1.2Ghz = No more powerusers!

[Recoil]

If apple comes out with a lame 1.2Ghz G4 then I'm going to the dark side!

Hello dual opterons!

[rice_web]

Yeah, go ahead. Dual Opterons will be great. In fact, I'm saving up for a nice Opteron system right now. However, I'm still going to be buying Macs at the same time. I just can't let go of OS X.

[job]

Originally posted by Recoil
If apple comes out with a lame 1.2Ghz G4 then I'm going to the dark side!



....then it forever shall dominate your life...:D


What is a Opteron?

hitman

[rice_web]

You haven't heard of the Opteron? Oh, it's only the best processor from AMD yet.

It runs both 64-bit and 32-bit software on the same chip. It also runs about 30% at the same processor speed as previous Athlons, and it is supposed to scale much higher than the current batch of XPs. It is awesome.

It's also known as the Clawhammer.

[Catfish_Man]

... Apple could go to 10GHz tomorrow with very little effect unless the ram gets sped up. Stop focusing on processor clock frequency. Under ideal circumstances a 1.2GHz with DDR266 could be twice as fast as a 1GHz without DDR (if I remember correctly a 533MHz G4 is can max 133MHz ram, so a 1.06GHz one should be able to max 266MHz ram.). Obviously this isn't going to be the case in the real world, but it *should* be quite dramatic (it may not be, but if it isn't, something's wrong).
The Opteron is going to cost a fortune (compared to a desktop chip, it'll be cheap for its intended market). It is NOT a desktop chip. It's a multiprocessor workstation/low end server chip. The Hammer for the desktop is going to be under the Athlon name. It does kick ass. In fact, it looks to be almost exactly what I expect the G5 to be. I've been doing some research on the Motorola MPC 8540, which is basically an embedded only G5, and it appears to be quite similar. Here's a comparison (Hammer on the left, 8540 on the right):
Memory: On chip DDR333 controller/same
Communication to the rest of the motherboard: Hypertransport/RapidIO (similar)
Manufacturing process: .13 micron/.13 micron SOI
Overall design goals: bandwidth/bandwidth and modularity
Release date: Q4 2002 to Q1 2003/Unknown, but quite soon (lots of info on it already, including benchmarks)




To get a G5, here's my guess as to what you do:
1) get an MPC8540
2) take out the integrated embedded stuff (network controllers, serial controllers, etc...)
3) Add a floating point unit
4) Add Altivec
5) Add 256k L2 cache (optional)
6) Add L3 cache (optional)
7) Boost the voltage from 1.5V to 1.8V




Basically, the G4 is dying and we all know it. It's a great chip, but it was made to compete with the Pentium 3 (low bandwidth, medium clock frequency) and is now having to compete with Athlons and Pentium 4s (medium bandwidth, high clock frequency), and (if the G5 doesn't come out when I expect it to) soon it will compete with Hammers and Prescott (high bandwidth, high clockspeed). It's chronically memory starved, and on an out of date manufacturing process. If this next revision brings DDR and .13 micron manufacturing, I will be satsified. With those the G4 will be able to last until the G5 comes out, which I expect quite strongly to be at the beginning of 2003.

Any questions?

Linkage:
Motorola's MPC8540 page: http://e-www.motorola.com/webapp/sps/site/prod_summary.jsp?code=MPC8540&nodeId=01M98655

AMDZone (should have hammer info): http://www.amdzone.com

Sources:
BadAndy's posts on arstechnica
Motorola's website
Logic
AMDZone
CNET news

[porovaara]

Originally posted by Catfish_Man
Sources:
BadAndy's posts on arstechnica

BadAndy writes some of the best, most informative postings I've seen on the net since USENET fell so out of favour.

Ars just tends to rock all the way around.

[Durandal7]

Jeeze, these threads are as bad as those Apple PDA threads.

[rice_web]

How so?

[Durandal7]

Let me clarify, I mean those pointless rehashes of the iWalk. This thread is nothing more then someone pouting.

[Cappy]

Not to throw gas on the fire but one has to wonder that if AMD and Intel continue to leap ahead in cpu and system architecture than Apple and Moto then how long before a fast, useable emulator is available for an amd or intel box to run Mac OS X fairly decent? To me that's where it starts getting scary to even think about it as Apple would begin losing potential sales from those geek boys who want to be able play with a nice unix os but are not thrilled with ponying up the $$$.

Of course I'm sure Apple would have their first 3 crack commando lawyer units in position to take the developers down but it could do some damage still.

