I use both wintel and Apple Laptops and let me tell you one thing, my new PB 1.5 ghz is the fastest laptop i have ever had and use a pentium m at 1.6 ghz and it is not even close in performance or speed.
You may have some system problem then. The Pentium-M 1.6 GHz should be noticeably faster than the new 1.5 GHz G4. Could you do me a favor and go to this thread and run the CPU benchmark described therein? I would like to see the results for both Pentium-M 1.6 GHz and G4 1.5 GHz.
You may have some system problem then. The Pentium-M 1.6 GHz should be noticeably faster than the new 1.5 GHz G4. Could you do me a favor and go to this thread and run the CPU benchmark described therein? I would like to see the results for both Pentium-M 1.6 GHz and G4 1.5 GHz.
You are aware that this test is a very narrowly defined fp task and not in any way indicative of overall cpu performance?
Actually the Pentium m 1.6 ghz is not superiori technologically or physically because the g4 used in the 1.5 ghz powerbook is a totally redesigned processor like the pentium m but is using a dual core design setup (altivec on top) rather than a single core pentium m, that should basically answer your question because the pentium m really is not more powerful and it has been known for a while that macs run at faster speeds when clocked at the same mhz as their pentium counterparts.
Matt
It really would not be an accurate benchmark and the only real way to test the machines is to use digital photography utilies and editing utilities such as final cut and photoshop but final cut is a mac only solution
Matt
Quote:
Originally posted by PB
You may have some system problem then. The Pentium-M 1.6 GHz should be noticeably faster than the new 1.5 GHz G4. Could you do me a favor and go to this thread and run the CPU benchmark described therein? I would like to see the results for both Pentium-M 1.6 GHz and G4 1.5 GHz.
Actually the Pentium m 1.6 ghz is not superiori technologically or physically because the g4 used in the 1.5 ghz powerbook is a totally redesigned processor like the pentium m but is using a dual core design setup (altivec on
No it's not, by any reasonable definition it's single core.
Also Altivec is a SIMD FPU, just like SSE2 as found
on the Pentium M, it's not some secret Apple voodoo.
FWIW the 1.6 GHz Pentium M has a better SPECint 2000
base score (1206) than the 2.0 GHz PPC 970. You may
not believe it but the Pentium M, and Dothan in particular,
is an excellent CPU. An on-die 2MB L2 cache and a 400 MHz
FSB, compared to the anemic G4 should make that blindingly
obvious to all but the most partisan. Intel's not planning
on using Pentium M follow-ons for their future desktop
CPUs just for the hell of it.
Quote:
top) rather than a single core pentium m, that should basically answer your question because the pentium m really is not more powerful and it has been known for a while that macs run at faster speeds when clocked at the same mhz as their pentium counterparts.
Matt
Absolute rubbish, Coppermine PIII had an excellent IPC.
In fact the whole range of Intel processors with the P6
microarchitecture delivered good to great performance
per cycle.
Posting the kind of things you did gives the rest of us
Mac owners a bad name. Macs are great but can the BS.
Spot on Matsu! I wish bare feats (or another web site) might do more of this type of testing.
I would like to see how a PB 1.5 stacks up against a Centrino 1.6/1.7 on like you said a set of tasks on various types of apps, with price the common denominator.
Not just a few cpu and gaming fps comparisons?. At least then well know for sure if we need to be more spec whores or not.
PS. If the PB doesnt do well against the 1.6 Pentium M, then its not gong to look good for Apple when we start seeing the Dothan 2G models start shipping soon.
Actually, the Centrino or Pentium M is a single core with SSM, but the G4 processor was designed and constructed on a two core process and broken down into two aspects the g4 Microprocessor and Altivec, Altivec is not code program like MMX or SSE2, it is actually a piece of hardware that runs on top of the g4 processor and is not some sort of mysterious apple voodoo, and actually specint 2000 are very inaccurate in benchmarking machines, especially because the powerpc 970 is a more powerful processor than the centrino and it is funny that IBM power4 chips have consistently beat centrino chips in performance, i wont speak about dothan because not enough performance reviews have been posted anywhere. I will agree that the bus on the g4 1.5 ghz leaves something to be desired, but lets not forget that the powerpc 970 which is based on the ibm power4 core which definetely beats the pentium m processor(which by the way is a mobile processor, not even capable of achieving the same bandwidth/performance and Specint despite the fact that it is a bad benchmark and that no true benchmark exists between pc's and mac's, measures laptop processors different than Desktop processors, taking into account that regardless of supposed performance, desktop processors are more powerful especially when being compared to lower powered, lower speed pentium m chips, now it may feel good for you to say that wow the pentium m chip is something special therefore it is better than apple processors, but i will say that 1) Despite the facts hindering bus speed, the G4 processor that runs at 1.5 ghz is equal to or superior to a Pentium M processor(dothan) at 1.6 ghz and at comparable speeds as well(1.5 ghz), the bus gives bandwidth superiority to the pentium, but the performance of the g4 microprocessor itself gives favor to the g4 itself. 2) Never analyze or bite more than you can support, in no registered or valid cases has any mobile processor, intel, ibm or motorola beaten a desktop processor at equal or greater speeds in benchmarks or whatnot because the comparable technology between a mobile and a desktop is completely different. 3) Never compare 32 bit and 64 bit, they are completely different, a 64 bit processor chews data in 64 bit chunks while a 32 does 32, bit by bit speed vs speed, a 64 bit processor can handle more computations per second than a 32 bit processor when running native 64 bit software. 4) THe Powerpc 970 is a superior processor to most or all Intel 32 bit processors and is based on a Scaled Down Dual Core IBM Power 4 processor which is a Enterprise level processor that sells for 2000-3000 for one processor alone with no computer, it is scaled down, but not greatly 5) despite all Intel's industry clout, IBM has more advanced fabrication technology and the factory in NY used to make the Power4 and Power5 Cores and thus the Powerpc 970 is the newest and most advanced fabrication plant used by the big three (amd, intel and ibm) and has chartered new territories with 65 nm, 90 nm and 130nm processes that churn out a processor that is superior to that of the larger process, 190, 230 nm still at use by intel. 6) Benchmarks are inaccurate, in no case has a accurate benchmark for performance been found to accurately correlate performance of a mac versus that of a pc and therefore any actual benchmarks will show no actual performance examples
7) Think before you write: I have been involved in the research and study field of processor technology for the past seven years and have done several projects(non major) that have constructed and tested the core performance of major processors, mainly amd vs intel, but have recently begin studying ibm processes and while the technological leaps have been great, i am a realist and i believe that accuracy must be maintained in presentation and the fact is that your statement saying that the Pentium M shows significant performance scores against G4 Mobile and especially the Powerpc 970 is absolutely invalid and based not on fact, but on opinion and represents the opinion and data supporting the opinion of you, and not fact, benchmarks are not an accurate survey and people know the performance of their machines, i have machines in both worlds i have several PC machines running from the original AMD athlon(love it to death), to the Pentium 4 and Pentium M at 1.6 ghz by the way, i also have a DP G5 at 2ghz and an Apple PB at 1.5 ghz and on these machines i have done extensive testing and in all cases i have seen better performance by my Apple machines, not only do they not crash, they take zero time to load up, zero install time and zero processing time, i get done what i want when i want, but when using my pc's i face processing times, crashes and stability issues with major programs that i find to be essential especially photoshop, spss and avid which the same or competiting versions work better on mac
Quote:
Originally posted by catbat
No it's not, by any reasonable definition it's single core.
