Linked by Thom Holwerda on Thu 19th Jan 2006 15:02 UTC, submitted by Richard Levangie
Apple "Since the first rumors of an Apple switch to Intel, everyone has been wondering about the potential speed of Intel-based Macs. Last week's announcement of the first shipping Intel-based Macs brought with it the promise of a major speed boost: Apple's Web pages suggest that the new iMac, powered by the Intel Core Duo processor, is twice as fast as its G5 predecessor. Macworld Lab's tests do show that the new Intel-based iMac is faster than the iMac G5 when running native applications. However, we found that those improvements are generally much less than what Apple claims is a 2x improvement in speed."
Order by: Score:
give it time.
by Anatomy4Artists on Thu 19th Jan 2006 15:14 UTC
Anatomy4Artists
Member since:
2006-01-09

give it time.

Reply Score: 1

RE: give it time.
by Kroc on Thu 19th Jan 2006 15:58 UTC in reply to "give it time."
Kroc Member since:
2005-11-10

indeed. I think OSX has a long way to go as far as code optimization. X.4.5, .6 etc may include recompiled system files to increase the speed on intel architectures.

Reply Score: 1

well duh
by thebackwash on Thu 19th Jan 2006 15:15 UTC
thebackwash
Member since:
2005-07-06

Apple's speed claims are always grossly exaggerated. My first thought upon seeing these 2x-3x speed improvement claims was to compare and compound over the years how much faster Jobs claims each new generation of macs is, and then compare with benchmark scores. It would be an interesting test. Even if there is some grain of truth in this statement (perhaps the processor is 2x-3x faster,) other components then become the bottleneck. (Looking at you, hard drive.)

This is no worse than any other company's overexaggerated claims about their product. If you have realistic expectations about any product's performance, and you know what you are going to buy, you should have no reason to be disappointed. If you believe every company's market speak, you should be broke, unhappy with your purchases, and pissed off that you could be so stupid to believe everything wholesale.

Reply Score: 2

RE: well duh
by TomB7 on Thu 19th Jan 2006 15:48 UTC in reply to "well duh"
TomB7 Member since:
2006-01-03

Actually, Apple has historically been pretty honest about benchmarking. On the other hand, MacWorld has always been very conservative.

Reply Score: 0

RE[2]: well duh
by n4cer on Thu 19th Jan 2006 18:24 UTC in reply to "RE: well duh"
n4cer Member since:
2005-07-06

Actually, Apple has historically been pretty honest about benchmarking. On the other hand, MacWorld has always been very conservative.

I encourage you to go check the SPEC numbers Jobs touted for the whole G5/world's fastest computer launch. They were debunked the day of the keynote along with the fastest claim. He severely underrated the P4 that he was comparing the G5 against. Much higher SPEC numbers (higher than what he reported the G5 as having) were found for the same P4 system on SPEC's website (not to mention even higher numbers for current AMD systems at the time).

Before the switch to fudged SPEC numbers, he used to fudge Bytemark numbers by using older, unoptimized versions of Bytemark on the PC versus newer versions on the Mac to show the Mac as being faster.

These examples and others have been well documented and should be available on the web.

Reply Score: 1

RE[3]: well duh
by rayiner on Thu 19th Jan 2006 20:59 UTC in reply to "RE[2]: well duh"
rayiner Member since:
2005-07-06

I encourage you to go check the SPEC numbers Jobs touted for the whole G5/world's fastest computer launch. They were debunked the day of the keynote along with the fastest claim. He severely underrated the P4 that he was comparing the G5 against. Much higher SPEC numbers (higher than what he reported the G5 as having) were found for the same P4 system on SPEC's website (not to mention even higher numbers for current AMD systems at the time).

The claims were "debunked" by idiots. There were two common criticisms of the benchmark:

1) That the P4 had hyperthreading disabled;
2) That the P4's results were lower than what was available on spec.org

The first point is ridiculous because the whole point of hyperthreading is to increase multithreaded performance by dividing up processor resources. That means single-threaded benchmarks (like SPEC) almost always show a performance loss when using hyperthreading.

The second point is ridiculous because Apple's benchmarks were taken with GCC, while the Intel benchmarks on spec.org are usually taken with Intel C++. That's why the SPEC results for the P4 were lower. By using GCC on both platforms, the compiler was taken out of the equation, leaving only the CPUs to be compared. Moreover, using Intel C++ on the P4 (or XLC on the G5) to demonstrate performance is stupid. Nobody uses these "super compilers" on real code. 99.9% of the Windows software out there is compiled with either GCC or Visual C++, which are (or at least were back when these benchmarks were taken) comparable in performance. Testing GCC on the G5 and ICC on the P4 would've been unfair, since it would be comparing the average case on the G5 with the absolute best case on the P4.

