Marvell and Mai Logic working towards PPC970 support?

[HyperionMP]

You will never be able to implement full AGP compliance with an AGP slot attached to a PCI-X slot.

I find all this Mai bashing also very amusing considering every Pegasos out there is using the old Articia and not even the new revision.

And not even using it fully BTW as the FSB is stuck at 100 Mhz for most boards.

[olegil]

Uhm, but a blade isn't a box, it's a board.

Forking out 2k USD for a motherboard with chips on it and having to fork out ANOTHER 4.5k USD just to get a box to put it in isn't THAT fun, really ;-)

If you have the box, PPC970 blades sounds excellent, though ;-)

[zacman]

>You will never be able to implement full AGP
>compliance with an AGP slot attached to a PCI-X
>slot.

Please post your ArticiaS AGP2x compliance test
results. Thanks.

[Zorro]

Ben Hermans wrote...

I find all this Mai bashing also very amusing considering every Pegasos out there is using the old Articia and not even the new revision.

If we (all the amiga community and, firstly, the vendors/developers) don't clear this and
all other questions (fud?) regarding the possible hardware choices that we have, a big part of the possible customers will never buy the products...

 :-(

I, for one, look forward for a united community with clear and strong targets...

[ksk]

@takemehomegrandma

AFAIK, AGP in the Marvell based Pegasos II will be inplemented pretty much in the same way as in the Pegasos I/Amiga One.

Do not believe all of the propaganda.

ArtisiaS has AGP 2x bus built in.
Marvell seems to require a external glue logic to support AGP 2x.

IMO: Perhaps some principles are the same, but the implementation will be totally different.

[olegil]

The principles aren't the same either.
Marvell: supports PCI-X, need external glue logic to get AGP
Articia: supports AGP, well duh.

AGP uses only 32 bit and 66MHz clock which is single, dual, quad or octal pumped for higher bitrate.
PCI-X uses 64 bit and 133MHz clock to achieve higher bitrate. At the very least you'll end up with higher latency because of the MPx/60x/DDR to PCI-X to AGP conversion in two steps.

Please tell me how "connecting some pins together" and "converting from 32 bit 66MHz x 1/2/4/8 to 64 bit 133MHz x 1" will be the same? ;-)

[HyperionMP]

Is that you Charles?

[downix]

@ksk

I don't know about this ArtisiaS chip you're speaking of, but the Articia-S does not have proper AGP support built-in, it's a modified 64-bit 66Mhz PCI slot.

That's why on the A1 if you plug the wrong kind of PCI card into PCI slot 0, the AGP slows down.

[downix]

You will never be able to implement full AGP compliance with an AGP slot attached to a PCI-X slot.

A good half *if not more* of the AGP slots out there don't have full AGP complience either.  Hell, I've yet to run across a single AGP 8x vendor that has full complience.  (don't forget, AGP 8x complience means AGP 2x or slower cards no longer work, so if the slot can run an AGP 2x card, then it's not fully 8x complient)

The goal here is not to do one upsmanship by throwing off-handed comments about some complience standard that nobody bothers with anywhere else.  It is to support AGP cards, to have enough compatability in the chipset to run the AGP cards needed.

[HyperionMP]

A modified 64 bit 66 Mhz PCI slot would mean AGP x1 whilst the Articia S does AGP x 2.

The reason why it drops down to AGP x 1 is because the bandwidth is then shared between that PCI slot and the AGP slot.

[olegil]

64bit? You sure about that?
Please try to point it out in the specification:

http://www.mai.com/products/BRA660R2.0.pdf

I'm waiting.
Apart from this, Ben is right. If you have 32bit, 66MHz PCI, you just need a few control signals to have 1xAGP. 2xAGP requires double-pumping that clock yet again (effectively 32 bit, 133MHz). So if you have 133MHz and 64bit, like the Marvell controller, you could use external glue logic to tie it up to a 4xAGP device, but you would SURELY add latency compared to the solution with 2xAGP plugged right into the controller. This is because of the half buswidth but double clock rate conversion you need.

[Van_M]

PPC970 support
yeah right....
since we have a ton of modern applications that involve hardcore hardware (powerfull Cad, a lot of amazing games, a lot of modern multimedia software), it's now time to move forward!!!! Get serius......
 8-)

[Hammer]

A good half *if not more* of the AGP slots out there don't have full AGP complience either.

Did you forget Intel's 875 chipset class (e.g. MSI 875P Neo)?

Hell, I've yet to run across a single AGP 8x vendor that has full complience.

One can only say this statement IF one doesn’t keep track of X86 world (specifically Intel Corp).

(don't forget, AGP 8x complience means AGP 2x or slower cards no longer work, so if the slot can run an AGP 2x card, then it's not fully 8x complient)

Depending on the standard e.g. “Type 4” vs “Type 5” vs "Type 6".

Referencing Intel's AGP 3.0 Interface Specification Rev 1.0. Page 39.  

There are 6 types of AGP equipped motherboards;
1. AGP 3.3 V Motherboard, 3.3 V keyed, Supports only AGP 3.3 V signaling. Available speeds 1x, 2x.

2. AGP 1.5 V Motherboard, 1.5 V keyed, Supports only AGP 1.5 V signaling. Available speeds 1x, 2x,
4x.

3. Universal AGP Motherboard (UAGP)
Universal (UAGP) Supports both AGP 1.5 V and 3.3 V signaling. Available speeds 1x, 2x, 4x.