Also Altivec is a SIMD FPU, just like SSE2 as found
on the Pentium M, it's not some secret Apple voodoo.
FWIW the 1.6 GHz Pentium M has a better SPECint 2000
base score (1206) than the 2.0 GHz PPC 970. You may
not believe it but the Pentium M, and Dothan in particular,
is an excellent CPU. An on-die 2MB L2 cache and a 400 MHz
FSB, compared to the anemic G4 should make that blindingly
obvious to all but the most partisan. Intel's not planning
on using Pentium M follow-ons for their future desktop
CPUs just for the hell of it.
Absolute rubbish, Coppermine PIII had an excellent IPC.
In fact the whole range of Intel processors with the P6
microarchitecture delivered good to great performance
per cycle.
Posting the kind of things you did gives the rest of us
Mac owners a bad name. Macs are great but can the BS.
in a few weeks i'll have the 1.33 PB (768MB RAM) to test against a thinkpad with a 1.6 pentium M (but less RAM). i might be able to do some testing there to see which is a better, but i'll need to find some common applications between the two and i won't have adobe photoshop, premier, etc or any games...
by late august i should be able to test the same PB 1.33 against a different thinkpad with equal ram and 1.6 Pentium M. there, i should have a bunch of adobe programs and possibly a game or two. if i'm still interested maybe i'll test them out
That might have been the most horribly formatted thing I've seen in quite a while here.
Quote:
Originally posted by mpopkin
i have seen better performance by my Apple machines, not only do they not crash, they take zero time to load up, zero install time and zero processing time
While the stability I have no dispute with, taking zero time to do anything seems a bit excessive, don't you think?
Also, while a 1.5 GHz G4 may be comparable to a 1.6 GHz Dothan, I'm not so sure that it would be that comparable to a 2.0 GHz Dothan.
Quote:
Originally posted by mpopkin
in no registered or valid cases has any mobile processor, intel, ibm or motorola beaten a desktop processor at equal or greater speeds in benchmarks or whatnot because the comparable technology between a mobile and a desktop is completely different
Are you kidding me? You're telling me that a 1.6 GHz Pentium M wouldn't beat a 1.8 GHz desktop P4?
No it's not, by any reasonable definition it's single core.
Also Altivec is a SIMD FPU, just like SSE2 as found
on the Pentium M, it's not some secret Apple voodoo.
FWIW the 1.6 GHz Pentium M has a better SPECint 2000
base score (1206) than the 2.0 GHz PPC 970. You may
not believe it but the Pentium M, and Dothan in particular,
is an excellent CPU. An on-die 2MB L2 cache and a 400 MHz
FSB, compared to the anemic G4 should make that blindingly
obvious to all but the most partisan. Intel's not planning
on using Pentium M follow-ons for their future desktop
CPUs just for the hell of it.
Absolute rubbish, Coppermine PIII had an excellent IPC.
In fact the whole range of Intel processors with the P6
microarchitecture delivered good to great performance
per cycle.
Posting the kind of things you did gives the rest of us
Mac owners a bad name. Macs are great but can the BS.
I think what some have been trying to say here,maybe not successfully,is that altivec is an independant unit whereas mmx/sse are wound up with the fpu pipelines.Same idea,much better implimentation on the ativec side.And by the way,these simd units handle integer as well as floating point code.
As for the p6 core having a high ipc,well it all depends what you compare it to.Certainly it has compared to the p4.
The rest of your post highlights just how much bs misinformation and downright paranoia there is in the mac community regarding x86 processors and their performance.
Let me make this as clear as possible-Spec is a highly suspect benchmark as it is but the scores intel reports useing their trick compiler are a joke.It's a public secret that intels compiler is tweaked to get unreal spec scores-why do you think AMD uses it to perform their own posted scores? They are not in any way related to real world performance.Let me quote from a cnet article out last year:
"Peter Glaskowsky, editor-in-chief of Microprocessor Report, said a company could get better benchmark results using a Dell machine with Intel and Microsoft compilers than with a Linux machine and GCC compiler. However, he also noted that Intel's chips perform disproportionately well on SPEC's tests because Intel has optimized its compiler for such tests. "
But lets play a game of reductio ad absurdum.Lets assume these scores are correct and actually mean something.Apple reported a score of 800 in specint.Intel claims a score of 1206.Thats over 50% faster for a chip that is only 80% the clock rate.Ok,at this point I'm LMAO so hard I can barely type,a chore for me at the best of times.You do see what I mean? This is as rediculous as if they claimed it was made of moon cheese!
So what is it that is supposed to make this uberchip (modified pIII actually) so damn fast.Well,its not the L2 cache.Check out places like tomshardware.com.They benchmark the new dothan chips and found little to no increase in performance from the doubled L2 cache.Actually only a couple of tests benefited and that was only a 5% increase over
the clock increase.If you look at comparisons between amd64 chips with 512k and 1m of cache the difference is about the same.So L2 cache cant cut it.I want everyone to look at those benchmark comparison I just mentioned because I have seen more than one poster here with eyes bugged out over the new 2m L2 cache in the centrino.Dont fall for it,it doesnt make the difference you think.
What about the bus? This is pointless to discuss.Centrino has a 100 mghz (that 400 mghz in intel market speak) bus vs. a 500 mghz (1 ghz in apple market speak) bus with independent read/write traffic for the G5.The G5 beats the centrino here so bad that it is obviously not this that accounts for the insane and unbelievable difference.
That leaves the core itself and its interesting that they share some features in common.G5 has a 16 stage alu and while the length of the banias pipelines have not been published,it is something above the 12 stages of the p6 core it was based on,so similar to the G5.Both have duel pipelines,though g5 also has a special unit for handeling internal "book keeping" instructions.Banias has a micro op fusion feature that sounds similar in principal to the group formation of the G5.Both have advanced branch prediction,though the G5 has a massivly larger bht,three actually.What really kills the x86 architecture is the dearth of registers-only 8 vs. the 32 of ppc.There is nothing here that could make it that much faster.
So whats causing all this? Magical fairy dust? Or maybe its just that intel cant be trusted? Yes I think it might be.
Ok,I'm getting tired now,so lets get to the punchline.If those spec scores represent the actual real world performance of these two processors,then IBM must be the most inept and worthless cpu designer in history.I mean,come on.How can you bring this intel propoganda here and then accuse someone else of giving mac users a bad name?
1) There are almost no reviews on the major review sites for the Pentium M Dothan, so your information there is invalid
2) Dothan vs Powerpc 970 who wins, powerpc totally, not only is it bit by bit a faster, more stable processor, it is faster and next generation, you need to remember that benchmarks are different for both mobile and desktops, i will also say that owning a pentium m the only advantage i like about it is its battery life
3) Despite all premonitions, 64 bit is next generation and has shown significant performance differences even when dealign with 32 bit operations except in the case of the itanium from intel, why would Intel be changing and cancelling all its processor lines for the next year to focus on the development of a 32/64 bit processor x86-64 format which it will follow Amd's lead for a change?
