Performance of G4 v. G5 v. P4

Jump to First Reply
Posted:
in Current Mac Hardware edited January 2014
I have become rather confused lately as to the difference of the G4, G5, and P4. Before the G5 came out it was obvious that the G4 was lagging behind, but I remember seeing benchmarks and the dual 1.42 G4 was holding its own with the 3Ghz P4. Now that the G5 is out people are all talking about how much faster it is than the G4. Then I see benchmarks saying that the G5 is really not that much faster than the top of the line PCs. I know a lot of the G4 speed hype was marketing by apple, but how much of it was really? I guess what I want to know is how much of a difference there really is between these CPUs.



A G4 at 1ghz = a P4 at ?

A G5 at 2ghs = a P4 at ?



I have a 1ghz Tibook and it seems just as fast as my friends 2.4ghz P4. It encodes music faster actually. I'm just really confused. I hear from some people that the G4 is slow and old and needs to be replaced by the G5, but It seems fine to me. Then I started thinking that if a 1Ghz G4 is indead considered slow to the PC world how much faster a G5 must be.



Then I hear all this stuff about the slow 167mhz maxbus the G4 has and how it is so bandwidth strived and all. The G5 1Ghz Bus and the DDR400 RAM and how the G4 has all this slow crap. I used a 1.8Ghz G5 the other day here and my University, Auburn, and it was fast but it really didn't do anything much faster than my TiBook. I mean I'm sure it is faster but not as much faster as the numbers would lead you to think. I just want to know what the real differences are between these CPUs.



The new iMacs are they not comparable to the P4s? I mean if they are not and they are actaully over priced and slow... I mean WTF. I heard that the 1.6Ghz G5 was = to a 2.4Ghz P4. But I have used 2.4Ghz P4s and they are really not fast or anything.... actually my dads 867Mhz PowerMacs seems faster.



I'm just confused and want to know WTF is going on. cause if it still takes a 2000 dollar 1.6Ghz G5 to = an 800 dollar 2.4Ghz P4 then apple has not gotten any where and they are still way behind. And if that is the case the bandwidth strived iMacs and PowerBooks must really suck ass. And I'm not even going to talk about the iBook.



I don't really care I will always by Apple. I just want to know what is really going on all BS aside.

Comments

  • Reply 1 of 17
    So far it looks like the G5 is about two times faster than the g4 in most things, and many more times faster in others.



    it is generally on par or right ahead/behind of P4 systems, and Xeon systems.



    But it's too early to decisively tell, because the g5 still needs software that can take advantage of it's architecture.



    there is some stuff out now that is optimized for the g5, but it will still take time before developers learn exactly how capable the machine is.
     0Likes 0Dislikes 0Informatives
  • Reply 2 of 17
    It's almost impossible to answer to your question Algol .

    You canno't say that a P4 is x % as fast as a G4 or G5. It depends on the type of task you ask the computers, and it depends on the degree of optimisation of the software used to perform these tasks.
     0Likes 0Dislikes 0Informatives
  • Reply 3 of 17
    chagichagi Posts: 284member
    Quote:

    Originally posted by Algol

    I have become rather confused lately as to the difference of the G4, G5, and P4. Before the G5 came out it was obvious that the G4 was lagging behind, but I remember seeing benchmarks and the dual 1.42 G4 was holding its own with the 3Ghz P4. Now that the G5 is out people are all talking about how much faster it is than the G4. Then I see benchmarks saying that the G5 is really not that much faster than the top of the line PCs. I know a lot of the G4 speed hype was marketing by apple, but how much of it was really? I guess what I want to know is how much of a difference there really is between these CPUs.



    A G4 at 1ghz = a P4 at ?

    A G5 at 2ghs = a P4 at ?



    I have a 1ghz Tibook and it seems just as fast as my friends 2.4ghz P4. It encodes music faster actually. I'm just really confused. I hear from some people that the G4 is slow and old and needs to be replaced by the G5, but It seems fine to me. Then I started thinking that if a 1Ghz G4 is indead considered slow to the PC world how much faster a G5 must be.



