Upcoming Intel Details Leak

[Nemesis]

Originally posted by wizard
I hate to break the news to you but all indications are that the latest chipsets from Intel, those support the 800MHz FSB are some of the best that Intel has come out with in a long time. Many are reporting very easy over clocking of the FSB. So it may not take a new chip set just an uprated device for themain stream market. Also realize that Intel quad pumps the bus so in realty the MHz values are actually lower that the Apple/IBM Bus.

While Apple/IBM have an interesting implementation of a FSB, it is by no means perfect. Also a fast FSB will not make up for all CPU short comeings.

Dave

Uf, uf! This FSB thing is interesting because it shows 2 nice things:

1) No more "monthly GHZ updates" from Intel's side. If they want their procs to run smoothly, they should make GHz leaps according to FSB speed: N x 800 MHz (or whatever the speed of FSB) = processor speed. Faster the FSB, larger the step. Voila!

2) Apple didn't announce everything, they have many aces in the pocket too. Like that with 3GHz G5 they announce 1.5 GHz FSB too ... ;-)

 

[nuckinfutz]

Well the argument is pretty easy to make once you realize that Floating Point isn't high on everybodies list of requirements. The G5 clock per clock does not perform all that much better than a G4. So running a G5 at 1.2 GHz when faster G4's are available does not make sense. This is especially the case if those G4's are low power. Now if Apple/IBM can deliver a low power 970 at 1.6 or 1.8 GHz that would be another story.

The only problem with this is that NO ONE has any definitive benchmarks to either prove or dissaprove your statements. The only thing I've seen is the NASA Benchmarks which show that Alitvec on the PPC 970 is superior primarily because of the increase in clock. I haven't seen anything that shows that Integer performance follows the same pattern. FPU is important to those in Content creation that run Maya or Scientific apps. FPU is a strong part of the P4 as well.

Canterwood is a very interesting chip set, and form all appearance to date is one of Intels better efforts. Much of Apples advantages are with support of Hyper Transport.

Both motherboards attempt to remove bus contention. Intel has their Communications Streaming Architecture(CSA) where they put Gigabit LAN and perhaps other tech. Apple does the same by using their system controllers

Of primary interest is the 7) the High Performance I/O which handles all I/O tasks and uses the 16bit Hypertransport links. It's basically just like CSA. System Controller 4) links the Memory and PCI-X(in some models). Nice because I have the extended bandwith to run my vertical market cards without affecting my basic I/O.

Both Architectures are actually more similar than they are dissimilar. I think it points to a saying of mine which is "The Hardware will generally be close, it's the software in which you will make your mark"


Dave I can definitely agree with many of your statements. We truly won't know more until the computers are shipping but I respect your classy responses. All in all both platforms have grown well and this is a benefit to all consumers.

 

[Dave K]

Originally posted by nuckinfutz
The only problem with this is that NO ONE has any definitive benchmarks to either prove or dissaprove your statements. The only thing I've seen is the NASA Benchmarks which show that Alitvec on the PPC 970 is superior primarily because of the increase in clock. I haven't seen anything that shows that Integer performance follows the same pattern.

Actually, the initial numbers from that NASA tests in the 100's are for non-Altivec scalar math and showed that when using that specific code optimized for the G4 a single 2 Ghz 970 comes out about 142% faster than a 1 Ghz G4.

Roughly linear performance gains for Altivec were expected because, in general, it's the same between the two machines and the 970's not introducing a new "Altvec Extreme" version.

The catch here being that the test was done using data that would fit rather comfortably inside a G4's 1MB L3 so it's bandwidth cap isn't a tested variable.

 

[MasterX (OSiX)]

Is it possible that you're overlooking something? A few things I believe.

Primarily, any large jump in clock accompanied by an INCREASE in per cycle efficiency is a great accomplishment. The fact that the g5 is only marginally faster per mhz should be impressive to you, not under-satisfactory.