4) i am not denying that dothan or any other version of the pentium m processor are bad, but clock vs clock the g4(mpc7447a processor is a redesigned processor and tweaked core, right now we are looking at the last generation of mpc74xx series g4 processors and they are definetely at their highest end and as i am typing off a pb 1.5 right now, i have been nothing short of impressed with its performance in gaming, final cut pro, photoshop cs and dvd studio, applications which require and support dual processor support and are getting great performance from my SP g4 and there is no doubt that the speed performance between this and my older, but not old pb 1.25 is much greater than stated
5) benchmarks are still not accurate and nothing can change that when comparing pc to mac
6) if there are few or no dothan reviews, how can you say it is the leader in the processor world
7) Intel is not going to be implementing centrino tech in desktops, it is switching over to larger bus speeds and a 32/64 bit processors, due out late next year
8) yep, you aare right, power4 is on its way out, but lets not forget that the next generation of g5 is not going to be power4 based, but power5 and speculation has that old steve jobs might toss it up to g6, but that is a whole different battlefront
9) IBM will be at 65nm by years end, where did you say intel will be? 90nm
big difference
10) no disagreements, i would love to have a g5 powerbook and i will have one when it releases and the g4 is a long lived processor no doubt about it, but lets not forget that every processor has new generations, the g4 has several and intel does as well the pentium m has banias and dothan and probably another 2 or 3 down the road.
Finally: G4 is superior clock vs clock in performance, but not as a necessarily mobile processor, the dothan will kicks its butt in the category that matters, battery life and that is important believe it or not, it will also win on heat, FSB, it will lose in performance, cache, and altivec(believe it or not, altivec is a superior hardware enhancement that because it is implemented is what makes it special, while INtel in its most recent Press Release announced that its going the way of the Dual Core processor to minimize heat/power consumption, so who is the follower now?
Quote:
Originally posted by catbat
I really shouldn't have bothered posting here. That said:
At least Matsu had a point about doing fair
tests of systems as a whole.
mpopkin: While Altivec is perhaps overall better than
SSE2, calling it dual core flies so far from standard
practice in computer architecture I hardly know where to
begin.
SPECint 2000 is the best of a bad bunch for cross-platform
benchmarks. Be definition if you wish to compare different
platforms you need to use cross-platform benchmarks.
Power4 beats Centrino on tasks which are sensitive to
cache (Power 4 systems have huge caches) and I/O but
I'd bet a 2.0 GHz Dothan could beat a Power4 on quite a
few smaller benchmarks. SP FP might be one.
1. Your saying the G4 is faster than Dothan?
Looking at STREAM results a Pentium III is close to a G4,
Centrino's bus is 3 times faster. On top of that it has a larger
cache, much higher clock, a whole bunch of tricks on top,
reported performance comparable to many high-end desktop
chips. If a G4 was a fast as Dothan a G5 would be the fastest
chip in the world, and yet where we can compare the two,
G5 to Dothan, Dothan wins. STREAM is a benchmark
useful for estimating sustained floating point performance
and Dr. McCalpin works on POWER development, so don't
call it biased.
2. Go read some Dothan reviews, it kicks the butt of all
other x86 from Intel at the same clock.
3. Your 64/32 bit comparison is funny. The only case where
a 64 bit chip would outperform a 32 bit chip is where all your
data is 64 bit integers, and even then it would equal out due
to cache pressure and memory bandwidth. 64 bit processors
have never shown substantial performance advantages due
to their register size. If you used packed data, it'll have no real
advantage at all. I suppose if you do crypto all day it'd be worth
it though...
4. It's cut down a fair bit, but the Power4 isn't that great anyhow,
it's getting replaced by IBM now. Power4 performs well because
of the large L3 caches and I/O in the MCM, both absent from PPC 970.
It's great as a server processor but was designed from the outset
to be packaged 4 chip/8 processors at a time. Stand alone it's
competitive but not class leading.
5. So wrong it's not funny. You claim that Intel is using 190 and 230 nm,
never have but I guess those were typos for 180 and 250. FWIW
the vast majority of Intel's production is 130 nm, with 90nm ramping
up now. Both Prescott and Dothan are 90 nm. The majority of Intel's
production will be 90 nm by the second half of this year.
6. Cross-platform benchmarking is difficult but SPECmark and STREAM
are as fair as such benchmarks come.
7. For someone who claims to be involved in research of processor
technology you made a lot of errors, IMHO. FWIW I have no crash
problems with my computers (Mac or PC).
I stand by what I wrote, I like Macs, I'm posting from an iBook right now,
but I don't hate PCs, and I recognise that Centrino (Banias and now
Dothan) is an excellent processor for laptops. It's turning out so
well that we'll see derivatives on the desktop next year. Some of
the performance projections I've seen suggest that those of us who
want a fast but relatively cool and quiet PC will be very happy with
the desktop variants. YMMV.
Performa636CD: If you do some benchmarks it would be good
if you'd post them here.
cuneglasus:
I'd agree that everyone tries to win SPEC benchmarks, but it's still
the best indicator out there, and in many cases things that are good
for SPEC benchmarks also benefit real applications. AMD has submitted
results using other compilers recently, gcc and Pathscale.
Your point about clock-rate against performance seems rather ironic on
a Mac forum. FWIW Itanium2 delivers the best floating point performance,
better than even the P4 at 3.4 GHz, yet runs at 1.5 GHz. Clock is a poor
indicator.
5% for a doubling of cache is about right, cache miss rates generally halve
with doublings, so if the rate was 90% with 1 MB you'd hope for 95%
with 2 MB, 97.5% with 4 MB, 98.75% with 8 MB, and so on. In simple
terms you get diminishing returns as you invest in more cache. 5% going
from 1 MB to 2 MB is quite good IMHO.
There's a difference between physical and architectural registers, the
Centrino may have 8 architectural registers, but the physical number
is greater. IIRC it was 32, might be higher.
To summarise: The G4 is long in the tooth, has a slow FSB, and the
performance is not as high as contemporary mobile x86 processors.
Worse still for Apple (at the moment) is that Dothan is competitive with
current PowerPC desktop processors. This will likely change though
as 970fx ramps up, 975 and 750vx are releases, and the new
e600 and so on arrive.
When the new processor for Apple laptops arrives, Mac 'group-think'
will change and people will admit that the G4 was past its prime.
6) if there are few or no dothan reviews, how can you say it is the leader in the processor world
Here is perhaps the first review of the new Dothans. If you don't know french, you can go directly to the benchmarks and architecture comparisons. It is clear that Dothan performs like an Athlon64 but for much less power consumption.
Quote:
7) Intel is not going to be implementing centrino tech in desktops, it is switching over to larger bus speeds and a 32/64 bit processors, due out late next year
This is much discussed lately. There are many indications that Intel indeed plans to introduce a desktop Pentium-M based chip:
To me, the ease of use advantage that Mac OS X enjoys over Windows has kept me away from Intel (and AMD) chipsets. So I know very little about Intel's chipsets, because it doesn't matter to me how well they work. But as much as I enjoy learning about Intel's new mobile chipset(s), can we please get back on topic?
IMO, G5 PowerBooks are not as far off as it might sound. After all, the 90 nm G5-based Xserve is now shipping. And because a G5 in a PowerBook would presumably be clocked lower than its equivalent in a desktop PowerMac, and could be clocked down further for mobile use, I don't think power consumption or heat are insurmountable issues. I seriously don't think we'll wait for a PowerBook G5 as long as we waited for the Titanium PowerBook G4 to arrive. But, until Apple actually ships a PowerBook G5, who knows?