The SPEC results Apple posted originally for the G5 are actually quite realistic. Based on the SPEC scores of the Opteron using a similarly old version of GCC, they peg the 2.0 GHz G5 at roughly 75% the integer performance of a similarly clocked Opteron, which is quite consistent with my own experience. If anything, the FP score is actually rather low, probably due to the fact that GCC wasn't very well-optimized for the G5 at that point.

Reply Score: 2

RE[4]: well duh
by n4cer on Thu 19th Jan 2006 21:24 UTC in reply to "RE[3]: well duh"
n4cer Member since:
2005-07-06

How do any of those points discount Apple using their SPEC results as the basis for promoting the G5 as the fastest PC? This still couldn't be justified with their SPEC numbers and wasn't true when the claim was made. You had single CPU systems with higher SPEC numbers than Apple's dual CPU systems even when using compilers like Visual C++.

Reply Score: 2

RE[2]: well duh
by fmaxwell on Thu 19th Jan 2006 20:52 UTC in reply to "RE: well duh"
fmaxwell Member since:
2005-11-13

Was that intended as sarcasm or a joke? Apple has consistently been completely dishonest about benchmarking. They touted the G5 as being the fastest personal computer on the market -- even though no standardized benchmarking software supported that claim.

Reply Score: 2

RE[3]: well duh
by tomcat on Thu 19th Jan 2006 20:56 UTC in reply to "RE[2]: well duh"
tomcat Member since:
2006-01-06

Apple has consistently been completely dishonest about benchmarking. They touted the G5 as being the fastest personal computer on the market -- even though no standardized benchmarking software supported that claim.

Look on the bright side: Once Apple retires PPC for good, then these "our processor is fastest" debates will disappear for good. Users will be able to compare for themselves by running a copy of OS X on a legacy Intel box and then comparing it to the newly upgraded hardware. 'Course, that won't stop Apple from claiming that updates to OS X yield unbelievable perf gains (which the fanatics will buy -- and skeptics will roll their eyes over).

Reply Score: 1

Steve Jobs
by vondur on Thu 19th Jan 2006 15:30 UTC
vondur
Member since:
2005-07-07

Only said in the keynote that the processor was 2x faster in terms of SPEC benchmarks, not actual usage. He even mentioned that the machines would not overall be 2x faster.

Reply Score: 5

RE: Steve Jobs
by tbostick78 on Thu 19th Jan 2006 17:13 UTC in reply to "Steve Jobs"
tbostick78 Member since:
2005-12-16

Did you notice though that the advertised SPEC results are "estimates"? How can you have an estimated benchmark? That's absurd.

Reply Score: 1

RE[2]: Steve Jobs
by rayiner on Thu 19th Jan 2006 20:42 UTC in reply to "RE: Steve Jobs"
rayiner Member since:
2005-07-06

It's standard nomenclature. You can only cite official SPEC runs as "SPEC results". All non-official runs are called "estimates". Pre-production machines and processors always come with SPEC estimates. It just means that Apple hasn't yet submitted, and SPEC hasn't yet accepted, an official run from Apple.

Reply Score: 2

RE: Steve Jobs
by somebody on Thu 19th Jan 2006 17:46 UTC in reply to "Steve Jobs"
somebody Member since:
2005-07-07

Only said in the keynote that the processor was 2x faster in terms of SPEC benchmarks, not actual usage. He even mentioned that the machines would not overall be 2x faster.

Now be a good boy and tell that to their web pages
http://www.apple.com/
http://www.apple.com/imac/
http://www.apple.com/macbookpro/

Reply Score: 1

RE: Steve Jobs
by taos on Thu 19th Jan 2006 19:09 UTC in reply to "Steve Jobs"
taos Member since:
2005-11-16

[vondur] Steve Jobs Only said in the keynote that the processor was 2x faster in terms of SPEC benchmarks ...

Even that is misleading.

He deliberately used _only_ SPECint_rate2000, not SPECint2000.

See Q15. What is the difference between a "rate" and a "speed" metric?
http://www.spec.org/cpu2000/docs/readme1st.html#Q15

Since the Core Duo is dual-core processor, of course the "throughput" will almost double if the number of tasks are doubled.

