Yes and no.

It doesn't matter what particular core you have - they all work the same. Some have better instruction scheduling capabilities, some have more cache. But they all operate the same, the same number of units, the same AltiVec instructions, same datastreaming, cache coherency protocol.

The only difference between the "G4" and "e600" is the package you get it in. The e600 is a little different to the one in the 7447A, because it has improved AltiVec, and support for ECC caches, and suchlike. But the difference is as noticable as the difference between the 603e and e300 - they're absolutely identical from the point of view of running code and generating code for them.

It's all just names. The e600 "brand" implies that this is the new SoC core, though, and differentiates the chips from the old discrete CPUs. Yes, there is a difference between the 7447A core (and even the 7448 core) and the "e600" in the MPC8610 and MPC8641D, but, it's not worth debating.

So, foot down. Let's nip this circular discussion in the bud and talk about interesting and relevant things. Don't make me get the tank out :D

_________________
Matt Sealey

One last question,

why is the 7448 a "e600"? It isn't a SoC, right? But it doesn't really matter.

And is the e700 canceled?

There has been some discussion about how no one needs 64-bit or multi-core systems, because there is no support or application for such things. I realize that I don't speak for everyone here, but that isn't actually true, at least for me.

From a desktop standpoint, multi-core systems are extremely valuable, especially for people who are likely to be developers. You can continue using your computer while compiling without slowing down compiles or computations, for instance. And the number of multi-threaded applications is quite large, and growing as multi-core x86 systems become ubiquitous.

As for 64-bit, there is the advantage of being able to address more RAM, of course. But there is also the advantage of having 64-bit words for computations. I'm a physicist -- I use computers to compute things, and native 64-bit types are a major advantage for me, and probably for at least some other people too.

So I would urge you not to just decide that such things are irrelevant hype -- though there is some business advantage to the hype, as well. Even if the 8610 is much better than the old G4s (which is good to hear), it is hard to convince people of this, as we've seen here.

The PA Semi chip got a lot of press at the time of the Apple switch, and people are intrigued by it -- see the number of posts about it on the forums here. It is obviously not equivalent to a several year old Mac Mini, and, since it is an SoC, shouldn't be much harder to integrate than the 8610, leaving aside the need for discrete graphics. And the Freescale CPUs are seeing most of their use in networking applications, too.

My point is that I, and I am sure others, *do* see value in 64-bit and multi-core designs and would be willing to pay substantially more to acquire them. And even though some people want them out of hype, some of us really do need them, and both sets of people will pay Genesi.

I think it is all about usage what specs the hardware need.

For an pc that only has to perform some basic office things there is no need for dual core and or 64 bit. e.g. an efike with some more memory and integrated graphics could easily do the job.

If you are a developer a multicore might be handy and lots of memory.

If you need to do a lot of high precision calculations you need a 64 bit platform for the extra performance.

For an gamer the videocard is more important then the mainboard. However with more simulation in games like physics a 64bit multicore platform is welcome.

For an basic server an efika might even be just enough.

Yes, this is absolutely true. My point was that I hope Genesi doesn't completely ignore the set of people who need high-performance desktops, especially given the draw to the PowerPC platform of its historically excellent vector and floating point performance. And especially since that set of people includes me...

Probably because the 7448 and 8641(D) was at the same level of core evolution when they launched, and probably because of marketing reasons when launching the 7448 to "push for the new", and because it's only a name anyway and they wanted to be consistent during the "transition" so that they could make comparison charts like this.

If you look at the freescale picture that I embedded above, you will see in that picture that the "G4" is described as *74xx* (i.e. including the 7448), while the "e600" is the heart in the e600 platform, *86xx*.

As you have already seen above, today they (back)label the cores inside all the old CPU's of the same family "e600" as well. The reason can only be that they are fundamentally the same (even if they naturally are weaker in comparison the older they are).

As years go by things improves, altivec performance, cache performance, CMOS fab techniques, etc. I bet that quite a lot of things in the core improved over the years between the 7400 and the 7447A ("Max" -> "Nitro" -> "Voyager" -> "Apollo 6" -> "Apollo 7"), but without changing the *fundamentals* that defines this family of cores.

Now some more time has passed since the release of the 8641D, and as Neko said elsewhere, it's an important fact that the 8610 is *at the very latest* step of core evolution. :)

I would like to know that too. And what is it supposed to be? What about it's specs? I have seen references of it being 4 cores at 3 ghz, 4 MB L2 cache, can run in either 32-bit or 64-bit mode, etc. Is this true?

The "e700" is still mentioned in lots of freescale road map docs etc, but I can find very little info about it...

This means? Input, please :D
Latest means "latest and greatest" or "last"?

And if the 86xx CPU is the last with AltiVec ... doesn't need FreeScales customers AltiVec anymore? Hard to believe … everyone puts a VMX unit in it, at least IBM and FS kicked it out ...

And about the 8610 EFIKA, is there any release date given?

I don't know if it's cancelled, I only heard (from a good source) that Freescale have put their focus on multi-core (not necessarily 64-bit, just multi-core). I don't disagree with nwhitehorn (I'm a physicist as well, btw :) in that multi-core and 64-bit are very important for many sorts of things, be it developer or user oriented.

It's just that I would really like a multicore chip WITH Altivec as well. Sth like Intel is doing. Each core has its own SSE unit (or at least so I understand, if not then it's a load of crap).

Plus, there are some operations that are actually *slower* on multi-core rather than a SIMD unit (eg. matrix multiplication is definitely one of them) or any kind of stream processing unit (like the SPUs on the Cell. Multi-core is great for generic stuff, a SIMD unit is for processing streams of data.

The 8610 and the 8641D is probably as far as Freescale will go in the foreseeable future in advancing the e600 core. That is the "highend" today. There is nothing more currently.

That said, we have a great group of people handling these parts. We have known them now for more than a few years and they are very supportive.

R&B :)

_________________
http://bbrv.blogspot.com

Yeah, I'm looking forward for both of them, I didnt' want to sound as criticizing Freescale. I just like AltiVec so much :)

in past, i liked AltiVec a lot... but with time... as it's not supported in MorphOS yet, i'm not that interested anymore. anyway, the multicoring... the idea seams to drive our future, let's build a multicore pegasos/efika, run MorphOS on it (please!!!), and let's develop multicore applications... i'd love to do so!

bye, MarK.

How do you expect to exploit multi-core operation on an operating system which architecturally obviates the use of more than one processor?

This is an OS with it's roots in 1984, before SMP existed on the architectures it was meant to support. There is absolutely no way you can drop in a second CPU core to the ABox model and have it work right.

_________________
Matt Sealey

I would love to see X on such a machine (8610).
Maybe someone can hack it, like X x86.

Well, the lack of SMP support in MorphOS should not also mean the lack of multicore machines. Linux supports it very well :-) And FreeBSD on PPC is also gearing towards SMP support. A developers machine or a LAMP server can take advantage any number of cores.

A 8610 based machine, which has "only" one core, would be still very welcome. The greatly improved memory bandwidth benefits any server operations, and there are now Altivec optimized software RAID routines which could make very good use of this CPU on a NAS.

_________________
CzP
http://czanik.blogs.balabit.com/