Secondly, Apple has succeeded where AMD has failed. Hailing 64-bit as the end all for Intel. I never cared enough to check, but from what people have told me the Opteron has had it's ass kicked in places where it needed to shine to become a desktop competitor. While I'm sure that in the server world where there exists a market for highly per-system optomized code and software, competing in the desktop space is far more difficult.

Third, Apple hasn't yet finished a fully 64-bit version of MacOS X. Many computations in OSX are (thankfully) passed on to code created by Apple. This is how I believe Apple has managed to squeeze out so much performance, specifically through use of altivec acceleation in system frameworks, compiler, etc. In the future when we have a full 64-bit compiled AND OPTIMIZED version of Panther we can expect even more superior performance.

Fourth, you can argue Apple's tasks are biased. If you play WWDC again you will notice the P4 does well and dandy until it gets to scaling down this massive TIFF of a whale. If I remember right, 64-bit chips should do best in massive computations of that type. It's clear that given the task of optimizing Photoshop for the g5 in order of what is used most and where the best performance would come from, scaling would be high in that list. But like the infamous debates over Quake3 driver tweaks, technically (I believe) that optomizing for something like scaling in phohshop then that task being the difference in winer and loser in apple's P4 vs G5 showdown is a fair fight. Who's to say that that isn't what has been killing performace on large photoshop projects all this time?

I don't know if there is a way to write for 64-bit specifically, or if all the 64-bit tweaks are done in the compiler. But if there is a way to code with 64-bit in mind that will be another example of increased performance to look forward to in the future. In my opinion it's not a question of if this current g5 line up will beat the new P4, it will. But I'm afraid Apple will no longer win on SPEC tests unless some crazy stuff happens inside the new P4s. Still If you remember from the days of "slow g4s" Apple was forced to very heavily optimize their code, as did 3rd parties. Now since most of that optimization took the form of Altivec we can reap the benefits on the g5 immediately, due to it's highly advanced Altivec support.

Outstanding

 

[Cubeboy]

A few things to note

1) Prescott will almost certainly be the Pentium 5 as it's a major enhancement in nearly every respect to the Pentium 4 (I'll list all the enhancements later). We should also consider the span of time that's passed since the original Pentium 4, considering about two years have passed, it's time for a new Pentium.

2)How high a cpu scales depends on it's architecture and the chip makers manufacturing capabilities. In the case of the Pentium 4, a move from 180 nm process to 130 nm process resulted in about a 50% increase in clock rate. Moving from the current 130 nm process to Prescott's 90 nm process well result in another 50% increase in clock rate allowing Prescott to scale to nearly 5 GHz. Of course thats only assuming moving to a smaller process, Prescott will also be the first cpu to utilize Strained Silcon which should be able to allow for even higher scaling in clock rates. Exactly how much higher is currently unknown.

3) Prescott will be a modified Pentium 4 core, but thats not to say there won't be any significant gains in performance, the 1 mb L2 cache alone should be able to boost application performance at least 5-10% (moving from 256k to 512k on the Pentium 4 resulted in a 7-17% increase) and SPEC by at least 10%. The Larger Data Cache, which maintains the same latency as the previous Pentium 4 should be able to offer a even more significant boost than the L2. Almost every other aspect of the Pentium 4 has also been enhanced, the Larger Trace cache increases the number of uOps it can issue every clock cycle (about 33% higher IPC), twice the number of entries in it's BHT gives it much better branch prediction, it can handle more instructions in flight, it can hold twice as many entries in it's load and store buffers, latency on integer multiples is lowered, larger integer and floating point register files, improved hyperthreading, SSE3 and probably a few minor core revisions (ie additional write buffers and what not). It might also have 32/64 bit capabilities, and feature a faster FSB which might be why motherboard makers have begun to show off motherboards that support 1.2 GHz Quad Channel RDRAM modules.