By this logic, we shall put Sun's UltraSPARC T1 in a desktop/laptop machine and claim it's 5x - 10x faster?

No surprise in Macworld Lab's benchmark result, since none of them is measuring "througput", which is certainly NOT a key issue on desktops.

It's all about marketing.

Reply Score: 2

RE[2]: Steve Jobs
by tomcat on Thu 19th Jan 2006 20:53 UTC in reply to "RE: Steve Jobs"
tomcat Member since:
2006-01-06

Even that is misleading.


What do you expect? Jobs has been spewing misleading marketing BS for years about performance...

Reply Score: 2

speed
by Thom_Holwerda on Thu 19th Jan 2006 15:35 UTC
Thom_Holwerda
Member since:
2005-06-29

I just came back from a press conference at Apple's HQ office here in the Netherlands, and while there I played with the new MacBook as well as the Intel iMac. They felt faster (esp. iLife '06 and Front Row) than the current iMac G5 I recently reviewed for OSNews, but noted must be that the machines the guys from Apple demo'd today all had 2 GB of RAM, while my own experiences with OSX limits to 512MB in my own iBook, and 1 GB in the iMac G5 I reviewed.

So, it's hard to say.

Reply Score: 5

RE: speed
by kamper on Thu 19th Jan 2006 16:29 UTC in reply to "speed"
kamper Member since:
2005-08-20

, but noted must be that the machines the guys from Apple demo'd today all had 2 GB of RAM, while my own experiences with OSX limits to 512MB in my own iBook, and 1 GB in the iMac G5 I reviewed.

So, it's hard to say.

It shouldn't be too hard to say. How much ram do you use on a day-to-day basis? How much was being used on the demo machines. If neither goes above 1g, then it's pretty obvious the amount of ram didn't have much of an impact (at least from your iMac).

Reply Score: 1

RE[2]: speed
by Thom_Holwerda on Thu 19th Jan 2006 17:02 UTC in reply to "RE: speed"
Thom_Holwerda Member since:
2005-06-29

It shouldn't be too hard to say. How much ram do you use on a day-to-day basis? How much was being used on the demo machines. If neither goes above 1g, then it's pretty obvious the amount of ram didn't have much of an impact (at least from your iMac).

OSX is a resource hog. This is no troll, it's commonly known among non-zealous Apple fans like myself. This means that you will CERTAINLY notice a difference in speed between two machines that are all equal except for memory (one having 1GB, the other having 2GB).

I did notice a speed improvement when using iLife (esp. iPhoto) and Front Row on both the Intel iMac (both models) and the MacBook Pro (which was a pre-production model, by the way, and hence I was not allowed to take any pictures, and the guy from Apple Europe didn't like it when I closed the lid to test sleep/wake; hence I think the MacBook Pro might not be as stable yet when it comes to sleep/wake), compared to the iMac G5 I used for that review (the latest 1.9 Ghz G5 model, specced with 1GB of RAM). However, how much of that speed improvement can be accounted for by the increase in RAM (twice as much in the demo models today), and how much can be accounted for by the processor & co?

In other words, if those Intel iMacs were to have 1GB (not unlikely; Apple sells the things with 512MB) would they still be noticably faster (as in, delivering a faster user experience) than the iMac G5 1.9Ghz?

THAT is hard to say.

Edited 2006-01-19 17:05

Reply Score: 5

RE: speed
by tbostick78 on Thu 19th Jan 2006 17:15 UTC in reply to "speed"
tbostick78 Member since:
2005-12-16

My truck feels faster after I wash it.

Reply Score: 2

Benchmark software using Rosetta
by Sabon on Thu 19th Jan 2006 10:55 UTC
Sabon
Member since:
2005-07-06

Come on. Get real people.

1) Steve Jobs said the CPUs, not the whole computer was 2 to 3 times faster.

2) The benchmark program (I forget its name) runs in Rosetta which is NOT an intel native app. So the benchmarks only give us the performance of a program running in Rosetta. The fact that it is as a PowerPC in that mode is pretty amazing to me. Especially since it isn't dual core aware.

3) Read #1 then read #2 again. Geez!!!!!

Reply Score: 3

Aaron1 Member since:
2006-01-19

Come on. Get real people.

1) Steve Jobs said the CPUs, not the whole computer was 2 to 3 times faster.


Yes, but look at www.apple.com. "4x faster." and "2x faster". No qualifications of CPU only. Apple should get real.

