Quote:

Well I think so. For instance, this one uses a 7447A @ 1.4GHz AFAIK, and as far as I can tell, the main thing holding back a 7448 CPU card is the lack of production level 7448's in volume (same as for the blade server).


Hmm, now I am going to speculate a little bit, so don't take this too seriously, OK?

1. The "public versions" of the 7447A and 7448 are both kind of new AFAIK. I doubt the 7447A was available to b-plan at the time the Pegasos2 was developed, and the 7448 even less so. I am absolutely no HW person (as you might spot by the following ), but even if they follow the same standard on buses and interface, maybe there will be some *slight* individual differencies between the CPU models when it comes to sending/receiving signal strenght, timing, tolerance, etc?

2. B-plan has been working with the Freescale HPC I/II boards. They have ported their firmware to it (and according to a post by Neko at morphos-news.de, they even helped debugging them). In forum posts it has been said (by BBRV I think) that many experiences and conclusions were drawn from this. (This can of course be all about firmware.)

3. Since the Pegasos2 was released, a few new northbridges has been introduced, like Discovery III, Tundra TSI108, and also a Discovery V has been announced. The two latter is explicitly specified for 7448 usage in their marketing materials. (This may of course mean nothing else than it fully supports the FSB speeds of that CPU or whatever.)

4. The 1.4GHz CPU's are not sold separately to current users. Also, BBRV has spoken loosly about a new upgrade program for the 7448 (which can mean just about anything). (Not selling those CPU-cards could of course be just a way of adding value to the deal, compared to the ODW being sold in a regular way.)

In the spirit of pure speculation , I want to outline two scenarios:

A. The "old" revision of the Pegasos2 with its Marvell II implementation works poorly together with the new CPU's (maybe the CPU's act/react slightly different on signal strenghts or whatever, especially on the Peg2's way of putting the CPU's on a card in a slot), and the "Free Workstation" Pegs is of a new revision and has some minor changes on the motherboard to cater for the new CPU's (I'm not necessarily talking about a new northbridge, rather some other, minor changes). Bottom line: In order for us "old-timer" Peg2 owners to use the new CPU's, we will have to exchange the motherboard as well, which will be handled through a trade-in program similar to the ones seen before.

B. The new CPU's works perfectly fine hardware wise with every single Peg2 made this far, and only requires a firmware upgrade in order to swap the CPU's (the new firmware versions already supports the 7448 according to Neko on morphos-news.de). The upgrade program that might come is only about CPU-cards. The reason to why Tundra TSI108 is used on the HPCII/7448 board is simply because it is able to utilize the FSB to the fullest, unlike the controller used in the Pegasos2.

For the record, I place my bet on the "B" option, but I don't mind either of them as long as the price is fair!


Hmm, just another thought. I know that replacing the Marvell Discovery II with a Tundra TSI108 is of monumental work and basically like designing a new motherboard (like the migration away from the Articia S). Maybe it wouldn't be worth the effort. However, it would bring two positive things - the 7448 CPU can be used to its fullest potential, which might incite sales of both the CPU and the motherboard more than otherwise, and using the same components in the server blades and Pegasos should help improving economy of scale of *both* products in the future.