    Then I hear all this stuff about the slow 167mhz maxbus the G4 has and how it is so bandwidth strived and all. The G5 1Ghz Bus and the DDR400 RAM and how the G4 has all this slow crap. I used a 1.8Ghz G5 the other day here and my University, Auburn, and it was fast but it really didn't do anything much faster than my TiBook. I mean I'm sure it is faster but not as much faster as the numbers would lead you to think. I just want to know what the real differences are between these CPUs.



    The new iMacs are they not comparable to the P4s? I mean if they are not and they are actaully over priced and slow... I mean WTF. I heard that the 1.6Ghz G5 was = to a 2.4Ghz P4. But I have used 2.4Ghz P4s and they are really not fast or anything.... actually my dads 867Mhz PowerMacs seems faster.



    I'm just confused and want to know WTF is going on. cause if it still takes a 2000 dollar 1.6Ghz G5 to = an 800 dollar 2.4Ghz P4 then apple has not gotten any where and they are still way behind. And if that is the case the bandwidth strived iMacs and PowerBooks must really suck ass. And I'm not even going to talk about the iBook.



    I don't really care I will always by Apple. I just want to know what is really going on all BS aside.




    All that I can really say is that perceived speed is very relative.



    For example, having low frame rates in a game can render it unplayable, but once you hit 60FPS +, another 40FPS is meaningless as far as what you can actually perceive.
     0Likes 0Dislikes 0Informatives
  • Reply 4 of 17
    pbpb Posts: 4,255member
    Quote:

    Originally posted by Powerdoc

    It's almost impossible to answer to your question Algol .

    You canno't say that a P4 is x % as fast as a G4 or G5. It depends on the type of task you ask the computers, and it depends on the degree of optimisation of the software used to perform these tasks.




    Exactly. Take as example Apple's technology overview of the new powerbooks; especially in BLAST and if we believe the numbers reported are accurate, the competition is completely destroyed. But remember, only for this particular test. Really impressive however if we take into account bus bottleneck in G4. I wonder how things would be if the compilers provided by Apple had autovectorization, at least to help developers in altivec hand-coding.
     0Likes 0Dislikes 0Informatives
  • Reply 5 of 17
    algolalgol Posts: 833member
    Yea, I know that it is impossible to tell the difference between chips and that some chips are better at some things than others. There has to be some general rule though. I mean generally speaking what equals what.



    When there was no G5 people were saying that a G4 is equal to a P4 at like twice the speed, but then I hear that a 1.6Ghz G5 is equal to a 2.4Ghz P4 that is not very impressive. So that means that the G4 must actually be a lot slower. The G4 has to be faster than an equal Mhz P4 cause it has less than half the pipe line length and it has Altivec.



    So when some one goes and buys a 1.25Ghz iMac for 1799 what is the power it has compared to a P4 at what speed. I'm trying to figure out whether Apple has actually got their act together.



    So if you buy a 1800 PC will the iMac be about the same in the benchmarks?



    I just want some idea here. Apple needs to let people know what they are getting. I have been using computers for a long time now and I'm confused. Imagine how confused a new user would be.
     0Likes 0Dislikes 0Informatives
  • Reply 6 of 17
    Quote:

    Originally posted by Algol

    So when some one goes and buys a 1.25Ghz iMac for 1799 what is the power it has compared to a P4 at what speed. I'm trying to figure out whether Apple has actually got their act together.



    What makes this tricky is that the tasks that take the longest (multimedia stuff) are the same tasks that are well-optimized for the G4.



    Many tasks seem to scale about with processor clock speed from G4 to Pentium, but many really don't. Floating-point math intensive stuff will probably be much stronger per clock speed on the G5 than on the P4, and media stuff (MP3 encoding for example) that can be vectorized for AltiVec will run MUCH faster.



    As an example, I've been starting to do a little benchmarking using Maya, which (for those who aren't very familiar with it) is a high-end integrated 3D package that runs on OS X, Windows, and Linux.



    I've been comparing my PB G4 (DVI) 800 to my Pentium 4 3.2 GHz machine and many tasks (particularly rendering and solid body dynamics) seem to be in the 4-5 times faster range on the P4, which is about what I'd expect.



    However, another test I've tried (involving a wrap deformer for you Maya-heads) is only half the speed on the G4, which is excellent performance. Since wrap deformers are often used in setting up 3D characters and can be very slow, it's a good place for the G4 to be disproportionately fast.