Reply Score: 2

kamper Member since:
2005-08-20

Yes, but look at www.apple.com. "4x faster." and "2x faster". No qualifications of CPU only. Apple should get real.
Um, have you never been exposed to marketing before? You don't have to like it, but did you really expect them to say: "2x faster (in theory, but tests indicate it'll only feel like 1.142x faster)"?

Reply Score: 1

BigJimSlade Member since:
2006-01-19

Wrong.

Check it out... www.apple.com/imac. Notice the 'footnotes' and the reference to SPEC, everywhere?

I watched the keynote, too. He did say that it's faster using the SPEC benchmark, only. Then he went on to say that the whole machine will be faster, but that it's not going to be 2x-3x as a whole. He even loaded Photoshop and said that it's nowhere near native speed, but it may be 'good enough' for most people.

From www.apple.com/imac
=======================

Twin Power
----------

The Intel Core Duo and a whole new architecture give iMac up to twice the horsepower it had previously, accelerating your digital life. 1

iMac Performance Benchmark
--------------------------
SPEC performance: 2 to 3 times faster than 20-inch iMac G5 with 2.1GHz PowerPC. 1
2-3X Faster iMac G5
2.1GHz iMac
Core Duo
2.0GHz Δ
SPECint_rate_base2000
Integer calculation
(estimate)
10.2 32.6 3.2X
SPECfp_rate_base2000
Floating-point
calculation (estimate)
13.0 27.1 2.1X

Footnote
--------
1 # Testing conducted by Apple in December 2005 using preproduction 20-inch iMac units with 2GHz Intel Core Duo; all other systems were shipping units. All scores are estimated. SPEC is a registered trademark of the Standard Performance Evaluation Corporation (SPEC); see www.spec.org for more information. Benchmarks were compiled using the IBM compiler and a beta version of the Intel compiler for Mac OS.

Reply Score: 2

TomB7 Member since:
2006-01-03

" Benchmarks were compiled using the IBM compiler and a beta version of the Intel compiler for Mac OS."


HMMM--so the MacWorld/ ArsTechnica, etc numbers may be based on an older, possibly slower compiler?

Reply Score: 1

rayiner Member since:
2005-07-06

The MacWorld/ArsTechnica results are based on GCC, which is the compiler 99.99% of OS X software uses.

Reply Score: 1

stew Member since:
2005-07-06

You'd be surprised how much PPC software is compiled in CodeWarrior and not gcc.

Reply Score: 2

RE: Benchmark software using Rosetta
by -ujb- on Thu 19th Jan 2006 16:17 UTC in reply to "Benchmark software using Rosetta"
-ujb- Member since:
2005-10-21

The bench mentioned in the article linked above says Rosette gives half the perfromance of a current G5. x86 native apps run about 20% faster on the CoreDuo than ppc native do on current G5.

That's all not bad results, but far away from Steves exaggerated fanatsies at the keynote.

Reply Score: 1

kaiwai Member since:
2005-07-06

In respects to the second statement, the CPU spec and fpu benchmarks were using the best compilers available and compiled native for the chip.

With that being said, it is doubtful that any application would reach the saem performance level as shown by the benchmarks as I hightly doubt that an application vendor would spend the amount of time required tweaking large amounts of their code simply to squeeze every possible bit of performance.

You are right, I saw the clip myself and he made a big effort to emphasise that it was only benchmarks for the chip - he went on to say that the speed of the hard disk, the memory and so forth dictate the over all speed of the machine; so it isn't as though he made the correction as a small statement.

Reply Score: 1

Not too bad for 1st release
by Pasha on Thu 19th Jan 2006 16:14 UTC
Pasha
Member since:
2005-07-06

I am not a Mac fanatic. I do not even own one. Given it's the 1st try, well, it's very good.
Things can only get better.

Reply Score: 1

Steve said
by SlackerJack on Thu 19th Jan 2006 16:20 UTC
SlackerJack
Member since:
2005-11-12

The Intel duelcore CPU's were 4X faster than the G4 and 5X faster then G5 per watt. SPEC benchmarks claimed this as well which make sense, since the duel Intel's will show that in benchmarking applications.

As regards to application performance, there is no way that will photoshop(for example) will be 4 to 5 times faster than a G4/G5. We know that a Intel P4 is faster at rendering filters in PS faster then a G4/G5, so I say say PS optimised for duelcore will be at least X2/X3 faster. Thats my theroy anyway.