4. AGP3.0 Motherboard, 1.5 V keyed, Supports only AGP3.0 signaling. Additional electrical ID to
prevent AGP 1.5 V operation. Available speeds 8x, 4x.

5. Universal 1.5V AGP3.0 Motherboard (Universal 1.5V AGP3.0), 1.5 V keyed, Supports AGP 1.5 V and AGP3.0 signaling. Available speeds 1x, 2x, 4x, in AGP2.0 mode and 8x, 4x in AGP3.0 mode.

6. Universal AGP3.0 Motherboard (Universal AGP3.0)
Universal (UAGP) Supports AGP 3.3V, 1.5V and AGP3.0 signaling. Available
speeds 1x, 2x, 4x in AGP2.0 mode and 8x, 4x in AGP3.0
mode. This includes 3.3V support for 1x and 2x speeds.

Manufactures can select different solutions for their needs.    
 

(don't forget, AGP 8x complience means AGP 2x or slower cards no longer work, so if the slot can run an AGP 2x card, then it's not fully 8x complient)

You are only referring the Type 4 AGP motherboard.

"the terms AGP1X, AGP2X, AGP4X, and AGP8X refer to the speed of data transfers on the AGP Interface while AGP1.0, AGP2.0, and AGP3.0 refer to specific releases of the AGP Interface specifications." - page 11

Examples;
ASUS nForce2 (A7N8X) supports "1.5V AGP 8X Pro" e.g. "Universal 1.5V AGP3.0 Motherboard" (Type 5).
(nVidia nForce 2 Class).

MSI K7T Turbo-R (MS-6330 V3) supports AGP 2.0 e.g. AGP 1.5 V Motherboard (Type 2). (VIA KT133A Class)

MSI 875P Neo (MS-6758) supports 8x/4x at 0.8V (AGP 3.0) or 4x at 1.5V  (3.3V is not supported) e.g. AGP3.0 Motherboard (Type 4). (Intel 875 Class).

[JoannaK]

Electro:

Facts,, I have DataSheets of Marvell Northbridges, and I have some experience on reading and understanding those kind of papers.

Proving above. Considering various NDA restrictions I Won't send them to you... Neither can I send you any up-to-date reference of my recent works (specs+ numbers manufactured) beoynd fact that they are based on Motorola PPC. Pitty, but I rather keep  my work than try to win  these pointless net figths that seem to be going on and on.

[Elektro]

Yoanna:

If you're not ready to prove your claims then shut up and thats exactly why you and kennyr r now such obvious childish trolls. Sure articiaS had a bug and its now fixed. But I guess marvell has no bugs right? Also raving about peg2 how its gonna be the gazillion times better than a1 when the thing is not even out yet is just pathetic. ffs every time some1 mentions a1 or os4 or mai, theres tens of posts from ppl who claim that mai is crap, amigaone doesn't work and os4 will never come. Give it a rest. Even kids get bored of 'my toy is better than yours' tactics why the hell cant you ppl... As I said childish stuff.

[Kurt]

"The difference between FUD and non-FUD is fact. That makes you technically more than a fudder than myself. Have a nice day, soldier-boy."

Uhh ever thing I said was fact.
Same thing can't really be said by you because every thing you said was technically FALSE.

Well very few if anything posted by you Morph boy's has been factual. The amount of left handed compliments and innuendo you guy's spread is amazing I am beginning to wonder if you guy's have a crib sheet that you pass around with a bunch of false rumors and twisted spin of the facts. because you all seem to be on the exact same page. It might be the Wrong page but always the same page.

So let's have fun and tear your VERY NON-FACTUAL POST APART.
 
No it won't be the same as A1/Peg1 AtriciaS has AGP built in.
AGP IS ALWAYS BUILT INTO THE NORTHBRIDGE.
AGP on the ATRICIAS is built into the Northbridge.

PCI and AGP are not the SAME thing.

I have two very large books on the specs I will send them to you, if you give me your address.

But to be honest there is really no big speed advantage between AGP2X and PCI in most cases as long as your graphics card has sufficient memory. That has been proven in speed tests on various PC sites.

So if Thendic was honest and did not try to hack AGP onto the PegII it would not be the end of the world. Hacking AGP onto a PCI bus would be  messy and would be nothing more than a marketing ploy.
Because the resulting circuit would likely be slower than a straight PCI card.

But not to create more FUD please really read the Marvell specs please before you go all agog
Dual PCI means if you convert one to AGP and  used the other to talk to the south bridge means no functional PCI expansion bus to be used by the end user.

" and of course Mai's total unwillingness to admit any bugs until IBM stepped in."

Ohh the above statement is really untrue. BBRV BS at the fullest not to be taken serious by anyone with a functioning brainpan.

"What isn't irrelevant is Articia-S's fatal bugs"
" Regardless of the Articia-S features, its bugs make it useless crap anyway."

Are you saying that BBRV sold 600 NON-working Pegasos boards to endusers.
Is he going to refund everyone's money for selling them doorstops.
No of course not,  Anyone with a Peg one board will tell you that they are quite happy with them, and that shock they work just fine .

Also I am a Marine there is a world of difference!!

Kurt