    So, even in one application you can see disparate benchmark results.



    For many common tasks, the fact is that the G4 is going to be slower. A lot of this is that Motorola's engineering has simply not been able to keep up with their development roadmap, and the hope is that the G5 will solve this problem for the long run. Still, Apple's put a lot of effort into making the really processor-intensive mass-market stuff (movie and audio processing) run unusually fast on the G4 using AltiVec, so mostly it's really not a usability issue.



    High-end applications like Maya are really where the Mac has fallen behind in the past, but the G5 is, it seems, bringing it up to par. Hopefully when mine arrives I'll be able to see how true that really is.



    -- Mark
     0Likes 0Dislikes 0Informatives
  • Reply 7 of 17
    algolalgol Posts: 833member
    That was rather helpful. So the G4 is more specialized in the areas that it is fast in. As far as usibility the G4 may seem as fast as a much faster P4 and even be faster in some things like MP3 encoding and other processes that use Altivec a lot. However the P4 is faster all around but it has to run at a higher frequancy to be faster. A 1ghz G4 would blow away a 1Ghz P4 right? How fast would the G4 really need to be for it to be comparable to todays pentiums? I know form the benchmarks that the PowerBook G4 is comparable to the PC laptops. The G4 has always been a good portable chip. However i am worried that the iMacs are in a bad position. The 1.6Ghz G5 is 350Mhz faster and has a brand new chip in it but only costs 200 dollars more. Does this make sense?



    If the 1.6Ghz G5 is equal to a 2.4Ghz P4 than the 1800 dollar iMac is equal to what a 1.8Ghz P4 or something? Do they even sell those anymore?
     0Likes 0Dislikes 0Informatives
  • Reply 8 of 17
    Quote:

    Originally posted by Algol

    A 1ghz G4 would blow away a 1Ghz P4 right?



    It would be about even on integer operations and slower on floating-point operations (however the latter would almost entirely be because compilers for the P4 automatically use SSL, which is their AltiVec equivalent, while Mac compilers don't automatically use AltiVec.)



    As for the G5 vs. the iMac, remember that the G5 does not come with a built-in LCD display!!



    -- Mark
     0Likes 0Dislikes 0Informatives
  • Reply 9 of 17
    pbpb Posts: 4,255member
    Quote:

    Originally posted by Algol

    When there was no G5 people were saying that a G4 is equal to a P4 at like twice the speed, but then I hear that a 1.6Ghz G5 is equal to a 2.4Ghz P4 that is not very impressive.



    I think it is not fair to do comparisons like that with a G5, for the simple reason that the G5 is a 64-bit processor. Such a processor can address incredible amounts of memory and do intensive calculations without swapping to the disk, something that would result in huge (and I mean HUGE) perfomance hit in a 32-bit (P4 or G4 for example) processor when it reaches its memory limit. With the commonly used amounts of memory (512 MB or so), don't expect the G5 to show more of its real power than what is evident from its clock frequency and some architectural advantages over the P4 or the G4.
     0Likes 0Dislikes 0Informatives
  • Reply 10 of 17
    http://www.pcmag.com/article2/0,4149,1274138,00.asp



    Here is a benchmark test at PC mag.



    But why is Photoshop so slow on a G5? Perhaps it's time for Apple to get rid of adobe once and for all.
     0Likes 0Dislikes 0Informatives
  • Reply 11 of 17
    well, this is going to just be based off of usage, artilces i've read etc, and shouldn't be taken as gospel at all. HOWEVER:



    1Ghz G3 was about 15% - 25% faster than a PIII, iirc.



    a 1Ghz PIII = 1.4 Ghz P4 (at least initially, the P4 got better by now)



    G4 was about 10% faster than a G3, but with Altivec sprinkled in here and there while using it, probably closer to 25% of overall processor time.



    i have no idea on a G5, haven't used one yet.



    i think it comes out to a 1Ghz G4 being equal to a 2.1Ghz P4, of the initial batches at least. i think there were bus speed improvements, arcitechture improvements etc. that would have shifted that in the P4's favor.