Reply Score: 1

Weird
by rayiner on Thu 19th Jan 2006 16:23 UTC
rayiner
Member since:
2005-07-06

I have to defend Apple's benchmarking for once. "Up to 2-3x faster" refers to the CPU, not the system. Everyone and their mother knows that a 2x faster CPU won't increase many tasks, like those performed by MacWorld by near 2x.

On the other hand, here's an interesting statistic: http://www.gusmueller.com/blog/archives/2006/1/17.html#1415

This guy's 1.83 GHz iMac compiles his XCode project over 15% faster than his dual 2.7GHz G5.

See also this post on xcode-users:
http://lists.apple.com/archives/Xcode-users/2006/Jan/msg00680.html

> How is the compile speed on the new machines?

On an iMac Core Duo, a tick or two slower than a Quad G5.


(References stolen from a post I saw on Macrumors.com)

So compiling, at least, is 2-3x faster on the new IntelMac than on a comparably-clocked single G5. Admittedly, compiling is something of a worst-case scenario for the G5 (no floating-point, heavy integer computation, many branches, random access to memory), but for people who write code all day, this is good news ;)

Reply Score: 3

RE: Weird
by BlackJack75 on Thu 19th Jan 2006 21:49 UTC in reply to "Weird"
BlackJack75 Member since:
2005-08-29

I suspect GCC has been much for optimized for X86 over the years than the PowerPC version. Anyone want wishes to counter this claim?

Reply Score: 1

RE[2]: Weird
by rayiner on Thu 19th Jan 2006 22:23 UTC in reply to "RE: Weird"
rayiner Member since:
2005-07-06

The GCC code generator has been much more optimized for x86, in that it generates relatively better code, but the GCC code itself is rarely optimized for a specific processor. Indeed, its only recently that the 4.x series has been optimized for performance at all. GCC is mostly just good old general-purpose, generic, portable C code.

Reply Score: 2

Phillip.Fayers
Member since:
2005-12-14

The Intel CPU has more processing power than the old CPU in an iMac, but it uses two processor cores to get it.

Were all the benchmarks multi-threaded applications capable of making good use of two CPUs?

A lot of those benchmarks would appear to be disk limited anyway so you aren't going to see performance improvements due to the CPU.

Reply Score: 1

Damned lies and...
by gypsumfantastic on Thu 19th Jan 2006 16:50 UTC
gypsumfantastic
Member since:
2005-07-06

You know, given that the iMac G5 is identical to the Core Duo, except for the motherboard and processor, comparing a number of these things, especially IO-bound tests, and using them as some kind of "average measure" of the processor performance is ludicrous.

You know, when I am ruler of Earth, I shall outlaw all benchmarks and make them punishable by being hung, draw and quaterly.

You know what? IT DOESN'T MATTER.

Unless we're talking MacBooks of course, in which case, the memory-bandwidth-starved G4 is gonna get a cork popped in its uterus.

Reply Score: 1

RE: Damned lies and...
by skingers6894 on Fri 20th Jan 2006 04:11 UTC in reply to "Damned lies and..."
skingers6894 Member since:
2005-08-10

"the memory-bandwidth-starved G4 is gonna get a cork popped in its uterus"

Hmmm - this has my vote for bad taste analogy of the year...

Reply Score: 1

resonate
Member since:
2006-01-19

this site http://handbrake.m0k.org/ has a screenshot of imac core duo encoding full frame size h264 at about 30fps........not bad i say

anyone can test the handbrake perfomance on their G5 dual or maybe G5 Imac?

Reply Score: 1

Still...
by 1c3d0g on Thu 19th Jan 2006 18:49 UTC
1c3d0g
Member since:
2005-07-06

...it's a pretty good boost in performance for what is essentially just a switch in CPU Architectures - PowerPC to x86. However, if that performance gain is due to a better Arch, a better implementation of the underlying software or something else entirely is up for debate. I'll tell y'all one thing though, this is one of the smartest moves Apple executed in a long time.

Now if only the gaming consoles stayed with x86...*sigh*... ;)

Edited 2006-01-19 18:52

Reply Score: 1

RE: Still...
by TomB7 on Thu 19th Jan 2006 19:03 UTC in reply to "Still..."
TomB7 Member since:
2006-01-03

gaming consoles haven't BEEN X86 in a while.

Reply Score: 2

I just don't get the hoo haa
by shepherdr on Thu 19th Jan 2006 19:15 UTC
shepherdr
Member since:
2006-01-19

It's a Mac. Its a bit faster than the last one. Thats how Macs have rolled out for years. That it uses a different CPU of similar clock frequency - who cares? Why do people want to buy a Mac? And is this available with the intel Mac?

