Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Originally posted by ddtlm
Chryx:

IBM will have their own chipsets as they always do... I would be amazed if the nForce2 was even on their radar. People *love* to believe unsubstantiated rumors round these forums, but just cause the nForce2 exists, just cause it has two channels of DDR memory, and just cause it sounds like it could be cool... we have not been provided any reason to believe that it will ever show up in a Mac.


Absolutely, I think it's very unlikely that we'll see Nvidia chipsets in Macs at any point, I would like to see a DSP for audio purposes though, realtime 5.1 encoding would seem like a Digital Hub features if anything is :)

The Athlon's FSB "which is obviously the bottleneck" is not really standing in the way of progress, see http://www.tech-report.com/reviews/2002q3/athlon-333bus/index.x?pg=1
for some hard numbers. At anandtech there is an article http://www.anandtech.com/showdoc.html?i=1344 which shows that lower speed Athlons (up to 1.2ghz) do not benefit much from DDR vs SDR


The higher clocked Athlons (1.8Ghz / 2200+) ARE FSB limited, to the extent that if you overclock the FSB from 133Mhz to 166Mhz you get roughly two speed grades worth of performance at the same clockspeed.

133/166Mhz FSB article on Tech Report

And for the record, I saw a VISIBLE speed improvement switching from PC133 (cas2) to DDR (also CAS2) with the same AthlonXP 1700+, it wasn't just a "ooh, it benchmarks faster now" thing, games ran visibly smoother for example :) (same board BTW, SiS 735 based)

having a big fast L3 cache is all well and good, but it's not much use if your dataset is larger than the cache size.
(how fast is the L3 cache on a Quicksilver anyway, and not the raw cache speed, the speed it connects to the G4?)

Also, all the coders I've spoken to on the matter say that Altivec is bandwidth limited a lot of the time, a faster FSB and a suitably meaty memory subsystem would be of huge benefit to the Mac based on what they've told me.
 
sneed:

You bet, huge heatsink plus large, slow fan => heat "breezes" away. Very quiet. Sadly, I've seen things like that in some OEM PC's for a while, bout time Apple got on the ball.

mischief:

Ahh, yes seems I did. But I still don't agree with you. :)
 
Re: Re: Re: Not to dampen the parade, but,

Originally posted by bobindashadows

The heat sink wouldn't dampen the sound of the fans, well maybe a little if the fans are close enough to the heat sink, though i doubt Apple would release a machine that loud ;-)

Um, that would increase the turbulence... and hence the noise?...
 
BTW, what the heck does 8-way superscalar design mean? Does it mean 8 cores on a single chip or 8-stage pipeline or that it can clock like crazy?
 
Chryx:

Do you find it amusing that you linked exactly the same article that I did? In any case, the 25% increase in FSB sure did not yield anywhere near a 25% increase in performance, which I believe backs up my assertion that the Athlon is not especially bound by it's FSB speed.

My point about my Athlon on PC133 is that is so much faster than my G4 on PC133, which suggests that PC133 would not benefit the G4 much in situtions that I have encountered.

The L3 on current Apollo G4's is 1/4 clock speed and DDR, so it's effective 500mhz on the 1.0ghz machine. The data bus is 64 bits wide. This means the bandwidth is not that high, but the latency is still very good compared to main RAM. Even though 2mb is small compared to main RAM, most of the time it should do a very nice job of reducing main memory usuage.

I agree that AltiVec can use up all available memory bandwidth, but for most of us that just isn't an issue. I'd much rather have G4 clock speed than a pure DDR memory system right now.
 
A New G4 Too?

Adding to the speculation, what about the G4? IBM might be building one, but the 64 bit processor gets the attention. The last statement is, "... implements a system interface capable of up to 6.4GB/s." The system interface a weak link of current G4s, yes? Low end PowerMacs would likely use a G4, but current G4s may not be compatible with the G5 for motherboard functions. If IBM builds a G5 for Apple, it would be simple to do a G4 with the same system interface.
 
dongmin:

"Superscalar" => beyond-scalar => more than one instruction per clock cycle. They are claiming this "G5" can theoretically do 8 instuctions per clock cycle (in an ideal case obviously).
 
[/i]
Chryx:
Do you find it amusing that you linked exactly the same article that I did? In any case, the 25% increase in FSB sure did not yield anywhere near a 25% increase in performance, which I believe backs up my assertion that the Athlon is not especially bound by it's FSB speed.