Nice little post Matsu. Unfortunately only relavant to that handfull of people that run photoshop exclusively.
In any event it is a bit of retro defense like back when the G4 towers where all Apple had for the desktop. The G4 portables are being left behind for the same reasons a lack of significant clock rate increases. There is also the upcoming issue of a failure to shrink the process fast enough to stay a head on the basis of power usage.
Now maybe this is not as dark as the dark days of the G4 tower but it isn't getting any brighter. The problem is denying how far Apple is behind the curve does not make it any better. A couple of weeks ago I thought a 2GHz G4 or G5 in a laptop would be competitive, now I'm of the impression that Apple needs even more than that.
I've always said that benchmarks are meaningless. YOu use the system, the speed (or lack thereof) that you may experience is on that level only -- how long does it take to do said task -- and so that is the only relevant test.
CPU benchmarks may give some indication of speed differences on a common platform, but even that is less than ideal. If I have a system with a "faster" CPU against a system with a better disk system, more/faster RAM, a better GPU etc etc (both wintels!) and the slower CPU actually runs a system that gets things done faster, then you can't argue that X machine is faster than Y machine just becasue it has a faster CPU.
Nowhere did I say that Macs were as fast/faster, only that there's no way to establish that based on benchmarks.
You gotta line-up and test 'em out.
Which gets us to the question of mobile platforms. There are other factors that slow down a laptop besides the CPU. This isn't relevant just to photo shoppers, but any user. Get the relevant software in your chosen disciplines and run a timed test.
The results give a far better indication of which system is faster.
Funny, I was using my 2" thick Wallstreet today at the library desk, and a lady came up and asked if it was a new G5 PowerBook or an iBook that I had. I told her no, it's an older PowerBook G3, and that there weren't any PowerBooks with G5s in them yet. She seemed surprised since this was a design she had apparently never seen before.
I love this little thing! Well, BIG thing. Same footprint as a 14" iBook, and 2" thick. Never been asked if it was a G5 though.
Funny, someone asked if my power adapter was some kind of modem a few days ago. Nothing like a genuine Apple-designed machine to turn people's heads!
I know from experience with dealing with Intel development that they are switching over to dual core and amd64 bit designs. Here is the difference that you need to understand, we are dealing with the G4 and the Pentium M dothan, you brought in the powerpc 970 which in all cases beats the pentium m dothan in performance, speed and ability. No doubt there and nothing you can say will change that
There is no disputing that the dothan is a good chip and i have already outlined where its advantages are in powermanangement and battery life, but not performance, bit by bit it is still based on the same series of cores that make up the current generation of 32 bit intel processors. It is powerful and i have no doubt that the 2ghz dothan is probably faster than the g4 at 1.5 ghz, however the g4 processor 3 years ago had a 600 mhz equivalency factor with intel processors and now has a 100-200 mhz factor, this factor is basically the g4 runs at 800 mhz it is the equivalent of x mhz in pentium 4/m/III processors and amd as well, it has shrunk to almost non existence, but when i am reviewing a processor's performance do not forget that the 1.5 ghz g4 and the dothan m processor are both brand new and speed for speed running at equal clock, the g4(mpc7447a) will run faster and perform better, but burn more power and run hotter than the dothan, these are tradeoffs to me
3) DO not forget that the AMD64 and IA64 are totally different architectures, the AMD 64 is x-86-64, which gives native 32 bit as well as 64 bit processes, IA64 is strictly 64 bit and will never be seen in a consumer desktop because of price and cost difficulties
4) The Power5 is a significant jump in performance from the power4 core and has been completely redesigned, very much the way that the itanium 2 has been redesigned from the itanium 1, which has been around for a while
5) Do not forget that Intel does not have a large market for the Itanium because, whoa, Ibm is the market leader in Servers, Enterprise ready solutions and Business Servers and has named Linux and 64 Power5 processors as its next generation and primary processor that powers it. The powerpc 980 which is yet to be released is goign to be based on the power5 core, do not forget that the power4 and power5 are Dual Core processors that are significantly more expensive and more powerful than even the itanium 2,
6) Ibm is suffering losses at its fab in NY, because it is NEW and has yet been pumped to full volume, when you open a new fab it takes between 2-5 years for it to begin producing at full capacity with minimal failures/defects and this particular factory has only been open approx 2 years and is working at very diminished capacity.
Both processors are good, but clock vs clock the G5 crushes the dothan and the g4 beats it in some categories and loses in others, but the g5 loses in none to the dothan.
Quote:
Originally posted by catbat
My final post FWIW. We could undoubtably keep this up
Comments
Originally posted by mpopkin
I use both wintel and Apple Laptops and let me tell you one thing, my new PB 1.5 ghz is the fastest laptop i have ever had and use a pentium m at 1.6 ghz and it is not even close in performance or speed.
You may have some system problem then. The Pentium-M 1.6 GHz should be noticeably faster than the new 1.5 GHz G4. Could you do me a favor and go to this thread and run the CPU benchmark described therein? I would like to see the results for both Pentium-M 1.6 GHz and G4 1.5 GHz.
Originally posted by PB
You may have some system problem then. The Pentium-M 1.6 GHz should be noticeably faster than the new 1.5 GHz G4. Could you do me a favor and go to this thread and run the CPU benchmark described therein? I would like to see the results for both Pentium-M 1.6 GHz and G4 1.5 GHz.
You are aware that this test is a very narrowly defined fp task and not in any way indicative of overall cpu performance?
Originally posted by cuneglasus
You are aware that this test is a very narrowly defined fp task and not in any way indicative of overall cpu performance?
lol yeah seriously~ :P
Matt
It really would not be an accurate benchmark and the only real way to test the machines is to use digital photography utilies and editing utilities such as final cut and photoshop but final cut is a mac only solution
Matt
Originally posted by PB
You may have some system problem then. The Pentium-M 1.6 GHz should be noticeably faster than the new 1.5 GHz G4. Could you do me a favor and go to this thread and run the CPU benchmark described therein? I would like to see the results for both Pentium-M 1.6 GHz and G4 1.5 GHz.
Actually the Pentium m 1.6 ghz is not superiori technologically or physically because the g4 used in the 1.5 ghz powerbook is a totally redesigned processor like the pentium m but is using a dual core design setup (altivec on
No it's not, by any reasonable definition it's single core.
Also Altivec is a SIMD FPU, just like SSE2 as found
on the Pentium M, it's not some secret Apple voodoo.
FWIW the 1.6 GHz Pentium M has a better SPECint 2000
base score (1206) than the 2.0 GHz PPC 970. You may
not believe it but the Pentium M, and Dothan in particular,
is an excellent CPU. An on-die 2MB L2 cache and a 400 MHz
FSB, compared to the anemic G4 should make that blindingly
obvious to all but the most partisan. Intel's not planning
on using Pentium M follow-ons for their future desktop
CPUs just for the hell of it.
top) rather than a single core pentium m, that should basically answer your question because the pentium m really is not more powerful and it has been known for a while that macs run at faster speeds when clocked at the same mhz as their pentium counterparts.
Matt
Absolute rubbish, Coppermine PIII had an excellent IPC.
In fact the whole range of Intel processors with the P6
microarchitecture delivered good to great performance
per cycle.
Posting the kind of things you did gives the rest of us
Mac owners a bad name. Macs are great but can the BS.
Especially the benchmarking stuff.
To an extent, so is the CPU, the OS, the app...