1. They look cool - exactly the same
2. They run MacOS X - exactly the same
3. They run iLife '06 - exactly the same
4. They run Office - usable now under Rosetta faster when Universal and Microsoft "on schedule"
5. They run Apple's Final Cut Studio, Aperture & Logic - available in March
6. They run Adobe Creative Suite - Universal binary available ? when

So the REAL reasons why one would choose a Mac haven't changed - they are all still there except temporarily Adobe. All the other apps aren't why you would choose a Mac over a PC anyway. Nor was the speed. My new iMac experience will be the same and that's why my next computer will be a Mac and will join my other 4 Macs.

Reply Score: 2

Mac's getting dumbed down
by JustAnotherMacUser on Thu 19th Jan 2006 20:40 UTC
JustAnotherMacUser
Member since:
2006-01-08

The new Mactel's are dumbed down to Intels roadmap.

The G5 roadmap has been wonderful in terms of price for performance.

Right now a Quad core G5 will go for about $3500

A same Dell Xeon Quad will got for $1200 more.

Also G5's still have Altivec, not the Core Duos.

I'm not liking this Intel switch all that much, I hope Apple keeps selling G5 processors for quite some time.

I especially don't like the new hardware DRM that comes with these new Mactels either. Nobody wants to talk about that little nasty b*tch.

http://www.againsttcpa.com/tcpa-faq-en.html

Reply Score: 2

g5
by sp29 on Thu 19th Jan 2006 21:46 UTC
sp29
Member since:
2006-01-04

Proves the G5 chip is still a good chip, but not as fast as the new Intel.

I wonder if the New Intel Duo would stand up to a G5 Dual. G5 dual would probably whip it, being the Intel chip is made specifically for laptops right?

Reply Score: 1

RE: g5
by rayiner on Thu 19th Jan 2006 22:23 UTC in reply to "g5"
rayiner Member since:
2005-07-06

It depends on what you do. For integer code, the 2.0 GHz Core Duo will give anything short of the Quad a run for its money.

Reply Score: 1

MAC ON SPEED
by lz1kwk on Thu 19th Jan 2006 23:17 UTC
lz1kwk
Member since:
2005-11-12

I am really tired of these speed tests which tell you nothing about real life use.

I use a Windows Pentium4 2.4 GigHz for work. When I first got it, boot up was 5 seconds and shutdown was 3 seconds. After one year of use and even after many defrags It takes me about 20 minutes to boot up and it will not shutdown unless I press the power button. It runs as slow as a snail even with corporate anti-spyware and anti-virus.

Contrast this with every *nix machine I have used. Even if they are slow to boot up, performance remains the same with age and even if you do not reboot for 1`year. I have not used a Mac but I have no doubt that it's Unix soul will remain true to it's reputation.

My take on all this: Stop all these phony tests and "Cert" comparisons that tell you nothing about real life user experience

Reply Score: 1

RE: MAC ON SPEED
by smitty on Thu 19th Jan 2006 23:30 UTC in reply to "MAC ON SPEED"
smitty Member since:
2005-10-13

I use a Windows Pentium4 2.4 GigHz for work. When I first got it, boot up was 5 seconds and shutdown was 3 seconds.

It takes my BIOS about 5 seconds just to count my RAM, so I highly doubt your statement.

Reply Score: 1

RE[2]: MAC ON SPEED
by n4cer on Fri 20th Jan 2006 00:59 UTC in reply to "RE: MAC ON SPEED"
n4cer Member since:
2005-07-06

It takes my BIOS about 5 seconds just to count my RAM, so I highly doubt your statement.

Not all BIOSes are equal (and not all follow MS' suggested guidelines to speed up boot time). Depending on the OEM that made his P4, his statement could be true.

Reply Score: 1

G5 isn't bad.
by MikeekiM on Fri 20th Jan 2006 04:14 UTC
MikeekiM
Member since:
2005-11-16

First, most of us don't have a G5 2.0 ghz.
I was in the apple store before the release of the new machines and the IMac G5 was noticable faster then my G4 1.5 ghz laptop.

Had Apple been able to get a G5 2.0 into a laptop they wouldn't have NEEDED to go Intel Duo.

Having said that, the Duo is faster 10% - 50%,
faster is faster.

Reply Score: 1

RE: G5 isn't bad.
by rayiner on Fri 20th Jan 2006 04:22 UTC in reply to "G5 isn't bad."
rayiner Member since:
2005-07-06

Had Apple been able to get a G5 2.0 into a laptop they wouldn't have NEEDED to go Intel Duo.