Suggested Readings:
http://www.chip-architect.com/news/2003_04_20_Looking_at_Intels_Prescott_part2.html

http://www.theinquirer.net/?article=10424

http://www.theinquirer.net/?article=10367

 

[sososowhat]

Originally posted by ddtlm
mgargan1:
Pentium does not mean five, its a made up word that suggests "five" as well as a metal. (And other stuff, I'm sure.)

I was surprised when they went to Pentium II after Pentium.

286->386->486->Pentium->Sexium would have made more sense.

 

[rutabaga]

Originally posted by nickmcghie
On particularly interesting note was about heat. x86 CPUs already get hot and require considerable cooling, but this is getting worse and eventually it will hit a wall. A report by the publishers of Microprocessor Report indicated that Intel is expected to start hitting the heat wall in 2004. The article goes on to predict that the x86 will soon lose it's price advantage because the cooling systems required to keep these beasts from literally melting cost several times what the CPUs themselves cost.

Long live RISC!!

this is just a dandy bit of intellect. apple just came out with a tower which contains the most advanced/sophisticated cooling system ever employed for a desktop system. by a wide margin. don't you think this price is being passed on to the macintosh consumers as well? meanwhile, current pc towers employ a heat sink and a couple of glue on fans. don't you think the manufacturers are passing those savings onto their customers?

let us not make the mistake of underestimated the BRILLIANCE of intel's engineers. they have steadfastly overcome every obstacle that various prognosticaters have predicted would derail the cisc architexture. they will continue to do so for the forseeable future.

 

[SubXaero]

Re: A few things to note

Originally posted by Cubeboy
1) Prescott will almost certainly be the Pentium 5 as it's a major enhancement in nearly every respect to the Pentium 4 (I'll list all the enhancements later). We should also consider the span of time that's passed since the original Pentium 4, considering about two years have passed, it's time for a new Pentium.


Suggested Readings:
http://www.chip-architect.com/news/2003_04_20_Looking_at_Intels_Prescott_part2.html

http://www.theinquirer.net/?article=10424

http://www.theinquirer.net/?article=10367

the "tejas" revision will be the next P5

and for anyone whos interested here are intels roadmap:
Roadmap

all i can say is, if things stay on course is IBM is gonna be very busy

 

[LinuxGigolo]

Originally posted by wizard
It is really up to the consumer to decide what is fast enough. Even for home use there are somethings that would be best handled by extremely fast machines. Machines that honestly aren't on the market yet, nor will they be anytime soon.

Your statement about consumer machines smack of comunism. It makes about as much sense as the people who complain about pickup trucks and SUV's. Some of us can't fit into a Saturn literally and some of us need more that a G5 to get our work done.

So don't run around trying to fit a limited vision of reality on the rest of us. Faster PC's or Macs are needed by many of us at home and the reality is the professional applications are unbounded.

Dave

My point was actually less about the fact that macs are slow enough/fast enough and more about the fact that people complain the second something new comes out that it needs to be implmented across the line. This sort of reminds me of the whole iPod Software 2.0 complaint whereby people somehow suddenly stop liking their older iPods just because something newer came out. (I own an iPod that does not support software 2.0 and if I want it, I'll buy a new iPod... but just because a new one came out, mine didn't lose any of its functionality). Nothing is stopping home users who need the power of the G5 architecture from buying Power Mac G5s, just as nothing stops people who can't fit in Saturns or who want different/bigger cars from buying them. Thinking that Apple will put a G5 into an iBook/eMac anytime soon, though, just seems silly.

 

[Cubeboy]

Re: Re: A few things to note

Originally posted by SubXaero
the "tejas" revision will be the next P5

and for anyone whos interested here are intels roadmap:
Roadmap

all i can say is, if things stay on course is IBM is gonna be very busy

I doubt the author of that page knows whether or not Prescott will be the Pentium 5, especially considering Intel's official roadmap doesn't cover anything beyond a 3.2 GHz Pentium 4.

The most widely accepted "leaked" roadmaps (look up The Inquirer's or Extremetech's roadmap) point out to Prescott being the Pentium 5 and Tejas being the Pentium 6. Which considering the sheer number of modifications and duration of time passed from the Pentium 4, would be quite reasonable.