The only thing that matters is the integration.
Put Photoshop on the PB up against Photoshop on a PC. Select a series of actions, see which does better. Select more actions, repeat.
The timed test is the only thing that counts.
Furthermore, it shouldn't even be time test to app, but timed test to task.
If an App exists on one platform and not on another, a good test is the best available on each against the best of the other.
That's how you compare systems.
So,
propose 2d, 3d, video, audio, coding, etc etc suite tests
set up the two "platforms"
and test away.
it's the only meaningful answer to the question of which is faster, and by how much.
EVERYTHING factors in, CPU, GPU, RAM, disk speed, OS, application choice. It is utterly pointless to seperate any of it.
The only denominator is price, period. What do you get for what price.
Add it up, line up it up, time it = test.
The rest equals spec whore bullshit.
I would like to see how a PB 1.5 stacks up against a Centrino 1.6/1.7 on like you said a set of tasks on various types of apps, with price the common denominator.
Not just a few cpu and gaming fps comparisons?. At least then well know for sure if we need to be more spec whores or not.
PS. If the PB doesnt do well against the 1.6 Pentium M, then its not gong to look good for Apple when we start seeing the Dothan 2G models start shipping soon.
7) Think before you write: I have been involved in the research and study field of processor technology for the past seven years and have done several projects(non major) that have constructed and tested the core performance of major processors, mainly amd vs intel, but have recently begin studying ibm processes and while the technological leaps have been great, i am a realist and i believe that accuracy must be maintained in presentation and the fact is that your statement saying that the Pentium M shows significant performance scores against G4 Mobile and especially the Powerpc 970 is absolutely invalid and based not on fact, but on opinion and represents the opinion and data supporting the opinion of you, and not fact, benchmarks are not an accurate survey and people know the performance of their machines, i have machines in both worlds i have several PC machines running from the original AMD athlon(love it to death), to the Pentium 4 and Pentium M at 1.6 ghz by the way, i also have a DP G5 at 2ghz and an Apple PB at 1.5 ghz and on these machines i have done extensive testing and in all cases i have seen better performance by my Apple machines, not only do they not crash, they take zero time to load up, zero install time and zero processing time, i get done what i want when i want, but when using my pc's i face processing times, crashes and stability issues with major programs that i find to be essential especially photoshop, spss and avid which the same or competiting versions work better on mac
Originally posted by catbat
No it's not, by any reasonable definition it's single core.
Also Altivec is a SIMD FPU, just like SSE2 as found
on the Pentium M, it's not some secret Apple voodoo.
FWIW the 1.6 GHz Pentium M has a better SPECint 2000
base score (1206) than the 2.0 GHz PPC 970. You may
not believe it but the Pentium M, and Dothan in particular,
is an excellent CPU. An on-die 2MB L2 cache and a 400 MHz
FSB, compared to the anemic G4 should make that blindingly
obvious to all but the most partisan. Intel's not planning
on using Pentium M follow-ons for their future desktop
CPUs just for the hell of it.
Absolute rubbish, Coppermine PIII had an excellent IPC.
In fact the whole range of Intel processors with the P6
microarchitecture delivered good to great performance
per cycle.
Posting the kind of things you did gives the rest of us
Mac owners a bad name. Macs are great but can the BS.
by late august i should be able to test the same PB 1.33 against a different thinkpad with equal ram and 1.6 Pentium M. there, i should have a bunch of adobe programs and possibly a game or two. if i'm still interested maybe i'll test them out
Originally posted by mpopkin
snipped what appeared to be long gibberish
uh huh.
That might have been the most horribly formatted thing I've seen in quite a while here.
Originally posted by mpopkin
i have seen better performance by my Apple machines, not only do they not crash, they take zero time to load up, zero install time and zero processing time
While the stability I have no dispute with, taking zero time to do anything seems a bit excessive, don't you think?
Also, while a 1.5 GHz G4 may be comparable to a 1.6 GHz Dothan, I'm not so sure that it would be that comparable to a 2.0 GHz Dothan.
Originally posted by mpopkin
in no registered or valid cases has any mobile processor, intel, ibm or motorola beaten a desktop processor at equal or greater speeds in benchmarks or whatnot because the comparable technology between a mobile and a desktop is completely different
Are you kidding me? You're telling me that a 1.6 GHz Pentium M wouldn't beat a 1.8 GHz desktop P4?
http://anandtech.com/mobile/showdoc.html?i=1800&p=13
Originally posted by catbat
No it's not, by any reasonable definition it's single core.
Also Altivec is a SIMD FPU, just like SSE2 as found
on the Pentium M, it's not some secret Apple voodoo.
FWIW the 1.6 GHz Pentium M has a better SPECint 2000
base score (1206) than the 2.0 GHz PPC 970. You may
not believe it but the Pentium M, and Dothan in particular,
is an excellent CPU. An on-die 2MB L2 cache and a 400 MHz
FSB, compared to the anemic G4 should make that blindingly
obvious to all but the most partisan. Intel's not planning
on using Pentium M follow-ons for their future desktop
CPUs just for the hell of it.
Absolute rubbish, Coppermine PIII had an excellent IPC.
In fact the whole range of Intel processors with the P6
microarchitecture delivered good to great performance
per cycle.
Posting the kind of things you did gives the rest of us
Mac owners a bad name. Macs are great but can the BS.
I think what some have been trying to say here,maybe not successfully,is that altivec is an independant unit whereas mmx/sse are wound up with the fpu pipelines.Same idea,much better implimentation on the ativec side.And by the way,these simd units handle integer as well as floating point code.
As for the p6 core having a high ipc,well it all depends what you compare it to.Certainly it has compared to the p4.
The rest of your post highlights just how much bs misinformation and downright paranoia there is in the mac community regarding x86 processors and their performance.
Let me make this as clear as possible-Spec is a highly suspect benchmark as it is but the scores intel reports useing their trick compiler are a joke.It's a public secret that intels compiler is tweaked to get unreal spec scores-why do you think AMD uses it to perform their own posted scores? They are not in any way related to real world performance.Let me quote from a cnet article out last year:
"Peter Glaskowsky, editor-in-chief of Microprocessor Report, said a company could get better benchmark results using a Dell machine with Intel and Microsoft compilers than with a Linux machine and GCC compiler. However, he also noted that Intel's chips perform disproportionately well on SPEC's tests because Intel has optimized its compiler for such tests. "
But lets play a game of reductio ad absurdum.Lets assume these scores are correct and actually mean something.Apple reported a score of 800 in specint.Intel claims a score of 1206.Thats over 50% faster for a chip that is only 80% the clock rate.Ok,at this point I'm LMAO so hard I can barely type,a chore for me at the best of times.You do see what I mean? This is as rediculous as if they claimed it was made of moon cheese!
So what is it that is supposed to make this uberchip (modified pIII actually) so damn fast.Well,its not the L2 cache.Check out places like tomshardware.com.They benchmark the new dothan chips and found little to no increase in performance from the doubled L2 cache.Actually only a couple of tests benefited and that was only a 5% increase over
the clock increase.If you look at comparisons between amd64 chips with 512k and 1m of cache the difference is about the same.So L2 cache cant cut it.I want everyone to look at those benchmark comparison I just mentioned because I have seen more than one poster here with eyes bugged out over the new 2m L2 cache in the centrino.Dont fall for it,it doesnt make the difference you think.