It wasn't bound by it's FSB, but it's becoming that way, when a 33Mhz DDR bus speed increase gives you the same performance increase as a 200Mhz core increase?.. well, that should tell you something, and what it should tell you is that the core is starving for data :)

My point about my Athlon on PC133 is that is so much faster than my G4 on PC133, which suggests that PC133 would not benefit the G4 much in situtions that I have encountered.

oh, you just reminded me, the Athlon has a much meatier FSB than the G4, which means that it can get much closer to saturating the memory subsystem whilst it's communicating with other things in the system (eg, it has the "opposite" of the Xserve situation, the processor can service everything but the memory subsystem is easily maxed out), also I've heard from (fairly reliable sources, but don't quote me on this) that the G4's FSB is actually quite inefficent on a per-mhz basis.

Also, the Xserve benchmarks I've seen would indicate that DDR helps quite a bit... unless there's something going on in the Xserve architecture that nobody is telling me about :)

The L3 on current Apollo G4's is 1/4 clock speed and DDR, so it's effective 500mhz on the 1.0ghz machine. The data bus is 64 bits wide.

Yes, I know that, but is the actual interlink between the G4 core and the L3 running at 250Mhz DDR?

This means the bandwidth is not that high, but the latency is still very good compared to main RAM. Even though 2mb is small compared to main RAM, most of the time it should do a very nice job of reducing main memory usuage.
Indeed, cache is a good thing, but it can't make up for a slow FSB AND a slow memory subsystem all by itself, a balanced architecture is better IMO.

I agree that AltiVec can use up all available memory bandwidth, but for most of us that just isn't an issue. I'd much rather have G4 clock speed than a pure DDR memory system right now.

Personally, I'd rather it were clocked higher (substantially), AND had a DDR memory subsystem AND 4MB of DDR Sram, but I'm picky like that :)
 
Originally posted by sneed
I was under the impression that a heat sink reduces the need for fans.

Well it does, but i mean there's a pretty good size heat sink in the G4 tower and it still has a number of fans. I personally don't know the heat produced by the power4, but I don't think even a huge heat sink is going to eliminate the need for fans. Like i said, Apple isn't going to produce a loud computer anyway so why argue about it?
 
BTW ddtlm, when I was referring to the Athlon's FSB being a bottleneck, I was referring to it not being able to utilise all the memory bandwidth available, NOT to it being 100% memory bound.

If the cpu NEEDS that extra bandwidth or not is a different issue, the point is that having the bandwidth is fairly pointless unless the chip can actually use it :)
 
Re: Re: .1 micron? don't think so

Originally posted by Chryx


I think Intel have already got .09 stuff working in the labs?

Ah - well like i said i probably got the number wrong, but i'll talk to him again
 
Chryx:

As revealed on this page http://www.tech-report.com/reviews/2002q3/athlon-333bus/index.x?pg=3 of TR's review, the increase in core clock was in fact 133mhz in one case (2 speed grades) and 166mhz in the other, not 200mhz. Additionally, those 33mhz of FSB represented a 25% increase whereas the 166mhz represented only a 10% increase and the 133mhz represented only a 7.5% increase. As far as I can tell, in NOT ONE benchmark did the FSB increase yield more performance than the clock speed increase, although at ~ the same core clock, the faster RAM almost always meant a faster system. Looks pretty definate to me: the 333mhz FSB is no big deal.

The bus connected the L3 to the 1.0ghz G4 runs at 500mhz (250x2).
 
Originally posted by ibookin'@mwny

I think topicolo is right. I haven't seen any mobos with true HyperTransport technology yet myself.

both the Xbox and the Xbox derived Nforce boards have real Hypertransport buses on them...

If you mean CPU <> other component interlinks, then no, they haven't been seen outside of an AMD Hammer demonstration.
 
Originally posted by bobindashadows
... Apple isn't going to produce a loud computer anyway so why argue about it?
You must not have one of the 2002 Quicksilvers. I owned a dual 1ghz and the thing was so noisy I sold it. I hope Apple gets noise under control with the new case design. At least put in some higher quality fans that have a lower noise output. The new iMac G4 fans run virtually silent. Let's hope the new PowerMacs are quieter than the Quicksilvers.
 
.10 micron!

Oh if these are going to be built in their recently opened ny factory then they will be .1 micron because they announced this at the opening of the factory.
 
ibookin'@mwny:

The nForce does in fact have a hypertransport link between it's chipset components and it does exist. Note, that's the nForce (1) not (2), so it's not even all that new.

Hypertransport is no big deal IMO, won't replace the current FSB, won't replace AGP... won't replace PCI... it's an internal interface. It is designed for chip-to-chip comunication such as for AMD's upcoming SMP systems and nVidia's nForce.
 
Originally posted by ImAlwaysRight

You must not have one of the 2002 Quicksilvers. I owned a dual 1ghz and the thing was so noisy I sold it. I hope Apple gets noise under control with the new case design. At least put in some higher quality fans that have a lower noise output. The new iMac G4 fans run virtually silent. Let's hope the new PowerMacs are quieter than the Quicksilvers.