I don't really see how Tejas can be considered the Pentium 5 if Prescott isn't when their both are supposed to be modified Pentium 4 cores, we probably won't see a entirely new core until Nehalem.

 

[soggywulf]

Originally posted by LinuxGigolo
Thinking that Apple will put a G5 into an iBook/eMac anytime soon, though, just seems silly.

It's not silly at all, at least for the iMac and eMac. Unless we approach it from a sheltered Mac-only perspective.

If we want consumers to buy Macs, the consumer Macs have to stay relatively competitive with PCs at the same price points.

 

[fpnc]

Originally posted by nuckinfutz
...The only thing I've seen is the NASA Benchmarks which show that Alitvec on the PPC 970 is superior primarily because of the increase in clock. I haven't seen anything that shows that Integer performance follows the same pattern.

...and...

Originally posted by Dave K
Actually, the initial numbers from that NASA tests in the 100's are for non-Altivec scalar math and showed that when using that specific code optimized for the G4 a single 2 Ghz 970 comes out about 142% faster than a 1 Ghz G4.

Dave K, note that nuckinfutz was asking about integer performance. The scalar math in the NASA report is still floating point, not integer. This also highlights the one area where the G5 architecture is clearly superior to both the G4 and the Pentium. From that same NASA report you can see that MHz-to-MHz the G5 is about 23% faster than the G4 and over 32% faster than the Pentium 4 (when measuring scalar MFLOPS/MHz).

 

[Analog Kid]

Originally posted by nuckinfutz

Again...no empirical evidence other than the profound "I doubt". You have no technical info to back up your statements at all. The burden of proof is with you .

Not much empirical evidence on your side either, but let me put in some thoughts:

P=.5fCV^2
'C' close to doubles when you double the bit width of the bus. 'f' goes up by a factor of 6 or so. This is just the bus between the CPU and the memory controller.

Then there's that 128bit, 400MHz, DDR (800MHz data) bus to main memory.

Then pull a datasheet on PC3200 memory.

Then remember Job's comment that the memory controller chip was built on the same process that the 970 is built on which means it's already at .13 (and not likely to go to .09 so soon).

Then realize that compromizing any of these specs starts to starve a core that has no L3 cache support.

Then read the news stories about new smaller, faster G4s out of Mot, and the answer seems pretty clear...

It's not all bad-- Mot has been talking about a dual core G4 which would make a pretty sweet basis for a laptop in my mind...

 

[Dave K]

Originally posted by fpnc
Dave K, note that nuckinfutz was asking about integer performance.

Given that all the talk from the initial quote and the trailing on that was about FP and combined with the comment about the NASA stats only refering to Altivec-enhanced FP, I presumed the integer reference was a typo...

 

[nuckinfutz]

Given that all the talk from the initial quote and the trailing on that was about FP and combined with the comment about the NASA stats only refering to Altivec-enhanced FP, I presumed the integer reference was a typo..

Actually I'm interested in both. However my post tended to combine the two in a confusing way.

AK- Thanks for that explanation. I'm curious to see what the low voltage PPC 970 shows for speed but it looks like Apple may skip in entirely. We'll see. Hell a G4 would be pretty sweet with a Ondie Mem controller.

 

[shawnjackson]

Originally posted by AhmedFaisal
First thing I did on my dual 1.42 was switch the fans and especially the CPU cooling system to that of a 3rd Party, now its quiet like my baby son when he sleeps. At any rate, P4 is one lousy selection for that home entertainment system. Since Apple is still denying us a pizzabox headless model I look what I could find in the PC world and this is what I use: www.hush-technologies.com. Its a neatly designed machine with a VIA Epia M 10000 ITX board, a 120MB sound insulated seagate barracuda and a slimline combo drive. All passive cooled so no noisy fans involved and so small it fits perfectly to my stereo and looks absolutely cool too. Its plenty fast for DVD & DivX and AAC audio music files. Only thing I added extra was a M-Audio Audiophile since I prefer an audio card with CINCH output and the sound of the audiophile is simply pristine.
Cheers,