[iGav]

Originally posted by asurace
Let me clarify, I mean those pointless rehashes of the iWalk. This thread is nothing more then someone pouting.

I get ya........ ;) they wind me up a a treat as well........ :mad: :p

[xelterran]

Opteron's are only going to be used for servers from what ive read...

[Wry Cooter]

Originally posted by xelterran
Opteron's are only going to be used for servers from what ive read...

I think it pays to keep in mind that intel/AMD are not always performing as well as they promise either. Wasn't the itanium or xeon supposed to be the first x86 64 bit chip? And that pretty much wasn't considered very successful at all, except perhaps for keeping your coffee warm.

The realm of 64 bit architecture has been threatened for years from PPC competitors, yet nothing successful has made it to market. And why is this?

Are they holding back till Motorola puts one out first, so they can drown it in a marketing blitz? (how long do they want to wait anyway?)

Or are they merely having problems getting useable bugfree yields themselves?

In general, I like AMD, they got the goods several years ago with some patents they successfully cornered. But a lot of the 'not here yet' chips ARE merely smoke and mirrors, until you have one smoking on your desktop.

[Recoil]

I'm to the point were I just don't want to respond anymore . I will wait until mac world . If it's just 1.2Ghz then I'm out .

[Hemingray]

Originally posted by Recoil
I'm to the point were I just don't want to respond anymore . I will wait until mac world . If it's just 1.2Ghz then I'm out .

Yeah, and what if they don't even announce new PowerMacs? :p

[Postal]

Two things:

First is that it's likely that PowerMac announcements are delayed until August 13th (but are still coming), so that's the only time we'll know for sure.

Also remember that Thinksecret (who got the current rumour) worded it as "speeds in excess of 1.2 GHz;" that may just be the minimum speed for the top-end system, if Motorola can't manage better. We may see 1.33 GHz or (less likely) 1.4 GHz on one or more of Apple's systems.

Unless there's some truly concrete information revealed about the PowerMac before August 13th, it's all up in the air!

[Cappy]

Originally posted by xelterran
Opteron's are only going to be used for servers from what ive read...

That's where the focus is because of the lack of 64 bit software but AMD is planning on 64 bit across the entire line long before Intel gets there.

[mozez]

one of the great things about amd is they sell their processors openly so you can do what you want with them. you can buy dual or quad opterons and tyan or abit or asus tech will come out with a board to support them usually already having one at the release. but guys, look back at apple's sales charts. what are the big sellers, imac, huge (original) new imac, huge again, ibook, emac is climbing the charts, what you don't see on major sales is the powermac series, it's low on the list because it costs to much, the power book sells really well, but nobody is really complaining on that front. the powermac just isn't a top seller, doesn't get top prioity. companies spend 3 grand each for a top end mac, 3 months later, apple brings on something new makes their 3 grand look like crap, and the company is pissed at apple.
beyond all of this, why is nobody looking at history. you went from 400 to 500 and 550, thats 150mhz. next time, 550 to 667, and 733, that's 183hz jump, then you went from 733, to dual 800 and 867, not counting the dual because at the time only like 2 programs could even use the second processor and the os couldn't so, that's a 133 jump again. then from 867, to dual 1gigs, well, that's only a 133 again, jump, anybody seeing the connection here. why do you all think that apple is gonna jump 500 mhz, even 400mhz in clock speed when it never has before. i realize they need to, but given history, i'd say we look at 1ghz, 1.13, and 1.2, and that'd be wishful thinking cause really, at most i'd think it'd be, 1, 1.1, and 1.67, which would be apple's highest jump in speed.
the last thing is, and i know this is said alot, but, if you go back to a pc, you are still going to windows, and i'm sorry but no way in hell i'm going back full swing, i have a pc and it's fine, but compared to osx, and i'm running jaguar 6c87, i would never buy another pc, doesn't matter the speed, it'd still be windows, and sorry, but osx is so far beyond windows it doesn't matter to me anymore. a 1.2 with ddr ram would be a faster mac than anyone has ever seen, so why are you so sure you know exactly how it will perform and how well it will work. why did you even buy a mac to begin with if all you were conserned with was speed? i have always been able to build pcs that run faster than macs, always, for far cheaper too. but i like mac, that's why i use it, work play, who cares, it's what i like, why use something you hate just because it runs faster, and i put an emphisis on faster, but not better, nor as stable, or as easy.