What about the bus? This is pointless to discuss.Centrino has a 100 mghz (that 400 mghz in intel market speak) bus vs. a 500 mghz (1 ghz in apple market speak) bus with independent read/write traffic for the G5.The G5 beats the centrino here so bad that it is obviously not this that accounts for the insane and unbelievable difference.
That leaves the core itself and its interesting that they share some features in common.G5 has a 16 stage alu and while the length of the banias pipelines have not been published,it is something above the 12 stages of the p6 core it was based on,so similar to the G5.Both have duel pipelines,though g5 also has a special unit for handeling internal "book keeping" instructions.Banias has a micro op fusion feature that sounds similar in principal to the group formation of the G5.Both have advanced branch prediction,though the G5 has a massivly larger bht,three actually.What really kills the x86 architecture is the dearth of registers-only 8 vs. the 32 of ppc.There is nothing here that could make it that much faster.
So whats causing all this? Magical fairy dust? Or maybe its just that intel cant be trusted? Yes I think it might be.
Ok,I'm getting tired now,so lets get to the punchline.If those spec scores represent the actual real world performance of these two processors,then IBM must be the most inept and worthless cpu designer in history.I mean,come on.How can you bring this intel propoganda here and then accuse someone else of giving mac users a bad name?
At least Matsu had a point about doing fair
tests of systems as a whole.
mpopkin: While Altivec is perhaps overall better than
SSE2, calling it dual core flies so far from standard
practice in computer architecture I hardly know where to
begin.
SPECint 2000 is the best of a bad bunch for cross-platform
benchmarks. Be definition if you wish to compare different
platforms you need to use cross-platform benchmarks.
Power4 beats Centrino on tasks which are sensitive to
cache (Power 4 systems have huge caches) and I/O but
I'd bet a 2.0 GHz Dothan could beat a Power4 on quite a
few smaller benchmarks. SP FP might be one.
1. Your saying the G4 is faster than Dothan?
Looking at STREAM results a Pentium III is close to a G4,
Centrino's bus is 3 times faster. On top of that it has a larger
cache, much higher clock, a whole bunch of tricks on top,
reported performance comparable to many high-end desktop
chips. If a G4 was a fast as Dothan a G5 would be the fastest
chip in the world, and yet where we can compare the two,
G5 to Dothan, Dothan wins.
useful for estimating sustained floating point performance
and Dr. McCalpin works on POWER development, so don't
call it biased.
2. Go read some Dothan reviews, it kicks the butt of all
other x86 from Intel at the same clock.
3. Your 64/32 bit comparison is funny. The only case where
a 64 bit chip would outperform a 32 bit chip is where all your
data is 64 bit integers, and even then it would equal out due
to cache pressure and memory bandwidth. 64 bit processors
have never shown substantial performance advantages due
to their register size. If you used packed data, it'll have no real
advantage at all. I suppose if you do crypto all day it'd be worth
it though...
4. It's cut down a fair bit, but the Power4 isn't that great anyhow,
it's getting replaced by IBM now. Power4 performs well because
of the large L3 caches and I/O in the MCM, both absent from PPC 970.
It's great as a server processor but was designed from the outset
to be packaged 4 chip/8 processors at a time. Stand alone it's
competitive but not class leading.
5. So wrong it's not funny. You claim that Intel is using 190 and 230 nm,
never have but I guess those were typos for 180 and 250. FWIW
the vast majority of Intel's production is 130 nm, with 90nm ramping
up now. Both Prescott and Dothan are 90 nm. The majority of Intel's
production will be 90 nm by the second half of this year.
6. Cross-platform benchmarking is difficult but SPECmark and STREAM
are as fair as such benchmarks come.
7. For someone who claims to be involved in research of processor
technology you made a lot of errors, IMHO. FWIW I have no crash
problems with my computers (Mac or PC).
I stand by what I wrote, I like Macs, I'm posting from an iBook right now,
but I don't hate PCs, and I recognise that Centrino (Banias and now
Dothan) is an excellent processor for laptops. It's turning out so
well that we'll see derivatives on the desktop next year. Some of
the performance projections I've seen suggest that those of us who
want a fast but relatively cool and quiet PC will be very happy with
the desktop variants. YMMV.
Performa636CD: If you do some benchmarks it would be good
if you'd post them here.
cuneglasus:
I'd agree that everyone tries to win SPEC benchmarks, but it's still
the best indicator out there, and in many cases things that are good
for SPEC benchmarks also benefit real applications. AMD has submitted
results using other compilers recently, gcc and Pathscale.
Your point about clock-rate against performance seems rather ironic on
a Mac forum.
better than even the P4 at 3.4 GHz, yet runs at 1.5 GHz. Clock is a poor
indicator.
5% for a doubling of cache is about right, cache miss rates generally halve
with doublings, so if the rate was 90% with 1 MB you'd hope for 95%
with 2 MB, 97.5% with 4 MB, 98.75% with 8 MB, and so on. In simple
terms you get diminishing returns as you invest in more cache. 5% going
from 1 MB to 2 MB is quite good IMHO.
There's a difference between physical and architectural registers, the
Centrino may have 8 architectural registers, but the physical number
is greater. IIRC it was 32, might be higher.
To summarise: The G4 is long in the tooth, has a slow FSB, and the
performance is not as high as contemporary mobile x86 processors.
Worse still for Apple (at the moment) is that Dothan is competitive with
current PowerPC desktop processors. This will likely change though
as 970fx ramps up, 975 and 750vx are releases, and the new
e600 and so on arrive.
When the new processor for Apple laptops arrives, Mac 'group-think'
will change and people will admit that the G4 was past its prime.
2) Dothan vs Powerpc 970 who wins, powerpc totally, not only is it bit by bit a faster, more stable processor, it is faster and next generation, you need to remember that benchmarks are different for both mobile and desktops, i will also say that owning a pentium m the only advantage i like about it is its battery life
3) Despite all premonitions, 64 bit is next generation and has shown significant performance differences even when dealign with 32 bit operations except in the case of the itanium from intel, why would Intel be changing and cancelling all its processor lines for the next year to focus on the development of a 32/64 bit processor x86-64 format which it will follow Amd's lead for a change?
4) i am not denying that dothan or any other version of the pentium m processor are bad, but clock vs clock the g4(mpc7447a processor is a redesigned processor and tweaked core, right now we are looking at the last generation of mpc74xx series g4 processors and they are definetely at their highest end and as i am typing off a pb 1.5 right now, i have been nothing short of impressed with its performance in gaming, final cut pro, photoshop cs and dvd studio, applications which require and support dual processor support and are getting great performance from my SP g4 and there is no doubt that the speed performance between this and my older, but not old pb 1.25 is much greater than stated
5) benchmarks are still not accurate and nothing can change that when comparing pc to mac
6) if there are few or no dothan reviews, how can you say it is the leader in the processor world
7) Intel is not going to be implementing centrino tech in desktops, it is switching over to larger bus speeds and a 32/64 bit processors, due out late next year
8) yep, you aare right, power4 is on its way out, but lets not forget that the next generation of g5 is not going to be power4 based, but power5 and speculation has that old steve jobs might toss it up to g6, but that is a whole different battlefront
9) IBM will be at 65nm by years end, where did you say intel will be? 90nm
big difference
10) no disagreements, i would love to have a g5 powerbook and i will have one when it releases and the g4 is a long lived processor no doubt about it, but lets not forget that every processor has new generations, the g4 has several and intel does as well the pentium m has banias and dothan and probably another 2 or 3 down the road.