Ahmed

actually, my htpc system is not running windows... its running linux with mythtv and various other parts built into it. that mythtv system is pretty suite and kicks the crap of just about any pvr system i have used, including eyetv, windows xp media center edition, sagetv, and snapstream... its really a dream to work with

not a dream to install though

 

[TMJ1974]

Originally posted by shawnjackson
no you silly guys... they arent going to call it the pentium 5... its codename will be prescott (just like northwood, etc.)

it will be known as the "D" Revision of the Pentium 4

A - 400Mhz Bus
B - 533Mhz Bus
C - 800Mhz Bus
D - not sure on spec but I know its double the cache of the c and is 90nm

oh and anyone who keeps saying that intel chips run hot should actually use one. I have two dp 1.42, an athlon xp 2100+ and a 1700+ and a 3.0 p4 c...

the g4 and the athlon are probably tied for "hotness" at around 55 deg C at max load (using a thermaltake heat probe at cpu core.

the p4 runs load at about 38 with a nice quiet zalman fan. thats why its my home theater system.

Nice fleet of computers, I'm catching up to ya! Anyway, everything you said is correct....BUT, you forgot an important fact..."Prescott" uses a new instruction set, a successor to the SSE2. EVERYTIME Intel has changed the instruction set or other major technical features in such a major way the marketing name of the chip has been changed. If it were just the die size decrease or upping the cache, I'd agree with you. But, with the instruction set change, especially since the new one is specifically aimed at Multimedia authoring....it will be the Pentium 5

By the way, how do you keep the P4 at 38C....my 2.6C P4 runs at 50 with minimal load - I used the Arctic Silver and have really good case fans, I am using the "Intel designed thermal solution" which I know won't compare to other manufacturers CPU fans.

Either way....still cooler than my DP867....I had to move that to the basement, it would heat up the room upstairs way too much.

Tim

 

[Mav451]

I am running an Athlon (as you can see in my sig) and i maintain 38-40c's idle and 43c's full load.
If i dropped it to a lower clock (i.e. the default clock), probably closer to 36-38, and 40-41 load.

(P4's run a bit cooler already b/c they need only 1.5v of vcore compared to the AIUHB's 1.6)
I only have 2 exhaust fans. And i'm in my upstairs room--however, i'm running a swiftech mcx370-c

If you're using retail...you should prolly toss it (though I've heard that Intel's retail coolers are considerably better than AMD's).

If you even TRIED to run an athlon using a stock cooler, you'd hit a lousy 50+ c's on idle...C.O.P would probably shut it off if you even tried to open an intensive app (game/encoding software).

Case fans don't affect it as much if your ambient is already low enough.

 

[GroundLoop]

New Intel details

I haven't read through all of the posts, so sorry if this has already been stated. Intel's new Prescott processor will likely not have a compatible mother board until Q2 2004 due to voltage incompatibilities. Therefore, it seems that Prescott won't officially be useable until after the socket overhaul. Seems like a major oversight on Intel's part. I am thinking that there would be some kind of workaround, but I am sure that Socket 478 Precott sales just vaporized.

http://www.anandtech.com/news/shownews.html?i=20014

Hickman

 

[wizard]

I do not consider it to be silly either, on the other hand I don't expect them to use the same hardware they used in the G5. Between power usage and board space requirements I think they would have problems. That doesn't mean that they aren't working on a solution or for that matter have it already. It simply means that we would have a different architecture, probally with a lower performance.

Of course there is also the possibility of greater integration. Here a 970 with built in memory controllers and hypertransport could provide an implementation that would squeeze into the consumer lines. The problem with this is that I've seen nothing to indicate that Apple/IBM are working in this direction. It does look like there is a mad rush to get to the 980 revision as soon as possible. Probally so they will actually have a chip that is faster than the Intel/AMD offerings.