    now it's probably a G4 is approx. 60% - 80% faster per Mhz than a P4.
     0Likes 0Dislikes 0Informatives
  • Reply 12 of 17
    Quote:

    Originally posted by mark_wilkins

    however the latter would almost entirely be because compilers for the P4 automatically use SSE, which is their AltiVec equivalent, while Mac compilers don't automatically use AltiVec



    This is incorrect -- very few compilers do any auto-vectorization of with either SSE/SSE2 or AltiVec. The source of your confusion is probably due to the fact that along with SSE2 Intel added a new way of doing scalar floating point math because the original x86 FPU sucks. The Intel compiler will generate code for those instruction, but that makes the generated code not backward compatible with older x86s so it isn't used that often in practice.





    Comparing processor and system performance is a very tricky business. All processor designs are different and therefore behave differently on different problems, and the systems in which these processors exist impact the performance on different problems in different ways. Of the 3 processors mention in the title, the G4 is the least sophisticated, runs at the lowest clock rate, and has the slowest memory bus which means that on most tasks it will be slower. On things which use the AltiVec unit, or which leverage the large L3 very well, it will typically do very well compared to the P4 (even if the P4 uses its SSE unit because AltiVec is considerably better). The G5, on the other hand, is at about the same sophistication level as the P4 (better in some ways, lesser in others) and its performance tends to be comparable. Due to the details of how the processors are designed, the P4 tends to be better at integer math while the G5 will usually kick its arse at floating point and vector operations. With the best P4 machines memory bandwidth is comparable, although it looks like Apple's high end machine currently has the edge. To the user this means that which is faster will depend on what you are running. In my opinion most of the things that are processor bound are either memory intensive or floating point intensive (especially under MacOS X which uses the FPU more heavily than Windows does). In many situations you're not really waiting on the processor, you're waiting on the operating system... which is why Panther can improve machine performance.



    Even just comparing specific benchmark numbers is fraught with issues -- did the person running the test have the machines configured for optimal performance? Is the software properly optimized for the processor? The G5 is new and software is still being adapted to it.
     0Likes 0Dislikes 0Informatives
  • Reply 13 of 17
    thttht Posts: 6,018member
    Quote:

    Originally posted by Algol

    A G4 at 1ghz = a P4 at ?

    A G5 at 2ghs = a P4 at ?




    In general, for approximations, taking in all factors, from all the benchmarks, the way I feel it is, the equation I use is:



    1 GHz 7457 = 1.1 GHz 7455 (no L3)

    1 GHz 7457 = 1.7 GHz Pentium 4

    1 GHz 7457 = 1.3 GHz Pentium-M

    1 GHz 7457 = 1.4 GHz Pentium 3

    1 GHz 7457 = 1.1 GHz Athlon

    1 GHz 7457 = 0.9 GHz PPC 970 (G5)



    Of course, one can't actually invert any of these equivalences to say that a 1.6 GHz 970 = 3.0 GHz P4. Lots of things come into play. For G5, I'd use:



    1.6 GHz 970 = 2.6 GHz Pentium 4

    1.8 GHz 970 = 3.0 GHz Pentium 4

    2.0 GHz 970 = 3.4 GHz Pentium 4



    Quote:

    I have a 1ghz Tibook and it seems just as fast as my friends 2.4ghz P4.



    Yes, for the vast majority of things, the slowest part of a the computer is the user, then probably the network connection, then the disk access, then the amount of memory, etc. There won't be much difference felt until specific applications are used like games and media intensive tasks that take awhile to do. When it comes to these, specific parts of the processors come into play.



    Quote:

    It encodes music faster actually. I'm just really confused.



    It's normal for one processor to be faster at something compared to another processor. It just depends on what. For Apple, they've been relying on the Velocity Engine to keep them in the game for awhile. If it wasn't for the Velocity Engine, the G4 wouldn't be faster in many applications or operations at all.



    Quote:

    Then I hear all this stuff about the slow 167mhz maxbus the G4 has and how it is so bandwidth strived and all. ... I used a 1.8Ghz G5 the other day here and my University, Auburn, and it was fast but it really didn't do anything much faster than my TiBook. ... I just want to know what the real differences are between these CPUs.



    On a clock for clock basis with the G4, the G5 would be, should be:



    1.0x as fast on integer operations (Photoshop, Finder, etc.)