Finally: G4 is superior clock vs clock in performance, but not as a necessarily mobile processor, the dothan will kicks its butt in the category that matters, battery life and that is important believe it or not, it will also win on heat, FSB, it will lose in performance, cache, and altivec(believe it or not, altivec is a superior hardware enhancement that because it is implemented is what makes it special, while INtel in its most recent Press Release announced that its going the way of the Dual Core processor to minimize heat/power consumption, so who is the follower now?
Originally posted by catbat
I really shouldn't have bothered posting here. That said:
At least Matsu had a point about doing fair
tests of systems as a whole.
mpopkin: While Altivec is perhaps overall better than
SSE2, calling it dual core flies so far from standard
practice in computer architecture I hardly know where to
begin.
SPECint 2000 is the best of a bad bunch for cross-platform
benchmarks. Be definition if you wish to compare different
platforms you need to use cross-platform benchmarks.
Power4 beats Centrino on tasks which are sensitive to
cache (Power 4 systems have huge caches) and I/O but
I'd bet a 2.0 GHz Dothan could beat a Power4 on quite a
few smaller benchmarks. SP FP might be one.
1. Your saying the G4 is faster than Dothan?
Looking at STREAM results a Pentium III is close to a G4,
Centrino's bus is 3 times faster. On top of that it has a larger
cache, much higher clock, a whole bunch of tricks on top,
reported performance comparable to many high-end desktop
chips. If a G4 was a fast as Dothan a G5 would be the fastest
chip in the world, and yet where we can compare the two,
G5 to Dothan, Dothan wins.
useful for estimating sustained floating point performance
and Dr. McCalpin works on POWER development, so don't
call it biased.
2. Go read some Dothan reviews, it kicks the butt of all
other x86 from Intel at the same clock.
3. Your 64/32 bit comparison is funny. The only case where
a 64 bit chip would outperform a 32 bit chip is where all your
data is 64 bit integers, and even then it would equal out due
to cache pressure and memory bandwidth. 64 bit processors
have never shown substantial performance advantages due
to their register size. If you used packed data, it'll have no real
advantage at all. I suppose if you do crypto all day it'd be worth
it though...
4. It's cut down a fair bit, but the Power4 isn't that great anyhow,
it's getting replaced by IBM now. Power4 performs well because
of the large L3 caches and I/O in the MCM, both absent from PPC 970.
It's great as a server processor but was designed from the outset
to be packaged 4 chip/8 processors at a time. Stand alone it's
competitive but not class leading.
5. So wrong it's not funny. You claim that Intel is using 190 and 230 nm,
never have but I guess those were typos for 180 and 250. FWIW
the vast majority of Intel's production is 130 nm, with 90nm ramping
up now. Both Prescott and Dothan are 90 nm. The majority of Intel's
production will be 90 nm by the second half of this year.
6. Cross-platform benchmarking is difficult but SPECmark and STREAM
are as fair as such benchmarks come.
7. For someone who claims to be involved in research of processor
technology you made a lot of errors, IMHO. FWIW I have no crash
problems with my computers (Mac or PC).
I stand by what I wrote, I like Macs, I'm posting from an iBook right now,
but I don't hate PCs, and I recognise that Centrino (Banias and now
Dothan) is an excellent processor for laptops. It's turning out so
well that we'll see derivatives on the desktop next year. Some of
the performance projections I've seen suggest that those of us who
want a fast but relatively cool and quiet PC will be very happy with
the desktop variants. YMMV.
Performa636CD: If you do some benchmarks it would be good
if you'd post them here.
cuneglasus:
I'd agree that everyone tries to win SPEC benchmarks, but it's still
the best indicator out there, and in many cases things that are good
for SPEC benchmarks also benefit real applications. AMD has submitted
results using other compilers recently, gcc and Pathscale.
Your point about clock-rate against performance seems rather ironic on
a Mac forum.
better than even the P4 at 3.4 GHz, yet runs at 1.5 GHz. Clock is a poor
indicator.
5% for a doubling of cache is about right, cache miss rates generally halve
with doublings, so if the rate was 90% with 1 MB you'd hope for 95%
with 2 MB, 97.5% with 4 MB, 98.75% with 8 MB, and so on. In simple
terms you get diminishing returns as you invest in more cache. 5% going
from 1 MB to 2 MB is quite good IMHO.
There's a difference between physical and architectural registers, the
Centrino may have 8 architectural registers, but the physical number
is greater. IIRC it was 32, might be higher.
To summarise: The G4 is long in the tooth, has a slow FSB, and the
performance is not as high as contemporary mobile x86 processors.
Worse still for Apple (at the moment) is that Dothan is competitive with
current PowerPC desktop processors. This will likely change though
as 970fx ramps up, 975 and 750vx are releases, and the new
e600 and so on arrive.
When the new processor for Apple laptops arrives, Mac 'group-think'
will change and people will admit that the G4 was past its prime.
all day and never change one anothers mind.
I will read any reply you post though.
1. http://www.x86-secret.com/popups/art...dow.php?id=104
This review is very interesing as it shows how Dothan at 2.0 GHz
is already very competivive with P4C, P4EE and Athlon 64.
Check out the performance when it's overclocked to 2.4 GHz, a
speed that we will likely see desktop variants.
2. Disagree.
3. Disagree. Intel is doing both IA64 and AMD64. Intel recently
showed first silicon from Montecito the next generation Itanium.
The per-core performance is predicted to be twice that of todays
1.5 GHz 6MB L4 Itanium2. Some facts and figures:
1.7 billion transistors.
~ 600 mm^2 IIRC.
24 MB L3 cache. Yes twenty-four megabytes ON CHIP.
Dual core.
SMT.
3 x the bus bandwidth of today ( ~ 20 GB/s).
90 nm IIRC.
Due in 2005.
Virtually guaranteed to be the winner of every significant benchmark.
4. Your opinion, which I will not dispute.
5. It's difficult but not impossible, and the best benchmark
is the end-user's work load.
6. I never said it was the leader in the PC world, just that it
thrashes the G4 and is similar in perfomance to the PPC970, faster
in some areas and slower in others.
7. Two different issues, the team in Israel who developed
Banias and Dothan are working on desktop variants for 2005. Those processors
will probably be AMD64 compatible like Prescott.
easy thing to add, AMD said it only accounted for 5% of the logic in the
Opteron.
8. The major difference of Power5 to Power4 is the inclusion of SMT, just
like Intel has SMT in the P4, and future Itaniums. FWIW IBM seem to have
done a good job, but it's likely to add perhaps at most 30% on latency
sensitive apps, and more like 10% in general. Applications that can speed
up dramatically also tend to be very I/O sensitive. That said it will be
sold as NEW MAGIC SUPER DUPER THREADING.
9. That remains to be seen. FWIW IBM like Intel have had big trouble moving
to 90nm. Cheap 90nm over problematic 65nm will make Intel $$$ whilst IBM
continue to lose $$$ running their fabs.
Finally: Disagree. Dothan is an excellent processor and the follow on will
likely be competitive with any mobile or desktop processors. Above that
IBM will find the Itanium a very formiddable foe.