[Wry Cooter]

Its more than price that keeps Power Mac towers from being renewed as often, although reasonable pricing might have more people renewing them more often... its stability with customization and peripheals, on top of the segment that do buy pro towers intend to get multiple years of use from them (and they should be designed with that level of flexibility in mind, as they have)

Others are buying imacs and ibooks the way some people buy clothes, a new cell phone, or even furniture... its time for a change so they make that change, and move the old mac into a less crucial position, handing it down to another part of the house, another family member, the backup office, whatever.

Apple just needs to keep price parity in mind, and get flexible good designs out with a reliable CPU vendor. The market for towers itself isn't going to change in its needs that much, beyond the competition looking more enticing.

[Pin-Fisher]

Originally posted by iGAV


Opteron's are only going to be used for servers from what ive read...


__________________



Its not the hardware that actually MAKES it a server. Its the use of the machine. You could use a 386 for a "Server" if it serves up text files. I could use the new XServer as a desktop if I really wanted to. ....Its the OS (sometimes) and the applied use....

[ktlx]

Originally posted by Wry Cooter

I think it pays to keep in mind that intel/AMD are not always performing as well as they promise either. Wasn't the itanium or xeon supposed to be the first x86 64 bit chip? And that pretty much wasn't considered very successful at all, except perhaps for keeping your coffee warm.



The Itanium is not a 64-bit x86 chip. It's instruction set has nothing in common with the x86 instruction set. Someone may do the logical thing and make an x86 emulator for it, but the PowerPC probably has more in common with the x86 chips than the Itanium does. :D The Xeon is not a 64-bit chip. It is basically a Pentium III or Pentium 4 with a much larger cache.

Don't think of the Itanium in terms of the Pentium 4 or PowerPC but in terms of the first 32-bit processors. It is a first generation
processor that is significantly different than its predecessors. While it has yet to be shown that Intel's choices will pan out, if the problems can be fixed by throwing money at it, Intel will do just fine. :D



The realm of 64 bit architecture has been threatened for years from PPC competitors, yet nothing successful has made it to market. And why is this?


I am not sure what you are saying here. There are lots of 64-bit chips out there today. The Alpha, MIPS, PA-RISC, Power4, UltraSPARC are out and have been there for some time. They just are not on the desktop because no has figured out how to make a 64-bit processor cost effective for desktop use. Typically you go to 64-bit to increase memory access, not for blinding speed at 32-bit operations typical of games and graphics processing. A 64-bit processor does not normally execute two 32-bit instructions at the same time; it executes a single 64-bit instruction.

[ktlx]

Originally posted by Pin-Fisher


Its not the hardware that actually MAKES it a server. Its the use of the machine. You could use a 386 for a "Server" if it serves up text files. I could use the new XServer as a desktop if I really wanted to. ....Its the OS (sometimes) and the applied use....

I think what iGAV was getting at is server processors are judged by a different standard than desktop processors. While you can use a 386, Athlon or Pentium 4 for a server, the lack of 64-bit addressing and tiny caches limit their use to simple print, Web and file serving. That is not where the money is.

The money is in building hardware to chunk on large databases and process massive amounts of data. In this arena the hardware has a pretty high market value and the margins are pretty sweet. But you need a 64-bit processor with large caches (more than 1MB) in order to cut the mustard. This is where AMD's 64-bit chip will be targetted and not the legions of Doom players who make up AMD's current core.:p

[Pin-Fisher]

Dig it...good points......thanks!!

[jefhatfield]

Originally posted by Recoil
If apple comes out with a lame 1.2Ghz G4 then I'm going to the dark side!

Hello dual opterons!

being a microsoft techie, i thought the darkside would be bad

actually pcs are not bad when they are working and are faster than macs

the problem is that pcs break down a lot more and there is much more downtime

so when i log on at home, 60 to 70 percent of the time, i use my ibook instead of my pc laptop

with the pc, some pages won't open up all the way and i get fatal errors which kick me off the internet altogether...that never happens on the ibook that much

also, the pc has its printer settings inadvertently change on me and i have to defrag the thing once a week to avoid the major pc problems

in these days, i think both pcs and macs are completely useful machines, but the uptime on the mac makes it more of a tool and less of a burden

the pc was a burden during windows 95a and before, but now it's a almost perfect tool

while the mac is not perfect, it is much, much closer

...so if i had to have just one machine in my home (i have two right now down from five), i would choose a mac...i have plenty of time to tool with pcs and their problems when i work

[iGav]

I didn't say....

