iici! ([info]sendai) wrote,
  • Mood: busy

DREAMS

This morning; at approximately four AM, I awoke; as if from a nightmare. I'm a rather a deep sleeper, especially now that I work full time. I remained alert for a short while, keeping an ear for anything that could have woken me up; while trying to recall if it was a dream that woke me.

Later this morning, I went on my usual trawl of tech sites and was greeted with the news that Apple really are going to switch to Intel CPUs.

In terms of change of ideology, this is no-where near the calibre of the Bungie buy-out, this is more akin to stubbing your toe rather than being knee-capped and kicked in the guts while you're down. This is just an annoyance, an annoyance for developers (I hear a sweaty monkey somewhere in the background,) an annoyance for users (niggling compatibility issues for n-years,) an annoyance for the bean counters (software houses pushing upgrades for optimised software.)

Other than it being an annoyance, it makes sense for the portables; I'd love to see the P-M in Apple's PBs. The G4s are getting long in the tooth, while they remain a vector powerhouse clock for clock, they just don't have the clock speed to have any sort of parity in integer processing. The P-M, while based on the PIII (which the G4 trampled all over in integer and FPU, clock for clock), has a faster bus (with speed bumps on the way,) scales higher and is lower power (which against the G4 is quite an achievement.) Along with those advantages, Apple do own a fair bit of the IP associated with Altivec; I wouldn't be surprised if Intel CPUs end up with some variant of Altivec (it might be too hard to get SSE to be a full 128bit vector engine, and their twin 64bit units kludged to do 128bit work well enough for the moment.) I would cackle if Intel decided to chuck Altivec in their procs and modified the decode stage in their processors so SSE calls are handled by the Altivec unit.

On the desktop side. I don't get it. Intel are phasing out their P4 line (said so them-selves, too hot, not scaling, inefficient) and don't as yet have a clear product to replace it, their 64bit solution is AMDs and not yet implemented in all their lines and on the workstation side; Itanic. Sure the G5 isn't scaling as well as Apple originally expected, but compared to the P4, it's an amazingly elegant solution (I don't want to know what sort of cooling solution Apple designs for an SMT SMP DC P4 PM G61.) The P4 is a processor designed with marketing in mind, but it caught up with physics. It is now relying on marketing to push it along, Intel knows that marketing doesn't actually increase performance and are taking steps to improve the performance of their other processor line. I guess we will just have to wait and see what is available by the time Apple completes it's architecture switch.

Other than that, I have work to do...I've already spent too much time on this.

DREAMS

1. Simultaneous Multi-Threading Simultaneous Multi-Processing Dual Core Pentium 4 PowerMac G6
SSE: Streaming SIMD Extension
SIMD: Single Instruction Multiple Data
Altivec: Originally Moto/Apple's SIMD extension (Apple called it the Velocity Engine,) the unit in the G5 is a compatible subset designed by IBM.
G4: Macs based on the PPC 7400 series processors from Motorola/IBM/Freescale.
G5: Macs based on the PPC 970 series processors from IBM.

Anything else?
Tags: apple, hardware, marathon

  • Post a new comment

    Error

    Your reply will be screened

    Your IP address will be recorded 

  • 15 comments

[info]peon374

June 6 2005, 21:36:11 UTC 6 years ago

Which is the fastest and why?

[info]sendai

June 6 2005, 21:48:17 UTC 6 years ago

To which?
G5 vs. P4 [vs. ADM64]?
G4 vs. P-M?

[info]peon374

June 6 2005, 22:12:07 UTC 6 years ago

The sums

P4 = Rubbish marketing exercise. To much lost to give a number which can be sold

AMD64 = Good design. Good power management unlike their previous gear.

P-M = Good design. Some elegant stuff there but also some dung.

G[45] = Nice but not serious contender. To small a market to warrant the heavy development needed to keep up with the leaders.

Basically all brute force work is done on multiproc AMD systems and all low power stuff is better done on the intel suite of chips.

I have always loved the instruction set of the motorola/ibm power-pc processors. It is elegant and doesn't still smell like an old washing machine, but marketting is god, and it is dead. Maybe I will go make an Amiga in a vertex-fpga and sulk (yes I know modern procs have more in common with the RS6000 processors than 68K series but I am old school yo).

[info]sendai

June 6 2005, 22:41:38 UTC 6 years ago

Double Plus

That's what I said...

[info]indriya

June 6 2005, 23:54:43 UTC 6 years ago

[info]peon374

June 7 2005, 00:00:27 UTC 6 years ago

Byte my bumb

[info]sendai

June 7 2005, 00:09:28 UTC 6 years ago

I'll ban you!!!

[info]indriya

June 7 2005, 01:06:04 UTC 6 years ago

Oh em gee *hides from the scary sock-panda*

[info]peon374

June 6 2005, 22:15:45 UTC 6 years ago

Are they gonna drop the dodgy mach-kernel so they can get around the RBL and actually take advantage of multi-processors :-) ???

[info]sendai

June 7 2005, 00:17:21 UTC 6 years ago

Hah! Unlikely! They have kludged the the kernel to make it take advantage of multi-processors, they needed to do that because they were held back by the G4 for so long, they couldn't improve performance with speed increases, so they talked about the Mhz Myth and made practically every machine SMP.

Any how.

[info]sendai

June 7 2005, 00:07:30 UTC 6 years ago

Oh, right...I asked Anything else? And you asked the above...

Tad slow today...

[info]peon374

June 6 2005, 21:37:55 UTC 6 years ago

Hey did you see Transmeta are closing up shop? There is a company that went nowhere and proved that everything Linus touches doesn't turn to gold.

[info]sendai

June 6 2005, 21:46:54 UTC 6 years ago

Yah, I did. Sorta shame. Had they managed to get it clocking faster so that it could do it's on-the-fly modification faster. It could have gone somewhere, but as it is, there are heaps of other specialised processors that could do the job faster.

The same sort of idea is coming through again, pure FPGA based processors with firmware that does the re-programming...will see...

[info]sirkahless

June 7 2005, 00:51:48 UTC 6 years ago

I thought Intel was going to adpot the M architecture for desktop processing as it's a more efficient model?

[info]sendai

June 7 2005, 00:57:49 UTC 6 years ago

Eventually. But will it be out within a year and a half?

They have invested a large amount in the P4 architecture. And as-yet, the P-M is still hanging around with a 533FSB, no EMT64, HT or several other technologies that the P4 line has.
Create an Account
Forgot your login or password?
Facebook Twitter More login options
English • Español • Deutsch • Русский…