I see this a lot, but its not really true. Apple went to x86 at least in part because there was nothing in the PowerPC pipeline that made sense for them. Freescale is concentrating on embedded G4-architecture chips, which are outdated with regards to the desktop, and IBM's POWER6 and Cell lines are all high-clockspeed low IPC designs in the spirit of the Pentium 4. If they'd stayed on PowerPC, not only would they not have laptop G5s, but they wouldn't have any realistic processors to replace the G5 when the time came.

Reply Score: 1

RE[2]: G5 isn't bad.
by kaiwai on Fri 20th Jan 2006 05:19 UTC in reply to "RE: G5 isn't bad."
kaiwai Member since:
2005-07-06

On the discussion of the PowerPC G4 processor - I've always wondered what would occur, in terms of performance if the FSB was pushed up to 667Mhz.

Reply Score: 1

RE[3]: G5 isn't bad.
by rayiner on Fri 20th Jan 2006 06:42 UTC in reply to "RE[2]: G5 isn't bad."
rayiner Member since:
2005-07-06

The thing to remember about the G4 is that the core processor design is now seven years old. Its branch predictor is ancient, it has only two integer units, a single FPU unit, it only dispatches two operations per cycle, it has a very limited number of instructions in flight, etc. From a theoretical consideration, one could say that a G4 with a faster bus wouldn't perform very well at all relative to modern CPUs.

Edited 2006-01-20 06:43

Reply Score: 1

RE[4]: G5 isn't bad.
by kaiwai on Fri 20th Jan 2006 09:05 UTC in reply to "RE[3]: G5 isn't bad."
kaiwai Member since:
2005-07-06

The thing to remember about the G4 is that the core processor design is now seven years old. Its branch predictor is ancient, it has only two integer units, a single FPU unit, it only dispatches two operations per cycle, it has a very limited number of instructions in flight, etc. From a theoretical consideration, one could say that a G4 with a faster bus wouldn't perform very well at all relative to modern CPUs.

The reason why I ask is Niagra esssentially 8 UltraSPARC II cores - very basic in design when compraed to modern day standards.

What I am looking at is, something similar to that, well, maybe not as radical as 8 cores, but maybe a 4 G4 Cores thrown onto a 667Mhz bus.

Reply Score: 1

RE[5]: G5 isn't bad.
by rayiner on Fri 20th Jan 2006 21:26 UTC in reply to "RE[4]: G5 isn't bad."
rayiner Member since:
2005-07-06

Niagra is also going to be a terrible processor for traditional desktop/workstations. The level of parallelism available in most desktop/workstation software is barely adequate for a 4-core machine, much less an 8-core one. I know many people see massively multicore processors as the wave of the future, but I have to disagree with that.

Writing software for a parallel system is very hard, and it is the writing of the software is the most expensive and time consuming part of the whole process. Nobody is going to expend the extra effort unless its absolutely necessary. Heck, just look on the Mac platform --- its been dual-processor for years, yet most common programs still don't take advantage of multiple processors.

What this means is that the minute the manufacturing issues that are blocking current clockspeed increases are overcome (and Intel is making good strides in this area with its 65nm and 45nm processes), the focus on multiple cores will dissapear on desktop/workstation processors. The number of cores on such CPUs will thus stabilize at a number that are relatively easy to take advantage of, say two or four. Machines with such relatively small amounts of parallelism are easy to take advantage of, even just through regular multitasking.

Thus, a quad-core G4 would be a highly specialized chip indeed. Its inherent simplicity would make its single-threaded performance quite weak, and even with perfectly scalable software, it likely wouldn't be faster than a dual-core Opteron or the like, at least not at a realistic clockspeed. It would make sense perhaps in a blade server setting, where the low power usage would offset the low performance, but not in any market Apple is a player in.

Reply Score: 2

RE[6]: G5 isn't bad.
by kaiwai on Sat 21st Jan 2006 03:01 UTC in reply to "RE[5]: G5 isn't bad."
kaiwai Member since:
2005-07-06

Niagra is also going to be a terrible processor for traditional desktop/workstations. The level of parallelism available in most desktop/workstation software is barely adequate for a 4-core machine, much less an 8-core one. I know many people see massively multicore processors as the wave of the future, but I have to disagree with that.

But the assumption is based on the same optimistic assumption that technology will catch up to solve the problem, or as in the case of parallelism and threading, a technology which allows the creation of massive scaling applications alot easier.