Opteron's are only going to be used for servers from what ive read...

xelterran did...... :D

[barkmonster]

why is nobody looking at history. you went from 400 to 500 and 550, thats 150mhz. next time, 550 to 667, and 733, that's 183hz jump, then you went from 733, to dual 800 and 867, not counting the dual because at the time only like 2 programs could even use the second processor and the os couldn't so, that's a 133 jump again. then from 867, to dual 1gigs, well, that's only a 133 again, jump, anybody seeing the connection here.

I take it you didn't :D

AppleSpec (on Apple's site) (http://www.info.apple.com/support/applespec.html)

From that you can see it went like this, in release order, only going off CPU speed not dual configurations :

fastest models, the code number of the motorola chip is at the end :

450 Mhz / PPC7400

500 Mhz : + 50Mhz (+20%) / PPC7410

733 Mhz : + 233Mhz (+47%) / PPC7450

867 Mhz : + 133Mhz (+18%) / PPC7450

1000Mhz : + 133Mhz (+15%) / PPC7455


Entry level models :

350 Mhz / PPC7400

400 Mhz : + 50Mhz (+14%) / PPC7410

466Mhz : + 66Mhz (+16.5%) / PPC7410

733Mhz : + 266Mhz (+57%) / PPC7450

800Mhz : + 67Mhz (+9%) / PPC7455


When the G4 suddenly increased by nearly 50%, it was because increased pipeline stages (4 vs 7) allow for higher clock speeds while adding a slight speed hit on a Mhz for Mhz basis. The faster L2, faster FSB (100 vs 133MHz) and the large L3 cache that later became DDR all make up for the loss to some extent. After only 2 different ranges of powermac, we've gone from 733Mhz - 1Ghz, that's not bad as a speed gain percentage.

133Mhz Increments are not really something that matter, it's how much faster the new CPU is than the previous fastest chip that counts. for example, A 1.133Ghz G4 wouldn't offer the performance benefit over a 1Ghz that a 600Mhz G4 would offer over a 466Mhz chip, the added performance of the extra Mhz decreases as the clock speeds get higher, they need to increase in 200Mhz increments now we're at 1Ghz or we're going to see that performance increase dwindle to insignificance if we only get 133Mhz more every 5 months or so.

[Cappy]

Originally posted by barkmonster

When the G4 suddenly increased by nearly 50%, it was because increased pipeline stages (4 vs 7) allow for higher clock speeds while adding a slight speed hit on a Mhz for Mhz basis. The faster L2, faster FSB (100 vs 133MHz) and the large L3 cache that later became DDR all make up for the loss to some extent. After only 2 different ranges of powermac, we've gone from 733Mhz - 1Ghz, that's not bad as a speed gain percentage.

133Mhz Increments are not really something that matter, it's how much faster the new CPU is than the previous fastest chip that counts. for example, A 1.133Ghz G4 wouldn't offer the performance benefit over a 1Ghz that a 600Mhz G4 would offer over a 466Mhz chip, the added performance of the extra Mhz decreases as the clock speeds get higher, they need to increase in 200Mhz increments now we're at 1Ghz or we're going to see that performance increase dwindle to insignificance if we only get 133Mhz more every 5 months or so.

There are a couple of things to add and point out here:

1. That depends on the architecture besides the pipeline increases. In the Intel world when Intel came out with their Northwood core P4's at the same clockrate as some of their previous P4's, they saw huge performance increases. Much of that can be attributed to the L2 cache increase.

2. In the x86 market it is normal to see cpu's come out with only 100mhz(P4) and 66mhz(Athlon XP) increments. They just do it more often than Moto and IBM do.

Lastly on a somewhat different note I was just thinking about how the P4 has made so many people aware now about the pipeline increasing and how it affects performance. A few articles on that simply has now made everyone believe that this impacts performance more than anything. It can but just like mhz cannot be tied to performance strictly, the same applies to the rest of the architecture. There are other variables which Apple we've seen is exploiting by the addition of L3 cache.

[GPTurismo]

First, the power4 is ppc.

second OH NO MY 700 MEGAHERTZ SGI MACHINE IS LAME BECAUSE OF THE LOW MEGAHERTZ!!! EVEN THOUGH IT OUTPERFORMS EVERY WINDOWS MACHINE OUT THERE BY AT LEAST 2 FOLD!!!

Why?

It has no system bus, every device can talk to each other at FULL speed. I can't remember what they call it. interconnect something...

Folks, it's moronic to focus on something which only means you can brag to your pc friends.

It's like putting a 800 horsepower engine in a car with a crap transmission.

No Performance, but all brag.