Dave

Originally posted by soggywulf
It's not silly at all, at least for the iMac and eMac. Unless we approach it from a sheltered Mac-only perspective.

If we want consumers to buy Macs, the consumer Macs have to stay relatively competitive with PCs at the same price points.

 

[wizard]

Hi Brandon;

I more or less agree with what you have said below. After all if you look at my shop you will find a good number of my fathers tools and I love to recycle older furniture and stuff.

Understanding that I'm willing to state that this perspective does not apply to computers for me and a lot of other people. Due to the rather terrible rate of performance increase over the last couple of years in the Mac line I think your going to see a huge demand for consumer line performance similar to the new G5. Apple may be able to meet that demand with an uprated G4, but that G4 will have to be a stellar performer. The problem is that most users DO benefit from increased performance of their CPU's. Even if it is to run iChatAV, which I think has the potential to drive MAc sales.

I do not tihink that a G5 in an Imac or similar computer is silly at all. It won't likely be the same design as the G5 towers but it is a possibility. How quickly it happens will be a question of market demand. I could not honestly reccomend any of the consumer Macs now to a friend. Even if they had a new G4 the reccomendation would be hard to make unless it was clearly a significant perfromer.

You see the trouble with PC or Macs is that software often does eclipse the pefromance of the CPU. It is find to buy at the bottom of the line, if such PC has a specific application, but not for general purpose machines. In that case I generlaly reccomend getting close to the state of the art in performance, this allows the owner a competitive and functional machine for atleast a couple of years. Lets face it software is the driver when it comes to deterrmining if a computer is useful anymore. Once that CPU has problems with the users application it will be labeled a dog and will be relegated to some other duty.

I don't specifically like this as a closet full of old computer hardware is not very usefull. On the other side of the coin is the desire to have a machine that runs your desired applications properly. I would also be willing ot say that most people are still not satisfied with their computing experience and would love nothing more than a little more peformance and a lot more features.


Dave

Originally posted by LinuxGigolo
My point was actually less about the fact that macs are slow enough/fast enough and more about the fact that people complain the second something new comes out that it needs to be implmented across the line. This sort of reminds me of the whole iPod Software 2.0 complaint whereby people somehow suddenly stop liking their older iPods just because something newer came out. (I own an iPod that does not support software 2.0 and if I want it, I'll buy a new iPod... but just because a new one came out, mine didn't lose any of its functionality). Nothing is stopping home users who need the power of the G5 architecture from buying Power Mac G5s, just as nothing stops people who can't fit in Saturns or who want different/bigger cars from buying them. Thinking that Apple will put a G5 into an iBook/eMac anytime soon, though, just seems silly.

 

[wizard]

Hi nuck

The only thing that I have to reference with respect to integer performance is Apples own benchmarking and a scrap of information posted awhile back. Actually very little to go on. On the other hand Apple did a lot of bake offs that took advantage of the G5's best features, and this is fully expected as it is marketing. So we really have very little to go on with repsect to how these machines perform with code that is not heavly FP based or bandwidth limited. The fact that the Specmark integer scores where low and that we do not have a L3 cache could be worrisome. Time will tell though. The big frustration is that I may not be able to experience a G5 until much later in the year! Even worst is that I may decide to go the laptop route.

The reality is that I think that objective testing will show a mixed bag as far as performance goes especially on old optimized G4 code. Especially if that code is Altvec optimised as it appears that most of the limitations of the 970 revolve around handling vector instructions (prefetches).

Dave

Originally posted by nuckinfutz
The only problem with this is that NO ONE has any definitive benchmarks to either prove or dissaprove your statements. The only thing I've seen is the NASA Benchmarks which show that Alitvec on the PPC 970 is superior primarily because of the increase in clock. I haven't seen anything that shows that Integer performance follows the same pattern. FPU is important to those in Content creation that run Maya or Scientific apps. FPU is a strong part of the P4 as well.