    1.5x as fast on floating point ops (Matlab, CFD codes, etc.)

    1.0x as fast on cache bound velocity engine ops (code cracking, certain small media stuff)

    1.3x as fast on memory bound velocity engine ops (HD video?, large matrix ops?)



    It's actually more complicated.



    Quote:

    The new iMacs are they not comparable to the P4s? I mean if they are not and they are actaully over priced and slow... I mean WTF.



    Overpriced, yes. Probably about $200 more than they should be. Perhaps, the iMac form factor is worth the extra $200?



    Slow? That's complicated.



    Quote:

    I heard that the 1.6Ghz G5 was = to a 2.4Ghz P4. But I have used 2.4Ghz P4s and they are really not fast or anything.... actually my dads 867Mhz PowerMacs seems faster.



    This is an user experience illusion. Sort of like your 1 GHz Powerbook experience compared to a 2.4 GHz P4. If compared on specific applications, the 2.4 GHz P4 should be faster on a majority of things. But, if in what you do, the P4 does something in 2.1 seconds while the G4 does something in 1.6 seconds, it doesn't really matter that much. If you do it hundreds of time a day, it would, but that's probably rare for consumers.
     0Likes 0Dislikes 0Informatives
  • Reply 14 of 17
    Okay, I want to make one thing clear. I understand that different CPUs are better at different things and that it is hard to say for certain what's faster than what etc. I am trying to make a point with all this. Joe Bob goes to buy a computer. He looks at a 1400 Dell with a 2.4GHz P4, then he looks at the 2000 G5 with a 1.6GHz G5, then he looks at the 1800 iMac with a 1.25Ghz G4. I understand that the G5 is about as fast as the P4 and that the G4 is slower in most things but you get a lot of cool software and a superdrive etc which the PC doesn't have. That the iMac is fast enough for what its for and that it looks a lot better and has a beautiful screen etc. However, a lot of people try to get the best deal and if a normal consumer has no idea what = what he will have a hard time deciding what to buy.



    I am trying to point out that no one really knows for sure what does = what. All of you are coming up with ideas but none of you really know, and this is because it is very complicated with many different varables. However, I think apple should come up with a way of showing people what is = to what. Like on the box of the 1.6Ghz G5 it should say somewhere 1.6Ghz G5 about = to 2.6Ghz P4. On the iMac in should say 1.25Ghz G4 about = to 1.8Ghz P4 or something. They could make the numbers better than they are after all cause no one really knows, but it would help the new buyers understand what they are getting.



    I think that for the consumer anything over a 1Ghz G4 is about as fast as you can notice a difference while using OS X. When 10.3 comes out this may not be the case. Windows has always been slow because of the bloated code but I imagine that a PC user can't tell the difference between a 2.4Ghz and a 3.2Ghz P4.



    Anyway I've got to go eat... or I'll be late to class.
     0Likes 0Dislikes 0Informatives
  • Reply 15 of 17
    This is why Apple has been trying to market features rather than straight performance. Superdrive vs combo drive, for example. They have 3 machines, each faster and more capable. The iMac isn't sold based on performance, its sold based on its form factor. Not everybody should care about performance, and many shouldn't care about it at all -- other factors are more important. Unfortunately the PC world is all about MHz MHz MHz MHz.
     0Likes 0Dislikes 0Informatives
  • Reply 16 of 17
    Quote:

    Originally posted by Programmer

    This is incorrect -- very few compilers do any auto-vectorization of with either SSE/SSE2 or AltiVec. The source of your confusion is probably due to the fact that along with SSE2 Intel added a new way of doing scalar floating point math because the original x86 FPU sucks.



    So you're saying this was added along with SSE2 but is not itself part of SSE2? OK, fine with me, same difference.



    -- Mark
     0Likes 0Dislikes 0Informatives
  • Reply 17 of 17
    Quote:

    Originally posted by mark_wilkins

    So you're saying this was added along with SSE2 but is not itself part of SSE2? OK, fine with me, same difference.



    -- Mark




    It is part of SSE2, but its not part of the vector functionality and so not comparable to AltiVec.
     0Likes 0Dislikes 0Informatives
Sign In or Register to comment.