicates a growing trend in running Linux on PPC platforms.

@digitaldisaster

Quote:

Also Gigabit Ethernet requires CAT5e cables so only newer CAT5 stuff works. Just me being picky IDE may be going the way of AGP for hard drives but it is still needed for CD/DVD drives unless you use SATA->IDE converters. Like AGP it is still needed for legacy compatability with hard drives.

Right, well that's what I was getting at, but I guess that's not what I said... Most home users are going to have the 5e stuff anyways, as home networks have only become mainstream since the last couple of years.

And right again about the IDE drives. I haven't heard anything about SATA CD/DVD drives yet. And I'm pretty sure some will want to bring over and existing IDE drive as well.

@Herewegoagain

RE: Running Gig-E, in case someone doesn't know it Cat-4, Cat-5, Cat-5e, Cat-6 are all specifications. Wires are then tested to the specifications. For example, Gigabit Ethernet requires Cat-5e, however Cat-5 may work perfectly fine for you. It could be the cable meets or exceeds Cat-5e but was never tested with that standard in mind. But, let's face it if you're going Gigabit it's not much more to upgrade a couple of wires to just be sure it all goes smoothly.

Quote:

I haven't heard anything about SATA CD/DVD drives yet.

MSI XA52P and Plextor's PX712-SA are two of the first SATA DVD Writers on the market. XA52P isvery limited only working with Intel chipsets. SATA II has ATAPI support so once this standard starts making it to motherboards we'll probably see SATA DVD/CD drives then.

@digitaldisaster

1) Forget about the 'old technology, ie. serial, parallel, ps/2, ide/eide

2) usb, Firewire (perhaps), SATA (serial ATA) for H/D, DVD etc, pci-express, gigabyte lan

Most of the major compter manufacturers will have new machines by
March 2005 that only have the items listed in 2).
By the end of 2005, everyone who wants to survive will meet the standard in item 2).

Item 1) will be dead and only used in 'legacy computers'.

@marrandy

Quote:

Most of the major compter manufacturers will have new machines by March 2005 that only have the items listed in 2).

Note that nForce4(CK8-04) prototype (Q4 2004/Q1 2005) and ASUS/Gigabyte/MSI Intel 915/925 based motherboards still includes PS/2 ports, 1x Parallel and 1x COM. Unreleased nForce 4 prototypes still includes P-ATA. Legacy ports can be replicated via USB btw?

_________________

@olegil

Note that Realtek GigE doesn?t count as "processional" btw ...

_________________

@digitaldisaster

Quote:

No P.C's feature SATA II to my knowledge so lets add that - twice the speed of SATA at 3gbps

Refer to http://www.supermicro.com/products/motherboard/Xeon800/E7525/X6DAL-TG.cfm
This is a Intel® E7525 (Tumwater) base motherboard and it includes SATA II from Marvell. This is a similar motherboard that broke through ~21000 3DMarks 2003 via SLI NV40s.

Another SATA II enabled feature is from
1. http://www.amdboard.com/ali_uli_m1573.html
AMD K8 Socket 939, PCI-E generation

2.
http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=2137
http://www.nvidia.com/object/IO_13739.html
nVIDIA's SATA-II (SATA 3Gb/s) support from incoming NV's MCP chip e.g. nForce4 (2004 Q4).

Plans for SATA?s second coming (SATA II) is integrated in most X86 motherboard reference design road maps.

_________________

@HMK

Quote:

Here's an article on PPC on Linux which could be relevant to the discussion. It focuses mostly on how to run Linux on PPC, but also indicates a growing trend in running Linux on PPC platforms.

Some of Dan Burcaw statements are questionable e.g. His RISC vs CISC e.g.
_Modern X86 doesn't directly execute CISC instructions (variable length), they execute RISC (fix length) like instructions.
_His claims on cooling is another questionable issue...
One could write debunking article against Dan Burcaw 's Outside Intel book.

_________________