 

[wizard]

I don't believe it is a question of killing off laptops sales, its a question of having a solution to squeezing a 970 into a laptop. That is squeezing the processor in and getting anything that resembles a significant performance boost over a G4.

The problem with laptops is that the thermal issues revolve around all of the parts in the system. We do not yet have a good idea how the chipset in the G5 behaves thermally, it might be completely unsuitable for a laptop. We do know that high speed memory is a problem. In fact I'm willing to bet that if a 970 based laptop ever did come out from Apple it would be implemented with a differrent chip set.

Thinking has resulted in this evaluation. Every indication is that 970 based laptops are not going to be the next rev to the powerbooks. Do understand that I'd love to be wrong and find out that Apple has a solution that maintains all the good features of the current powerbooks. Number one being the ability to run on battery power for extended lengths of time. Who knows may be Apple has a norht bridge that gets rid of hyper transport and integrates much of the peripheral support chips. Now that would be a surprise.

Dave

Originally posted by Nemesis
Uff, do people really believe that Apple would kill their laptop sale by announcing G5 in a Powerbook in 3-4 months? Who would in hell buy anything before that?
Jeeez, some people are unbeleivable. Just use your heads, mates. Think! Think!

 

[ColdZero]

Originally posted by nuckinfutz
LOL. Shuttle?? Please I've heard plenty of stories about the Shuttles overheating .

Do you really want me to find more?

Don't write checks your ass can't cash. I've never heard of a G4 system shutting down due to overheating and I doubt a G5 will.

Really? Do you own a shuttle? Have you ever used one for extended periods? Have you ever seen one in person? Do you know anything about how the cooling system works in a shuttle? Probably not, so basically you have no idea what you are talking about. You're some ignorant zealot sitting at his godly mac that bashes anything that thinks different, so much for Apple's own slogan. I own a Shuttle, an iBook, a G4 and work on Xserves and the full range of Apple products every day at work. So I think I can make a comment based on actual fact and experiance rather than some google post I could get in 30 seconds. But while we're on the subject lets take a look at that post....that you didn't even read. And I quote from that very google post:

I had difficulties correctly applying thermal compound. When originally unpacking, I didn't see the thermal compound included with the SK41G, and bought some. I then carefully followed the instructions included with the compound I bought about applying a THIN layer to the heat sink and CPU. After putting it all together I booted up, and it lasted about 2 minutes before shutting down due to the heat. Took it apart again, having found the included thermal compound. I piled the whole package high on the CPU die, and haven't had a reset due to overheating since.

WOW lots of overheating problems, he didn't apply thermal compound correctly, THAT is why it was overheating. Do that to ANY processor and see if it can run without overheating.

Now lets look at some facts. You know those things the rest of the world likes to work on:

A comparison of internal case temperature
What do they say? That the shuttle best coped with the change in higher temperatures. Next:

A Low Temperature

Next:

CPU at a comfortable temp

So lets see, I have 3 sources that say that it handles temperature of the whole system well, you have one source where the guy admits that he applied thermal paste wrong and now that he has done it correctly, it works fine.

Why don't we look at some G4 overheating stories:

G4 Upgrades Overheating

TiBooks Overheating

Flaw in G4s cause overheating

I guess you just never heard of those. You've never heard of a G4 shutting down because it is overheating? I guess you don't know that the Cube has temperature circutry to do this very thing, along with every Xserve out there. But I guess you think the laws of physics don't apply to Apple computers. If 45 watts of power go into the processor, that much power needs to be dissipated in some other form of energy. Its called the law of conservation of energy, go to school and learn something.

[mod. edit - Personal attacks will get you banned. First and only warning.]

 

[de_la_saracen]

For Pete sake guys,
Who gives a toss about gains in heat and in an extra Ghz or so, and all that other techi nerdy geeky tripe! sod all that! XP can't touch OS X, nuff said! I use a Mac cos of OS X, not because of IBM or Moto or clockspeed!