Originally posted by mpopkin
6) if there are few or no dothan reviews, how can you say it is the leader in the processor world
Here is perhaps the first review of the new Dothans. If you don't know french, you can go directly to the benchmarks and architecture comparisons. It is clear that Dothan performs like an Athlon64 but for much less power consumption.
7) Intel is not going to be implementing centrino tech in desktops, it is switching over to larger bus speeds and a 32/64 bit processors, due out late next year
This is much discussed lately. There are many indications that Intel indeed plans to introduce a desktop Pentium-M based chip:
Intel to reunite notebook, desktop chip architectures
Dothan-based P4 appears on Intel roadmap
Intel to Cancel NetBurst, Pentium 4, Xeon Evolution
IMO, G5 PowerBooks are not as far off as it might sound. After all, the 90 nm G5-based Xserve is now shipping. And because a G5 in a PowerBook would presumably be clocked lower than its equivalent in a desktop PowerMac, and could be clocked down further for mobile use, I don't think power consumption or heat are insurmountable issues. I seriously don't think we'll wait for a PowerBook G5 as long as we waited for the Titanium PowerBook G4 to arrive. But, until Apple actually ships a PowerBook G5, who knows?
Escher
In any event it is a bit of retro defense like back when the G4 towers where all Apple had for the desktop. The G4 portables are being left behind for the same reasons a lack of significant clock rate increases. There is also the upcoming issue of a failure to shrink the process fast enough to stay a head on the basis of power usage.
Now maybe this is not as dark as the dark days of the G4 tower but it isn't getting any brighter. The problem is denying how far Apple is behind the curve does not make it any better. A couple of weeks ago I thought a 2GHz G4 or G5 in a laptop would be competitive, now I'm of the impression that Apple needs even more than that.
Dave
Originally posted by Matsu
[B]That's all meaningless.
Especially the benchmarking stuff.
CPU benchmarks may give some indication of speed differences on a common platform, but even that is less than ideal. If I have a system with a "faster" CPU against a system with a better disk system, more/faster RAM, a better GPU etc etc (both wintels!) and the slower CPU actually runs a system that gets things done faster, then you can't argue that X machine is faster than Y machine just becasue it has a faster CPU.
Nowhere did I say that Macs were as fast/faster, only that there's no way to establish that based on benchmarks.
You gotta line-up and test 'em out.
Which gets us to the question of mobile platforms. There are other factors that slow down a laptop besides the CPU. This isn't relevant just to photo shoppers, but any user. Get the relevant software in your chosen disciplines and run a timed test.
The results give a far better indication of which system is faster.
I love this little thing! Well, BIG thing. Same footprint as a 14" iBook, and 2" thick. Never been asked if it was a G5 though.
Funny, someone asked if my power adapter was some kind of modem a few days ago. Nothing like a genuine Apple-designed machine to turn people's heads!
but last details are important
I know from experience with dealing with Intel development that they are switching over to dual core and amd64 bit designs. Here is the difference that you need to understand, we are dealing with the G4 and the Pentium M dothan, you brought in the powerpc 970 which in all cases beats the pentium m dothan in performance, speed and ability. No doubt there and nothing you can say will change that
There is no disputing that the dothan is a good chip and i have already outlined where its advantages are in powermanangement and battery life, but not performance, bit by bit it is still based on the same series of cores that make up the current generation of 32 bit intel processors. It is powerful and i have no doubt that the 2ghz dothan is probably faster than the g4 at 1.5 ghz, however the g4 processor 3 years ago had a 600 mhz equivalency factor with intel processors and now has a 100-200 mhz factor, this factor is basically the g4 runs at 800 mhz it is the equivalent of x mhz in pentium 4/m/III processors and amd as well, it has shrunk to almost non existence, but when i am reviewing a processor's performance do not forget that the 1.5 ghz g4 and the dothan m processor are both brand new and speed for speed running at equal clock, the g4(mpc7447a) will run faster and perform better, but burn more power and run hotter than the dothan, these are tradeoffs to me
3) DO not forget that the AMD64 and IA64 are totally different architectures, the AMD 64 is x-86-64, which gives native 32 bit as well as 64 bit processes, IA64 is strictly 64 bit and will never be seen in a consumer desktop because of price and cost difficulties
4) The Power5 is a significant jump in performance from the power4 core and has been completely redesigned, very much the way that the itanium 2 has been redesigned from the itanium 1, which has been around for a while
5) Do not forget that Intel does not have a large market for the Itanium because, whoa, Ibm is the market leader in Servers, Enterprise ready solutions and Business Servers and has named Linux and 64 Power5 processors as its next generation and primary processor that powers it. The powerpc 980 which is yet to be released is goign to be based on the power5 core, do not forget that the power4 and power5 are Dual Core processors that are significantly more expensive and more powerful than even the itanium 2,
6) Ibm is suffering losses at its fab in NY, because it is NEW and has yet been pumped to full volume, when you open a new fab it takes between 2-5 years for it to begin producing at full capacity with minimal failures/defects and this particular factory has only been open approx 2 years and is working at very diminished capacity.
Both processors are good, but clock vs clock the G5 crushes the dothan and the g4 beats it in some categories and loses in others, but the g5 loses in none to the dothan.
Originally posted by catbat
My final post FWIW. We could undoubtably keep this up
all day and never change one anothers mind.
I will read any reply you post though.
1. http://www.x86-secret.com/popups/art...dow.php?id=104
This review is very interesing as it shows how Dothan at 2.0 GHz
is already very competivive with P4C, P4EE and Athlon 64.
Check out the performance when it's overclocked to 2.4 GHz, a
speed that we will likely see desktop variants.
2. Disagree.
3. Disagree. Intel is doing both IA64 and AMD64. Intel recently
showed first silicon from Montecito the next generation Itanium.
The per-core performance is predicted to be twice that of todays
1.5 GHz 6MB L4 Itanium2. Some facts and figures:
1.7 billion transistors.
~ 600 mm^2 IIRC.
24 MB L3 cache. Yes twenty-four megabytes ON CHIP.
Dual core.
SMT.
3 x the bus bandwidth of today ( ~ 20 GB/s).
90 nm IIRC.
Due in 2005.
Virtually guaranteed to be the winner of every significant benchmark.
4. Your opinion, which I will not dispute.
5. It's difficult but not impossible, and the best benchmark
is the end-user's work load.
6. I never said it was the leader in the PC world, just that it
thrashes the G4 and is similar in perfomance to the PPC970, faster
in some areas and slower in others.
7. Two different issues, the team in Israel who developed
Banias and Dothan are working on desktop variants for 2005. Those processors
will probably be AMD64 compatible like Prescott.
easy thing to add, AMD said it only accounted for 5% of the logic in the
Opteron.
8. The major difference of Power5 to Power4 is the inclusion of SMT, just
like Intel has SMT in the P4, and future Itaniums. FWIW IBM seem to have
done a good job, but it's likely to add perhaps at most 30% on latency
sensitive apps, and more like 10% in general. Applications that can speed
up dramatically also tend to be very I/O sensitive. That said it will be
sold as NEW MAGIC SUPER DUPER THREADING.
9. That remains to be seen. FWIW IBM like Intel have had big trouble moving
to 90nm. Cheap 90nm over problematic 65nm will make Intel $$$ whilst IBM
continue to lose $$$ running their fabs.
Finally: Disagree. Dothan is an excellent processor and the follow on will
likely be competitive with any mobile or desktop processors. Above that
IBM will find the Itanium a very formiddable foe.