Personally, I think its going to be alot more complicated than just more cores, or more speed; what is required is for the CPU to do more work per clock cycle - make it as efficient as possible - in terms of cores, make it easier to take advantage of those extra cores by lowering the bar required to parallelising code.

With that being said, will be come a time where a complete radical change in direction will be required - not a completely new ISA or architecture, but a radically new way of approaching computers and solving problems.

Reply Score: 1

RE[7]: G5 isn't bad.
by rayiner on Sat 21st Jan 2006 07:24 UTC in reply to "RE[6]: G5 isn't bad."
rayiner Member since:
2005-07-06

I think you may have misinterpreted my statement. My point was that massively multicore CPUs like Niagra are great for their target market, but have very limited potential as CPUs for running general purpose code.

As you point out, the current approaches to solving problems do not lend themselves well to parallelism. The basic problem is a lack of a sufficiently mature body of theory that can be used as the basis for writing parallel code. Moreover, almost all current programming languages are based on two formalisms (the Turing Machine and the Lambda Calculus), neither of which treat parallelism explicitly. When you have no theoretical understanding of parallel programs, and when you cannot formally express parallel algorithms in your language, all you have left are ad-hoc means (like threads), which makes things very difficult indeed.

Reply Score: 1

Buy plenty of ram
by Snooks on Fri 20th Jan 2006 04:49 UTC
Snooks
Member since:
2006-01-10

Mac OS X likes ram. Not a thing wrong with that since it uses extra ram very efficiently unlike XP which leaks memory like a sieve. The intel iMac likes ram even more than usual though, probably because of the caching used by Rosseta. The iMac really comes into it's own with 1 GB of ram or higher. Mine has 1.5GB of ram and universal apps fly while emulated apps are noticeably faster. With 512MB I only had 20 to 30 MB free which really hurts performance. Apple should incude a minimum of 1 GB of ram with intel Macs and allow for more than 2 GB max. Also OS X in general feels more fluid with dual procs and even though dual core isn't quite the same as dual procs it's pretty darn close.

Reply Score: 1

They are faster
by Snooks on Fri 20th Jan 2006 04:57 UTC
Snooks
Member since:
2006-01-10

"In other words, if those Intel iMacs were to have 1GB (not unlikely; Apple sells the things with 512MB) would they still be noticably faster (as in, delivering a faster user experience) than the iMac G5 1.9Ghz?
THAT is hard to say."

No, its very easy to say. They are faster in universal apps, in some cases much faster. There are benchmarks all over the web proving it as well as the experiences of users like myself who replaced a 2.1 ghz G5 iSight iMac with the intel iMac.

Reply Score: 1

HT
by Snooks on Fri 20th Jan 2006 05:01 UTC
Snooks
Member since:
2006-01-10

"1) That the P4 had hyperthreading disabled;
The first point is ridiculous because the whole point of hyperthreading is to increase multithreaded performance by dividing up processor resources. "

Hyper-threading does NOT make a PC faster, it can slow things down by alot so Apple did them a favor by turning it off. Thats why Dell has the option to turn hyperthreading off on new workstations before they are shipped.

Reply Score: 1

Can always switch to...
by nii_ on Fri 20th Jan 2006 07:01 UTC
nii_
Member since:
2005-07-11

"Intel-Based iMacs Are Fast, But Gains Don't Match Apple's Claims"

How about a switch to the IBM/Sony/Toshiba Cell Processor? ;) *wink wink*

Or, Apple can always switch to Sun's Niagara (May be a a little expensive though?).

Reply Score: 0

RE: Can always switch to...
by kaiwai on Fri 20th Jan 2006 09:06 UTC in reply to "Can always switch to..."
kaiwai Member since:
2005-07-06

Or stupid, try wedging a 70watt chip into laptop.

Reply Score: 0

RE: Can always switch to...
by rayiner on Fri 20th Jan 2006 21:27 UTC in reply to "Can always switch to..."
rayiner Member since:
2005-07-06

What good would Cell do for Apple? The damn thing is the equivalent of like a 1GHz PIII on integer code. Who wants that in a modern Mac?

Reply Score: 2

2X faster AGAIN...
by bigbadguy on Fri 20th Jan 2006 20:44 UTC
bigbadguy
Member since:
2005-08-31

Two years before, Apple already delivery super-computer performance with G5. Today, they double it up!

Thanks Jobs on re-defining the word: Super-Computer

Reply Score: 1