It could be loud as hell and I wouldn't care. As long as it's fast! ;)
 
Originally posted by Joshlew
64-bit G5 by IBM.........

Excellent............:cool:

and I wrote in another thread (in part)

Originally posted by Rocketman


The fact that future Macs will not even boot on OS9 is your indicator of things to come. Unix kernal is OS agnostic and processor agnostic. A today Mac for example can boot OS9 and thus VCP running every major flavor of Wintel OS on a "virtual machine".

OSX classic is OS9 on a "virtual machine". In short very son now your main OS will be Unix/OSX and any legacy applications will run on virtual machines smoother, faster, more reliably than they did on legacy hardware!

This is a good thing. Furthermore this whole environment will run on whatever chips happen to be hooked underneath. Right now the debate is PowerPC v Intel v AMD. But in the near future it will be Sparc v Power v BioChip(tm).

This is the very leading edge of the "new era" in computing. On the lower end the CPU agnostic OS allows use in embedded systems like phones and pagers and dishwashers.

We are witnessing the first round of an infection perhaps more pervasive and long lasting than any windows variant, which itself was lisenced from, um, APPLE!.

Rocketman

This also means servers can run 64 bit applications right now (many web server applications have been compiled for 64 bit already). This is NOT a consumer chip and even might not be a prosumer chipset. It is a server.

The bottom line is this. When a superchip comes out, it can have a low level kernal written for it to chat with OSX and is immediately useable. I suspect 64 bit Power chips will have 32 bit and perhaps even 16 bit modes for compatibility during transitions.

The scary thing is this announcement will start the rumor mill about 128 bit chips. "Not that there's anything wrong with that."

The rest of the computer cannot keep up with the chips now. Hypertransport is one bottleneck reduced, but I/O is going to require a Firewire 2 main drive bay protocol as well and possibly some sord of solid state drive to keep up with it all.

I used to have a solid state drive hooked to my Mac+ and to be honest my G4 is not all that much faster.

Rocketman
 
Just my two cents worth. Opteron will be launching shortly. I reckon that we will see G5 much much sooner than later. Please don't flame me for saying this but i would not be surprised if we see one of these babies announced at the end of this month with production and shipping to begin in October. Apple probably have one of those agreements with IBM as they did with MOTO where they get to demonstrate the newest processor first.
Also for what ever its worth i live in Ireland. There is a huge apple manufacturing plant down in the south of Ireland in a place called Cork. I rang apple today because i was not so sure whether or not i should place an order for a DP 1GHZ ... i was worried that it might go out of date. The sales rep on the phone (very nice to deal with) said that i should not worry.... it was said to me that i should not worry because if i buy a G4 dual Proc this week .. if apple were to launch a G5 in the next few weeks.... and she would not commit on knowing anything officially about a g5 launch ...soon after they will upgrade it to the G5 for me at cost difference.
The very fact that she did not say "listen there s no such thing as a G5 and if there was it would be months if not years away" on top of the fact that IBM have this wonder proc.... makes you think. Just my two cents worth. I don't claim to know anything!
Regards,
i_wolf
 
Silent Fans?

Originally posted by ImAlwaysRight

You must not have one of the 2002 Quicksilvers. I owned a dual 1ghz and the thing was so noisy I sold it. I hope Apple gets noise under control with the new case design. At least put in some higher quality fans that have a lower noise output. The new iMac G4 fans run virtually silent. Let's hope the new PowerMacs are quieter than the Quicksilvers.

What I can't understand is that we have the technology to make computer chips with 90nm pipelines, but are unable to come up with a silent fan :confused:

I know the make some noise owing to air turbulence, but maybe the blades need redesigning? I'm not a mechanical engineer, but is there some physical reason why it's not possible to make a fan quieter?
 
Re: Silent Fans?

Originally posted by StuPid QPid

I know the make some noise owing to air turbulence, but maybe the blades need redesigning? I'm not a mechanical engineer, but is there some physical reason why it's not possible to make a fan quieter?

A fan can only push so much air. You can make the blades bigger and make it spin slower, make the blades smaller and make it spin faster, etc. It's cost vs. noise. Most people side with cost. Those that have no regard for cost go with water cooling. :)
 
Originally posted by dongmin
Yeah, Altivec is here to stay. IBM must be licensing it from Moto. Or maybe Apple bought out the PPC assets from Moto. It's one aspect of the G4 that really shines. So it's good to see Apple keeping it around and committing to it more. Hopefully developers are hearing this and following suit.
I agree that Altivec is here to stay (in some form or another).

But, there is no way that Apple has bought out the PPC assets of Motorola. If they had, there would have been press releases and SEC filings. Publicly traded corporations can't make large deals like that without announcing it. More likely, if the IBM chips do have Altivec, is that Apple has persuaded IBM to license Altivec from Motorola or that IBM engineered their "own" version of it.
 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.